A review of the GWInstek 1054B

[whited]

Yes, only a couple daydreams ago I was convinced 1054B was for me. And now I have this nagging feeling about memory segmentation and logic decode from full memory 
But if price is plotted reasonably, on log scale, with a few beers... maybe it's happy holidays as you say.

The 1054B is <$366 street price via Tequipment not including the eevblog discounts and possibly other promotions.

At the published bandwidth and list price they are about the same proportion, 50 MHz vs 200 MHz is still 4x more at 4x the price as well as serial decodes and segmented memory.

However, its in the extreme end, the usable bandwidth were you can get more.  Its likely to still go past 500 MHz based on the curve slope at its more sensitive vertical amp setting.

So the big question is do you need that much bandwidth to pay more for it?  Or in the end its all part of happy holidays?

Saturation, I have a little more time now to see your numbers; seems 1050B has ~70 MHz bandwidth instead of its rated 50 MHz. It is not too surprising but nice to see in a budget scope. I wonder if all GDS 1000B and 2000E have similar conservative bandwidth rating?

Maybe ntcnico can give another datapoint on frequency response? ntcnico I read your review thread; I hope I didn't miss it. Nevermind I found it here on post #67 https://www.eevblog.com/forum/testgear/gw-instek-gds2204e-(200mhz-4-channel-dso)-review/?all
It's a good measurement showing 2204E roll off (rated 200MHz bandwidth)
-3dB @ 290MHz with 50mV/div, more steep curve
-3dB @ 210MHz with 10mV/div, more gradual curve

I am personally thinking 2070E is a sweet spot @ $920... but 2204E is so close @ $1256. Can I convince myself?
I know this is a 1054B thread - I try to keep it relevant.
Thanks to you all.

[saturation]

Single rise time,  Pulse waveform from Tek 284 rated at 70 ps, 50 ohm termination




Stats while in STOP mode



Snapshot of live buffer calculated rise time, in RUN mode

Sorry, mea culpa, I have them but didn't have time to take it out; will replace that photo with the screen grab.

A screenshot captured on a USB stick is much more helpful and clear for those following than something like this:
 

[saturation]

Some added performance tests accumulated over the past holidays using a "calibrated" RF generator HP 8640B.  The 3B point of the 1054B is about 120 MHz versus my prior conservative test.



This is the maximum usable frequency of the unmodded scope.  At near n/2 at the Nyquist limit the onboard counter quits just over 400 MHz then waveforms are wholly unstable at 430MHz.   Down - 30dB.



IMHO, the practical limit of the scope, Down - 23dB.  This is were all the math functions and on board calculations still are accurate and stable.  YELLOW single channel, 363 MHz.  Red math channel calculating derivates of CH1.  Note Vpp and calculated RMS value, rise time of 821ps,  Cursors are measuring rise time manually at 10-90 which is 1.15ns, or by 0.35/Tr ~ 304 MHz or 426 MHz using the automated rise time.  Calculations were unstable on ungated mode, so these are based on using the 1054B gated to screen mode.



The stat function of the 1054B to show the level of stability at 363 MHz.



Close up of 1054B doing derivate of CH1 in real time.  It will calculate cleanly even at 1kpts.  Derivate of sin(x) is cos(x) but while the appearance is correct the phase shift is a bit prolonged.



Close up of the derivate of the rise time.  This is also called the Dirac delta function.



Live 2 channels and math channel.  Source is a Tek 284, Yellow CH1 is the pulse output and Blue CH2 is the trigger output, they are delayed by 5ns as shown on the cursors. 

[saturation]

New firmware released, untested by me.  The changelog suggests a number of bugs I did not report or experience.  Gwinstek now includes a changelog copied here:


Version 1.08

1.Fixed the system crash randomly which caused by the ¡§Advance Math¡¨ function
2. Fixed the displaying issue of XY mode under certain circumstances
3. Fixed the time base error when the ¡§AUTOSET¡¨key pressed under ¡§Window Zoom¡¨ mode
4. Fixed the AD malfunction under certain circumstances

Version 1.09
1.Added German & French icons and help menu
2.Adjusted the maximum value of the waveform brightness


Version 1.11
1.Added Korean icons and help menu
2.Added Russian icons
3.Fixed the residue waveform issue
4.Fixed the system freeze bug after the QR code menu activated
5.Added the DVR calibration under distributor mode

Version 1.12
1.Added Russian help menu
2.Fixed the initial value for the ¡§Advance Math default¡¨ expression
3.The output polarity of the Go/No-Go has been adjusted which same as the GDS-2000A series

Version 1.13
1.Fixed the Trigger Filter bug.

Version 1.14
1.Changed the display method for waveform interpolation.
2.Fixed the pictbridge's connectivity issues.
3.The Trigger out function is activated now. (Press UTILITY -> Probe Comp. -> Frequency)

Version 1.15
1.Add Multilanguage: Spanish

Version 1.16.001
1.Fixed the trigger level offset issue, if the vertical position was shifted out off the screen range.
2.Fixed the ¡§Advance math¡¨ algorithm which caused the system crash.

Version 1.17.001
1.Added USBDelay command in order to solve the USB data leaks for WIN10.
2.Fixed the ext trigger of roll mode which may cause the system crash.
3.Fixed the zoom state malfunction for horizontal position command.
4.Changed the value range of trigger hold off command: the minimum value changed to 4ns from 10ns.

[Fungus]

1.Fixed the system crash randomly which caused by the ¡§Advance Math¡¨ function

 No comment. 

[saturation]

Yes, I stand corrected.  The maker knows more than I.  Its been almost a year and so far I have experienced 3 events that I cannot reproduce. 

Recently, a locked up happened with a long equation using the advanced math function requiring a power down, but since then the full equation or individual functions within haven't caused a lock up. Another the scope control response slowed on 4 channels that didn't return when 3 other channels were deselected. it required the default button to be pressed [ soft reset to factory settings].  Once the FFT had no output after being deselected and then reselected, it reappeared after selecting advance math then FFT again, setting preserved.  The only uniform event was the scope was on for several hours monitoring a circuit, or  I was waking the scope from backlight sleep mode, but I've done all 3 issues many times since and can't get it to repeat.

So yes, in summary there are bugs.

1.Fixed the system crash randomly which caused by the ¡§Advance Math¡¨ function

 No comment. 

[Fungus]

So yes, in summary there are bugs.

Good to know, thanks!

[MrAl]

Hello there,

Very nice thread.  Lots of helpful information for someone who is looking at new scopes and does not want to spend a fortune.  Special thanks to 'saturation' for doing so much work on this.

It's kind of hard to weed out the bad ones with little information on the practical use in real life, so it is nice to see all this.

I see one thing in particular that is interesting but not sure how exactly to interpret.  The rise time test showed about 3.8ns and i wondered what setting the scope was on to get that.  I noticed some of the scopes have a 1ns setting, so i wondered if we could get a raw ADC view on the scope.  If in fact they use a true 1GSPS ADC then we should be able to see, somehow, about 0v at say 10ns and say +5v at 11ns (10ns plus 1ns).  Is that the way these DSO's can work, or do they obtain that 1GSPS sampling rate by doing several scans of the same waveform coming in in real time?   Or, do they do some intermediate work on the wave which would bring it up to 4ns instead of 1ns?  BTW what is that scope rated for in rise time?
I am hoping that they use a true 1GSPS ADC chip of some kind combined with fast memory when they advertise as 1GSPS of course.

I am rather new to the DSO's as i had and have used many CRT scopes in teh past including 10 thousand dollar CRT storage scopes, but only used two DSO's so far and only for a short time, so i dont know much about what to look for and what to watch out for.

Thanks for any additional info like this.

[MrW0lf]

Test request due to this thread:
https://www.eevblog.com/forum/testgear/opinions-on-gw-instek-scopes/msg1131021/#msg1131021
Please check analog bandwith with Sin(x)/x ON|OFF, with all 4 channels ON to minimize sampling rate.

[saturation]

Hi, that's a good idea worth detailing.  You cannot turn off Sin(x) interpolation, so for comparison I'll use dot mode vs vector mode and automated measurements.  I no longer have access to the 8640B, so the comparisons are against the -3dB point, ~ 120 MHz using an RF Explorer Signal Generator as an RF source.



Baseline measurement -3dB at 120 MHz is 9.88mVpp.  Automated measures uses whole buffer for calculation, so set as GATING OFF.



At 2 CH or 500MSa/s the -3dB point is reduce to 110 MHz.



At 3 CH or 250MSa/s the frequency is read correctly but the low sampling causes more  error in the automated measures of Vpp.



No difference at 3CH or 4 CH since the sampling rate is the same.



At 120 MHz, just shy of the Nyquist n/2 limit, the waveforms are unstable and the frequency counter gives erroneous count.



At 250 MSa/s the automated readings are near the same 1 GSa/s at <= 80 MHz.



Baseline 1 GSa/s at 80 MHz to compare against 250 MSa/s.
 

Test request due to this thread:
https://www.eevblog.com/forum/testgear/opinions-on-gw-instek-scopes/msg1131021/#msg1131021
Please check analog bandwith with Sin(x)/x ON|OFF, with all 4 channels ON to minimize sampling rate.

[saturation]

Comparing dot vs vector modes.  Using dot mode reduces or removes bias in measurements caused by the Sin(x)/x interpolation.  'Gating' means automated calculations are based on the screen contents, when "OFF" calculations are on the entire input buffer.  80 MHz was selected as the test frequency as this was the highest frequency where the error at 250MSa/s using automated measurements at gating off, produced the least error versus 1GSa/s.






Baseline for all comparisons.  1GSa/s 80 MHz vector vs dot mode and against manual measurement.  Here I used gating to screen.






500MSa/s 80 MHz vector vs dot mode and against manual measurement.  Here I used gating to screen.





250MSa/s 80 MHz vector vs dot mode and against manual measurement.  Here I used gating to screen.

Comments:

There is little difference between measurements made between dot and vector modes.  Automated measurements concur very well with manual measurements when using gating to screen, since manual measurements are only based on screen data.

At 250 MSa/s, the lower sampling rate increases the measurement error, compared against the baseline 1GSa/s.  To reduce this error use 'gating off' which then uses the entire buffer contents to calculate values, increasing the data size for analysis.  As shown in the prior post, this causes the difference between 1GSa/s and 250MSa/s rates to be substantially reduced.

[saturation]

Yes, this thread exists just because there really isn't anyone else on the net making a detailed review.  If someone better came along, I would link this thread to it and move on.

There are many good videos and documents about basic DSO function.  It'll help to review them to make an informed decision.

Rise time is a standard measure developed since the 1940s the only difference today is whether it begins say in DC, from 0V + 10% and ends at 90% of peak or 20% and 80% or other end points.  It can be understood in a simple way or it can get very complex. 

https://en.wikipedia.org/wiki/Rise_time

The 1000B series was torn down in detail by Dave and can be found in the eevblog forum.  Details of it working can be found in the datasheet of a Hittite HAD1511 ADC particularly in relation to interleaving, and the main CPU is a Zynq Z7010 SoC.  Enjoy.

Hello there,

I see one thing in particular that is interesting but not sure how exactly to interpret.  The rise time test showed about 3.8ns and i wondered what setting the scope was on to get that.  I noticed some of the scopes have a 1ns setting, so i wondered if we could get a raw ADC view on the scope.  If in fact they use a true 1GSPS ADC then we should be able to see, somehow, about 0v at say 10ns and say +5v at 11ns (10ns plus 1ns).  Is that the way these DSO's can work, or do they obtain that 1GSPS sampling rate by doing several scans of the same waveform coming in in real time?   Or, do they do some intermediate work on the wave which would bring it up to 4ns instead of 1ns?  BTW what is that scope rated for in rise time?
I am hoping that they use a true 1GSPS ADC chip of some kind combined with fast memory when they advertise as 1GSPS of course.

I am rather new to the DSO's as i had and have used many CRT scopes in teh past including 10 thousand dollar CRT storage scopes, but only used two DSO's so far and only for a short time, so i dont know much about what to look for and what to watch out for.

Thanks for any additional info like this.

[saturation]

News:

Mid-May 2017 Tequipment.net stocked 55 1054B and today it had 14 remaining.

Its also sold on Amazon at the same price as tequipment, so maybe it the source.

https://www.amazon.com/GW-Instek-GDS-1054B-Oscilloscope-4-Channel/dp/B011IFQUQ8/ref=sr_1_1?ie=UTF8&qid=1498244799&sr=8-1&keywords=1054b


There is also a marked increase in the total number stocked by tequipment for other 1000B models

http://www.tequipment.net/Instek/GDS-1054B/Digital-Oscilloscopes/


The increase of unit sales/week was about less than one a week in 2016 to ~ 1.5/day in YTD 2017 for the 1054B.

[nctnico]

News:

Mid-May 2017 Tequipment.net stocked 55 1054B and today it had 14 remaining.

Make that 4 remaining!

[saturation]

Yes, checking today from Amazon stock 9/17 remaining since I last posted.  Its unlikely Tequipment as the seller since their remaining stock is far lower.

For 1054B, 3/4 since your post
1074 no stock
1104 1/20 remain since your post

Given the Siglent DSO using Zynq and the issues discussed on the thread it shows how well the GWInstek managed their version.

Not much more to post on my end as the DSO just works, and at this time no bugs to complain out, no new firmware for this model YTD, last is    1.17    en    22.1 MB    2017/03/17.

News:

Mid-May 2017 Tequipment.net stocked 55 1054B and today it had 14 remaining.

Make that 4 remaining!

[saturation]

By chance I found out new firmware released this week.  Other DSO in the Zynq line could check if they too have new firmware for their models.  Have not yet time to test: V 1.18

In packet:
No changelog
No update instructions

[bitseeker]

Saturation, it's great that you've been able to maintain this thread. As you said, you're probably the only one able to provide so much ongoing info on the 1054B.

[saturation]

Currently change log and firmware:


Version 1.08

1.Fixed the system crash randomly which caused by the ¡§Advance Math¡¨ function
2. Fixed the displaying issue of XY mode under certain circumstances
3. Fixed the time base error when the ¡§AUTOSET¡¨key pressed under ¡§Window Zoom¡¨ mode
4. Fixed the AD malfunction under certain circumstances

Version 1.09
1.Added German & French icons and help menu
2.Adjusted the maximum value of the waveform brightness


Version 1.11
1.Added Korean icons and help menu
2.Added Russian icons
3.Fixed the residue waveform issue
4.Fixed the system freeze bug after the QR code menu activated
5.Added the DVR calibration under distributor mode

Version 1.12
1.Added Russian help menu
2.Fixed the initial value for the ¡§Advance Math default¡¨ expression
3.The output polarity of the Go/No-Go has been adjusted which same as the GDS-2000A series

Version 1.13
1.Fixed the Trigger Filter bug.

Version 1.14
1.Changed the display method for waveform interpolation.
2.Fixed the pictbridge's connectivity issues.
3.The Trigger out function is activated now. (Press UTILITY -> Probe Comp. -> Frequency)

Version 1.15
1.Add Multilanguage: Spanish

Version 1.16.001
1.Fixed the trigger level offset issue, if the vertical position was shifted out off the screen range.
2.Fixed the ¡§Advance math¡¨ algorithm which caused the system crash.

Version 1.17.001
1.Added USBDelay command in order to solve the USB data leaks for WIN10.
2.Fixed the ext trigger of roll mode which may cause the system crash.
3.Fixed the zoom state malfunction for horizontal position command.
4.Changed the value range of trigger hold off command: the minimum value changed to 4ns from 10ns.

Version 1.18.001
1.Removed the extra unwanted carriage returns symbol for *LRN? command.
2.Fixed the starting condition of autotrigger for Datalog function (APP funciton)
3.Extended the "Duration" until the 999 hours for Datalog function (APP funciton)

Version 1.19.001
1.Fixed the Japanese translations error for digital filter menu tree. (APP function)
2.Updated the simplified Chinese menu tree
3.Updated the License encoding rule


Note, the changelog is not entirely complete.  After 1.14 when the trigger out was activated it also increased the wfms/sec rate which is not mentioned officially except by my testing on this thread.


I have not found any functional change in 1.18 since I last posted.

[maxwell3e10]

I just got a new GDS-1054B, primarily motivated by deep memory per channel and the hope for a smoother user interface.
It seems there are missing bits in the 8-bit acquisition. Here is a screen shot of a transient from the probe calibration signal. I stop the acquisition and then increase vertical scale 2 steps. The display is set to dots. The small step between levels is consistent with 256 bits over 10 vertical divisions. But some steps are twice as large. It seems to happen on various time scales and sensitivity scales. I also checked the same problem exists if I save data to file, so its not a display problem. I don't think there should be missing codes in 8 bit ADC.

[saturation]

Hope it serves you well.  I agree with your insights of the screen capture but I cannot reproduce what you show, and I am not clear as to what this pulse represents.  Is this the scopes onboard cal out?   If so, its 2Vpp.  To get a 1mV/div read you have the scope scale at x1 with a probe set at x10? You are also looking at the very edge of a waveform given the trigger location, captured the waveform to memory then zoomed. 

Just quick thoughts:

Since you have a pattern in that waveform that is fairly consistent, it can be the slower ~1000 wfms/sec rate at 50us so  gaps could also be 'dead time'  even if sampling is 1GSa/s; not necessarily an ADC issue.

A dot mode image is purely a series of dots, we see lines.  This means the settings are distorting the data or its artifacts; stretching out what should be a dot.  Likewise the gaps seen by the vertical amp setting cannot be said to be free of artifacts either.

Since I can't reproduce your image to see it on my DSO, I can be wrong, but it hard to make clear inferences based on the settings shown.

[maxwell3e10]

Here is a more proper test using a linear ramp. One can see gaps in the data, which are also in the saved waveform. I believe the reason is that in the firmware they do some vertical scaling while keeping only 8 bit numbers.

Another problem I found is that on high sensitivity scales it does not have true 8-bit resolution. For example, on 1 mV scale, all bit values are spaced by 3. So they simply rescale the values instead of increasing the analog gain.

In terms of responsiveness, I found the horizontal shift knob has rather large lag.

[nctnico]

@maxwell3e10: I think your unit is somehow broken. Can you determine whether you are missing a specific bit in the ADC values?

[maxwell3e10]

It is a specific value missing, but it changes depending on how many channels are being recorded. I have a bit more info in this post.
https://www.eevblog.com/forum/testgear/missing-bin-values-in-gw-instek-gds-1054b/
I am more inclined to think its a firmware problem or perhaps firmware calibration is compensating for an analog defect.

[nctnico]

Just return the unit and get a replacement. I'm very sure this is a defect (there is a bit stuck somewhere) and not a bug in the firmware. If it where a bug it would have been found & reported long ago.

[saturation]

Great, will give it a test and will report back.  What firmware are you using on the unit?

Here is a more proper test using a linear ramp. One can see gaps in the data, which are also in the saved waveform. I believe the reason is that in the firmware they do some vertical scaling while keeping only 8 bit numbers.

Another problem I found is that on high sensitivity scales it does not have true 8-bit resolution. For example, on 1 mV scale, all bit values are spaced by 3. So they simply rescale the values instead of increasing the analog gain.

In terms of responsiveness, I found the horizontal shift knob has rather large lag.