Gw Instek GDS2204E (200MHz 4 channel DSO) review

[nctnico]

GW Instek GDS2204E review and testing

The reason I was looking for an oscilloscope is because I wanted an extra oscilloscope to bring to customers. No way I'm going to bring my Agilent DSO7104 to some factory floor! Especially since my other oscilloscope fell from a table during some tests. My basic requirements are 200MHz-ish, reasonably deep memory, serial protocol decoding (including CAN) and easy to transport.

So time to grab something cheap... I have been looking at other (used) oscilloscopes as well like the Agilent/Keysight DSOX2k or 3k series. I didn't like the R&S and Lecroy offerings because while the oscilloscopes are not terribly expensive you have to pay through the nose to get the serial decoding options. Rohde & Schwarz charges over 3000 euro (ex. VAT!) for UART, CAN, SPI and I2C for their RTM1500 series; Lecroy is slightly cheaper but still in the same ballpark. At least the Agilent/Keysight scopes have the possibility to liberate the options or spend around $800 on an option bundle. Rigol would have be an option too but lack of cheap 4 channel >=200MHz models, decoding only what is on screen (utterly useless!) and slow bug fixes put me off. The same goes even more for Siglent which has been  struggling for two years already to get their SDS2000 series oscilloscopes on the rails and there is still no light at the end of the tunnel.

So the search went on... After looking around on various websites the GW Instek GDS2204E caught my eye. The GDS2204E (which is the top model in the GDS2000E range) is a 4 channel 200MHz oscilloscope which comes standard with I2C, SPI and CAN/LIN decoding, LAN (ethernet), USB, segmented recording for around 1290 euro / $1450 (ex. VAT). Now that looks like a good deal! I am not going to take it apart because Dave already did some teardowns of some GW Instek scopes and I doubt the GDS2204E will be much different (A Xilinx Zync + DRAM board and analog front-end) . Besides that I have it for evaluation and at the first sign of serious problems it goes back! I got my GDS2204E from a local dealer called Eleshop. Oddly enough it got shipped to me from Good Will Instruments Europe directly which turned out to be located in the Netherlands. Unfortunately they did not install the latest firmware on it (V1.20 at the moment of receiving it). Also the calibration certificate is dated from January 2015 so this scope has been in their warehouse for a while! Updating the firmware is described in the firmware update guide but it is not very clear. After the firmware update the scope says it needs to be turned on/off but after that the GW Instek logo remains on the screen. After 15 minutes I decided to power cycle it again and this time it started. Pfew!



During my testing I ran into some issues and GwInstek has send me version 1.21b to test with. The update from 1.20 to 1.21b2 went smoother; the oscilloscope started by itself after a short while. After I found a few more issues they send met version 1.21b6 to test with. In the end this review went through a few rewrites to update it with the fixes.  Also the update manual is not very clear on whether the update procedure must be run twice (update the kernel and then the firmware itself) or not. Better safe than sorry so I loaded each new firmware file twice.

OK! Let's see how the Taiwanese build oscilloscopes...
First impression: lots of function keys around the screen and it also has a 'force trigger' button. Very handy to get some lines on the screen to help adjusting the trigger level. Others complained about the position of the 'Autoset' button. I think I'll have to agree with them because it is very close to the 'run/stop' and 'single' buttons. To add to the insult the 'Default' button is below the 'single' button. The top-right corner is definitely a danger zone for pressing buttons without looking and messing up the settings because there is no undo button. For the rest the layout of the buttons seems pretty standard. Each channel has it's own control but I don't really care whether each channel has it's own controls or you have to select a channel first. BTW most scopes need extra 'virtual channel' buttons to control math and bus (decoding) traces so you you still have channels without dedicated controls. Some rotating knobs are push buttons (channel offset, trigger level, horizontal position). Pushing them makes the control go back to zero. What GwInstek did right was putting a seperate select/enter button next to the rotating knob (called 'variable') for changing values. On other scopes pushing a rotating knob very often results in selecting a different option. On the GDS2204E the rotating knobs respond well and the accelleration is just right for selecting single items or change a large number quickly. Even the on-screen keyboard (which I tend to hate in general) for entering text and numbers is easy to use compared to other on-screen keyboards I have come across!

The GDS2204E also has the ability to choose whether to expand the signal from the trigger point or from the time-base offset. The selection is slightly hidden in the acquisition menu but it is a very handy feature to zoom in on part of a signal quickly without needing to scroll left & right to get the point where you want to look at on screen again. On other oscilloscopes the way to do this is to set the trigger point somewhere along the horizontal axis. On my DSO7104 I can choose between left, centre or right and on my Yokogawa DL1740 I can set it freely.

Another interesting point are the available languages, As far as I can see the scope supports English, Taiwanese, Chinese, Japanese, Korean and Russian. No German, French, Italian or Spanish which would be important to support for the European market since the people in those countries are notoriously bad at speaking other languages than their own!

Test plan
Based on past experiences I have made a test plan with items I wanted to test. I want to take the GDS2204E through it's passes thouroughly to make sure I don't get burned again by an incomplete and bug infested oscilloscope. Tests in random order:
- Bandwidth / aliasing
- Memory length versus number of channels enabled since this wasn't specified very clearly
- Samplerate versus channels enabled (again not specified very clear)
- FFT function
- Filtering
- Protocol decoding and especially how much of the memory is decoded
- Saving images and data
- Saving / recalling setup
- Cursors
- AM modulation signal distortion problem Dave spotted in the GDS2304A
- Operating / using the scope
- PC software
- Networking abilities
- Signal noise floor
- Fan noise level
- Peak detect and roll mode
- Segmented recording and decoding
- Retaining settings
- Storing waveforms
The secondary goal is to use the oscilloscope for a couple of hours in order to spot stability problems.

Memory / Operating / using the scope
The GDS2204E oscilloscope has 10Mpts per channel which isn't shared and not used for double buffering so each channel really has 10Mpts unless FFT is enabled which can be used up to 1Mpts (I have carefully avoided the word limited here because a 1MPts FFT is anything but limited!). The samplerate is 1Gs/s with up to (any!) 2 channels enabled and 500Ms/s with 3 or more channels enabled. The fan produces a low humming sound but it is not extremely loud but there is also a high pitched whining noise coming from it which could become irritating in a quiet environment. The fan noise level is pretty much similar to my Agilent DSO7000 and doesn't become annoying. Each channel has a 1-2-5 selectable probe division factor with a  range from 0.001x to 1000x (nice!) and the channels can be switched between Volts and Amperes. When set to Amperes the division factor also shows Volt per Ampere. There is also a dedicated function key to select the 10x factor directly as this is the most used value. After playing a little with the various functions and buttons it has become clear to me that the firmware is a general purpose one that GW Instek uses among other oscilloscopes. There are greyed out options for digital channels, function generator options, etc and it seems they can also add extra (paid) options to future firmware versions. The protocol decoders for example are listed as software options. Besides the decoders there are also applications for enabling digital filters on channels, data logging and a voltmeter. One neat feature is that this oscilloscope also has internal storage so you aren't required to bring a USB stick along for storing images or some data. The internal storage is 32MB which is enough for lots of screendumps but not for storing serious amounts of waveform data (10MPts results in a 244MB CSV file per trace). When browsing through files it shows a thumbnail of images. The firmware developers obviously had some time to spare on small details like these! When creating a screendump by pressing the hardcopy button the process is remarkably quick. Not some progress bar showing it is doing an image dump for several seconds. The GDS2204E pops up a message with the filename which has been saved almost immediately after pressing the hard copy button.

Another fun fact is that the frequency for the probe calibrator signal output can be adjusted. Perhaps they guessed it can also serve as a simple signal generator.

After the firmware update I ran the self calibration procedure. There is a calibration signal output BNC for this purpose which needs to be connected to an input when the oscilloscope says so. Additionally there is also an SPC (signal path compensation) procedure which can only be run after the scope has been on for at least 30 minutes.

One quirk so far: when in stop mode the oscilloscope goes into acquisition when the 'zoom' button is pressed. WTF? This doesn't happen if you press the 'pause/playback' button so the 'pause/playback' button is the safe route into zoom mode without starting an acquisition. Oddly enough this only happens with acquisition lengths set to 1Mpts or 10Mpts. According to GwInstek this is due to a limit in the hardware but since the 'pause/playback' button sometimes allows to get into zoom mode while staying in stop mode I'm not so sure... For example the workaround doesn't work when decoding is enabled though.

I also tested peak detect in roll mode and the GDS2204E has no problems showing a 16ns pulse in 2s/div so that is good. On the other hand the roll mode scrolls a bit jerky which gets worse with more channels enabled. The response to controls also slows down when it needs to do calculations on long acquisitions.

Screen layout
The screen layout on the 800x480 display isn't too crowded though I wish they went for a 12x8 or 14x8 graticule with squares instead of 10x8 rectangles; it would make better use of the wide screen. There is a dedicated menu-off button to get rid of all the menus so no drama there. What I like is the frequency counter in the bottom right corner which seems to be based on number of edges the trigger sees. It is always handy to have a quick idea about the frequeny of a signal. This frequency counter keeps working while in stop mode.

Retainging settings
I can be short about this: the settings are retained between power cycles. Even the input filtering settings are retained. Let's mark that as OK.

Bandwidth and acquisition modes
I measured the bandwidth for each channel with the reference level at 10MHz (the generator I used doesn't go lower) and using a BNC T with a 50 Ohm terminator to make the inputs 50 Ohm-ish. All channels have an equal bandwidth of 290Mhz but they start to roll-off from 70MHz so it seems the anti-aliasing filter isn't very sharp. OTOH a sharp anti-aliasing filter will cause more phase shift and peaking so pick your poison. When having 3 or more channels enabled (500Ms/s samplerate) the useable bandwidth is 200MHz. Above that frequency there is all kinds of aliasing but that is to be expected. But there is something odd: with peak detect enabled the 20MHz sine wave gets distorted when having 3 channels enabled! There is a 2ns (1/500MHz=2ns) delay between the distortions so perhaps it is trying to display 2 peak values next to each other as one waveform which isn't going to work. Actually I think the peak-detect mode should disable itself at certain time base settings where there are less samples than (horizontal) pixels on the screen!


When in stop mode and expanding a waveform with peak detect on (in the area of 5ms/div) there are aliasing distortions as if it doesn't do peak-detect at all. Are these phenomena to be expected? Yes, it all has to do with limits of sampling systems! GW Instek has send me some screenshots from a Tektronix scope they use and it shows exactly the same behaviour at low time/div settings with peak-detect enabled. When doing similar tests on my Agilent DSO7104 it also shows similar behaviour but to a lesser extend. I guess at some point a trade-off between screen update speed and going through all the memory has to be made.

Noise level
The minimum volts per division setting is 1mV and with the inputs open the noise is around 0.5mV (using a 50 Ohm terminator makes no difference). When a channel is set to GND the ADC value is clearly overwritten by 0.

[nctnico]

Decoding
As I wrote earlier the GDS2204E can decode UART, CAN, SPI and I2C. First I tried some CAN decoding for which a samplerate of 4 times the bitrate seems to be enough for the decoding to work properly. The decoding works as expected and the GDS2204E does decode the entire memory which can take a little bit of time when the memory length is set to 10Mpts. I think at memory lengths over 1Mpts it is necessary to do the decoding in hardware instead of using the generic CPU in the oscilloscope in order to keep the response snappy. Still it is 1000 times better to do decoding slow than not at all! Because everything is decoded you can stop the acquisition and scroll through the messages and enlarge signal sections where there is a problem with the partial message still on screen so you know where you are. The decoding also works in zoom mode. Excellent! A nice touch is that the decoded information (bus trace) can be moved along the vertical axis with the variable button. I also tested triggering on certain CAN messages based on the ID and that also works as it should be. The same goes for I2C decoding. It is even possible to trigger on certain (multi-byte!) data patterns in the I2C stream with per-bit wild cards. Suddenly I had to debug an SPI problem so instead of the DSO7104 I pulled out the GDS2204E and found SPI decoding works as it should be. There is one catch though: SPI decoding must have a chip select signal! For fun I also looked at the UART decoding and tested how much oversampling is required. It turned out a 38400bps stream needed at least 500ks/s to decode properly (that is 13 times). There is also a packet mode in the UART decoding which concatenates characters into a packet. The end of a packet can be 0x00, CR, LF, space or 0xff. Similar to the I2C triggering the UART triggering can be based on up to ten bytes so it is easy to trigger on specific messages which works as advertised. But unlike the other protocols the UART decoding can also show the messages in ASCII (readable text) and it is also possible to setup the trigger using ASCII. Take notes Rigol & Siglent! However I did find a problem with the packet mode decoding; at some time/div settings the text is placed at the wrong location on the screen. Also the packet mode decoding disables viewing the event table. I'm a bit dissapointed by this because having messages in an event table would make it easy to go through a conversation between 2 devices because every message is on a single line.

Persistence
Persistence can be adjusted from a 16ms to 4s in doubling steps, set to infinite or disabled. In the initial firmware this didn't work properly so I reported this back to GwInstek and they fixed it in the V1.21b6 update!

Cursors
The cursors work slightly different than the oscilloscopes I'm used to. There are two modes: one with only horizontal controls where the vertical values track the selected input and one with seperate horizontal and vertical cursors. Selection between modes (off, horizontal, both) happens by pressing the cursors button repeatedly. There is an on-screen display which shows the values of the cursors. The units can be selected between seconds, Hz, percentage and phase. Additionally to showing the delta between the cursors the dV/dt is also shown so it is easy to see the slew rate of a signal. Unfortunately the cursors cannot be placed outside the screen. In zoom mode the cursors can be used outside the screen but their positions don't track the signals exactly as if the resolution of the cursors is somehow limited so it is not possible to enhance the horizontal cursor resolution much by using the zoom mode. IMHO GwInstek should fix this!

Averaging and high-res
The GDS2204E does not have a high resolution mode but it does have averaging and/or the input filtering to clean up noisy signals. A side note: high-res mode is tricky to use because oversampling a signal while producing a meaningful result requires having enough Gaussian distributed noise and the ADCs to be very linear. On my Tektronix TDS500/700 series scopes I could get the weirdest high-res results with the bandwidth limiting on. My primary use for high resolution and averaging are to clean a signal up to allow for more accurate cursor measurements because of the (much) thinner trace. More about input filtering later on. Unfortunately averaging is only available for 10kpts to 1Mpts memory lengths and together with zoom mode averaging seems to be limited to 100kpts max.

Automated measurements
Ofcourse the automated measurements are there. The gating (which data is used) can be choosen between entire record length, screen or between cursors. Statistics can also be enabled so all in all it is pretty complete.

Input filtering
The input filtering can be selected between low pass or high pass filtering. When enabled the waveform capture rate slows down visibly with long record lengths so the input filtering is obviously a post-processing step performed in software. Still the filtering is very useful to clean a signal up. For example: if a signal has a lot of high frequency components or 50/60Hz hum then the filters can be used to get rid of it. The filters can be adjusted in steps of 1 Hz from 1Hz to 500MHz. Band-pass filtering would have been nice as well but maybe I'm pushing my luck here. On other scopes I use high-resolution mainly to clean up noisy signals in order to make accurate cursor measurements but the input filtering can be used as a much more precise tool for this purpose. And the cool thing is that since input filtering is a post processing function you can change the filter settings in stop mode so you can apply different filter settings to a signal without recapturing it.

FFT
FFT is under the menu which pops up by pressing the math button but FFT is seperated from the other math functions. The FFT works OK and it is easy to adjust. The long (up to 1Mpts memory for FFT) allows to get a fine grained FFT result. It is not a problem to look at an FM modulated signal with a 20MHz carrier and a 100kHz frequency swing. It would have been nice if there was an averaging mode on the FFT signal to stabilise it a bit more but that is a very uncommon feature. So far I have seen this only on the Tektronix TDS500/700 series. To be honest the speed at which the FFT is calculated is pretty quick. I think (part of) the FFT is accellerated in hardware; it certainly feels it is doing 1Mpts FFT faster than my Agilent DSO7000 is doing 128kpts FFT.

Math
There are two math functions. The first is a simple one with +. -. * and / operators while the second one is a complex math function which allows to create a trace based on a math expression. GW Instek has taken this part quite seriously because there are quite a lot of possibilities here to create complex math expressions. I have not ventured deeply into this feature but it looks useful to take on more demanding signal analysis tasks. To start an expression can use data from the traces, reference memory or 2 variables. Besides the usual operators  (+, -, *, /, >, <, etc) the trigoniometric functions (sin, cos, tan, etc), integrator, differentiator and the automatic measurement functions are at your disposal. The two variables I mentioned earlier can be changed from the menu by using the variable knob.

LAN (ethernet) connection and software
The GDS2204E comes standard with an ethernet connection so let's see how useful it is. After enabling DHCP the oscilloscope obtains an IP address immediately. First of all I tried the Openwave software. As usual this doesn't work; it produces a 'can't find device' message or is this only supposed to work with USB? How 2001... The software does not have an installer and looks pretty basic. Just a few buttons in a screen. Perhaps it can be used to do some screen captures??? Actually I didn't expect anything else because remote control software is often just a checkbox item for this kind of equipment (keep some software engineering intern busy for a while between getting coffee for everyone). So lets forget about the Openwave software. The oscilloscope also doesn't have a built in webserver/website. This would have been nice to do quick screen captures and/or offer a platform independent remote control. In 2016 you should be able to control your scope from your smartphone! Then again the build-in webserver in my Agilent DSO7000 would be just as useful if/when Agilent left it out.

In the ethernet settings menu there is also an option to enable a socket server. If I enable it I can telnet to the oscilloscope and it responds to the "*IDN?" SCPI command to query a SCPI compliant device for it's name and other information so I guess this is an SCPI over ethernet interface. I have not tested it further but it could prove useful for automated testing. Through the applications (APP) button it is also possible to mount to a network share. Interesting! But.. what kind of network share? The manual doesn't tell! Trying a Windows networking share works first time and mounting/unmounting a disk for the second time also works (with a Linux server so it if doesn't work with Windows blame uncle Bill!). I managed to save screendumps and a dump from a waveform (a 244MB CSV file for 10Mpts) to the network drive without problems and it is not slow either. Unfortunately a USB stick is no longer recognised after mounting/unmounting a network share. I had to power cycle the scope to get the USB stick back to work again. After some more fiddling it turned out I stumbled onto a very specific sequence involving removing the USB stick and mounting the remote drive which goes wrong. Please no remarks about fools and their ability to break stuff which is supposed to be unbreakable. This is also reported back to GwInstek.

Signal distortion?
Dave shows us a GDS2203A in 'EEVblog #474 - GW Instek GDS-2000A Series Oscilloscope Unboxing & First Impression' ( ) from May 2013. With Dave's AM modulated signal he managed to get all kinds of weird signals on the screen (from 45:00) so I tried the same on the GDS2204E and guess what!? I managed to reproduce them (10MHz carrier + 10kHz 95% AM modulation)!  I informed the dealer that this is a bug I can't live with; in turn they informed GwInstek and a couple of days later (despite it being Christmas time!) GwInstek produced a new version (1.21b2) which fixed this problem. So extra points for Gw Instek for being able to resolve firmware issues quickly!

Segmented recording
I also tested segmented recording. This works as you would expect. There is also the possibility to calculate statistics over a range of recorded segments based on an automatic measurement. My primary interest however was bus decoding combined with segmented recording and see how this would work. Bus decoding combined with segmented recording is an ideal way to debug intermittent failures in communication between devices over a long time period. As expected the bus decoding works when stepping through the segments. However it is not possible to display an event table from all the segments in segmented recording mode. The event table only shows events from the current segment. That is a bit of a let down because a full event table would allow to store all the decoded data in one file for analysis on a PC. It is much easier to go through a lot of data on a PC than on an oscilloscope.

Reference traces
Last but not least I played a bit with the reference traces. Up to 4 reference traces can be stored. These come in handy every now and then to compare an old (reference) trace with a new acquisition or to have various acquisitions on the screen simultaneously. In the GDS2204E the reference traces only work up to 1Mpts because they are stored in flash memory and therefore can be recalled after a power cycle (this is common behaviour for many oscilloscopes). The time/date, length and samplerate of the reference trace can also be recalled. Because the reference trace is written into flash memory it takes a while (5 seconds or so) before a 1Mpts reference trace is saved though. Storing 10Mpts probably was too slow because I can't imagine the flash is too limited to store 40MB of data for 4 reference traces. What is very useful is that the reference traces can be moved and scaled both horizontally and vertically. I don't recall seeing that before on any oscilloscope I have used so far; both my Agilent DSO7000 and Yokogawa DL1740 can't do that. Additionally the reference traces can also be saved to a waveform file, read from a waveform file and/or used for math/FFT functions. All in all the reference traces allow to use the oscilloscope for doing (re)analysis on the data later on. However the protocol decoding and signal filtering can't use the reference traces as a source but I'm already very happy with the possibilities the reference traces offer.

Bugs / short comings
A short recap on the bugs / short comings I have found:
- signal distortion with peak-detect on at 100ns/div or less and with 3 or more channels enabled.
- cursors don't track the signal in zoom mode accurately
- zoom mode has some limits regarding memory length and switching between zoom-mode and normal mode so you have to decide whether you want to use zoom mode before doing a single-shot acquisition.
- Whining noise from PSU

Conclusion
I think the GDS2204E is a very good deal for an oscilloscope in this price class especially since everything works as advertised and it offers a very complete package with lots of clever solutions. The oscilloscope is also very straightforward and easy to use; I did not need the manual at all. The GDS2204E also has a lot of areas where the firmware engineers went the extra mile. Take the dedicated x10 probe function key for example, the possibility to move the bus decoding trace vertically or to be able to use the reference traces for more than just a marker on the screen. A lot of the operations are performed (mainly) in software but it is not particulary slow. I like the fact GwIstek didn't choose to limit functions so everything works snappy but instead they went for getting the maximum functionality from the hardware even if it means the user has to wait a little while in some cases. In the end you want a tool which gets the job done!

Anyone considering buying a Rigol, Siglent, Hameg (R&S) or other economy oscilloscope should really REALLY look at GwInstek's offerings first as it seems that GwInstek has a good grip on their software and offers good value for money! I'm surprised Rigol and Siglent have so much traction on this forum; GW Instek is almost like a hidden gem. With my software engineering hat on I have the impression the engineers at GwInstek started with realistic goals for the firmware so it could be completed and tested within a reasonable timeframe. Better have 9 functions implemented correctly than 10 functions being buggy/incomplete! I'll admit there where some bugs in the firmware when I got it initially so firmware testing could be better but it seems GwInstek is able to come up with fixes quickly once they know what is wrong. They send me 2 updates in a 4 week timeframe. I also think the GwInstek GDS2204E gives the lower bandwidth Keysight DSOX2k and DSO3k series a run for their money due to the much larger memory and standard decoding options. For example a serious downside of Keysight's scopes in general is that in a common usage scenario their memory is only one quarter of what is specified due to the memory being shared between channels and cut in half again by double buffering. Worse: according to the DSOX2k user manual the memory gets even shorter when reference traces are turned on. You never really know what kind of memory depth your Keysight scope is actually using. The GDS2204E on the other hand is very open about this and has enough memory under the hood to store a minimum of two 10Mpts segments. When debugging a serial bus problem you really want as much memory (segments) as possible. I'm wondering if/when GwInstek comes up with a version of the GDS2000E series with digital channels.

Now the main question is: do I keep it? The answer is yes (due to the V1.21b6 firmware!). I have tested it thouroughly and some of my comments are close to nit-picking because my other (A-brand!) oscilloscopes turn out to be just as bad. Also this oscilloscope packs a few clever features which will come in handy for certain measurements. Two thumbs up! 

Note: I had to cut the posting in two parts because appearantly the limit for one posting is 20000 characters. No, I did not type all of this just now; this posting has been written during the past 6 weeks.

[wraper]

Anyone considering buying a Rigol, Siglent, Hameg (R&S) or other economy oscilloscope should really REALLY look at GwInstek's offerings first as it seems that GwInstek has a good grip on their software and offers good value for money! I'm surprised Rigol and Siglent have so much traction on this forum; GW Instek is almost like a hidden gem.

Thanks for the detailed review. I had GW Instek scope in the past and used their another equipment at my previous job. Now I own 3 Rigol devices and certainly I don't quiet get all that worshiping on them. They certainly have some issues including the build/design quality. Those GW Instek devices certainly felt better. Nice to read that someone finally made a good review on a modern device from them.

[wblock]

That is an excellent review!  It is really encouraging to hear that they have fixed bugs that quickly, and the shared codebase means more than one model benefits from the fixes.

Is the "VPO" (DPO) what did not work initially but was fixed by one of the firmware updates?

My Instek power supply has PWM fan speed control, but it is choppy enough to make the fan noise very noticeable.  There is a Microchip app note on reducing that: http://ww1.microchip.com/downloads/en/AppNotes/00771b.pdf.

[markone]

Decoding
-snip
There is one catch though: SPI decoding must have a chip select signal!

This means that the two channels scopes in this family are not able to decode an SPI interface even in the monodirectional form (DATA LINE + DATA CLK), could you please confirm this ?

Another question : looking at the Dave's video of GDS2000A @50:00 we can see that there is no screen intensity gradient applied for AM waveforms on the screen, is also the case for your instrument (so 2000E family )?

IF negative, could you post a screen that shows intensity gradient capabilities ?   

Anyway, you are doing a great debugging job for them !

[wraper]

Another question : looking at the Dave's video of GDS2000A @50:00 we can see that there is no screen intensity gradient applied for AM waveforms on the screen, is also the case for your instrument (so 2000E family )?

What I've seen on the GDS2000A video does not go along with what Dave commented. IMO probably he just set the the persistence from min to max very quickly, so it was hard to see that actually it is there. And most of the time it just was set to max.
In that moment (48:22) he just set the persistence from max to min in about 1.5 seconds and said there is none.

[markone]

What I've seen on the GDS2000A video does not go along with what Dave commented. IMO probably just set the the persistence from min to max very quickly, so it was hard to see that actually it is there. And most of the time it just was set to max.
In that moment (48:22) he just set the persistence from max to min in about 1.5 seconds an said there is none.

I'm not familiar with GW settings for persistence and intensity gradient but if you go to 50:00 the AM modulated signal appers to be like a "solid block" without any gradient, dunno if due to erroneous settings.

So i'm wondering if it's a different matter for the series 2000E, as it seems to be for the 1000B looking at the the Dave's @37:10.

[wraper]

I'm not familiar with GW settings for persistence and intensity gradient but if you go to 50:00 the AM modulated signal appers to be like a "solid block" without any gradient, dunno if due to erroneous settings.

Because persistance was set to 100%, at least at that 48:22 moment he was adjusting it from 100% down to the bottom. If you set it to 100% on rigol, it will be the same. Probably on Instek default setting is 100% (I don't have one to verify). On Rigol DS2000A which I have, the default intensity setting is 50%.

[nctnico]

Decoding
-snip
There is one catch though: SPI decoding must have a chip select signal!

This means that the two channels scopes in this family are not able to decode an SPI interface even in the monodirectional form (DATA LINE + DATA CLK), could you please confirm this ?

Another question : looking at the Dave's video of GDS2000A @50:00 we can see that there is no screen intensity gradient applied for AM waveforms on the screen, is also the case for your instrument (so 2000E family )?

IF negative, could you post a screen that shows intensity gradient capabilities ?   

Anyway, you are doing a great debugging job for them !

According to the manual SPI decoding won't work on 2 channel versions. There is intensity grading but it starts from 50% intensity and then goes up. This means it doesn't simulate an analog scope but has the practical purpose of showing a signal very clear when there is a signal. IMHO this is a good thing; I don't want a scope which shows signals in a way they are hardly noticeable. I have used my fair share of analog scopes and one of the problems I ran into repeatedly was dim (hardly visible) signals.

Regarding the debugging: I would have liked to avoid that but since they came up with a fix so quickly it was impossible to say 'too little, too late' 
I didn't got a message from the dealer indicating when Gw Instek is going to release a new official version but I guess that will be after the Chinese new year.

[markone]

There is intensity grading but it starts from 50% intensity and then goes up. This means it doesn't simulate an analog scope but has the practical purpose of showing a signal very clear when there is a signal. IMHO this is a good thing; I don't want a scope which shows signals in a way they are hardly noticeable. I have used my fair share of analog scopes and one of the problems I ran into repeatedly was dim (hardly visible) signals.

Many thanks, now it's clear how it works.

As far as i understood you owned a Siglent SDS2000, in which areas do you think that this one is much better ?

[Hydrawerk]

Thanks for the review.

[nctnico]

There is intensity grading but it starts from 50% intensity and then goes up. This means it doesn't simulate an analog scope but has the practical purpose of showing a signal very clear when there is a signal. IMHO this is a good thing; I don't want a scope which shows signals in a way they are hardly noticeable. I have used my fair share of analog scopes and one of the problems I ran into repeatedly was dim (hardly visible) signals.

Many thanks, now it's clear how it works.

As far as i understood you owned a Siglent SDS2000, in which areas do you think that this one is much better ?

First of all: the GDS2204E has mature firmware which works and the SDS2000 doesn't so there isn't really any sensible comparison to make 
But I'll try.... Compared to the SDS2000 the GDS2204E is cheaper, has an impressive amount of processing power under the hood, has way more FFT points (1 million versus 1000), decodes the entire memory, can average much longer traces, has input filtering, more extensive math functions, saves images as PNG (instead of uncompressed BMP), decoding is included and more extensive protocol triggering to name a few.

[markone]

First of all: the GDS2204E has mature firmware which works and the SDS2000 doesn't so there isn't really any sensible comparison to make 
But I'll try.... Compared to the SDS2000 the GDS2204E is cheaper, has an impressive amount of processing power under the hood, has way more FFT points (1 million versus 1000), decodes the entire memory, can average much longer traces, has input filtering, more extensive math functions, saves images as PNG (instead of uncompressed BMP), decoding is included and more extensive protocol triggering to name a few.

Well, lot of important points, ie " decodes the entire memory" and 1M FFT's point will be important for me, i will be tempted to buy the 100Mhz 2 channels version but things like not shielded & whining PSU and overall not so great build quality worry me a bit.

Many thanks for your prompt answer.

[wraper]

Well, lot of important points, ie " decodes the entire memory" and 1M FFT's point will be important for me, i will be tempted to buy the 100Mhz 2 channels version but things like not shielded & whining PSU and overall not so great build quality worry me a bit.

I'd prefer non shielded Instek over siglent rust. Also, from what I have read, it seem to use quasi-resonant SMPS topology, so the need for the shielding should be significantly reduced.

[markone]

I'd prefer non shielded Instek over siglent rust. Also, from what I have read, it seem to use quasi-resonant SMPS topology, so the need for the shielding should be significantly reduced.

Watching to Dave's 1000B teardown video @31:25   you can see some quite strange periodic noise stuff, i'm wondering if it comes from PSU, that seems to be identical to the one mounted on GDS2000E family.

Regardless of the switching PSU's topology you always have high dV/dT circuit traces, avoid shieldind is not an option without implications, also for safety reason.

If some capacitor blows out, with open frames solution like that the whole instrument is involved.

[nctnico]

I'd prefer non shielded Instek over siglent rust. Also, from what I have read, it seem to use quasi-resonant SMPS topology, so the need for the shielding should be significantly reduced.

Watching to Dave's 1000B teardown video @31:25   you can see some quite strange periodic noise stuff, i'm wondering if it comes from PSU, that seems to be identical to the one mounted on GDS2000E family.

Regardless of the switching PSU's topology you always have high dV/dT circuit traces, avoid shieldind is not an option without implications, also for safety reason.

If some capacitor blows out, with open frames solution like that the whole instrument is involved.

It's not like every capacitor explodes at some point (especially since they have safety vents). Also the input circuitry is well shielded in a metal box so there really is no noise visible from the PSU with the inputs open on my GDS2204E. It is hard to tell where the artifacts Dave shows come from. First of all there are signal display issues in the firmware of the scope Dave is showing and secondly he noticed the circuitry is slightly different so that could also be a source of noise coupling into it. Furthermore it seems Dave never has opened a Tektronix TDS500/600/700 series scope; in those scopes Tektronix has also used ribbon cable to distribute the power (a lot!) to various parts of the scope. For as long as you stay within the maximum voltage and current rating of a connector it is fine to use whatever connector to distribute power.

@Achmed99: I think the letter E is for Economy.

[markone]

It's not like every capacitor explodes at some point (especially since they have safety vents). Also the input circuitry is well shielded in a metal box so there really is no noise visible from the PSU with the inputs open on my GDS2204E. It is hard to tell where the artifacts Dave shows come from.

Yes, sure, luckily not all capacitor explodes at some point but ...just in case.

Times ago it happened to me that a brand new Panasonic fax machine had a PSU capacitor suddenly failed spreading electrolyte liquid through the vents everywhere inside the box, ruining everything it touched, quite toxic smell inside the room !.

Of course it was swapped under warranty.

So you can confirm that your scope is not affected from the same strange noise that Dave spotted in its review ?
And, just out of curiosity, have you bought your from TME ?

[nctnico]

I have no weird noise on my scope. I have attached some screendumps with the inputs open and set to 1mV/div at various time/div settings. Acquisition mode is set to sample and the variable persistence set to 240ms. But note that I have (newer) beta firmware in my scope!

[tequipment]

E is the economy version but really its  just a new scope version.  In some ways it has better specs.

Also it had a 20% price drop yesterday on our site www.tequipment.net

Thanks
Evan Cirelli

TEquipment.NET

[nctnico]

It seems the beta firmware got finalised into version 1.24 which was released a couple of days ago together with a new feature (seperate download) which shows the FFT result and the traces in a seperate window. I have not had the time to see if there are more new features in this firmware release.

I did however made a slight modification to the fan. It started resonating into the casing which made an annoying sound. Instead of going for a new fan I choose to change the mounting using rubber universal fan mounts: http://uk.farnell.com/duratool/dtrfmm-1/fan-mount-universal-4-5mm-pk50/dp/2472730 These soft rubber mounts decouple the fan from the casing so that inbalances in the fan do no get amplified by the casing.




I used the remaining pieces of the fan mounts to create stand-offs for the other 2 holes of the fan to keep it a bit more in place. The end result is a much quieter scope. And the best part is that I have 48 of these fan mounts left for other projects/equipment!

[markone]

I did however made a slight modification to the fan. It started resonating into the casing which made an annoying sound. Instead of going for a new fan ... -snip

Those ADDA fan are cr@p, i changed mine with an 60x60x15 Sunon Maglev (MB60151V3-000U-A99), much less noise and much more air flow, actually internal scope's case temp dropped a lot.

The original 20mm deep fan falls to close to analog front-end metal shielding and does not work properly, the resulting air flow is much lower than spec sheet, the 15mm model has instead enough room to work good. 

[nctnico]

The Sunon fan you choose is rated at 25dB where the original Adda fan is rated at 18.4dB which translates into the Sunon fan being 4 times louder but I guess the Sunon fan also moves a lot more air. I'm sorry but I don't see how the Sunon fan can be more quiet than the Adda fan when you just compare the fans as they are by themselves. Mounting the original Adda fan with the rubber mounts has lowered the fan noise a lot! All in all I don't think the original Adda fan is a bad choice perse if you look at the specs but it is not a high quality fan and any unbalance can easely couple/resonate into the casing. I did look at fan replacements but the good & low noise fans from Papst etc start from 25 euro (ex. VAT) and there is not a large improvement where it comes to noise: 16dB versus 18.4dB.

[markone]

The Sunon fan you choose is rated at 25dB where the original Adda fan is rated at 18.4dB

I worked a lot with fans for forced air cooling systems in power electronic and i soon learned that the last thing that you can trust is the noise rating on spec sheet, especially if you compare them among different brands.

In addition if the application implies high static load / flow, when air "sounds" through metal fins and holes, those numbers become utterly useless.

In this scope it's even worst, the fan placement is wrong for a 20mm deep device, as you can check in your scope the intake side falls about 9mm from the nearby metal shielding that literally suffucate it, you can forget the promised CFM and the alleged dBA, flow is much lower than expected and noise is much higher.

The 15mm model has additional 5mm air space in between that let it work a lot better, no sh!t.

Now facts are that if i put my hand in proximity of scope's fan exhaust grid i feel a much stronger air flow than before and i have LOT LESS NOISE 

Back side bnc connectors now are almost at Tamb while before were about 15 °C higher, also case top in proximity of Zynq chip is now colder.

I would say it's a huge improvement, especially because this room temp can easily exceed 30 °C in summer time (34 °C in 2015 !) and there is still room to further decrease fan speed (and noise) with a series resistor.

Sunon MB60151V3-000U-A99 comes for cheap (1/3 papst cost), has right size for the purpose, much less bearing noise and no vibrations at all.

[nctnico]

Hmm now I'm tempted to give that Sunon fan a try and compare... where did you buy it?

[markone]

Hmm now I'm tempted to give that Sunon fan a try and compare... where did you buy it?

From TME http://www.tme.eu/en/details/mb60151v3-a99/dc12v-fans/sunon/mb60151v3-000u-a99/ but now it's no more available, i think you can only order the slower model (i fear too weak) ha60151v4, available for cheap also from Reichelt.

My suggestion is to go for the V3 model than adjust rotation speed with a series resistor and avoid normal sleeve and ball bearing fans, are much more noisier.

Of course shipping cost does not justify an order only for one fan.

[nctnico]

I have been wanting to do an interesting experiment for a while now to see if there is a (practical) limit to the SPI decoding speed. I had to wait though for an FPGA project to come along so I didn't have to build a test setup for one simple test. I created a simple SPI signal generator which just outputs incrementing values (LSB first) with an SPI clock frequency of 125MHz and fed it into the GDS2204E. It turns out it can decode and trigger SPI at this speed just fine even though the clock signal looks like a sine wave.

[rickv14623]

Nice job on the review.

[MT]

Great review, from where did you say you bought it?

[nctnico]

I got it from a Dutch dealer called Eleshop.nl

[pascal_sweden]

Pity that the review is mostly text, and hardly any screenshots!

[saturation]

What screenshots would you like, maybe the OP or others with similar models can add to it?

Pity that the review is mostly text, and hardly any screenshots!

[nctnico]

I did make quite a few screenshots but the problem is that this forum makes it very hard to put a large posting together with the pictures sitting in the text nicely.

[Kilrah]

Very easy if you host the images externally.

[nctnico]

Very easy if you host the images externally.

Yes, I started doing that recently but if Flickr ceases to exist then my postings will have to do without pictures. I used to use Imageshack but they have already purged all of the pictures I used on other fora.

[Kilrah]

I use Flickr for actual photography, makes little sense to use it for random forum stuff.
I use Photobucket for forum stuff, the interface is a bit slow but what I uploaded back in 2005 is still alive and kicking.

[System Error Message]

well you beat me to a review about it. trying to learn as much as i can with electronics.

One effective way to remove noise is to capture the wave and subtract it but this only works to a maximum of 1M memory, not available for 10M. I noticed that when you have subtract/divide the math works slower, i dont know if this is because of eliminating the noise so the wave seems to vary less or if it is actually slower. It seems that after some use the input noise gets very significant, around 2mV with nothing connected and a -0.2V deviation of voltage values. SPC doesnt fix this but i hope calibration will, i managed to get calibration started by using 2 probes and making sure ground is also connected though i will fully calibrate when i find a way to keep the 2 probes touching without any issue.

If i put the probes on a wooden surface i get other noises too but this isnt a scope issue.

I subtract input wave with probe connected before connecting so that i can see the right waveform and noise coming from what i measure.

The ability to capture 1M worth of wave and perform math with it is quite a nice feature and useful when you are comparing. I wish the scope was heavier, would make it easier to play with all the knobs and buttons.

[dcarr]

Thanks for posting a very thorough review.

David

[pascal_sweden]

Or why not make a video review on YouTube? If this oscilloscope is really that great, how come that there is no single video review about it on YouTube? (Besides the silly videos from the manufacturer itself)

Appreciate if you can make a video review about your brand new spanking oscilloscope!

What about MSO functionality? Can it be expanded with an MSO probe? How many digital channels? 8 or 16?

Which distributor did you buy it from? Is there a distributor in the Benelux?

[nctnico]

I guess nobody bothered to do a video review. My problem with a video review is that I don't want to dump a longwinded piece of raw footage on Youtube. I'm quite sure I lack Dave's ability to come up with a coherent story of the cuff like he does. I would like to try and make a video but I'm skeptic about my skills to do that succesfully. I'd need a script, preparation and lots of video editing which all translate in a lot of time I don't have. The text review I posted to begin this thread took me a whopping 6 or 7 weeks to complete.

About your questions: unfortunately no MSO function although it seems the firmware could deal with it because the MSO function buttons are disabled. It appears GW Instek uses one firmware sourcecode for all their scopes which is good because if they fix a bug, they fix it for all models (unless it is model specific) and need less effort for testing. I bought my GDS-2204E from Eleshop.nl. GW Instek has a European office in the Netherlands BTW (no, I don't work there).

[pascal_sweden]

That's nice to know about the European office. So it means good warranty procedures.
I am originally from Belgium, so definitely consider that as a plus.

GW-Instek used to have modular slots in the GDS2000A series. It's a pity that they took out that concept in the GDS2000E series.

But do they have an MSO model in the GDS2000E series, or is that only in the GDS2000A and GDS3000 series?

[nctnico]

So far no MSO in sight for the GDS2000E. The casing does seem to be prepared for an MSO plug so at least someone at GW Instek has played with the idea to add an MSO option to a similar model.

[Kilrah]

I'd have bought one already if it there was an MSO...

[nctnico]

If there was a version with MSO and 10Mpts I would have bought that one but OTOH I also have an MSO on my bench for nearly 2 years and I have not yet used the digital channels. 4 analog channels take you a long way and for serious digital debugging I use my logic analyser.

[pascal_sweden]

Which MSO and which Logic Analyzer do you have?

[nctnico]

Agilent MSO7104A and Tektronix TLA715 logic analyser.

[AR]

Hello nctnico,

Stupid question , what the most used scope on your bench. Secondly has the GDS2204E got quieter inputs when they are shorted with the same bandwidth compared to The MSO7104A. 

The local Instek distributor has finally got a GDS2000E series sample for me to play with.

[nctnico]

I'm using the MSO7104 the most and that is mainly due to the much larger screen.

[eeguy]

E is the economy version but really its  just a new scope version.  In some ways it has better specs.

Also it had a 20% price drop yesterday on our site www.tequipment.net

Thanks
Evan Cirelli

TEquipment.NET

Is it really true that the GDS2204E is newer, has better specs and cheaper than the GDS-2204A?

[eeguy]

Where can I find more videos about the GDS2204E? On youtube, I only found six video under the name GDS-2000E. There are more under GDS-2000A. I want to make sure that I like the respond and UI of the GDS2204E.

[JPortici]

try and ask GWi for a loaner?

[wraper]

Is it really true that the GDS2204E is newer, has better specs and cheaper than the GDS-2204A?

Yes it's true, except 1GSa/s vs 2GSa/s sampling rate as 2000E are budget series. Also 2000A does not include options by default, and need them purchased separately.

[eeguy]

Is it really true that the GDS2204E is newer, has better specs and cheaper than the GDS-2204A?

Yes it's true, except 1GSa/s vs 2GSa/s sampling rate as 2000E are budget series. Also 2000A does not include options by default, and need them purchased separately.

Thanks. Sorry, what do you mean by "not include options by default"? What options?


From the specs:
http://www.gwinstek.com/en-US/products/Oscilloscopes/Digital_Storage_Oscilloscopes/GDS-2000E

Under Real Time Sample Rate: Max: 1GSa/s (4 ch model), Per channel 1GSa/s (2ch model)
What does that actually mean? Does that mean on a 4 ch GDS-2204E model, the max sample rate of each channel is 1GSa/s even when all four channels are used?

Under Trigger, it states that EXT* only on dual channel models. Does that mean I cannot use an external trigger such as a button to trigger on/off the recording on the GDS-2204E?

[wraper]

Is it really true that the GDS2204E is newer, has better specs and cheaper than the GDS-2204A?

Yes it's true, except 1GSa/s vs 2GSa/s sampling rate as 2000E are budget series. Also 2000A does not include options by default, and need them purchased separately.

Thanks. Sorry, what do you mean by "not include options by default"? What options?

For example:

I2C/SPI/UART/CAN/LIN Serial Bus Trigger and Decoding Function

Also full memory and other things fully enabled by default. In most of the scopes those need to be purchased separately and often will cost more than the scope itself.

Under Real Time Sample Rate: Max: 1GSa/s (4 ch model), Per channel 1GSa/s (2ch model)
What does that actually mean? Does that mean on a 4 ch GDS-2204E model, the max sample rate of each channel is 1GSa/s even when all four channels are used?

It means that 2ch will always run at 1GSa/s, 4ch version will drop to 500 MSa/s if more than 2 of any channels simultaneously enabled.

Under Trigger, it states that EXT* only on dual channel models. Does that mean I cannot use an external trigger such as a button to trigger on/off the recording on the GDS-2204E?

Vast majority of 4ch scopes do not have separate external trigger input. Therefore you need to sacrifice one input channel for this purpose if you need one.

[eeguy]

Is it really true that the GDS2204E is newer, has better specs and cheaper than the GDS-2204A?

Yes it's true, except 1GSa/s vs 2GSa/s sampling rate as 2000E are budget series. Also 2000A does not include options by default, and need them purchased separately.

Thanks. Sorry, what do you mean by "not include options by default"? What options?

For example:

I2C/SPI/UART/CAN/LIN Serial Bus Trigger and Decoding Function

Also full memory and other things fully enabled by default. In most of the scopes those need to be purchased separately and often will cost more than the scope itself.

Under Real Time Sample Rate: Max: 1GSa/s (4 ch model), Per channel 1GSa/s (2ch model)
What does that actually mean? Does that mean on a 4 ch GDS-2204E model, the max sample rate of each channel is 1GSa/s even when all four channels are used?

It means that 2ch will always run at 1GSa/s, 4ch version will drop to 500 MSa/s if more than 2 of any channels simultaneously enabled.

Having full memory and I2C/SPI/UART/CAN/LIN Serial Bus Trigger and Decoding functions enabled by default seems to be a good deal.

Given that I am told 1GSa/s should be sufficient for my use, is it bad to have the sample rate dropped to 500 MSa/s if more than 2 channels of the 4 channels GDS2204E is used?

From the Keysight's specs for the 2000X Series:
http://literature.cdn.keysight.com/litweb/pdf/5990-6618EN.pdf?id=2002854

The maximum sample rate is 2 GSa/s half-channel interleaved, 1GSa/s per channel

The 3000T X Series:
http://literature.cdn.keysight.com/litweb/pdf/5992-0140EN.pdf?id=2545408

The maximum sample rate is 5 GSa/s half channel interleaved, 2.5GSa/s all channel

Does that mean for the 4-channel 2000X ones, when all four channels are used, the sample rate for each channel is 1GS/s while for the 4-channel 3000T X ones, when all four channels are used, the sample rate is 2.5GSa/s?

[saturation]

When sampling rate drops the key problem is aliasing error.  With mutiple channels engaged and I'm concerned about possible aliasing artifacts I intermittently turn one or several of the CH off; this raises the sampling rate instantaneously in real time so I can compare waveforms very quickly, flipping between full and reduced sampling.   Its a simply ploy when using low cost DSO at your home lab versus better gear at work.

Having full memory and I2C/SPI/UART/CAN/LIN Serial Bus Trigger and Decoding functions enabled by default seems to be a good deal.

Given that I am told 1GSa/s should be sufficient for my use, is it bad to have the sample rate dropped to 500 MSa/s if more than 2 channels of the 4 channels GDS2204E is used?

....

Does that mean for the 4-channel 2000X ones, when all four channels are used, the sample rate for each channel is 1GS/s while for the 4-channel 3000T X ones, when all four channels are used, the sample rate is 2.5GSa/s?

Yes.  There are extensive reviews and teardowns in the forum of both series DSOs but I can't recall the details, Keysight DSO do have more ADC to sustain the sampling rate.

[nctnico]

I did some testing with the GDS2204E and aliasing is minimal because it has anti-aliasing filters in place. With a 200MHz bandwidth the 500Ms/s is more than enough to show the signal. You have to keep in mind that in theory the minimum sampling frequency is twice the bandwidth. Everything more is overkill but in practical situations you need some margin for the anti-aliasing filters.

[marmad]

You have to keep in mind that in theory the minimum sampling frequency is twice the bandwidth. Everything more is overkill but in practical situations you need some margin for the anti-aliasing filters.

It has nothing to do with "overkill"; it has to do with design. Out-of-band signals may only be attenuated by 10 dB beyond the band-edge of the instrument, and it doesn't matter what kind of anti-aliasing filters the DSO has, they absolutely don't roll-off infinitely fast and some buffer room is needed.

[nctnico]

That is why I wrote: some margin is required. Also the typical passive probing solution will load a high frequency signal way more so there is already filtering before the signal gets into the scope.

[marmad]

Reliable sampling and reconstructing frequencies at exactly the Nyquist Rate is physically not attainable, and thus a higher sampling rate is not "overkill"; your precise descriptor which I was responding to.

[nctnico]

Again: I wrote 'margin' so don't insist on reading something different!

[marmad]

Again: I wrote 'margin' so don't insist on reading something different!

Sheesh... I'm not reading anything other than your English - which, admittedly, is poor. You can't "overkill" based on a theory, and by "practical situations", you actually mean "real-world applications". It's ridiculous to write posts which might make some people believe that sampling above the Nyquist Rate is irrelevant (i.e. "overkill").

[Wuerstchenhund]

Again: I wrote 'margin' so don't insist on reading something different!

Sheesh... I'm not reading anything other than your English - which, admittedly, is poor. You can't "overkill" based on a theory, and by "practical situations", you actually mean "real-world applications". It's ridiculous to write posts which might make some people believe that sampling above the Nyquist Rate is irrelevant (i.e. "overkill").

I understood nctnico's post so that he's saying that, while in general anything over Nyquist-Shannon doesn't provide any benefit, some provisions have to be made for signals above the rated analog BW beccause the true BW might be higher and because of the filter properties which isn't infinite steep. And that is absolutely correct, and a sample rate in excess of what's required to cover that frequency range is, indeed, "overkill" (or do you want to say a 20GSa/s sample rate would make sense for a 100MHz scope?).

On a side note, I have to say nctnico's English is absolutely fine and much better than what I've seen from many 'native' speakers, and on a personal note I find your (repeated) rather condescending attitude towards non-native English speakers pretty poor. Especially when your own grasp of the use of words hasn't been exactly stellar.

[eeguy]

I did some testing with the GDS2204E and aliasing is minimal because it has anti-aliasing filters in place. With a 200MHz bandwidth the 500Ms/s is more than enough to show the signal. You have to keep in mind that in theory the minimum sampling frequency is twice the bandwidth. Everything more is overkill but in practical situations you need some margin for the anti-aliasing filters.

Thanks. Do you mean it should be OK but if I could get an oscilloscope (100MHz-200MHz) with 1GS/s per channel, that would be better?

[marmad]

I understood nctnico's post...or do you want to say a 20GSa/s sample rate would make sense for a 100MHz scope?

I understood his post, and although you may accept at face value that "with a 200MHz bandwidth, the 500Ms/s is more than enough to show the signal" on a low-cost DSO, I'm a little more skeptical. In any case, there have been repeated discussions here (and much talk by Dave during some of his reviews) on this subject, and whether or not you feel that a sampling rate that is 2.5x the bandwidth is "overkill" to reliably reconstruct complex, high-frequency signals with sin(x)/x, I don't. In that regard, I feel it's worth debating whether it's the correct term in this context for prospective GW-Instek (or any DSO) buyers.

...or do you want to say a 20GSa/s sample rate would make sense for a 100MHz scope?

You believe there's an equivalency between a sampling rate 2.5x higher than the BW and one that's 200x higher than the BW?

...and on a personal note I find your (repeated) rather condescending attitude towards non-native English speakers pretty poor.

On a personal note, I find your use of the word "condescending" to describe any other poster on this blog rather humorous, and while I'll fully admit that I'm sometimes (either intentionally or unintentionally) condescending, somehow I doubt you'll follow suit.

And I've certainly never claimed my English is "stellar", and I know I've made plenty of incorrect or awkward word choices while dashing off posts. And yes, nctnico's English is normally fine and I was being overly harsh, but I believe "overkill" to be a misleading description in this particular example.

[jjoonathan]

I don't think it reflects negatively on peoples' English skills, but I do think that many of the remarks in this thread -- "rolloff is not infinitely steep" and "some margin for the anti-alias filters" understate the role of rolloff.

It's typical for modern scopes to devote half (or more!) of their Nyquist band to rolloff rather than sticker bandwidth (i.e. samp rate >= 4x BW). Gentle rolloffs make for better step responses, so Oscilloscope designers prefer them when possible, but the tradeoff is that a huge part of your Nyquist band must then be dedicated to rolloff in order to promise decent attenuation of aliases. Not 10% or 20% of your Nyquist band, 50% or 60% of your Nyquist band, and even then the alias attenuation isn't spectacular. For the front end on my personal scope, a Rigol DS4014 hacked to 500MHz (see "Rigol MSO4024-500" line on the graph below), here is the relationship between sample rate and the alias attenuation they can promise given the gentle rolloff of their front end filter:

SampleRate	AliasAttenuation
2BW	-0.5dB (!)
2.5BW	-2dB (!)
4BW	-13dB (actual sample rate, 4 channels)
8BW	~30dB (actual sample rate, 2 channels)

I am quite pleased they did not try to sample at 2.5*BW. The benefit of going from 4*BW to 8*BW sampling (-13dB aliases to -30dB aliases) is more debatable, but -13dB attenuation isn't actually all that great -- it's still 20% of the original voltage amplitude, which is a considerable amount of worst case waveform fuzziness. It's comforting to have the option to bump up to 8*BW sampling (aliases at most ~3% of original amplitude).

With credit to Altemir:


Of course, we are discussing the GDS2204E, not the Rigol DS4000 series, but I could not find frequency response plots for the GDS2204E so it would be difficult to comment on the tradeoffs made by GW Instek. Here is the Bode plot for the DS1054Z (credit to ankerwolf) which might be more representative:

[nctnico]

When it comes to aliasing you also have to realise that the roll-off mirrors at half the sampling frequency so higher frequency signals past the Nyquist frequency still get attenuated as if the response continued.

[jjoonathan]

True, I should have made it more clear that the numbers I gave were the worst case, where you have a signal just above the Nyquist frequency aliasing to just below the Nyquist frequency. It's an edge case to be sure, but like many edge cases it punches above its weight. It pops up right at the theoretical limit of what a scope can measure, and given that bandwidth costs money and both oscilloscope bandwidths and clock/data frequencies gravitate towards "nice" round numbers, it's an edge case that I could see someone hitting if they weren't careful.

[nctnico]

Since there are no other graphs available I measured the bandwidth of my GDS2204E starting at 10MHz with 1 channel enabled at 1Gs/s. I noticed that switching to a more sensitive range changed the amplitude more than I expected so I measured twice using a 0dBm signal and a -20dBm signal and plotted the results in a graph (Y-axis in dB, X-axis in MHz):


The 50mV/div graph shows the -3dB frequency is around 290MHz but for 10mV/div the -3dB point is around 210MHz.

[Merlot1970]

I unboxed my new GDS-2204E yesterday and fired it up..  About 15 minutes into using it, I was able to find a problem on a circuit I"ve been working on..It appeared on screen to be a high impedance power supply causing a glitch at the starting point of my circuit.  My Tek analog scope did not see this, but the Instek picked up the glitch immediately.  I'll see If i can post a photo.


First photo is the captured glitch...When the circuit initially conducts current, it wigs out.

Second photo is the corrected problem.

[saturation]

Good to hear, would be interested in why the Tek couldn't pick it up and what the Instek had that found it quickly.

I unboxed my new GDS-2204E yesterday and fired it up..  About 15 minutes into using it, I was able to find a problem on a circuit I"ve been working on..It appeared on screen to be a high impedance power supply causing a glitch at the starting point of my circuit.  My Tek analog scope did not see this, but the Instek picked up the glitch immediately.  I'll see If i can post a photo.

[Merlot1970]

Photos posted, Saturation.

[nctnico]

Just wanted to show a useful application of the input filtering. In a project I'm currently working on I need to do some digital signal processing. Usually I have a PWM output + an RC filter as a monitor output instead of a real DAC. With the GDS2204E the PWM output suffices when combined with the input filtering:

Original signal:


Filtered signal:


Other purposes can be looking at what is output from PWM motor drives and class D amplifiers.

[nctnico]

Recently firmware version 1.26 has been released. This version now has band filtering besides low and and highpass input filtering which is an excellent addition 
I also did some testing with UART and CAN decoding regarding the baudrate tolerance because earlier firmware versions where quite strict. From my tests it seems the UART decoding allows a deviation from the baudrate of +/-3%. The CAN bitrate can now be set freely. Also the bitrate tolerance of the CAN decoding exceeds +/-2%.

[saturation]

Yes, it seems when they update one model they update the others.  Good to know since there are no announcements from Instek. 

[JanJansen]

If you mess with the fan do you loose the warranty ?

[nctnico]

If you mess with the fan do you loose the warranty ?

Technically yes but there where no warranty void stickers on mine  So just store the original screws if you replace them with rubber fan mounts.

[wraper]

If you mess with the fan do you loose the warranty ?

Technically yes but there where no warranty void stickers on mine  So just store the original screws if you replace them with rubber fan mounts.

Yep, just keep original fan and screws, to put them back if need arises. If your hands are straight, there is nothing to worry about. If I do something like this, I also wear ESD gloves to avoid leaving any fingerprints inside  .

[Fungus]

Anybody here feel up to doing the new "Wolf" test?

Apply a square wave to an input, display "rise time" on screen, then zoom in and out.

The idea is to see if the rise time is computed with screen pixels or with sampled data.

If the displayed rise time varies then it's using screen data. 

[MrWolf]

"Wolfs test" thread is here:
https://www.eevblog.com/forum/testgear/testing-dso-auto-measurements-accuracy-across-timebases/

It should be interesting since GW Instek seems to have excellent engineering team. Just bit lacking on marketing side 

[NorthGuy]

May I ask you if 10Mpts setting is per channel or overall. E.g. if you sample 4 channels, do you get 10Mpts for each (40Mpts total) or is the overall amount spread around channels (e.g. you get 5Mpts or even 2.5Mpts when you have all 4 channels on).

[wraper]

May I ask you if 10Mpts setting is per channel or overall. E.g. if you sample 4 channels, do you get 10Mpts for each (40Mpts total) or is the overall amount spread around channels (e.g. you get 5Mpts or even 2.5Mpts when you have all 4 channels on).

10Mpts for each (40Mpts total)

[nctnico]

Actually more: With 10MPts you can still get 2 segments which means there is 20Mpts of memory per channel if you calculate the memory like Keysight does it. 

[whited]

Ntcnico, would you buy the gds-2204E again today? Instek 2204E is $1256 USD right now. I won't spend over$1300.

I've ruled out siglent 2000x series. Don't think R&S, Tek, BK, nor Rigol have anything competitive with 4 channels in my budget.

I don't want to ignore a strong competitor. What are your thoughts?

[tautech]

I've ruled out siglent 2000x series.

Could you say why ?

[nctnico]

Ntcnico, would you buy the gds-2204E again today?

Definitely yes. Currently it sees more use than my very expensive Agilent scope.

[whited]

They seem very similar in specs. Siglent has 2gsa/s but instek has more FFT points. Siglent has more memory but Instek seems to have better firmware support. Instek serial decode seems robust in context of DSO (over Siglent).

Between their differences I personally like what I see on Instek more and my fuzzy logic after much reading and some sleeping is pushing me strongly to Instek.

I've ruled out siglent 2000x series.

Could you say why ?

[whited]

Yes this is so good for my wallet 

Ntcnico, would you buy the gds-2204E again today?

Definitely yes. Currently it sees more use than my very expensive Agilent scope.

[nctnico]

It seems firmware version 1.27 has been released and on GW Instek's site it is tagged as 'GDS-2000E / MSO2000 series firmware'. I have not tried it yet.

[saturation]

New version of the GDS 2000E series, the MDO2000E series with AWG option.

http://www.gwinstek.com/en-global/news/28/140236

MSRP not published by Instek to date but there are vendor prices posted for the EU market for those who stock it.

[nctnico]

New version of the GDS 2000E series, the MDO2000E series with AWG option.

Interesting! Also the 10Mpts for the MSO part is nice. Ofcourse this extension was to be expected because the firmware for the GDS2000E already had references to digital channels for a long time.

It seems the price for the most complete model is around 1900 euro ex. VAT (including decoding, MSO probe, etc).

[Asymmetric13]

@nctnico @saturation May I propose splitting the discussion on the new MSO series into a new thread

https://www.eevblog.com/forum/testgear/gw-instek-announces-new-mso-2000/

and keep this thread focussed on the "original" GDS-2000E (unless specific comparisons arise).

Thank you

[wer]

Does anyone actively use the remote mount feature?  I used it for  a few weeks, and it was nice to offload images instantly.  It no longer works, and doesn't attempt to connect.  I'm unable to erase the settings, or get any further debug.  I reached out to support months ago, but they didn't offer a solution.  Just curious if anyone else has seen this problem.

My other ask from support (besides a way to wipe it clean), was for login info.  Has anyone gotten that yet? I tried swapping out hash's but something about ubifs must validate the bits changed and reject the patched firmware.  I didn't try too hard and don't know much about ubi.

My ongoing disk mount Error is "Mount remote disk fail.", it's persisted through all firmware updates and resets and attempts to factory default, app removal re-installs.  It kills me that I can't just log in and fix it.  Especially when there is a damn ssh port listening on it's network interface.

Otherwise I'm pretty happy.
Thanks for all the info in this thread.  Hope it's cool I posted this here.

-=wer

[nctnico]

Did you check the file server logs? Maybe something changed on the system you are trying to connect to. Also try another network cable (even if the cable you use now seems to work) because many network problems are caused by cables.

[wer]

It's working again. 
Thanks for the thought.  I had run tcpdumps and all kinds of things previously, and indeed it's all my fault.  caught it trying, and fixed it.
-=wer

[dfnr2]

Am I overlooking something, or is there no user-definable baud rate for the UART serial decode?

I have access to a 2204E, and was trying it out on a system with a UART stream generated in programmable logic with nonstandard bit rate, and was only able to select standard baud rates.

This comes up every once in a while, and it's an oversight not to include user definable/adjustable bit rates for serial decodes in general.  This seems like either an oversight in the decode module or in the UI (making it hard to find.)

[nctnico]

You can do this for sure (maybe this has been added in a recent firmware release). If you select a baudrate you can turn the 'variable' knob to select a custom baudrate.

[digsys]

OK OK You talked me into getting one !! I have to spend less time in these forums !!

[dfnr2]

You can do this for sure (maybe this has been added in a recent firmware release). If you select a baudrate you can turn the 'variable' knob to select a custom baudrate.

It must be in a firmware update.  What's the latest version?

I have v1.24, and the baud button allows me to select standard baud rates with the 'variable' knob, but there is not an obvious way to select a nonstandard rate.

Also, does anyone have a working link for firmware updates?  It doesn't seem to be on the GW instek page.

[nctnico]

The most recent version is 1.27 (30 Jan 2017). You have to create a login in order to see the download on GW Instek's website. I know:  and 

[fishandchips]

As I recall this scope was released about a year ago. How likely will Gw Instek release an updated version this year?

[JPortici]

after just one year? a scope ain't no smartphone

(they did release the mso version a couple of months ago though)

[dfnr2]

The most recent version is 1.27 (30 Jan 2017). You have to create a login in order to see the download on GW Instek's website. I know:  and 

Yep.  I downloaded 1.28 (25 apr 2017), and indeed, the baud rate menu is easier to use, allows more different "standard" settings, and once selected, allows fine tuning of the baud rate.  Much, much better.

As a second scope for portable use, this is a really nifty one, and for a new scope, pretty good bang for the buck.

BTW, it looks like they do have a newer offering.  It doesn't seem to be a new iteration, but rather it offers the MSO and signal generator options. 

[nctnico]

Hmmm they released a new version. With 2.7 they already made a lot of input easier if you press the select button again. Today I needed a 1 second trigger hold-off and that was much easier to do with the list of presets than spinning the variable knob ad nauseam.

[thanasisk]

Considering to buy the 70mhz version for audio preamp/opamp etc work.

However the manual lists the fft horiz scaling as 10khz/div minimum, which would mean that audio freq response would be squashed within 2 horizontal divisions! Making fft unusable for audio frequencies..

Is this true or am I interpreting this wrong?

[nctnico]

Considering to buy the 70mhz version for audio preamp/opamp etc work.

However the manual lists the fft horiz scaling as 10khz/div minimum, which would mean that audio freq response would be squashed within 2 horizontal divisions! Making fft unusable for audio frequencies..

Is this true or am I interpreting this wrong?

I just tried and it is wrong. The minimum frequency depends on the samplerate so you can go down to less than 1Hz/div on long time/div settings.

[CC555]

Hello,

I bought this oscilloscope (GDS-2204e 1050€ on tme.eu) to try to solve some rs232 issues, I like it but I don't know if rs232 decoder is working properly (firmware version 1.32).

I connect the GDS-2204e and a DS1054Z to the same signals and I get different decoder results.

I am triggering the oscilloscopes at yellow signal's rising edge and capturing an uart communication (57600 8N1).




When I look at the same communication packets on the oscilloscopes I see different decoder results:

Rigol




GW




If I look at the first byte, I think rigol is decoding the signal right, at 57600 this is 02 (if I am not wrong  ):





Am I doing anything wrong with the GD-2204e or is it a bug?

If the GW doesn't decode the RS232 properly I will return it. Any good and cheap.. oscilloscope recommendation to work with serial buses (I2C, SPI, RS232 and full memory decoding)?.

Thank you and excuse my english.

[tautech]

You need to set a falling edge trigger on the GW.

[CC555]

Hello tautech,

thanks for your reply, setting a falling edge (channel 1) on both oscillocopes I get the same result.

[2N3055]

Try setting polarity normal (high = 0)..

[wraper]

As I remember someone wrote that he got wrong decoding on instek and fine on Rigol or maybe some other scope. Turned out that actual baud rate was off but Rigol was more tolerant to that. And after firmware update instek became more tolerant to inaccurate baud rate.

[tautech]

Hello tautech,

thanks for your reply, setting a falling edge (channel 1) on both oscillocopes I get the same result.

You need to take a step back, reset to defaults and set up again with thought and not fast button bashing.
Use only the channels needed and keep the channels and Trigger position on the display.
As 2N3055 mentions the idle setting for the data stream must also be set correctly.
Take screenshots in a way to show us as much info as possible for members to best help you.

[CC555]

Yes, polarity may be wrong specifying the start bit  , thanks for your help, I am going to try.

[nctnico]

As I remember someone wrote that he got wrong decoding on instek and fine on Rigol or maybe some other scope. Turned out that actual baud rate was off but Rigol was more tolerant to that. And after firmware update instek became more tolerant to inaccurate baud rate.

First make sure to have the latest firmware. Secondly the decoding thresholds must also be adjusted to match the signal (for each channel). Triggering doesn't matter at all.

[CC555]

Latest firmware V1.32.

I have to trigger with channel 1 because I want to look the communication at that event.

Default polarity on both scopes (start bit at falling edge, 0 = high) triggering with yellow signal.




Rigol decodes right (0xFD should be)


GW


At 57600 bps (57600 8N1): 1 bit = 17.36 us, so 0xFE is wrong, the 0 (logic 1) begins at 34.2 us.


The baudrate is ok, 17.36 us * 9 (start bit + 8 data bits) = 156.24 us


the start bit is at the rising edge instead the falling edge.

1V threshold levels, what am I doing wrong?

Thanks

[nctnico]

First of all set the polarity back to 'inverted (h=1)'.

[CC555]

Yes, that was my initial setting. The result is the same.

[2N3055]

I agree with nctnico, set polarity back..
Values in the first photos with original polarity on GW instek are shifted to left ....

hex 1 -> hex 2
hex 2 -> hex 4
hex 19 -> hex 32

you must check if both are exactly 1 start bit, 8 data , no parity and 1 stop bit... 10 bits total.
It seems GW Instek is stuffing additional bit in the beginning  (LSB)

[CC555]

Settings are ok: 57600 8 N, I have not found the stop bit option, there is an end of packet option (00 NULL by default).

[nctnico]

I did some testing and I've found the problem. At 2Ms/s your samplerate is too low. I was able to reproduce the problem but it went away at samplerates of 10Ms/s or more.

[2N3055]

Good info Nctnico.. 

[CC555]

Thanks for your help, I am going to try with a bigger sample rate, I thought 2Ms/s was enough for 57600 bps, Rigol is sampling at 1Ms/s

[CC555]

nctnico is right, It was a sampling problem:




10M per channel but you can only capture 1 second of the signal for a 57600 bps rs232, a DS1054Z can capture several seconds.

GW instek advertises it in its datasheet like "Serial bus design and debugging" and I may be a bit disappointed with this point after this test

I like the oscilloscope (what I have seen in two days) but I don't know what to do, if keep it or to look for another one.

Thanks for your help, my doubts with this issue have been solved.

[nctnico]

I'm a bit surprised too. From my initial testing it seemed the GDS2000E series need 13 times oversampling for UART decoding. Needing 10Ms/s for 57600 is more like 173 times. Maybe some optimisation in the decoding in later firmware versions doesn't quite pan out. I've contacted GW Instek technical support about this issue.

[rhb]

Gad!

The Rigol has a crap FFT, but decodes UART well.  The Instek is just the opposite.  Whatever is a person to do?

The really maddening part is given source code, I know how to fix this sort of thing.  But despite the GPL, OEMs are not complying. So one does not have  even the parts they did not write.

[digsys]

The Rigol has a crap FFT, but decodes UART well.  The Instek is just the opposite.  Whatever is a person to do? 

I'd chose FFT ability over RS232+ decode any day of the week. There are plenty of far more powerful stand-alone data decoders out there,
for very cheap prices. Nothing to stop you connecting both the DSO and a data decoder together, if you need to look for signal faults as well.

[CC555]

Hello,

similar problem with I2C  .

100 kHz clock signal, channel 4: SCL, channel 2: SDA, triggering at channel 1 event:

At 2 Ms the oscilloscope is not decoding:


a higher sample rate and it decodes properly:

[tautech]

Wouldn't you normally trigger on the clock ?

[CC555]

Wouldn't you normally trigger on the clock ?

The problem would be the same, when you select 10 Mpts the oscilloscope needs a higher sample rate to decode the signal, if you select less memory it decodes at 2Ms/s iirc (I will try it later).

In this case I want to capture the bus at a specific event in my pcb, so I select the channel 1 to trigger the oscillocope on that event.

Thanks for you reply.

[rhb]

*&#@!!!

I bought an MSO-2204EA precisely for this task!

[nctnico]

I have not heard back from GW Instek's technical support in the EU but I'm pretty sure they are closed for the holidays.

[CC555]

with less memory the scope decodes at lower sample rates:



but if 10 Mpts is selected:

At 10 MS/s it doesn't decode the signal (clk = 100 khz):


It needs 20 MS/s, that is a 0.5M buffer  :


I don't know if my settings are 100% right but I would say this oscilloscope has a problem decoding serial buses with the memory and sample rates selected.

*&#@!!!

I bought an MSO-2204EA precisely for this task!

Me too, uCs and serial buses.

[nctnico]

I think the firmware engineers at GW Instek cut a corner by skipping (too many) samples when decoding deep memory. This should be easy to fix for them though at the expense of speed.

[rhb]

My SWAG is that it is a consequence of Object Oriented Programming.  Instead of decoding a short buffer and logging data to a long buffer it's trying to decode the entire buffer whatever the size.

I may be wildly wrong as I've not looked into the matter yet.  All I know is what's been posted here.  But I've seen similar effects from OOPS, for example drawing 6 curve traces by drawing 1, erasing it and drawing 2, erasing them and drawing 3, etc.  Looks damn stupid if you're drawing 6-10 curves.  Intelligent users tend to giggle a lot looking at that.

The decoder should be operating on a small window as data comes in.  That window should slide along the buffer as samples come in.  The observed behavior suggests that it is always starting the decoding at the start of the buffer.

This is a clear argument for open source scope application code.

[tautech]

I don't know if my settings are 100% right but I would say this oscilloscope has a problem decoding serial buses with the memory and sample rates selected.

What would concern me more is the displayed signal integrity ? Not so much the scopes ability to portray it but the glitches and the variances in the data idle voltage of ~1/2/div. 

Something's maybe not right with your digital circuit or probing technique.

Jump back out of Decode and see what waveform shapes and amplitude you get then.

[CC555]

Hello tautech,

I think there is not any problem with these glitches, they are real and always show up when clock is low during a communication so It is the slave who is controlling the data line and can generate them.




And I have to check if the decodification is in the right place in the pictures. Anyway I am going to use a saleae logic to decode the bus.

[nctnico]

I don't know if my settings are 100% right but I would say this oscilloscope has a problem decoding serial buses with the memory and sample rates selected.

What would concern me more is the displayed signal integrity ? Not so much the scopes ability to portray it but the glitches and the variances in the data idle voltage of ~1/2/div. 

Something's maybe not right with your digital circuit or probing technique.

Jump back out of Decode and see what waveform shapes and amplitude you get then.

The shape of the signals isn't related to decoding. There are probably resistors between the various I2Cs devices to prevent one part of a circuit blowing up the rest of the circuit (safety barrier). Alternatively the value of the pull-up resistors may be too low so some devices may not sink enough current to pull the I2C lines completely low. Either way I've seen similar I2C signals before.

[tautech]

I don't know if my settings are 100% right but I would say this oscilloscope has a problem decoding serial buses with the memory and sample rates selected.

What would concern me more is the displayed signal integrity ? Not so much the scopes ability to portray it but the glitches and the variances in the data idle voltage of ~1/2/div. 

Something's maybe not right with your digital circuit or probing technique.

Jump back out of Decode and see what waveform shapes and amplitude you get then.

The shape of the signals isn't related to decoding.

If thresholds are set correctly.
Runts like in the middle of this screenshot will throw a cat amongst the pigeons.

[CC555]

Hello tautech, thx for you reply.

I would say those spikes are not important, i2c communication is working properly in my device.

It is a 100 khz clock so I think there is not any signal integrity problem (i didn't measure its rise time), the frequency is too low.

In the end, I am not going to send the oscilloscope back, in the 1000 € range I think is a good choice.

[CC555]

I have not heard back from GW Instek's technical support in the EU but I'm pretty sure they are closed for the holidays.

Hello,

any news from GW Instek?

[nctnico]

I have not heard back from GW Instek's technical support in the EU but I'm pretty sure they are closed for the holidays.

Hello,

any news from GW Instek?

I got a reply saying they are working on it. I have not checked if a new firmware version has been released.

[CC555]

Thx for your reply,

there is a new firmware version, v 1.34

Version 1.34
1.Added a new ¡§Frequency Response Analyzer¡¨ APP function which can be downloaded from our website.
(This APP only available for MSO-2000EA & MDO-2000E series)
2.Updated the Japanese menu tree translations for SA function.
3.Updated the Japanese menu tree translations for digital filter function.
4.Updated the simplified Chinese menu tree translations for some functions.
5.Updated the Russian menu tree translations for DMM function.
6.Fixed the EVENT table display and storage error issue for UART bus decoding.
7.Fixed the window zoom display issue for I2C bus decoding.
8.Fixed the I2C bus decoding error.
9.Fixed the type of executable for DEMO function.
10.Fixed the return value of SA:MEMory SCPI command.
11.Fixed the return value of <x>:MEMory and :DIGital:MEMory SCPI command.
12.Fixed the setting value of TRIG:DEL:LEV SCPI command.
13.The digital filter function can be implemented now after the SEGMENT data caught.
14.The firmware upgrade procedure is now ONLY support the ¡§Safe mode¡¨ firmware upgrading from this firmware version.

[nctnico]

It seems the biggest change is the network analysis feature for the models which have an internal waveform generator. Unfortunately GW Instek didn't fix the oversampling needed for UART decoding 

[DougM]

How well does this 'scope handle input overdrive?

The reason I'm asking is that I quite often want to look at a smallish voltage range of a larger signal. Right now, I've found that my Rigol DS1054Z is truly dreadful in this respect. For example, I have a square wave at 1kHz between 0v and 40V. The Rigol becomes useless at 2v/div let alone at higher gain... (I have some screenshots on the DS1054Z thread). 

Some people may suggest that this isn't a "proper" use of a digital scope, and that I should be keeping the whole signal on the screen and digitally zooming. I, respectfully disagree, and think the input stages should be capable of placing the limited digital resolution on the interesting part of the voltage range. (Especially with only 8 bits of ADC)

That being said, how does the GDS2204E perform in this respect?  I have excellent scopes available at work, but for home use, I think I need something better than the Rigol...

Best regards,
Doug

[nctnico]

I did a test with a 20Vpp 1kHz square wave:



The offset range is specified in the datasheet. At 500mV/div this seems to be +/- 25V. At more sensitive settings (200mV/div and less) the offset range is +/-5V.

[2N3055]

For DS1000Z:

Offset Range (Probe ratio is 1X)

1 mV/div to 499 mV/div: ±2 V
500 mV/div to 10 V/div: ±100 V

As far as I could see, DougM wants to look at part of 40V P-P signal at better than 1V/DIV or less..
No digital scope I know of will have dynamic reserve of 200x of full screen sensitivity... And probably only +-8 to 10 DIVS...

If A/D converter full scale input for instance is 2V, so on 100mV/DIV  input, your scope  full dynamic range is +- 5 DIVS.  DC offset will be only helpful if you have 100 mV AC riding on top of 12V offset, so you dial in offset in opposite direction to "pull it back" to screen.. Or you AC couple it..
If you have 40 V AC P-P, you need 40V dynamic range signal path for it all to fit in. Without mentioning amplification.. That is with unity gain.

I don't know of any digital scope that can do it. If someone know one I would like to learn about it. It would be good info.
Also, no matter how quick they recover from overload, it ain't pretty...

I believe David Hess mentioned sampling scopes (old analog diode bridge types) in other thread.
Diode bridge sampling scope are direct sampling devices. They don't have input amplifiers and are basically S/H circuit that is disconnected from input all the time, except at the very short time of sample window (aperture). By choosing time at which to sample and if needed adding DC offset, you can look at any part of waveform with no problem of overshoot.
They are repetitive sample devices (random or sequential) and are useful only for repetitive and stable signals..

Modern approach to more detail on digital scopes is hi res scopes (more than 8 bit), keeping signal within dynamic range, and zoom in in software.. 12bit scope will have 4x more detail (pixels) in vertical than 8 bit... But, it costs money...

Regards
Sinisa

[nctnico]

For DS1000Z:

Offset Range (Probe ratio is 1X)

1 mV/div to 499 mV/div: ±2 V
500 mV/div to 10 V/div: ±100 V

As far as I could see, DougM wants to look at part of 40V P-P signal at better than 1V/DIV or less..
No digital scope I know of will have dynamic reserve of 200x of full screen sensitivity... And probably only +-8 to 10 DIVS...

That depends entirely on how the input circuit is designed and it is not 200x but 5x (40V/8V full display range @ 1V/div = 5x). If the oscilloscope allows a large offset then the input circuitry should be designed to handle a signal which swings between the maxima of the offset. It seems the Rigol has a design flaw and thus has overdrive problems.

[2N3055]

That depends entirely on how the input circuit is designed and it is not 200x but 5x (40V/8V full display range @ 1V/div = 5x). If the oscilloscope allows a large offset then the input circuitry should be designed to handle a signal which swings between the maxima of the offset. It seems the Rigol has a design flaw and thus has overdrive problems.

He said that he would want to look at 10mV/DIV if he could... I simply stated that what he expects is unrealistic.. The fact that other scopes might recover a bit better or will allow one more step on a vertical atten before going crazy is not gonna give him what he wants to do... Every overdriven amplifier have overdrive problem.

Also I really doubt ANY digital scope will have full 5 screens below and above actual screen of dynamic range.. Maybe a screen but not much more.
Like I said, if you know one that has such a great dynamic range I would like to know..

Old CRT scopes had signal path with much higher voltage on power supplies .. That gave them much higher inherent dynamic range.. They were much better in that regard..
On DS1000Z input path is powered by +-5V.. if they used +-15V, followed by limiter to protect A/D converter, it would be much better... I agree. But not much better.
Maybe one more step on attenuator..

Don't get me wrong. I'm not defending Rigol. I would be very interested in a test how other scopes would fare in this respect. I suspect not much better.
How much quantitatively, needs to be tested....

Regards
Sinisa

[DougM]

That depends entirely on how the input circuit is designed and it is not 200x but 5x (40V/8V full display range @ 1V/div = 5x). If the oscilloscope allows a large offset then the input circuitry should be designed to handle a signal which swings between the maxima of the offset. It seems the Rigol has a design flaw and thus has overdrive problems.

He said that he would want to look at 10mV/DIV if he could... I simply stated that what he expects is unrealistic.. The fact that other scopes might recover a bit better or will allow one more step on a vertical atten before going crazy is not gonna give him what he wants to do... Every overdriven amplifier have overdrive problem.

Also I really doubt ANY digital scope will have full 5 screens below and above actual screen of dynamic range.. Maybe a screen but not much more.
Like I said, if you know one that has such a great dynamic range I would like to know..

Old CRT scopes had signal path with much higher voltage on power supplies .. That gave them much higher inherent dynamic range.. They were much better in that regard..
On DS1000Z input path is powered by +-5V.. if they used +-15V, followed by limiter to protect A/D converter, it would be much better... I agree. But not much better.
Maybe one more step on attenuator..

Don't get me wrong. I'm not defending Rigol. I would be very interested in a test how other scopes would fare in this respect. I suspect not much better.
How much quantitatively, needs to be tested....

Regards
Sinisa

No, I don't think I asked for 10mV per box (though that would be excellent!).  With a 10x probe the Rigol can only handle 5V/div on a 40V pkpk signal, which is very disappointing. (I will try with the probe at 1x later today and see what I can do).

I'd like better than 1V/division for sure, and 100mV would be nice. I'm coming to realize that this may not be a very mainstream use case.

Doug

[2N3055]

No, I don't think I asked for 10mV per box (though that would be excellent!). 

No you didnt, but I kinda figured you would like it :-))

I'm coming to realize that this may not be a very mainstream use case.

No it's not so much. But it would be very useful.  On several occasion during years  I wish I had a scope that could do that....

Regards,

Sinisa

[tautech]

For DS1000Z:

Offset Range (Probe ratio is 1X)

1 mV/div to 499 mV/div: ±2 V
500 mV/div to 10 V/div: ±100 V

As far as I could see, DougM wants to look at part of 40V P-P signal at better than 1V/DIV or less..
No digital scope I know of will have dynamic reserve of 200x of full screen sensitivity... And probably only +-8 to 10 DIVS...

That depends entirely on how the input circuit is designed and it is not 200x but 5x (40V/8V full display range @ 1V/div = 5x). If the oscilloscope allows a large offset then the input circuitry should be designed to handle a signal which swings between the maxima of the offset. It seems the Rigol has a design flaw and thus has overdrive problems.

Yep, the devil is in the detail.
For the new Siglent X-E series:

Offset ranges 1x input.
500uV~118mV/div: ±2V
120mV~1.18V/div: ±20V
1.2V~10V: ±200V/div

2V/div will give sufficient offset for DougM's needs and using the minor divisions ~250mV resolution.

[nctnico]

That depends entirely on how the input circuit is designed and it is not 200x but 5x (40V/8V full display range @ 1V/div = 5x). If the oscilloscope allows a large offset then the input circuitry should be designed to handle a signal which swings between the maxima of the offset. It seems the Rigol has a design flaw and thus has overdrive problems.

Yep, the devil is in the detail.
For the new Siglent X-E series:

Offset ranges 1x input.
500uV~118mV/div: ±2V
120mV~1.18V/div: ±20V
1.2V~10V: ±200V/div

2V/div will give sufficient offset for DougM's needs and using the minor divisions ~250mV resolution.

That only works if the Siglent doesn't have overdrive recovery problems just like the Rigol. Designing for a DC offset is easier than for a large signal swings.

[DougM]

Even though my excuse for getting a better 'scope for home use has been undermined by finding that my Rigol works a lot better at 1x probe setting, I've read this entire thread and am being sorely tempted to get one of these scopes!

Thanks for the detailed review and discussion.

Doug

[2N3055]

For DS1000Z:

Offset Range (Probe ratio is 1X)

1 mV/div to 499 mV/div: ±2 V
500 mV/div to 10 V/div: ±100 V

As far as I could see, DougM wants to look at part of 40V P-P signal at better than 1V/DIV or less..
No digital scope I know of will have dynamic reserve of 200x of full screen sensitivity... And probably only +-8 to 10 DIVS...

That depends entirely on how the input circuit is designed and it is not 200x but 5x (40V/8V full display range @ 1V/div = 5x). If the oscilloscope allows a large offset then the input circuitry should be designed to handle a signal which swings between the maxima of the offset. It seems the Rigol has a design flaw and thus has overdrive problems.

Yep, the devil is in the detail.
For the new Siglent X-E series:

Offset ranges 1x input.
500uV~118mV/div: ±2V
120mV~1.18V/div: ±20V
1.2V~10V: ±200V/div

2V/div will give sufficient offset for DougM's needs and using the minor divisions ~250mV resolution.

Tautech, thanks for the info..

It is not about offset. That was mentioned in passing, not relevant for original question. It is a good reminder to the fact that all scopes are different, though..
For instance from 500mV to 1.2V Rigol has 100V offset compared to Siglent 20V. If that is of significance to anybody...

But that is only relevant if you have a small AC component riding on top of a large DC component. In which case you can use AC coupling too.

Most of the time, DC offset is simply used to move channels vertically on screen for visual separation.

Problem is something else.

If you have 10 vertical  DIVs on screen... you put in 40V AC P-P, and put it on 5V/div you see nice +/-4 divs signal.
Then you realize a small something, maybe 100mV anomaly at zero cross. So, you put it on 50mV/div so anomaly would be at least +/-1 div (2 div P-P) vertically so you can have a better look. At this point, full screen dynamic range of input is +/-500 mV (1V P-P).. You are overdriving it 40X... On top and bottom.. not where you are looking.   

Old analog CRT scopes had different circuitry, working on much higher internal voltages (even full solid state integrated ones). They had to actuate electrostatic plates on CRT tube. That gave them much larger linear dynamic range and also many times a more graceful clipping and recovery... They were better at this. 

Regards,

Sinisa

[rhb]

Heating of the input attenutator network, especially the capacitors, is a big deal.  Someone at Tektronix derived an analytic solution which was held as a closely guarded trade secret for a long time.  There was a reason "scope" and "Tektronix" were synonymous for a long time.

I have an article about it in a book somewhere, but it was not in the ones by Jim Williams and Bob Pease where I expected to find it and my library is a bit too large to justify a search.  As I recall it was a Bessel series expansion.  I remember thinking when I read it the article what a tremendous amount of work (i.e. months) it must have been to derive.

Each setting in the attenuator network is expecting a particular voltage range.  So if you drop off the higher voltage attenuators the remaining stages get severely overdriven.  Unlike analog scopes which had to generate the deflection plate voltages, a DSO reduces the voltage before feeding the signal to the ADC if it is above the ADC range.  An amplifier is only needed for voltages below the ADC range and to buffer the input. 

Sadly despite the fact that component level repair would be much easier on a single board SMD DSO than something like a Tek 465, the tradition of providing proper service information to the user has died.  As a consequence, few people are aware of the details of DSO circuit designs.

Getting a believable presentation of something like high frequency ringing after a 40 V step is not easy.  The only way I can think of to do that would be to apply an external bias,  AC couple the scope and use an external comparator to switch the signal on and off.  Otherwise the opposite swing would cause heating and non-linearity.  This would be rather difficult to implement if the step had a fast rise time.

[seronday]

For anyone who may be interested, it is possible to install a license file in the GDS-2000E series to upgrade the bandwidth.

I have recently used an experimental license file generator created for the GDS-1000B series by EEVblog member,   wgoeo  ,  to upgrade a GDS-2074E (70Mhz) to 200Mhz bandwidth.
The Info on the license file generator and how to use it, is in this thread :-
     https://www.eevblog.com/forum/testgear/possible-gw-instek-gds-1000b-hack/

When the license file is installed, an additional entry appears on the " Utility - System - System Info " screen.

Also a new file called  "  OptionConf  ", is created in the GDS-2000E file system under   /home/dso .
The option license can be removed by deleting or renaming this file.
Note that in the current FW ver 1.34,  ssh access has been disabled, so it is necessary to install an earlier FW version to gain access to the internal file system.

The attachments show the additional info on the System Info screen and the frequency response before and after the 200Mhz license file installation.

Regards

[rhb]

How did you collect the data for the plot?

[seronday]

Using the DSO voltage measurement function to measure the displayed signal level from a calibrated RF signal generator at various frequencies.
The reference is the level measured at 1Mhz.

[CC555]

I can't update the oscilloscope to v1.35: https://www.gwinstek.com/en-global/products/detail/GDS-2000E

it seems the .upg file is not valid for my oscilloscope (gds-2000e)  , upgrading by safe mode.

[wraper]

I can't update the oscilloscope to v1.35: https://www.gwinstek.com/en-global/products/detail/GDS-2000E

it seems the .upg file is not valid for my oscilloscope (gds-2000e)  , upgrading by safe mode.

Are you sure you did not download wrong or corrupted file? Sounds like you want to brick your scope.

[CC555]

Thx for the reply,

I think the .upg file is good, I downloaded it several times.

Did anyone update the oscilloscope to v1.35?

Current version in my oscilloscope v1.34.

Thanks

[rhb]

I just updated my GDS-2072E from 1.35b4 to 1.35 using "safe mode".  The MD5 checksum is:

9a38e230f4a2168fd602aaef8352a71a  FW_GDS-2000E_V1_35.upg

I had the USB drive inserted when I turned it on.  I had only one file on the drive.  In "safe mode" the update file *must* be the first file in the output from "/bin/ls".  If it's not, the program to update the FW can't find the .upg file. I ran into that issue at some point, so I now just have a single file on the USB.  It's a practical use for a 512 MB SD card.

Has anyone sorted out how to generate the CRC for a .upg file?  It doesn't seem to match any of the CRC programs on the scope.

[CC555]

Thanks rhb, I am going to try to update the oscilloscope again.

[CC555]

I did the update using a 8 GB memory instead a 16 GB memory, with the 16 GB memory didn't work.

[rhb]

I decided to fool around with the BW hack from the hack thread.  It works.  I've got a 40 pS pulser from Leo Bodnar.  Excellent piece of kit!

 I reverted to 1.28 so I could log in via ssh, applied the 200 MHz  BW upgrade. The 300 MHz is BS as that's above Nyquist of 250 MHz. I did the 200 MHz and then the 300 MHz just to see if there was a difference in single channel mode when Nyquist is 500 MHz.  No change.  All the other option hacks from the 2000B are already installed.

Interestingly, the 2074E as delivered had the same BW as the "100 MHz" after the hack.  It's actually about 85 MHz by the traditional rise time metric, so  not better than 70, but not quite 100.

I then updated the 2072E to 1.35 again and saved screenshots from the 2204E and the 2072E.

I need to quite fooling around like this and start working on FOSS FW for these.  That's what I bought the 2072E for.  I'm not about to risk bricking my MSO-2204EA.

[nctnico]

I did some experimenting with the license generator from this post on my GDS-2204E:
https://www.eevblog.com/forum/testgear/possible-gw-instek-gds-1000b-hack/msg1256089/#msg1256089
I downgraded to version 1.21 because version 1.34 doesn't accept the licenses but version 1.21 does. The power analysis options doesn't seem to enable anything but the 300MHz option does work! The bandwidth went from 288MHz to 314MHz and now there is a 200MHz bandwidth limit option. Sure the 300MHz option isn't useful when using 4 channels but for that you can still enable 200MHz. All in all not bad! 

Edit: the 300MHz remained when I installed firmware version 1.35 (currently the most recent).

[rhb]

Interesting.  I wonder if they removed the 300 MHz option between 1.21 and 1.28.

How did you measure the BW?

[nctnico]

Interesting.  I wonder if they removed the 300 MHz option between 1.21 and 1.28.

How did you measure the BW?

I installed version 1.35 afterwards and the 300MHz option is still there. I measured using an RF generator and a 50 Ohm feedthrough. I referenced the bandwidth from the signal amplitude at 20MHz.

[Mortymore]

On the 2074E (70MHz) with v1.21b6 the 300.lis is accepted but it does nothing that the 200.lis hasn't done yet.
After reboot it still says "Bandwidth upgraded to 200MHz. Which is just fantastic per se.
Maybe the 300MHz.lis works on the 100MHz scope. In the 200MHz is confirmed.

This is just me guessing...
Because there's an option in the hack to enable "Power Analysis"... isn't this option directed to the 3000 series scopes?
Any of the 1000 and 2000 doesn't has the ability to enable this tool, right?
Correct me please if I'm mistaken.

[nctnico]

I tried to install the power analysis license as well but it is not accepted. I'm not aware of any power analysis option being available. Can you check if you have a 200MHz bandwidth limit in the channel bandwidth menu? If it is there I'd expect that the 300MHz option is active.

[Mortymore]

Hi nctnico!

There's no 200MHz bandwidth limit option available. Only "Full", 20MHz and 100MHz.

About the "Power Analysis" option available in the lis.html... if it's there... my suspicion is that the hack may also be directed to the 3000 series. I think those have that tool as optional.

[rhb]

With all channels enabled, they're being sampled at 500 MSa/S. You can only do 300 MHz on two channels of a 4 channel, or one of a two unless all you want to do is look at it.

I was quite amazed to discover that a 2 GS/S scope channel can go as high as 1.5 GHz, a mere 50% above Nyquist.  It works.  It's a LeCroy and it gives  220-230 pS with one of Leo's pulsers.  Both sampling at 2 GSa/S and 4 GSa/S.   But the overshoot is "<20%"!!!

[Mortymore]

For those who haven't noticed, there's a new firmware version 1.37 for the GDS-2000E series (released at 2018-11-28)

https://www.gwinstek.com/en-global/products/detail/GDS-2000E

Version 1.37
1.Fixed the  crash issue for  the "bus decode" function
2.Fixed the "peak table" data storage issue for the SA mode.
3.Added the new "SA:LIST? SCPI" command.
4.Changed the data transmission format of "SEGMents:MEASure:TABle:LIST?" (SCPI command) to  the "block data" transmission format.
5.Changed the data transmission format of " LISTer:DATA?" ( SCPI command) to the "block data" transmission format.
6.Improved the measurement high-low(histogram) calculation method.
7.Fixed the correspondence problem of frequency value for the  Math FFT Detail CSV of "Save All Waveform" function.

Have a great 2019

[Mortymore]

This scope doesn't stop!

Yesterday I updated to the latest firmware version v.1.38 (2019-01-26) and manage to activate a licence for the Spectrum Analyzer option.

After fooling around a bit with the SA functionality, I realize that I have a 70MHz oscilloscope (IDS-2074E), "upgraded" to 300MHz (EDIT: upated from 200MHz), "upgraded" again to a Spectrum Analyser with a frequency span of 500MHz 

Attached 3 screen captures in SA mode. 2 of them from a 60MHz sinusoid (SDG1032X) with a span of 500MHz, and another from a 1MHz sinusoid with a span of 10KHz. Neat 

Stretching the rope, I wish Santa could make FRA APP work with an external function generator so it could run on the GDS scopes.

[nctnico]

Hmm, you didn't install the 300MHz license key? 

[Mortymore]

Hmm, you didn't install the 300MHz license key? 

It didn't work on my 70MHz scope. But I'm happy with the 200Mhz anyway.

PS: What is the frequency span of the SA on your stretched 300MHz scope?

UPDATE: Now I manage to have the 300MHz working. But had to clear first the file OptionConf to clean the 200MHz BW licence instaled

[nctnico]

I have not tried but I'm sure the span will be the same because 1Gs/s is the maximum samplerate either way.

[rhb]

Stretching the rope, I wish Santa could make FRA APP work with an external function generator so it could run on the GDS scopes.

It would be nice if the FRA app were usable on the units *with* AWGs.  A scan from 13-14 MHz at the maximum sampling of 90 samples per decade produces two points.  So no way to find the resonances of a crystal with that.

I was able to sync it with  a 33622A but not consistently.  Despite several hours of fiddling, I was never able to get the series response of a crystal such as an 8560A SA w/ TG will produce.  It's probably because the various operations are not operating in the same clock domains.

[Mortymore]

Oh well!!... An SDS1104X-E to marry with the SDG1032X would solve the Bode plot problem   

[rf-loop]

Oh well!!... An SDS1104X-E to marry with the SDG1032X would solve the Bode plot problem   

Yep.

SDG1032X -- Xtal---SDS1104X-E


Span 5kHz


Span 500Hz

[nctnico]

Stretching the rope, I wish Santa could make FRA APP work with an external function generator so it could run on the GDS scopes.

It would be nice if the FRA app were usable on the units *with* AWGs.  A scan from 13-14 MHz at the maximum sampling of 90 samples per decade produces two points.  So no way to find the resonances of a crystal with that.

I was able to sync it with  a 33622A but not consistently.  Despite several hours of fiddling, I was never able to get the series response of a crystal such as an 8560A SA w/ TG will produce.  It's probably because the various operations are not operating in the same clock domains.

You should be able to do this if you have a generator which can step through a range of frequencies with a certain dwell time. Most RF generators support this.

[Mortymore]

You should be able to do this if you have a generator which can step through a range of frequencies with a certain dwell time. Most RF generators support this.

Are you referring to something old school with sweep mode and trigger from a generator?
That actually works a bit better than I recall of my analog scope, with this digital 2000E, if the display is set to infinite persistence, because it will keep painting the screen, but that's to basic and calculation tedious comparing to the neat Bode plot of the Siglent scope (for instance).

Just for fun, placed a simple RC low pass filter on a bredboard with a cut-off frequency of 4KHz.
The generator 1032X sweeps from 10Hz to 100KHz on CH1, and on CH2 outputs a square wave of 1Hz, for triggering the scope on CH2.
Made 3 screen captures, one with a linear sweep, and the other 2 with logarithmic sweep, and the persistence in the last one was set to infinite, so after a minute or so the area to plot is filled. The vertical position on CH1 on the scope was deliberately set so roughly only the positive half of the wave is shown on screen... Bode plot alike.

This is just a "poor man" Bode plot. 

PS: installed the FRA APP but it kicks me off right from the start, stating that I don't have a generator in my scope, so I can't even take a look at it. Uninstalled the APP, since was of no use for me.

A side note: I came from a long "dark age" in electronic equipment and only recently acquired a decent digital scope and generator with some bells and whistles that many of you master here. Be forgiven if I say something idiot or don't quite get the picture here and there.

[nctnico]

You should be able to do this if you have a generator which can step through a range of frequencies with a certain dwell time. Most RF generators support this.

Are you referring to something old school with seep mode and trigger from a generator?

No, use the SA mode in peak hold mode (which can be set to averaging).

[Mortymore]

Having played a bit with the Spectrum Analiser option, after "re-upgrading" my 2074E from 200 to 300MHz, found that the higher frequency that can be set is 750MHz, but the span remains the 500MHz (so it has to start at 250MHz). Though it may seem a nice feature having the ability to shift the frequency range 250MHz up, it actually has no use since there's nothing shown in the display above 500MHz (see picture attached: GW2000E_SA_750MHz.PNG). Am I doing doing something wrong? Is that a bug? Or am I missing some useful point for this to be like it is?

Something that I find annoying with this scope is the "supra imposed windows", like with the feature "statistics" or a "table". Seems that Rigol and Siglent (series 1000) manage to re-arrange the windows so things fit or having some degree of transparency, don't block much the signal to be viewed. GW just place a opaque board, blocking everything (GW2000E_SA_Table.PNG). 

Even in the XY mode were half of the screen is empty, the window for statistics has to fell on top of what you are trying to see (GW2000E_XY_statistics.PNG). 

Just for comparison, attached 2 random images that can be found on this forum, posted by other members (hope they don't mind) from a Rigol and a Siglent (series 1000). I think they serve the purpose to show a better arrangement than in the GW2000E scope.

Even harder to get is the fact that the GW has an 8" screen and the other scopes on comparison are 7" (though with the same resolution), besides the fact that the Rigol has the icons associated to the side buttons always visible, stealing screen area. I believe there's some margin for improvement in the GW UI. But by all means, correct me if I'm wrong.

I love this scope, but I'd like it to be as prefect as possible. 

[nctnico]

GW Instek could improve the screen layout but then again the fonts used by Rigol and Siglent are much smaller which may affect readability.

BTW if you look at the SA screen you'll see it stops measuring at 500MHz.

[Mortymore]

GW Instek could improve the screen layout but then again the fonts used by Rigol and Siglent are much smaller which may affect readability.

Yes, the fonts in GW look written in Bold format. Good for those like me that start to need glasses to read. But even then, cases like the screen in XY mode could have better display placement, some windows without border, some transparency or unnecessarily big size of the window. Cases that only a line of data is represented on a table, and the windows fills most of the screen. Well, maybe it's just me being picky, but i like to see data side by side with an image to get a better interpretation of graph vs data. Even an HP39GS graphing calculator has similar feature.

BTW if you look at the SA screen you'll see it stops measuring at 500MHz.

That's exactly my point. What's the purpose of extending the SA up to 750MHz if it's useless above 500MHz? I would call it a bug. Reminds me a speedometer with a scale up to 220Km/h in a car that only reaches 160. 

[rhb]

Having played a bit with the Spectrum Analiser option, after "re-upgrading" my 2074E from 200 to 300MHz, found that the higher frequency that can be set is 750MHz, but the span remains the 500MHz (so it has to start at 250MHz). Though it may seem a nice feature having the ability to shift the frequency range 250MHz up, it actually has no use since there's nothing shown in the display above 500MHz (see picture attached: GW2000E_SA_750MHz.PNG). Am I doing doing something wrong? Is that a bug? Or am I missing some useful point for this to be like it is?

The 750 MHz label is a FW bug.  It cannot go above the 500 MHz Nyquist frequency.  And the anti-alias filter limits it to around 350 MHz *if* they designed it properly.  It gets quite difficult to devise test cases. 

FWIW the step response using a 40 ps rise time square wave is exactly the same for the 200 & 300 MHz settings.

I need to do some more testing, but at the moment I'm playing with a Tek 11801 sampling scope in preparation for measuring actual clock skew in the Zynq vs what Vivado calculates.  The SD-26 heads sitting at the post office have a 20 GHz BW. So I'll be able to look at the skew of all the 8 data lines in a PMOD port to a resolution of 2-3 ps.

[nctnico]

BTW if you look at the SA screen you'll see it stops measuring at 500MHz.

That's exactly my point. What's the purpose of extending the SA up to 750MHz if it's useless above 500MHz? I would call it a bug. Reminds me a speedometer with a scale up to 220Km/h in a car that only reaches 160. 

It is not a bug. How can you get the maximum frequency in the center of the screen otherwise? Regular spectrum analysers usually allow similar settings.

@rhb: I have measured an increase in bandwidth on my GDS2204E after the 300MHz upgrade (IIRC from 270MHz to 320MHz) but it is probably too small to really see in the step response.

[rhb]

BTW if you look at the SA screen you'll see it stops measuring at 500MHz.

That's exactly my point. What's the purpose of extending the SA up to 750MHz if it's useless above 500MHz? I would call it a bug. Reminds me a speedometer with a scale up to 220Km/h in a car that only reaches 160. 

It is not a bug. How can you get the maximum frequency in the center of the screen otherwise? Regular spectrum analysers usually allow similar settings.

@rhb: I have measured an increase in bandwidth on my GDS2204E after the 300MHz upgrade (IIRC from 270MHz to 320MHz) but it is probably too small to really see in the step response.

My 8560A sets the center frequency  to 2.9 GHz, but issues an error message that 30 Hz RBW is incompatible and basically does nothing.  It's an analog instrument, so a bit different.  An upper limit above Nyquist on a DSO is just simply silly.  It's neither possible nor useful.  Like the Instek FRA app, it's a demonstration that the programmer didn't have a clue what they were doing.

Were those the -3dB points?   I've not tested it with my 8648C calibrated against my 438A.

[nctnico]

These are the -3dB points I measured (the numbers are from the top of my head so please look up a previous post to verify).
And again I don't think it is silly to end the trace half way the display. On spectrum analysers the center frequency is often used to have the frequency of interest in the middle of the screen so it is very easy to see where a frequency component is at. If the trace would be stuck to the right side of the screen you'd have to 'count squares' or or use the cursors to figure out what frequency you are looking at. The way GW Instek has implemented it is just convenient to use. The scope isn't trying to measure anything over Nyquist.

[Mortymore]

Thanks for your clarification nctnico.

Let it be this then:

.. Or am I missing some useful point for this to be like it is?
...

[Krisztián]

It would be nice if the FRA app were usable on the units *with* AWGs.  A scan from 13-14 MHz at the maximum sampling of 90 samples per decade produces two points.  So no way to find the resonances of a crystal with that.

Last weekend my buddy and me measured a 9 MHz crystal in FFT mode with the help of a BG7TBL noise source. Persistence was set to infinite. I think you'll not get the necessary resolution until switching to 1M record length. Spectrum averaging by oscilloscope would be nice. That was a GDS-2104E oscilloscope.

[nctnico]

If you use the key generator (see the GDS1054B thread) then you can use the SA mode which does support averaging multiple FFT traces.

[Krisztián]

In SA mode you'll get only 5kHz resolution at 9MHz because record length is fixed at 10k.

[nctnico]

Not on my GDS2204E. It seems in spectrum analyser mode the record length is fixed to 100kpts.


This image was created using the max-hold function and using an RF generator to step through the frequencies. Yes, this is a slow process... but it works.

Edit: the record length may be bound to the RBW.

[rhb]

I used an AWG (don't recall if it was the 2204EA or the 33622A) to sweep a 40 MHz 3rd overtone xtal from 13.2 to 13.6 MHz.  The Instek SA function displays the series notch as a peak. 

I'll add a screen shot to this post later.  I *think* I saved one, but I'd have to root around to find it and it was fairly tedious to set up if I have to do it over.

The Instek record length is fixed at 100 Kpts.  The RBW then depends upon the frequency span.

I plan eventually to get a Siglent SDS1104X-E and do a series of head to head comparisons for the same measurements with the Instek MSO2204EA, Owon XDS2102A and Rigol DS1102E.  All accompanied by the correct answer using my HP and Tek gear.

But at the moment I am having far too much fun with a Tek 11801 using the calibrator and an SD-26 sampling head.  In particular, I've been doing TDR on cables, connectors, terminators etc. I'm amazed at how much detail I can see with 20 GHz of BW.

[Mortymore]

Something that I find annoying with this scope is the "supra imposed windows", like with the feature "statistics" or a "table". Seems that Rigol and Siglent (series 1000) manage to re-arrange the windows so things fit or having some degree of transparency, don't block much the signal to be viewed. GW just place a opaque board, blocking everything (GW2000E_SA_Table.PNG). 

Actually GW-Instek have windows transparency, but in the 2000A series.

It can be seen at 6m30s on this video when the statistics option is enabled. Or for instance in the submenu options, scroll down windows. (buttons on the right side of the LCD, when they have submenu options)



On Dave's video "EEVblog #474 - GW Instek GDS-2000A Series Oscilloscope Unboxing & First Impression" it can be seen also throughout the video.

So why not on the 2000E?

[Mortymore]

New firmware update to Version 1.39 (7 Nov 2019)

Version 1.39
1.The SPC function is supported under roll mode now.
2.Added new erase disk function that can empty all the saved image files which located on the DSO.
3.Fixed UART parity bugs.
4.Fixed ¡§Adv math¡¨ operation issue.

EDIT: Why "Erase Disk" and "1 of 3" has a different font?

[Mortymore]

I was to put this info here earlier, but only now it goes.   

Some posts ago I complained that there's a lot of free space in the screen when in XY mode. It turns out that it can be filled if cursors are used.

BC547C: hFE=575



Did some experiments using the XY mode to trace some BJT transistor curves, taking advantage of a 2 channel AWG.

TUT had a 100k base resistor and a 50 (100//100) at emitter.
Generator CH1 connected to the TUT collector and X (ch1) on the scope.
Generator CH2 connected to 100k base resistor.
TUT emitter connected to 50 and Y (ch2) on the scope.

For NPN
CH1: Ramp up, 8kHz, 10Vpp, 5V DC offset
CH2: Stair up, 1kHz, 4Vpp, 2V DC offset


BD139: hFE=292



For PNP
CH1: Ramp down, 8kHz, 10Vpp, -5V DC offset
CH2: Stair down, 1kHz, 4Vpp, -2V DC offset
EDIT: CH1 and CH2 on scope were inverted

BD140: hFE=298



Took advantage of the fact that it can be selected a current probe for input, and so on  channel2 (Y) of the scope, current values can be shown, instead of Volts.

EDIT: Edited to replace images along the post

[Vytautas]

Hi, guys. Do we have a high resolution image of the AWG board used in any of the GDS-2000E scopes? Could anyone post it, pleeeaaase...

[Mortymore]

New firmware v1.40

1.Added German menu
2.Added Portuguese menu (I didn't ask for this) 
3.Fixed UART minor bug

Haven't tested yet. But would like to see at least a uniform font style.

And since I'm here, I'll ask a couple of things that I probably now the answer for, but...
Is there a GW scope/AWG pair to do a Bode Plot being the instruments independent, like Siglent has?
And following what I think is more or less the Vytautas idea: It could be added an internal AWG board to a GDS-2000E? (I realize that the case has no holes for the output BNCs)

EDIT:
They changed the font of "Erase Disk" 



But as I suspected, the menus in Portuguese... 



In some cases they had to change the font size to fit, in other, almost out of the box, some text was not translated as the "Erase Disk" (Apagar Disco)... I never intended to change the language from English, but it's good to see the effort to reach other markets.
Does anyone knows if GW has a good market on Brasil? When I see stuff like this in Portuguese language, it's never actually because of the tiny market in Portugal, but because of an eventual much bigger market on Brasil that justifies the effort of translation.

EDIT2:

Wile browsing through the scope I don't recall I've seen the Function Module "Fixed 4 Measurement" before.
Have I been distracted or something?  What does this do? Anyone has this unlocked?



EDIT3: The font type and style of the information above the horizontal menu buttons (yellow lettering) is different in the English to the Portuguese layout. In this regard I prefer the arial (maybe) font and not being bold, as seen in the Portuguese layout.
GW, can we have that font style in the English layout in the next firmware version? 

[Ruggo]

Hidey Ho.
Just got myself a gds-2074e.
When I tried the hack using the html code generator, all options failed.
Scope cam with F/W V1.34

Am I right in assuming that I need an older F/W and if so, any ideas where i can get it from?

TIA

[Mortymore]

New firmware for GDS-2000E 2020-04-07

Version 1.41

1.Fixed Datalog error at low frequency signals.
2.Fixed the Portuguese translations.
3.Fixed the Japanese translations.
4.Fixed Russian menu bug.
5.Fixed the SA menu freeze bug

https://www.gwinstek.com/en-global/products/detail/GDS-2000E

[Mortymore]

OpenWave v1.05 2020-04-13

A "continuous" option for capture was added.

[PeterKlop]

Based on what i learned on this forum and got from both his thread and the GDS1000B thread, i recently bought a GW Instek GDS-2074E. I tried the license generator and it worked easy and flawless. Thanks to you all!

I had no idea what to expect from trying to 'hack' this scope with the tools that are provided in these threads.
For the people comming after me i want to tell what i experienced.

The scope came with firmware 1.34. I download the lis2.txt from the gds1000b thread, renamed it to *.html to turn it into a webpage and then loaded it into my browser. Then i put the serial number from my scope into the page, clicked "generate" and after that i could save the license files from that webpage. I put those on a usb stick.

With the scope on, i put the usbstick into the scope, then wend into the menu 'Utility' > 'File Utilities'. The scope showed the license files that i had on the usb stick. I then selected one of these (the bandwidth upgrade) and the scope installed it. Then i had to turn of the scope. I repeated this for the other license.

The license for 300 Mhz is installed and also the Spectrum Analyser. I did not check the bandwith yet but the license is visible on the System Info page. The spectrum analyser works very good and is accessible via the Option button.

After this i download the latest firmware from the GW Instek website (version 1.41). I did put that file on a otherwise empty usb stick. As instructed in the manual i wend on to install the formware by turning the scope on while rotating the Variable knob. The scope couldn't read the firmware file, eventhough it had read the license file earlier (i had those removed from the stick now). I tried another usb drive and this one could be used. After the installation of the new firmware the licenses remained and everything worked fine.
It was really that easy! )

Oh wait, forgot to mention; I did by the way first make a mistake with entering my serial number and because of that the license files didn't work. I had misread a G for a C.

On to another but related topic;
I was thinking about either buying this GW Instek or the Siglent SDS2000X Plus. I'm sure the Siglent is a great scope and it has better specs (because it is a more recent scope). I do however own a few pieces more GW Instek gear and have gotton to appreciate the brand. The people behind it are also very responsive, both here in The Netherlands and the people from Taiwan, and i think that is of great value.

Anyhow, i also like the interface/screen/layout on this scope very much. That was an important reason to choose this one (i do this for fun after all). I do find the scope to be even much nicer to look at in person, then what i got from the pictures and videos. It's very enjoyable.

Anyhow, you guys saved me a lot of money and helped me with all your oscilloscope discussions. I read a lot of them and had fun doing that )

[Mortymore]

New firmware released Version 1.42 - 5/5/2021

Version 1.42
1.Fixed a freeze issue that could occur when switching to PictBridge printing.
2.Fixed a bug for the BUS trigger mode.
3.Fixed the Portuguese translations.
4.Add Multilanguage : Polish.
5.Add Multilanguage : Italian.
6.Fixed an issue when saving the "ALL CSV" data in Roll mode.
7.Adjusted the OCP maximum value for the Power supply function of the MDO series.
8.Adjusted the LCD panel readout of  vertical scale to the second decimal place.

[tv84]

@Mortymore, is it you that have been fixing the portuguese translation? 

[Mortymore]

@Mortymore, is it you that have been fixing the portuguese translation? 

No I haven't.
I don't even use it. 

EDIT: I don't use... the interface in Portuguese. Not talking about the scope

[piramida]

I read the thread with pleasure. Good oscilloscope. But there are points that can be improved.
1. One mathematical channel is not enough, at least two are needed.
2. Make it possible to select a math channel as a source for FFT and SA and for another math channel.
3. Make it possible to display only the necessary traces on the screen. For example, to be able to leave only a trace of the math channel without traces of the source channels.
4. Vertical zoom.
5. And from the category of fantasy. Make it possible to combine two input channels into one differential input.
It would be a mega device!

[2N3055]

I read the thread with pleasure. Good oscilloscope. But there are points that can be improved.
1. One mathematical channel is not enough, at least two are needed.
2. Make it possible to select a math channel as a source for FFT and SA and for another math channel.
3. Make it possible to display only the necessary traces on the screen. For example, to be able to leave only a trace of the math channel without traces of the source channels.
4. Vertical zoom.
5. And from the category of fantasy. Make it possible to combine two input channels into one differential input.
It would be a mega device!

Check Siglent SDS2000X+.

1. Yes
2. Yes
3. Yes
4. Yes
5. No and, correct that is a fantasy! 
6. It has 2Gsa/s sampling rate.
7. Much bigger screen. Touch screen.

[kcbrown]

5. And from the category of fantasy. Make it possible to combine two input channels into one differential input.

5. No and, correct that is a fantasy! 

I do have to wonder how hard it would be to make it a reality.  The scope already has hardware that is able to apply voltage offsets to the signal prior to amplification.  How hard would it be for the offset to include the voltage supplied by one of the channels?

I do agree that it's unlikely to become a scope feature (possibly because it's harder than it sounds?  The offset input mechanism itself would have to have a bandwidth that equals or exceeds the channel bandwidth for this to work, and that alone might be enough to nix the idea).  It'd probably be more expensive at a minimum, and it's probably better to spend the money on making the math trace fast enough and flexible enough that the absence of the differential input feature wouldn't be noticed.

[piramida]

... and it's probably better to spend the money on making the math trace fast enough and flexible enough that the absence of the differential input feature wouldn't be noticed.

Yes, a good solution.
Make it as a separate math channel, with vertical sensitivity adjustment with one knob. Solved by firmware. But all the same, I would like to be able to choose this channel as a source for FFT, SA or other mathematics.

Check Siglent SDS2000X+.

  I am also considering this device ... as well as the picoscope 5000 series. 

[nctnico]

5. And from the category of fantasy. Make it possible to combine two input channels into one differential input.

5. No and, correct that is a fantasy! 

I do have to wonder how hard it would be to make it a reality.  The scope already has hardware that is able to apply voltage offsets to the signal prior to amplification.  How hard would it be for the offset to include the voltage supplied by one of the channels?

For that you use a differential probe or pre-amplifier which you can put much closer to the circuit as well.

But still... there are oscilloscopes out there that have differential inputs. Some older Yokogawa DSOs for more specialised purposes come to mind.

[ibuski]

Hi, I have been just a reader of this forum's contents so far.  There are lot of valuable contents and I could gain much knowledge from your discussion.   In that manner you can call me a free rider.   I registered myself now just to say thank you to all members who enabled Hacking of GW Instek Oscilloscope, tested it, shared it and reported some successful experiences.   Actually I don't own GW Instek brand Oscillo, but Iso-tech IDS-2072E.   From its appearance, it was clear that they share the hardware.  Of course it's possible that some software modification differentiate them, but from my experiences, usually they don't spend such efforts.   So I tried GW Instek Hack on my Iso-Tech.   It worked without any issues!!   I could enable 200 MHz BW setting on my 70 MHz model.  Spectrum Analyzer application is now active as below.   For those who is wondering which digital oscilloscope to buy, I can suggest that Iso-tech works as same as GW Instek and you can hack it if it's the reason for you to buy GW Instek.   I guess probably you can buy Iso-Tech Oscilloscope with less money than GW Instek's equivalent model due to the brand's less popularity.

[Mortymore]

Hi ibuski!

Welcome to the forum.

I have an Iso-Tech also, and as you figured out, it's a GW-Instek rebranded.
You cant take your scope up to 300MHz if you wish. At first, I also "upgraded" mine only up to 200MHz, thinking that if it that was the max frequency commercially available, it should be the limit for the scope, but it's not.

For further reading about my adventure with the BW:
https://www.eevblog.com/forum/testgear/possible-gw-instek-gds-1000b-hack/msg2298513/#msg2298513

Have fun with your scope

[ibuski]

Hi Mortymore,

Thanks for your reply and welcome.
I recall there was somebody who verified compatibility (or identity) of GW and ISO, that's one reason I could try, but didn't find the particular post.  Now I see that was you!

I have only70 MHz probe and probably it doesn't make much sense to enable 300 MHz.   I work only on Audio signal and that's another reason.   Thanks for your comment anyway.

[Mortymore]

New firmware released V1.43 - 2023/06/14

Version 1.43
1.Updated Russian translation.
2.Fixed the issue that the data is not synchronized when the "DataLog" function saves pictures and waveforms together.
3.Fixed the FFT calculation when the memory length is 1k.
4.Fixed the related SCPI return string of measurement function.
5.Improved internal memory usage.
6.Fixed an issue with the "Math" function in calculating values containing Deskew value.
7.The Vertical scale display changed to 3 digits (supplemental change)
8.The channel's deskew range are expanded to +/- 1us.
9.Fixed DVM function screenshot problem.