-
Note to reader: go to the bathroom and get something to drink and/or some snacks first. This review is long!
Rohde & Schwarz RTM3000 review
Apparently I hit Rich's nerve and he had the Rohde & Schwarz branch in the Netherlands lend me one of these for a very long while to pick it apart:
After a short trip to the Rohde & Schwarz office in the Netherlands there is finally some R&S gear in my lab again. The previous bit of R&S gear I had was this RF generator:
Anyway, back to the RTM3000. There is already a thread about the RTM3000 series but here is a quick summary with the most important specs:
2 or 4 analog channels
1 external trigger input
2.5Gs/s and 40Mpts memory with all analog channels enabled
5Gs/s and 80Mpts memory with half the analog channels enabled
400Mpts in segmented recording /history mode
100MHz to 1GHz (software upgradeable)
64000 waveforms/s peak
10.1" 1280x800 TFT display with capacitive touch screen.
Stackable math functions (stackable)
10 bit ADC
Options:
16 channel MSO option (2.5Gs/s and 40Mpts memory with all MSO channels enabled, 5Gs/s and 80Mpts memory with half the MSO channels enabled)
Segmented recording up to 400Mpts in total (700k segments/s peak)
Protocol decoding & triggering
Waveform & pattern generator
Spectrum analysis mode
Power analysis
All in all on paper the RTM3000 looks like it is a very versatile instrument and a big step up from the RTB2000 series. R&S seems to position this scope in the segment where the Keysight 3000T, Lecroy's Wavesurfer 3000z and Tektronix' MDO3000 series are also part of the regular crowd. I think R&S did well to have way deeper memory in the RTM3000 compared to the direct competing devices (especially Keysight's 3000T series). Also the RTM3000 has 10 bit ADCs and the highest display resolution of all so you can get more details of a signal on screen (which is one of the primary functions of an oscilloscope after all).
By the way: the '10 bit ADC' text on the badge is photo-shopped into the picture above. Real units don't have this text so don't be alarmed if the text is missing.
The RTM3000 comes standard with 500MHz passive probes and when the MSO option is ordered with two MSO probes. I would have liked it when a front cover is also included. The 4 channel 100MHz base unit (RTM3004) retails for around 4,500 Euro. The fully optioned one (1GHz + MSO and all other available options) is in the 18,000 Euro ballpark. There is a reasonably priced option bundle (the RTM-PK1 option which costs 2500 Euro) which enables all options except for Audio (I2S, LJ, RJ, TDM) triggering & decoding and the MSO. A 4 channel 350MHz model including the option bundle sets you back by around 9000 Euro. In my opinion the RTM-PK1 option bundle is a no-brainer and besides being useful it probably also increases the resale value of the RTM3000 when/if it is to be sold or traded in. It seems R&S has listened to the complaints that their options are too expensive compared to the competition (including the aggressively priced oscilloscopes from Chinese B brands). BTW it seems Batronix has a special offer where you get the new R&S FPC1000 spectrum analyser for free if you buy an RTM3000 series with a bandwidth of at least 350MHz.
I'll be using the term 'RTM3000' to refer to the RTM3000 series and 'RTM3004' when I refer specifically to the 4 channel model or the 4 channel unit I have here.
Test plan
In order to get a good feel for what the oscilloscope is capable of I like to do a whole bunch of tests to try and get to the limits. I don't think I can test everything though. The user manual has over 770 pages and it is not long winded at all so there is a lot of functionality packed in the RTM3000!
Tests in random order:
- Bandwidth / aliasing
- Fan noise level
- Low level signal triggering
- Signal noise floor
- Overdrive recovery
- Math
- MSO inputs
- Protocol decoding, required over sampling and how much of the memory is decoded
- Long I2C messages
- Deep memory and decimating long traces onto the display
- Automatic measurements
- FFT function
- Spectrum analysis mode
- DMM application
- Search function
- Saving images and data
- Saving / recalling setup
- Cursors (outside screen?)
- Operating / using the scope
- Remote control abilities
- 10 bit usefulness
- Peak detect and roll mode
- Segmented recording and decoding
- Retaining settings
- Storing and manipulating reference waveforms
- Waveform generator
- Pattern generator
- Mask testing
- Self calibration/ self test
- Power analysis
- XY mode
- Waveforms/s
The secondary goal is to use the oscilloscope for a while to get used to the operation and in order to spot stability problems / usability issues.
Firmware version: 01.300 (although most of the initial testing has been done with version 01.200)
First impression
Despite the big 10.1" wide screen the oscilloscope is still compact. It has the same form factor as the RTB2000 series and the front cover seems to be exactly the same. With 39cm wide and 22cm high it doesn't take up too much desk space (the size is on par with my GW Instek GDS2000E and similar sized scopes like the Keysight MSOX3000T and Tektronix TBS2000 series). The biggest space saving comes from not having buttons around the screen and not having separate controls for each channel. The buttons are spaced relatively close together but not too close. Because the RTM3000 can be fully operated using the touch screen the buttons are more like a short cut keys to menus. I've found that most items showing a number (like the horizontal and cursor positions) can be tapped which then shows up a dialog which allows entering a number. That works way quicker than spinning a knob. The adjustment knob (in the analysis section of the front panel) and the horizontal position can use a better acceleration algorithm but due to the possibility to enter a number it doesn't matter that much. I rather enter a number anyway. Speaking of short cut keys: The top left corner in the display has eight short-cut-key positions which can be selected from 25 functions.
I would like to be able to have a choice from all the menu items here as well because some of them are useful to have as a short cut button on screen. On top of that you can also drag items from the main menu onto the screen creating even more short cut buttons although at the cost of screen space. This means the user interface can be highly customised to your own needs/wishes and maximise productivity. I'd almost say the physical buttons/controls are there to make the RTM3000 look like an oscilloscope but you don't really need them :-) I find myself using the touch screen the most and I'm only using the physical vertical and horizontal control knobs.
Despite the top menu bar and bottom status bar a large part of the display is taken by the signal display area. Horizontally there are 12 divisions, vertically there are 10 divisions. Due to the small dot pitch of the display vertical signal traces can become very narrow. The traces are fattened when the trace intensity is set to 100%. Still I like this way better than artificially fattened traces because fattened traces hide signal detail and waste screen space.
In case of multiple screens (like a list, FFT or measurements section) the height of the screen is freely adjustable so you can either have a big signal display area or show something else with more detail. I like this feature because it gives you a lot of freedom. I have been missing this on other oscilloscopes.
A while ago I had the chance to play a bit with an older (now obsolete) R&S RTO1000 series scope. This also has a touch screen but I found the user interface (screen layout and physical buttons) cluttered and disorganised compared to the RTM3000. It is clear R&S has created a new lean & clean user interface from scratch for the RTM3000 (and RTB2000).
The display itself is crisp with a lot of contrast. I have about 800 Lux blasting down from an overhead lamp. That doesn't make the display unreadable but I can see myself clearly in the display if I focus my eyes differently. At first I was afraid this would distract from using the oscilloscope but it turns out it isn't a big problem. A display with a matte finish would be better though. Another option would be to use a dark-on-light colour scheme. I have modified some of my previous scopes to have a dark on white colour scheme. After all a TFT screen doesn't have the disadvantage of masking dark areas like a CRT screen does. It may take some getting used to but hey, we all switched from light on dark to dark on light on our PCs a few decades ago!
What is nice is that a channel is automatically enabled and setup properly when a probe is plugged in (of course this only works with probes which can be detected using a probe ID pin). After using an input with a 50 Ohm source the channel reconfigures itself automatically when a 1:10 probe is attached. However there is a downside to this feature but more about that later.
As expected there is also a fan inside to keep the RTM3000's circuitry cool. R&S choose to use a relatively large fan (I estimate 92x92mm) running at a low speed. The noise isn't a nuisance even in a very quiet office. The BNCs on the front do get warm but not alarmingly hot. From the tear-down Mike (mikeselectricstuff) has done it is clear that the construction is different compared to the RTB2000 series. The main board is mounted parallel to the front and the BNCs are bolted to the chassis. The latter prevents plugging/unplugging probes and cables to flex the circuit board and also protects the circuit board in case the scope is dropped or bumped. Unfortunately these kind of accidents do happen and it is cheaper to bend a BNC back in shape than needing to replace the entire board and/or cause the oscilloscope to develop an intermittent fault.
Vertical
The vertical controls allow setting an offset based on divisions or absolute voltage. A voltage offset keeps the signal centred on the display when changing input sensitivity. A division offset keeps the zero level of the signal on screen. The voltage per division can be adjusted in 1-2-5 steps or variable.
The traces can be moved across the screen by dragging it using your finger. Before moving a trace it must be selected by tapping it. If the scope can't figure out which trace you are pointing to by itself it presents the user with the following question:
One of the interesting things I have found during this test are the over range indicators. When the signal goes outside the range small markers pop up in the channel status bar indicating overshoots and/or undershoots. The analog Hameg oscilloscope I used at my first employer some 20 years ago had the same feature. I used it to find invisible spikes in signals. I'm pleasantly surprised to find this very useful feature has trickled up into the RTM3000! No I really can't write 'trickled down' here!
This signal has a small spike going outside the screen at the bottom:
One of the things added to version 1.3 is an auto scale button to the channel menu. I don't like it because tapping it by accident takes a while to execute and it messes up the channel sensitivity settings.
Horizontal
The time per division can be adjusted in 1-2-5 steps (the minimum time/div is 500ps) or freely variable. Having a freely variable time base is helpful to display a signal across the full width of the screen. The horizontal reference can be set to left, centre or right. I like to have the zero reference (=trigger point) at the left because in most cases the interesting part of a signal is after the trigger.
There is also a setting to adjust when the RTM3000 switches over to roll-mode. For some reason the selection at what time/div the roll mode is switched on is in the 'acquire' menu and not in the 'horizontal' menu. I would expect this setting to be in the horizontal menu.
Going over the specs in the datasheet I'm missing a trigger jitter specification.
Probes and accessories
The RTM3000 comes with probes for each channel (2 or 4 pieces). The probes are fixed 1:10 and have tips with a pogo pin (a spring loaded pin). Accessories include a BNC adapter, colour coding rings, ground spring, ground ring isolator and an extra tip.
All probes adjusted to show a nice square wave:
However... after putting the colour coding rings onto the probes it turns out I'm missing an indication for the channel colour near the input BNC connectors.
Bandwidth / aliasing
Time to hook it up to my RF generator using some good Huber+Suhner RG223 coax to see what the frequency response and actual bandwidth is. With only channel 1 enabled (sample rate=5Gs/s) the bandwidth appears to be precisely 1GHz and not one Hertz more. I also measured the bandwidth on the other channels and channel 3 on my RTM3004 has a maximum bandwidth of 990MHz. The rise time on channel 1 is around 330ps so it seems the BW=0.35/rise time formula applies to the RTM3000 when the signal is sampled using 5Gs/s. The sin (x)/x reconstruction holds up nicely up to 1.1GHz at 2.5Gs/s which is what I'd expect so no drama here.
I've made some bandwidth graphs:
During testing with higher frequencies I also found another interesting phenomenon. When I input a signal above 1GHz (say 1.7GHz) the amplitude stays the same in the 500uV/div to 20mV/div vertical sensitivity ranges! I can change the sensitivity on the oscilloscope or amplitude on my generator: it doesn't make any difference. What the hell is going on here? At 505 MHz (odd frequency to rule out aliasing effects) everything seems to be normal.
And no, the problem isn't sin(x)/x reconstruction (note the sample rate!). The same happens in linear interpolation mode and my RF generator is also OK (checked to make sure).
According to R&S this effect is caused by a limited slew rate in the pre-amplifier stage. I assume this pre-amplifier is necessary to amplify low level signals so they match the ADC's input range.
Because of this effect I wanted to investigate a bit deeper to see if this would have an effect on actual measurements. I decided to feed a 275MHz square wave with sharp edges from an ADF4351 based generator with a 50 Ohm output into the RTM3000. A 275MHz square wave has strong harmonics at 825MHz and 1.375GHz but still looks like a square wave when using a 1GHz bandwidth. I used a 200mVpp and a 10mVpp amplitude. The latter is the 200mVpp signal attenuated 20 times (26dB) to simulate using a 1:20 low-Z passive probe.
200mVpp:
10mVpp:
To have some comparison with another oscilloscope I also measured the same signals on my Agilent 54845A (1.5GHz bandwidth):
All in all it doesn't seem like a big problem after all.
Acquisition modes
Compared to other oscilloscopes the RTM3000 has a lot of different acquisition modes. Besides sample, peak-detect, average, high-res and envelope mode there are also combinations of acquisition modes like high-res + average, peak-detect + envelope and envelope + high-res so you can really tailor the acquisition mode to your needs. Note that envelope and peak-detect are closely related. So related that on Japanese scopes peak-detect is usually called envelope mode but that is a different story. On the RTM3000 peak detect lets the ADCs run at full speed and you get the minimum and maximum in the sample interval. Envelope mode on the other hand uses multiple acquisitions (the number of acquisitions is adjustable - say it is X -) and shows the minimum and maximum of the last X acquisitions. I like that the acquisition modes allow some mix & match. I've been missing that on other oscilloscopes every now and then.
A good test when it comes to acquisition modes is to check whether peak-detect works in roll mode. For this I'm using 16ns pulses which are output once every second. The RTM3000 has no problem catching these even when set to a short record length (5kpts). Of course the exact shape of the pulses gets lost but at least you can see something is there and investigate further.
In average mode there is an issue though. When using averaging mode the averaged trace disappears when changing the time base in stop mode. I did not expect that but I have seen it before on other oscilloscopes so I guess it is something software developers working on oscilloscope firmware tend to overlook. This will need to get sorted though.
Triggering
In short: the trigger works very well. With 1 channel enabled I can get a stable trigger on a 2GHz sine wave. Using a 10MHz signal and bandwidth limit set to 20MHz the RTM3000 is able to trigger on signals as small as 350uVpp with the trigger set to the most sensitive setting. By the way: the trigger status also shows how much time has elapsed since the last trigger. Not bad!
When in auto-mode there is a short delay before the RTM3000 goes back to free-running mode after a signal caused a valid trigger. This used to be a typical Tektronix feature but other manufacturers seem to copy it nowadays.
What I'm missing is a small trigger frequency counter somewhere in a corner of the screen. There is a trigger frequency counter (which also shows the period) available but that sits in a big windows which overlaps a significant part of the display.
Signal noise floor
Signal noise floor is important to look at when it comes to oscilloscopes intended for general purpose use. Some older high bandwidth scopes have so much noise that they become less useful for looking at lower frequency signals especially when it comes to small details of a signal or when making accurate cursor measurements.
I'm not looking for how to get the lowest noise here but just to get an idea of what to expect in a normal usage scenario. For this test I selected a 1Mpts record length in order to get a lot of samples.
500uV/div 50 Ohm full bandwidth:
500uV/div 1M Ohm full bandwidth:
500uV/div 50 Ohm 20MHz bandwidth:
500uV/div 1M Ohm 20MHz bandwidth:
This looks all nice and dandy but 500uV/div isn't a setting you'd use daily so let's take a look at the noise level with full bandwidth at a more useful V/div setting with a probe attached:
At least that doesn't result in a wide band of noise across the screen so thumbs up!
10 bit usefulness
One of the cherries on top of the RTM3000 is the 10 bit ADC. Now how useful is having a 10 bit ADC instead of an 8 bit ADC? One of the limitations of getting a large dynamic range from an ADC is the sample clock and sample&hold jitter. The higher the sample rate, the higher the signal frequency and the larger the number of bits the smaller the jitter must be. To get 10 bits at several Gs/s the jitter needs to be in the femtosecond territory. All in all it is unlikely you can get 10 bits at 5Gs/s. So far the bad news. Now on to the good news. If the signal is bandwidth limited the clock jitter isn't that important and you can get more bits more easily. To show this I have measured similar signals using the RTM3004 and a GW Instek GDS-2204E.
The signal comes from a 125 kHz RFID write operation which has a high dynamic range because the write pulses are very large compared to the received data.
Zoomed horizontally on a write and a response:
Now zoomed in vertically:
I think the screen shots speak for themselves. The small part of the signal can be enlarged to study it in greater detail on the RTM3004 but on the 8 bit scope zooming in vertically to see more becomes futile quickly.
The deep memory also makes a measurement like this easy. There is no need to really trigger on a specific part (shape) of the signal. Just press stop when you see the signal of interest and you have all the details you need.
Overdrive recovery
It's always interesting to see what happens when the amplitude exceeds the input range. For this I used a 3.5Vpp square wave from the generator:
At 100mV/div I can still get a nice waveform but at 50mV/div (I can hear a relay click when switching between 50mV/div and 100mV/div) the signal gets severely distorted.
Display modes
An interesting feature of this scope is that the colour grading method (or default trace colour) can be selected separately for each channel. I find this a very good choice. On other oscilloscopes where the colour grading setting applies to all channels it is easy to lose track of which channel is which.
There are several colour palettes to choose from (temperature, fire and rainbow) and by setting a trace to inverse colour seldom occurring events can be highlighted.
Of course the persistence can also be adjusted. There is one odd thing though: the persistence continuous to fade out when the stop button is pressed and the signal should be frozen. Also the infinite persistence trace disappears when the cursors are enabled/disabled. That shouldn't happen! What if you let the scope run for several hours and you forgot the enable to cursors?
Saving images
Saving images can take up 10 to 15 seconds but it is not consistent. When it takes long to save an image the signal update rate slows down a few times during the operation so it seems saving an image is a heavy burden on the CPU. I have definitely seen faster image saving and I would expect saving an image to be done in a few second tops on a modern oscilloscope. I've used the same USB stick (from Kingston so no cheap crap) I use for all my test equipment so that can't be the problem. After further usage it appears that doing a screen shot works fastest (a few seconds) with the acquisition stopped. After a screen shot a message appears on the screen which can be tapped (clicked on a touch screen?) away.
Remote control abilities
The RTM3000 has a USB and Ethernet connection. USB can be used in three different modes: TMC (test & measurement class), VCP (serial port) and removable storage using MTP (media transfer protocol). TMC and VCP allow doing remote control using SCPI commands. I like the fact that R&S also included VCP (serial port) because a serial port is among the easiest and hassle free interfaces to use. Unfortunately the removable storage doesn't work out of the box on my Windows XP and older Linux system. On Windows 7 and the newest Debian Linux it does work out of the box and it allows to access the RTM3000's internal flash storage, an on-demand storage (where files contents like a screen dump is created on-the-fly) and a non-volatile storage to transfer temporary files.
The Ethernet connection has several functions as well. It can be used for VXI and SCPI over telnet for remote control but the RTM3000 also has a built-in web server. I played a bit with that and with Firefox (version 59 on Linux) I have no problem whatsoever to use the web interface. The web interface allows making screen shots, controlling the RTM3000 remotely through a virtual front panel and last but not least testing SCPI commands. The latter is very useful for testing commands while developing your own software. IMHO R&S made the right decision to go for a web interface and not develop a separate software package. A web interface works on every reasonably modern computer with a recent browser. I can even control the RTM3000 from my Android based Smartphone (within the limits of a small display of course but it works).
The user manual has a comprehensive list with SCPI commands and also shows a few examples on how to set the RTM3000 up remotely and retrieve data.
Measurements
Like many modern DSOs the RTM3000 has a wide variety of measurements. When selecting a measurement there are clear thumbnails showing what the measurements do. After adding a measurement the result is shown in the bottom left area of the screen. By touching a measurement you can change the type and the source channel. That is a huge improvement compared to other oscilloscopes I have used so far. On those existing measurements need to be removed and recreated to -for example- use a different source channel. On the RTM3000 you can simply change an existing measurement. Statistics are also available which show the average, minimum, maximum and standard deviation. With the menu at the right side collapsed (tap the R&S logo) there is a button to reset the statistics at the right side of the statistics box.
When the gating is enabled a blue bar is drawn over the part of the acquisition which is used for the measurement.
When doing some other testing it seems the measurements are taken on sub-sampled data and not directly from the acquisition memory. Only when I zoom in the information becomes available. In this case I scroll through a frequency sweep and use the gating to find the frequency for the section currently under the gating bar. This seems like a perfectly legitimate use to me but I get no reading from the measurement.
Something similar happens for a rise-time measurement on a square wave. At some point the measurement value is no longer available (after being marked as inaccurate). At least the measurements don't show bogus values.
Zoom mode
The zoom mode allows to zoom in both horizontally and vertically (even below 500uV/div) so every detail of a signal can be shown.
MSO inputs
The 16 digital inputs are divided over two 8 bits pods. The MSO pod is connected to the RTM3000 through a sturdy but flexible ribbon cable with 29 pin HDMI type B connectors on each end. Even though the connectors are on the side of the oscilloscope they don't stick out very far. The pod itself is a separate box and because it gets warm it seems to be an active probe. An LED indicates which channels (0-7 or 8-15) are connected to the logic probe POD. Here is a picture with the RTM3000 MSO pod and the output pod from a pattern generator.
The connectors are standard IDC connectors so they fit well on standard 2.54mm (0.1") header pins but are less suitable for thinner pins or wires.
A set of micro grabbers (good down to SOIC packages) is also included.
Using the menu the digital signals can be set to small:
... or full screen:
However the digital channels can be scaled in smaller steps using the 'Scale' knob and the vertical position knob can be used to move the digital signal. It would be nice if the larger/smaller buttons in the menu could do several steps instead of only small or large.
One thing that I was wondering about: are the MSO probes hot-pluggable? HDMI connectors seem to be designed for hot plugging and there is no explicit warning against hot-plugging the MSO probes in the manual. Only one way to find out: try it! And YES hot plugging the MSO probes works just fine! Perhaps this seems logical to many people and not a big deal but I have come across quite a few pieces of equipment which don't allow hot-plugging active probes/pods so I'm a little cautious.
A nice feature is the adjustable hysteresis. One of the problems I had with an MSO from Agilent was that it would create false pulses because the hysteresis was too small to deal with a slow edge from an I2C bus which in turn (annoyingly) disrupted protocol decoding.
One question remains though: where can one buy HDMI cables with type B plugs? Just for kicks it would be interesting to try and see if the MSO cables can be replaced with off-the-shelve cables or not. Longer or shorter cables could be helpful in some cases depending on how your work bench is setup. On my workbench I have the oscilloscope at the right side of my work area because I didn't have enough depth in front of me for the vastly deeper larger oscilloscopes I used to have. But for this setup the MSO cables are slightly short.
Cursors
When it comes to cursors there seem to be two ways of doing things: keeping the cursors confined to the screen (the Keysight scopes I have used so far do this) or allow the cursors to be placed outside the screen. The latter allows doing way more precise measurements on signals. The RTM3000 falls in the category where the cursors can be placed outside the screen so all is well. Tapping the cursor positions on screen allows entering a position (time offset from the trigger point).
Decoding
The RTM3000 has 4 decoding busses in total where each bus can decode one stream. For example: decoding the RX and TX from a UART means using 2 decoding busses. For this test I pulled out my new toy: a digital pattern generator. With its maximum bit-rate of 250Mbit it can easily create a 62.5MHz I2C signal:
In order to tell the I2C bytes apart I added a small delay between each byte.
I2C decoding
Triggering on an I2C start seems to be no problem. When I screw up the I2C signal on purpose the RTM3000 spots a problem and marks the signal red.
One clock pulse missing:
Also decoding long I2C messages aren't a problem. The message is truncated in the list but the content is shown at the bottom of the screen as well. Of course the area is limited. If you want to see even longer messages then you'd have to save the bus list to a CSV file. I have not tried to upper limit but 64 bytes wasn't a problem.
Time for some over sampling tests... I generated a 1MHz I2C signal and checked at what sample rate the RTM3000 starts to decode properly. With a memory length of 40Mpts the minimum sample rate is 62.5Ms/s. That means an over sampling factor of 62.5. With a record length of 5kpts the minimum sample rate to get a steady decoding result is 41.7Ms/s. That is rather disappointing because the signal still looks like an I2C signal at much lower sample rates. I get that some filtering is required for protocol decoding to suppress false edges but the over sampling ratio could be much smaller (like 10 or so).
On the upside: according to the datasheet I2C decoding works up to 10Mbit/s but from my testing it seems I2C decoding works well up to 40Mbit/s (40MHz clock). By the way the digital signal used for the decoding can also be shown as part of the bus (enable/disable 'bits' item in display menu). The picture above shows that the digital signal is different compared to the actual (analog) signal so at least it is visible that the sample rate is too low for decoding the signal properly.
SPI decoding
On to SPI decoding. In order to decode both MISO and MOSI signals I need to configure 2 busses. Because I had bus 1 already in use for I2C I had to use bus 3&4 to get the MISO and MOSI data streams. When I selected bus 2 I could only get one data stream decoded. With the SPI rate set to 10MHz the minimum sample rate is 312Ms/s which means an over sampling factor of 31.2. When the sample rate is set to 2.5Gs/s the maximum SPI clock rate I can get to decode is 62.5MHz (on the digital inputs) which is much higher than the specified 25MHz.
Yes, I have enabled the gating on the measurement to show the data rate for the I2C or SPI bus! The blue bar shows which part of the signal is used to for the measurement. It is also possible to add labels to the digital channels. There is a whole range of predefined labels which you can choose from.
With both I2C and SPI decoding enabled let's see what the bus listing does with a continuous stream of back-to-back messages at the maximum memory depth... The listing shows both I2C and SPI messages. It seems we've got about 16700 I2C messages and 32000 SPI messages. Good! Not every scope I have come across can store all the messages it can decode from memory. It is no surprise an oscilloscope in this class decodes the entire memory. When browsing through the list a push on the Analysis button takes you to the start of the message on screen. A purple marker shows where the message starts. The amount of display space taken by the list display and the signal display can be chosen freely by swiping the white opposite arrows on the left up or down. Pretty neat!
Another neat feature is that non-ASCII characters are shown as hex:
UART decoding
Some screen shots from UART decoding:
The start and stop bits are clearly marked.
Of course the data can be displayed as ASCII (but also hex, binary and hexadecimal). The UART decode message list has an interesting feature: characters which are send back-to-back are displayed on the same line. I have not found a way to set a time-out though so it might not always work but it is a major improvement from other scopes I have experience with. A nice addition would be to set a frame terminating character (or sequence) but I'm not complaining.
CAN decoding
It is getting boring. CAN decoding also seems to work well with this less than ideal signal.
Note how the data gets rotated 90 degrees to fit the height of the bus decoding when the bus display is made taller. Nice!
Some remarks about decoding
Like on Keysight scopes you can't change the decoding parameters after doing an acquisition. The decoding seems to be done in parallel with the acquisition into a separate memory. I don't like this because you can't fiddle with the decoding parameters after doing an acquisition to see which settings work right. For situations where you have two different bit rates and/or parameters (yes, this does occur) you can use two decoder busses in parallel each set to a different bit rate and/or configuration.
With decoding enabled the maximum depth is 40Mpts and the maximum sample rate drops to 2.5Gs/s. I'm not sure whether I can agree with this choice. On one hand given the bit rate limits of decoding it doesn't make sense to have the 5Gs/s sample rate but more memory depth would have been nice. On the other hand the limits make it harder to look at a high speed signal in detail while doing decoding. A usage scenario could be testing a circuit which has a SERDES (where a high speed serial stream is converted to and/or from a parallel bus) and using the decode feature to print a list with data and/or trigger on a bad message. At 2.5Gs/s the upper frequency limit for having a signal visible is about 1.1GHz.
What I do like about the decoding is the 'find threshold' button. Just tap the button and the RTM3000 finds the right threshold. It even worked on the less than ideal CAN signal.
Something else I wanted to take a look at is how decoding works together with segmented recording. One of the things I liked about the Agilent MSO7104A I used to own is that it can list the decoded data from all segments. This means you can set a trigger for a specific message and cram as much relevant data into the memory as possible using segmented recording. Unfortunately the RTM3000 doesn't do that. In segmented recording (history) mode the 'bus list' only shows the decoded data for the currently selected frame and not data from all the recorded frames.
Note: keep in mind that the signals I used for testing SPI and I2C have picture perfect timing so in real circuits the maximum bit rate that can be decoded for SPI (I doubt I2C will have any issues) may be lower.
Math
All you need is math.... The RTM3000 doesn't have the free form math bonanza the GW Instek 2000E series offer but it still has a decent amount of abilities. There are 5 math traces (M1 through M5) which can be stacked sequentially. M2 can use M1, M3 can use M2 and/or M1, etc. Among the usual operators (add, subtract, multiply, divide, etc) the math traces can also do low-pass or high-pass filtering. By stacking the filters using two math traces it is also possible to create a band pass filter. So far the theory. I tried some math equations and I noticed that the math traces seem to use decimated (screen) data instead of the actual data. This leads to wrong results on some signals.
Filtering
Filtering can be very useful for developing DSP algorithms and cleaning up signals in general. Therefore I wanted to dig a little deeper into this feature. I applied the low pass filtering on a frequency sweep. Unfortunately there is some funny business going on when zooming in on the signal: the filtering stops working!
A filter initialisation artefact which could be hidden if the filtering started a few samples before the left side of the screen:
But when zooming in more the filtering stops working:
Changing the memory depth doesn't help.
Also the filter could roll off a bit sharper. However what is more problematic is that the way it sits now the filtering is not really useful mainly because the math trace uses decimated data but also because of the flaws. In my opinion R&S should move the filtering into the acquisition hardware somehow.
-
Part two:
Deep memory
When using deep memory every oscilloscope needs to do some sub-sampling on the acquired data to keep the display updates fast. This is the place where a DSO can show funny stuff and the RTM3000 is no exception. To test I have used a 1Vpp sine wave swept from 100Hz to 12kHz over a time of 100s so sub-sampling is likely to produce aliasing artefacts. With the scope set to 10s/div I get little over one sweep on the screen.
Sample mode shows some interesting inter modulation products due to sub-sampling:
A closer look at the signal as a check:
As usual high res mode is even more fun but the signal is definitely there!
But don't worry! All DSOs can be 'tricked' into showing this kind of behaviour. I just wanted to demonstrate the limits and show why it is important to have an idea what a signal looks like before pushing a probe onto a pin to measure a signal.
FFT and spectrum analyser mode
In FFT mode we can see both the signal and its frequency spectrum. To demonstrate how useful FFT can be for fixing EMC problems I used an old board which needed some tweaking to fix an EMC issue. After making the board 'bad' again I used an H field probe to get the frequency spectrum of the problem area:
The source of the excess radiation seems to be located near the rectifier diode of the DC-DC converter. Now the big question is: which signal and what part of it is causing the frequency spectrum? A frequency spectrum doesn't say that. What does help is to be able to dissect a signal and determine the frequency components of the various parts and that is where FFT comes in very handy! Here is the signal at the anode of the rectifier diode from the fly-back DC-DC converter which powers the board (measured with a >3GHz 1:20 low-Z probe for low tip capacitance so this was also a good opportunity to adjust the probe factor manually).
It turns out the diode reverse snap-off has a similar frequency content to that of the radiated noise. By sliding the window over the signal the frequency spectrum of a small portion of the signal can be examined. The window can be made smaller/larger by setting the frequency span and time/div. Unfortunately this does take some going back & forth to get it adjusted right. Surprisingly being able to set the span, start, stop frequencies and bandwidth (RBW) seems to make things harder instead of easier. This could use some improvement but the root of the problem is that an FFT with 128kpts is not very long.
After fixing the board the signal can be checked again to see if the problem has been solved:
This spectrum looks way better!
All in all the FFT is powerful and yet deeply intertwined with the sample rate and memory depth setting. The latter makes it hard to look at specific parts of a signal and/or get the maximum frequency resolution.
In spectrum analyser mode the span can be set to several Hertz. I had no problem looking at an FM modulated signal with a 100Hz deviation on a 900MHz carrier. The screen update rate is on par with the sweep rate you'd get from a real spectrum analyser. Another feature is the spectrogram which shows a waterfall graph with a frequency spectrum/versus time display.
There is also a display which can list the peaks:
Search function
The search function allows looking for specific occurrences in acquired data (yes, unlike math the search function does use the acquired data!). However it seems searching for events based on protocol decoding hasn't been implemented yet as the choice is greyed out.
Here I made the RTM3000 search for the time between edges of a clock and data:
Saving data
When saving data there is a choice between binary, floating point, CSV and text. CSV seems the most useful to me. Furthermore there is a choice between saving the screen data, acquired data or the history buffer. I like to have these choices because with deep memory just saving the display data is often enough (and it helps to select small chunks of data from a long record). Before saving the RTM3000 shows an estimate on how long it will take to save the data. It takes a few seconds to write a 3MB CSV file. A nice touch is that there is a warning shown when the file won't fit on the USB stick. The CSV file consists of two columns. One with a time stamp (in s) and one with the value. Unfortunately R&S choose to use the , (comma) as a separator instead of the more standard ; (semicolon). A comma is used as a decimal symbol in many countries. Still I had no problem reading the acquired data directly into GNU Octave to do some further analysis & prototyping for a signal processing algorithm.
FFT and math traces can be saved in the same way. There is also a choice to save the visible channels. When you choose this all enabled channels (thus not math traces!) will be saved.
Besides data the settings can also be saved so complicated measurement setups (think about MSO labels and bus decoding) can be saved.
There is a limit of 16 characters for the filename. The RTM3000 has an internal storage space of about 400MB which is useful to store data if you don't have a USB stick available.
Function generator
The internal function generator (included in the RTM-PK1 option bundle) is a basic generator with the usual features (waveforms and outputs). Interestingly the square wave is called rectangle. The maximum frequency for a sine wave is 25MHz. There is also a possibility to output a DC level and to add noise to the output. I have not seen the latter very often but it can be helpful to create a signal which is more like what a circuit can expect when deployed in the field. The maximum output swing is 10Vpp into a high-Z load which isn't bad at all. It at least allows to drive 5V logic or (opto isolated) inputs you typically find on PLCs and other industrial hardware.
The internal generator can also do modulation and sweep. Again the touch screen makes it very easy to operate and it doesn't feel like the function generator has been bolted on as an afterthought. However (like many other oscilloscopes) it is not possible to trigger on the sweep or the waveform internally. This feature seems quite obvious to me so why is this missing? When I created a very noisy signal the RTM3000 started to have problems triggering on the signal (no surprise). With an internal trigger it would have been rock solid despite the noise. Another example: I wanted to use the internal generator to determine the frequency characteristic of a circuit but I ended up using an external function generator which has a sync output connected to the external trigger input of the RTM3004. All in all the internal function generator is not a replacement for an external bench top function generator but it could be made much more useful if it could trigger on the function generator internally.
I also tested the log-sweep which seems to be difficult to get right in general and the function generator in the RTM3000 is no exception. Instead of a continuous waveform the log sweep turns out to be a stepped sweep.
Pattern generator
The pattern generator is part of the function generator / pattern generator option (and thus also enabled by the RTM-PK1 option bundle) and can be used to create digital signals. The output level can be set between 1.5V and 3.3V so no luck for digital circuits which run on 5V. 3.3V won't meet the logic high threshold for many logic families and I2C running on 5V. The most useful mode is the arbitrary mode which allows programming or loading a bit pattern and then repeating it infinitely or for a number of cycles. A pause in between can also be programmed. The pattern generator also allows generating I2C, UART, SPI and several other patterns but these modes are pretty much useless because you can't set the data (content) for the messages and the bit rates are also limited. I had high hopes to be able to (for example) send an initialisation sequence to a chip over I2C. This would be extremely useful for doing tests on a board without needing firmware. Still it is possible to write a small program in your favourite programming language to create the patterns and load them into the RTM3000.
DMM application
The DMM application shows a simple meter which is capable of showing DC, AC(RMS) and AC(RMS) + DC. It is nice to get a quick reading on the level of a signal but don't expect great levels of accuracy. Remember oscilloscopes don't have stellar accuracies when it comes to signal amplitudes. The RTM3000 is specified for a DC accuracy of 1.5% so that will also be the maximum accuracy you can expect from the DMM application. Besides all that I don't really see why a DMM application is useful on an oscilloscope which has measurements as well because a measurement can show exactly the same.
Segmented recording / history mode
R&S has combined history and segmented recording into one function called history mode (option RTM-K15; part of the option bundle). What history mode does is keep the previous acquisitions available for viewing. I have owned scopes before with this feature but I have never used history on those. The number of acquisitions which are stored depends on the acquisition length. The shorter the length, the more acquisitions can be stored. The RTM3000 has a whopping 428Mpts (maximum) of memory per channel available. So even at 80Mpts it is possible to record 5 acquisitions.
It is also possible to use the history mode to do segmented recording for a fixed number of frames. This requires tweaking some values. First the number of segments or record length needs to be set:
Now we need to make the RTM3000 to do 5 acquisitions. This is set from the acquisition menu in the Nx Single field.
After pressing the 'single' button the RTM3000 will start recording the acquisitions (in this case 20). A progress bar shows how many segments where recorded.
In run mode the history mode works as a circular segmented recording function.
The difference between fast segmentation on/off is that when it is set to ON the acquisitions will be done with as little dead time between them as possible (according to the datasheet less than 1.5us). If fast segmentation is off then each acquisition will be shown on screen before capturing the next one. Each mode has its specific use case.
After the recording is done the acquisitions can be recalled and replayed. There is also the possibility to show and average, envelope and accumulation of all. I tried this with a frequency swept triangle wave.
There is something odd here: the signal accumulating functions only work after the play button has been pressed. It is possible to do a partial accumulation but it takes pressing play & pause and then step through the segments.
Also the back button in the history menu doesn't always respond. Tapping the R&S logo at the bottom right helps to collapse the history menu but the back button should do the same of course.
Reference traces
No digital oscilloscope is complete without reference traces. These are very useful to compare a new measurement against an old one. The reference traces on the RTM3000 can be scaled and moved both horizontally and vertically. This is not possible on all oscilloscopes! Being able to scale & move reference traces is useful to align a reference trace to a new measurement which is shifted or acquired using different settings. Furthermore the reference traces can be used for measurements and FFT. Unfortunately the reference traces cannot be used for math and protocol decoding.
It is also possible to save and load the reference traces to and from a file but the data which is saved is shorter than the actual trace data. I guess this is done for faster handling of the data but it will result in different FFT and math result compared to the original traces. This is a 40Mpts trace and the resulting reference trace.
Power analysis
I don't really know where to begin here because there is so much to show. The power analysis package (which is part of the RTM-PK1 option bundle!) offers a huge number of analysis options for all kinds of power supplies. There are four different groups of measurements:
1) input which is about how the circuit loads the mains. The power factor, energy consumption and harmonics can be measured.
2) output deals with measurements like ripple, harmonics and transient response
3 and 4) deal with aspects of switching power supplies like efficiency, switching losses, modulation, dissipation in transistors, transistor SOA (safe operating area), etc.
When going through the various measurements a clear schematic is shown on how the voltage and current probes need to be connected. Most of these measurements do require a current probe, a differential probe (CAT rated when used on mains) and a de-skew fixture but let's see how far I can get with what I have available in my lab.
The first test subject is an HP 6024A 200W switching DC power supply from the late 80's. The reason I choose this is because PFC (power factor correction) was not mandatory back then and therefore the 6024A simply rectifies the mains and uses a few electrolytic capacitors as a buffer to get a 310V-ish DC rail. The result is a pulsating 50Hz current so it is ideal to see the power analysis at work when it comes to power consumption, real power, harmonics, etc. At the primary side I'm using a clamp-on current probe to measure the AC current and a differential probe to measure the AC voltage. The wiring of the 6024A makes this very easy. At the secondary (output) side I use my CP2100 current probe (my own product) and a regular probe to measure the output current and voltage.
The power analysis can do simple things like measure real power, power factor, etc and power consumption over time.
Things get more interesting when it comes to harmonics. There are several standards pre-programmed into the RTM3000 which is very useful for pre-compliance testing. No surprise the 6024A is a hard fail on some of them.
The power analysis is also able to determine the efficiency from input to output
The second test subject is a prototype of an 8A 2.5V synchronous DC-DC converter running at 295 kHz I have made a very very long time ago. I'm surprised it still works after over a decade! I choose this one because it is a relative safe circuit to work on while it can deliver a substantial amount of current. I've connected my CP2100 probe in series with the drain of the low side MOSFET with litze wire to reduce skin effect and use a normal probe to measure the voltage across it.
However before connecting everything I did a probe de-skew first because the switching times are getting into the nanosecond range. This officially requires a test fixture but since I don't have that I used the RTM3004's generator output with a 10Hz square wave and a 50 Ohm resistor in series with the current probe. By measuring both the voltage and current at the same time it is possible to de-skew the voltage probe and current probe so their delays are equalised.
Back to the circuit...
First I tried measuring the dynamic on resistance. This requires setting a window over the part of the waveform where the MOSFET is on manually. The dynamic on-resistance is the actual RDSon for a MOSFET. Depending on the drain current and gate voltage (Vgs) the RDSon changes so it is something interesting to verify in an actual circuit but unfortunately I can find very little on the exact definition of dynamic on resistance. If you look in the graph of a MOSFET you can see that RDSon isn't constant. At low Vgs (constant current area) this effect is the most extreme. With Vgs=3.5V the RDSon seems to be 1/27=37 milli-Ohm at 1V (27A) and 3/28=107 milli-Ohm at 3V (28A).
.
However the number the RTM3004 shows below can't be right. The actual RDSon is around 12 milli-Ohms and at 1A this would mean the voltage across the MOSFET will be around 12mV. Given the full vertical range in the picture below is 20V that gives about 20mV/bit (@ 10 bit resolution) not including noise. It is not possible to get a meaningful measurement for this MOSFET's RDSon this way.
Let's revisit this later on!
Secondly I tried to measure switching losses. Again this requires setting 4 different windows for each stage of the switching cycle (switch-on, conducting, switch-off and non-conducting). The number for the non-conducting part of the cycle seems to be complete nonsense. I can't find any relation between input power (6W), output power (5W), total loss (1W) and this number. The actual switching loss should be a couple of hundred milliWatts at most. I think this measurement suffers from the same problem as the dynamic RDSon. To measure the voltage across the MOSFET accurately enough the sensitivity must be much higher. OTOH that would make it impossible to measure the switching losses because these also need to have the full height of the switching edge.
Using the high resolution acquisition mode doesn't make any difference by the way.
I also played a bit with the safe operating area measurement but I'm not sure this is actually useful for designing circuit which use semiconductors for switching. What I'm missing here is the aspect of the duration of the volt * current product (which usually is a pulse when a semiconductor is used a switch). You can define a limit for the voltage versus current graph but the duration of the maximum power is equally important. In a SOA graph the maximum current and voltage say how much power can be dissipated in a device during a certain amount of time. Usually the shorter the time, the higher the maximum power (volt * current). In a small nutshell: the effect behind SOA graphs is that the power dissipation in a transistor isn't distributed evenly. There will be hot spots where more power is dissipated than in the rest of the die causing unequal heating and parts of the die going over the maximum temperature. The shorter the duration of the peak, the less hot such a 'hot spot' can become and thus more power (higher volt * current) can send through the device. However the RTM3000 only checks the maximum voltage and current (which are plotted) but not the duration of the peak power event. I think R&S should to change the SOA measurement to also use the width of the power peak. This would mean that a SOA graph needs to be imported into the RTM3000 so it can be used to test against. I know that this will be complicated to implement and it appears the competition did not went that far either.
The way it is now it is only useful to test against the DC (non-pulsed) limit in the SOA graph of a device.
I got feedback from R&S saying that they are aware of the limitations of the SOA graph but that they can't tell yet if and when they are going to add more features to it.
I also noticed that some power analysis functions change the time/div to get an optimal reading but others don't so you have to tweak the settings yourself. For example: I had some trouble to get the efficiency to produce any numbers. After changing the time/div so I got a visible waveform on screen I got results which seemed right. The power analysis feature uses math traces as input and these (as a rule of thumb) only work if you have the shape of the signal visible on screen. Remember the math traces on the RTM3000 use decimated data instead of the acquired data.
Dynamic RDSon measurement (again)
Because I wanted to show this effect in more detail I decided to create a small setup using a power MOSFET with an RDSon of around 300 milli-Ohm. I set the Vgs fixed to 10V so the MOSFET is fully on and stepped the current from 2A to 5.5A using a programmable power supply. According to R&S the dynamic RDSon is defined as delta V/ delta I.
By the way: the math trace which calculates Ohms (V/I) clearly shows that the RDSon isn't constant with a changing current.
Using the cursors the calculated value seems to be correct given the definition (delta V=1.2V, delta I=3.5A => 342 milli-Ohm:
By the way: I had to make a separate screen shot showing the cursors because the cursors cannot be enabled with the power analysis active.
Waveforms/s
A subject which always leads to a lot of debate so let's do some tests to find a glitch which doesn't happen very often. As a test I have created a signal (using a Tektronix pattern generator) which goes wrong one out of 50,000 cycles. I choose to create a 500 kHz square wave which usually has a 50% duty cycle but to mimic the glitch a 40% duty cycle has been thrown into the mix. 500 kHz means 2us per period so 50,000 cycles takes 100ms and thus the glitch happens 10 times per second. To verify I captured over 100ms of signal while triggering on a shorter pulse and used the search function. This shows that the signal has one shorter pulse per 100ms (50,000 cycles):
Let's see if we can catch some. When hunting for glitches you have to realise that you don't need to limit the display to just one cycle full screen. You can capture many and the high resolution display of the RTM3000 allows showing lots of cycles with great detail.
First let's set the memory depth to 40Mpts. The inverse colour display setting helps to make the shorted cycles stick out like a sore thumb. It doesn't take long for the anomalies to show up even though the trigger rate is around 115Hz. By having 60 cycles on screen the chance we catch the anomaly increases a lot!
When using the shortest memory setting (5kpts) catching the anomalies becomes a piece of cake and in a few seconds almost every cycle shows an anomaly. The inverse colour grading helps to highlight the cycles with a reduced duty cycle. The waveform update rate seems to be around 12 kHz (measured at the trigger out).
Even when looking at a single cycle it doesn't take long to capture a few anomalies within a few seconds:
The datasheet says the RTM3000 can display do up 64000 waveforms/s in dot mode. In my test I have sin x/x reconstruction enabled and dots disabled which means a lower update rate.
XY mode
It has been over a decade ago since I last used XY mode but that doesn't mean it is useless. For example XY mode is handy to look at very small phase changes between two signals. On the RTM3000 the XY mode can use one X channel and one or two Y channels. The XY can be combined with the various acquisition, colour grading and persistence modes.
Let's take two triangle waves with a 90 degree phase shift to draw a square which is rotated by 45 degrees:
As you can see the XY function shows both the XY result and the signals. The history/segmented recording can also be used in XY so you can step through several acquisition cycles and see the signals & XY plot. Seeing the signals as well is handy to check if you have at least one full cycle of the signals you want to see in XY mode.
But let's get funky and turn the persistence to 5 seconds, adjust the frequency of one channel slightly and set the signal colour to rainbow:
Psychedelic!
Mask testing
With mask testing an oscilloscope can compare a mask against an acquired signal. The RTM3000 supports to test one channel at a time and the acquisition mode seems to be forced to 'sample'.
First the mask needs to be defined. For that we need to capture a trace on the display:
It seems the RTM3000 uses the screen data to define the mask. Small details in the signal will be overlooked:
For a while now I have been wondering why mask testing in general is so limited on oscilloscopes and geared towards doing the mask testing at high speed. To me it would seem useful to be able to capture a long trace (several Mpts or more) and then compare it against a mask which is also several Mpts. That way you can send a sweep and/or some other (complicated) waveform through a circuit and do a functional go / no-go test quickly without needing a complicated automated setup. Sure the mask test would be slow but you can test a lot in one go.
Anyway, back to the mask testing on the RTM3000. When going over the mask testing menu I asked myself: Who is Captain Fails and what is he doing on an oscilloscope? It turns out 'Capt. Fails' stands for Capture Fails. Ahaaa!. R&S has tied the segmented recording into the mask testing. When a signal exceeds the limit (a fail) the RTM3000 can do several things like saving the waveform, take a screen shot but it can also save the failed signals into the segmented recording buffer. This way we can browse through the 'failed' signals only in an easy way. Because only on-screen data is used it is not necessary to use a very long acquisition record. This in turn means that the RTM3000 can record a huge amount of failed signals in its memory if necessary (of course this data will be lost during a power outage so be careful).
A failed signal:
Browsing through the recorded failed signals:
Self calibration / alignment
The RTM3000 can do a self calibration/alignment. IMHO a good thing (and expected from a modern oscilloscope) so you don't need external instruments and/or software for aligning the internal circuits.
Retaining settings
No major problems here. Every time I powered the scope up it started where I left. However the RTM3000 enables every channel with something connected. Apparently it is using the probe pins to sense a probe is connected but the RTM3000 can also sense the BNC outer shell is connected to a ground somewhere. Pretty clever but I don't like it (sorry to the person who worked on this!). I leave unused probes connected all the time and it is bit of a nuisance to turn all unused channels off after turning the scope on.
ther features
There is an educational mode which allows disabling the measurements and auto-set. Apparently students are not supposed to cheat while learning the basics on how to use an oscilloscope!
Issues
Issues found so far:
- Averaged trace disappears when changing time base in stop mode
- Infinite persistence trace disappears when the cursors are enabled/disabled.
- Persistence fading doesn't stop in stop mode
- Power analysis shows wrong numbers in switching losses.
- The back button in the history menu doesn't always respond
- Segmented mode segment accumulation only works after pressing 'play' button.
I would like to see a shorter list here and some things like the averaged trace disappearing really need to be fixed but all in all the list isn't that long given the number of features the RTM3000 has.
Some thoughts
The user interface could use some minor improvements. For example when I want to change a math trace I can open a pop-up menu at the bottom of the screen to enable/disable math traces. If I want to change the math definition I have to tap on 'menu' menu item instead of the menu item for the math trace itself. Tapping on the menu item for the math trace itself to go to the math definitions is much more intuitive to me. Other menu's like those for the channel could use a similar improvement. I feel like I'm tapping the screen too much to get things done.
At the start of my journey with the RTM3004 I was impressed with the configurable menus and ability to drag menu items onto the screen so you can customise the user interface a bit. As it turns out I'm not using these at all. Everything is very easy to reach from the main menu, acquisition settings and channel settings.
A feature which seems missing is frequency response analysis (FRA). This is the hot new oscilloscope feature 'du jour'. The 10 bit ADC could give an extra edge over the competition for having more dynamic range.
Conclusion
After doing lots of test and around 250 screen shots (I didn't include all the screen shots I made) I have covered a lot of ground but I'm pretty sure there is even more to explore.
Decoding could be improved with a lower over sampling factor so the memory can be filled to the brim with data for debugging rare protocol issues. Sometimes capturing as much data as possible, doing analysis on the decoded messages on a PC and trace the time stamp back to a part of a signal is the quickest way to find a bad message. Still the RTM3000 is able to capture and store a lot more decoded messages compared to other oscilloscopes I have used so far.
Also I have to conclude that the RTM3000 could use more processing power under the hood to deal with the long memory in all cases. Some corners had to be cut to make operations finish quickly and this pops up when doing measurements and math on traces with a lot of points (for example signal filtering). It means you have to be aware of the limits when using automatic measurements and math traces. Then again you'll need to look at even higher end oscilloscopes (usually powered by PC hardware) to have math & analysis based on multi-mega points of acquired data instead of decimated data.
Still I think the RTM3000 is a really nice & versatile oscilloscope and the touch screen makes it very easy to use. Keying in a number is often quicker than turning a knob around for a long time. Even writing to a USB stick which was full was handled well. Sure there are some areas that need attention but overall I'm impressed by the amount of features that have been packed into the RTM3000 and the versatility the RTM3000 offers. It can of course do the basics of a DSO but the RTM3000 can do protocol decoding, logic analysis, signal analysis, FFT, spectrum analysis, signal generation, etc. On top of that the user interfaces ties it all together neatly without giving the impression features where bolted on as an afterthought. The RTM3000 is easy to operate and the people at R&S did pay attention to a lot of small details. The small details (like the overshoot / undershoot indicators, collecting UART characters on one line or the time-elapsed-since-last-trigger display to name only a few) often don't make it into the datasheet but IMHO these make a piece of test equipment much easier and more productive to use. The enhanced productivity is what helps to justify the price tag. During my testing I did not need to consult the manual or on-line help but some things where located at unexpected places.
The RTM3000 is not a cheap oscilloscope but if you want to have more than 200MHz of bandwidth the choices to buy a good solid oscilloscope are getting slim. Even the B-brands are expensive so the relative difference in price is a lot smaller compared to the price differences seen in the low end segment.
According to the competitive comparison sheets R&S created they seem to want to compete with Keysight's, Lecroy's and Tektronix' 3000 series (what is with this 3000 number?). To me the RTM3000 seems like a very good choice. Keysight's MSOX3000T series has a very short memory (500kpts with all analog and digital channels enabled). Lecroy's Wavesurfer 3000z doesn't have peak detect which is a hard requirement for me to get rid of aliased waveforms when using less than the maximum sample rate. According to other forum users the Tektronix MDO3000 gets very slow when doing more complicated tasks. And let's not forget about the 10 bit ADCs in the RTM3000.
-
Very nice review Nico, and I just started reading. One thing I noticed is that the Flickr links do not work for me, is there anything you can do to correct that? They all give me 'Page Not Found'
-
Very nice review Nico, and I just started reading. One thing I noticed is that the Flickr links do not work for me, is there anything you can do to correct that? They all give me 'Page Not Found'
This seems like a (hopefully) temporary problem on Flickr. A freshly generated link for an inline image also doesn't work. -
Wow, great review. (Only just skimmed it for now, will get back to it later.) This thing looks sweet. Wish I could afford one, not that I need one.
-
Many thanks! Very good review.
Can you show the FFT in the audio band with the -90dBTHD or better generator? -
A very nice review of very nice 'scope. I'm busy working on one myself for an Element14 road test. It's taking a while as there are loads of great features and there's a lot to cover.
By the way: the '10 bit ADC' text on the badge is photo-shopped into the picture above. Real units don't have this.
Mine does have this text printed on the case. Maybe some last minute design changes. -
One thing i didn't immediately see : how MANY of those converters are there in this machine ? I hope the answer is 4 , not 2 or 1 that is shared amongst channels. ( OK to share at top speed as in we use 2 to read 1 channel , not as in 2 channels on 1 converter that we alt or chop. i never want to deal with scopes that alt or chop again in my life.)
-
AWESOME review! Thank you!
-
Nico - I want to publicly thank you for the very thorough and incredibly detailed review. I'm honestly amazed at the amount of time and effort you put in to it and all of us at R&S are greatly appreciative for the feedback, suggestions and guidance you've provided already. As you hopefully saw in 1.30, some of it already made it in, with more to come.
-Rich
-
Fantastic review, thanks.
There’s a lot in there that I wish I had on my 3000T. Looks like Keysight has a lot of catching up to do. -
Great review!
I wanted to do a similar one (but maybe not as comprehensive, e.g. I don't care about FFT), but I didn't get selected for Element14 RoadTest.
I'm glad R&S picked you to do the review, having a competent person do this makes the review much more meaningful and valuable.
When I was writing my application for the RoadTest and I was reading on RTM3004 one thing struck me - this seems like a strange oscilloscope segment, not only for R&S.
Let me explain. When below 100 MHz we can use cheap probes and B-brand scopes and all is fine. When above 500 MHz only active probes make sense and then it is more of a serial data analyzer than a scope.
RTM3004, Keysight 3000T and similar are in this strange segment where it supports active probes, but bandwidth and sampling is not enough for even simple stuff like USB 2.0 HS and there is no analysis functionality and decoding for such stuff as well (eye diagram analysis etc).
Therefore it seems such scopes are rather expensive "daily drivers" like the B-branch cheapos with passive probes, but when one needs to, for example, debug 50 MHz DDR SPI for SPI Flash one can use an active probe and decode the SPI. That seems to me is pretty much it, anything else (e.g. USB, MIPI etc) is either too fast or there are no analysis and decoding tools to match (e.g. one could poke at Ethernet lines, but there is no decoding anyway) or both.
Maybe it is just my "digital" perspective, maybe analog and RF people really use this 1 GHz bandwith without the need for analysis and decoding. -
Quote
One of the interesting things I have found during this test are the over range indicators. When the signal goes outside the range small markers pop up in the channel status bar indicating overshoots and/or undershoots. The analog Hameg oscilloscope I used at my first employer some 20 years ago had the same feature. I used it to find invisible spikes in signals. I'm pleasantly surprised to find this very useful feature has trickled up into the RTM3000! No I really can't write 'trickled down' here!
very useful indeed
Picoscopes have it too. You are looking at a signal that apparently is inside the window but you get the Out of Range message... there you can find nasty surprises
All scopes should have it. -
These are several pictures for direct comparison of noise characteristics: RTM and RTB. I hope the conditions are identical (except BW). Unfortunately the channels are not identical
The QuickAccess feature is a very convenient feature. Just add the button to the top line and there will be all the required menus.
Thanks again. I got a good lesson.
-
Thanks everyone for the thumbs up
One thing i didn't immediately see : how MANY of those converters are there in this machine ? I hope the answer is 4 , not 2 or 1 that is shared amongst channels. ( OK to share at top speed as in we use 2 to read 1 channel , not as in 2 channels on 1 converter that we alt or chop. i never want to deal with scopes that alt or chop again in my life.)
I didn't test for that specifically since I assume your misfortune was a singularTektronixDanaher f**k up.RTM3004, Keysight 3000T and similar are in this strange segment where it supports active probes, but bandwidth and sampling is not enough for even simple stuff like USB 2.0 HS and there is no analysis functionality and decoding for such stuff as well (eye diagram analysis etc).
1GHz bandwidth does allow to look at high speed signals but more for 'simple' signal integrity jobs. I think the biggest advantage of having 1GHz bandwidth together with low noise is that you don't need a second high bandwidth scope with less features (and more signal noise) compared to your daily driver.
Therefore it seems such scopes are rather expensive "daily drivers" like the B-branch cheapos with passive probes, but when one needs to, for example, debug 50 MHz DDR SPI for SPI Flash one can use an active probe and decode the SPI. That seems to me is pretty much it, anything else (e.g. USB, MIPI etc) is either too fast or there are no analysis and decoding tools to match (e.g. one could poke at Ethernet lines, but there is no decoding anyway) or both. -
So, yesterday nctnico posts a review of this fantastic new scope.
Today, Keysight sends out a marketing email reminding us that “the most important oscilloscope spec” is update rate.
Coincidence?
-
So, yesterday nctnico posts a review of this fantastic new scope.
Today, Keysight sends out a marketing email reminding us that “the most important oscilloscope spec” is update rate.
Coincidence?
Ask Rich and Daniel, who are probably having coffee together somewhere in Colorado Springs right now
-
Ha! Sometimes mountain biking, but I’m not a huge fan of coffee. I probably just lost some credibility huh?So, yesterday nctnico posts a review of this fantastic new scope.
Today, Keysight sends out a marketing email reminding us that “the most important oscilloscope spec” is update rate.
Coincidence?
Ask Rich and Daniel, who are probably having coffee together somewhere in Colorado Springs right now
-Rich -
Quote
Today, Keysight sends out a marketing email reminding us that “the most important oscilloscope spec” is update rate.
If I were a scope manufacturer I would be sorely tempted to jimmy one of timebases to get 1000001 wfm/s. -
So, yesterday nctnico posts a review of this fantastic new scope.
no.. the most important spec is 'price per feature' or simply... 'price'
Today, Keysight sends out a marketing email reminding us that “the most important oscilloscope spec” is update rate.
Coincidence?
@ntnico... whats the maximumm fft length? -
Having used the RTM3000 side by side with the 3000T for a while now, I find that by far the biggest downside of the RTM is the sluggish UI.
It just feels clunky and under-powered compared to the 3000T, and is the primary reason I always go to the 3000T first, and typically only use the RTM if I need the bigger screen, more flexible decode functions, more memory or just a second scope display.
I have little doubt that it could be improved significantly with some careful software optimisation, removing unnecessary transition effects, prioritising user input etc.
It is particularly disappointing that there was no improvement from the RTB2000 considering the higher price.
-
So, yesterday nctnico posts a review of this fantastic new scope.
no.. the most important spec is 'price per feature' or simply... 'price'
Today, Keysight sends out a marketing email reminding us that “the most important oscilloscope spec” is update rate.
Coincidence?
@ntnico... whats the maximumm fft length?
128k points.. -
That is in FFT mode. It seems the spectrum analysis mode can use a much larger number of FFT points.So, yesterday nctnico posts a review of this fantastic new scope.
no.. the most important spec is 'price per feature' or simply... 'price'
Today, Keysight sends out a marketing email reminding us that “the most important oscilloscope spec” is update rate.
Coincidence?
@ntnico... whats the maximumm fft length?
128k points..
@mikeselectricstuff:
I agree the user interface could do without the transition effects. Still it is not unusable slow. It doesn't really bother me but OTOH I have never used a Keysight 3000T. Is the 3000T much different compared to the older Agilent DSO6000 / DOS7000 series? -
I wouldn't call it anywhere near unuseable, just annoying
That is in FFT mode. It seems the spectrum analysis mode can use a much larger number of FFT points.So, yesterday nctnico posts a review of this fantastic new scope.
no.. the most important spec is 'price per feature' or simply... 'price'
Today, Keysight sends out a marketing email reminding us that “the most important oscilloscope spec” is update rate.
Coincidence?
@ntnico... whats the maximumm fft length?
128k points..
@mikeselectricstuff:
I agree the user interface could do without the transition effects. Still it is not unusable slow. It doesn't really bother me but OTOH I have never used a Keysight 3000T. Is the 3000T much different compared to the older Agilent DSO6000 / DOS7000 series?
3000T is as slick as the 6/7000 - you pretty much never notice any delay for anything ( apart from bootup!) - only exception is when zoomed out with serial decode decoding a ton of data it can't display -
Nico, excellent review; I am still reading all of it.
One detail: the top menu seems very large (consequence of the touch interface). Can it be collapsed to get more waveform realstate?
Somehow this menu reminds me of the Ribbon interface of MS Word...