SAG1021 vs SDG1032X a comparison

[HendriXML]

I'm currently the owner of a Siglent SAG1021 and have tested another one.

While exploring it's possibilities I ran into some issues:
Issue 1: It's USB powered and draws about 360 mA. A lot of this current will run through the BNC ground lead and causes a voltage difference of about 13 mV between the GND of the AWG and the DSO.
https://www.eevblog.com/forum/testgear/sds1104x-e-and-sag1021-unwanted-dc-offset

Issue 2: DAC stepping errors results in glitches (a sudden drop of 2 mV):
https://www.eevblog.com/forum/testgear/using-a-awg-and-a-scope-do-a-low-voltage-level-characterization-of-a-1n4005/msg2412279/#msg2412279

Issue 3: Both device have quite large offsets (about 30 mV), which depend on the range the device is operating on
https://www.eevblog.com/forum/testgear/using-a-awg-and-a-scope-do-a-low-voltage-level-characterization-of-a-1n4005/msg2422833/#msg2422833

Issue 4: In DC mode, going up 1mV means at a certain point going down 8 mV.
https://www.eevblog.com/forum/testgear/using-a-awg-and-a-scope-do-a-low-voltage-level-characterization-of-a-1n4005/msg2424117/#msg2424117

My seller offered me the option of a upgrade to for instance a SDG1032X, if I add the difference in cost. I like that solution if it leads to a device which I can do some precise DC testing with. My question is what to expect when choosing for a SDG1032X. Using the SAG1021 extensively gives me the feeling that it was not properly tested (and redesigned to solve issues). Being usb powered might also be restrictive in design options and choice of components. So I still have some hope the SDG1032X (I can't spend much more on an AWG) will perform better.

[Performa01]

My seller offered me the option of a upgrade to for instance a SDG1032, if I add the difference in cost. I like that solution if it leads to a device which I can do some precise DC testing with. My question is what to expect when choosing for a SDG1032. Using the SAG1021 extensively gives me the feeling that it was not properly tested (and redesigned to solve issues). Being usb powered might also be restrictive in design options and choice of components. So I still have some hope the SDG1032 (I can't spend much more on an AWG) will perform better.

Generally speaking, AWGs are no metrological devides and particularly are they not DC calibrators.

However, I highly appreciate your efforts to use a cheap AWG that way! 

I would say, the upgrade to an SDG1032X is a very attractive offer. Consider this:

• USB puts some limits on the SAG1021 (ground reference / offset issue, very limited output amplitude) - you won't have such issues with a standalone device like the SDG1032X
• It should be clear that the SDG1032X can only perform better
• The SDG1032X has the much higher output and dual channels which can be phase-locked, these are very useful features in my book
• The SDG1032X has more features in general
• The SDG1032X is a stand-alone device and can be operated independent of the DSO

I would most definitely do that upgrade, especially for your use case.

For instance, you could use the dual channels to build up an even higher resolution source by e.g. combining Ch1 + Ch2/256, using a stable precision signal attenuator + combiner. You could then make an automated calibration routine that gathers correction values for each output value of Ch1 and use this as an offset for Ch2 (using some very special arbitrary waveforms, or DC mode alternatively), where Ch2 performs the fine steps and needs no additional correction, because the absolute INL error there is already devided down quite nicely.

Having said that, an 8bit DSO would certainly not be the perfect tool for such a calibration (and high precision measurements in general), but you can still get more than 8 bit resolution with proper use of the offset feature (Y-position).

[HendriXML]

Thanks for your reply!

I did some extra research on the device as well and see the benefits you mention, like using 2 combined outputs.

I understand  almost every use case has it’s dedicated instrument, and I (as most of the enthusiasts on the forum) would like to own them all, but in practice there seems to be this limiting stuff called money. That makes it -especially in the consumer price range- important that products are well thought out and versatile.

The extra money seems to be justified, so I’ll take the upgrade option. 

[HendriXML]

Having said that, an 8bit DSO would certainly not be the perfect tool for such a calibration (and high precision measurements in general), but you can still get more than 8 bit resolution with proper use of the offset feature (Y-position).

I was on my way to do such measurements, thinking the DSO would be the weakest link, but the AWG ruined that party. So I’m quite happy with the DSO, I think using the offset and maybe with a bit of scaling and a better AWG the results will be very usable. For most purposes having a few hundred measurements is good enough. Thus a very thrust worthy 8 bit DAC would not be that bad.

Doing my tests on the AWG, I came to realize bits aren’t everything. Better to have lesser bits and no step errors for example. Or one can waste bits by digital scaling. Or mess up bits using wrong rounding logic. Or shift all negative values by using wrong interpretation of signed integers. Testing and thinking things through, that’s important! 

[Performa01]

I was on my way to do such measurements, thinking the DSO would be the weakest link, but the AWG ruined that party. So I’m quite happy with the DSO, I think using the offset and maybe with a bit of scaling and a better AWG the results will be very usable. For most purposes having a few hundred measurements is good enough. Thus a very thrust worthy 8 bit DAC would not be that bad.

Doing my tests on the AWG, I came to realize bits aren’t everything. Better to have lesser bits and no step errors for example. Or one can waste bits by digital scaling. Or mess up bits using wrong rounding logic. Or shift all negative values by using wrong interpretation of signed integers. Testing and thinking things through, that’s important! 

Using the Offset DAC (X-Pos) for high resolution DC measurements seems to work quite well. I think I have shown this in my SDS1104X-E review, where I was able to measure 205V at the 2V/div range with an absolute accuracy of better than 0.1%.

I agree that for your applications the number of bits is not nearly as important as the INL. But to achieve a near perfect linearity, you'll have to use the 2nd channel for step corrections indeed, which could then even overcome some minor calculation/rounding errors in the waveform data processing.

That being said, should the SDG1032X also show some weird behavior in this regard, we can (and should) request a fix. Such silly errors can happen, I've indentified some similar mistakes in early SDS X-series scope firmware as well (+/-1 uncertainty, improper handling of negative values) and they have been fixed at some point.

I think you made the right decision in upgrading to the SDG1032X. Two channels and the high output amplitude alone should be worth it in your case.

[HendriXML]

That being said, should the SDG1032X also show some weird behavior in this regard, we can (and should) request a fix. Such silly errors can happen, I've indentified some similar mistakes in early SDS X-series scope firmware as well (+/-1 uncertainty, improper handling of negative values) and they have been fixed at some point.

That’s also one thing I have considered, the SDG1032X has had firmware updates and can be calibrated to some extend.
As a software developer, I‘ve send my share of bugs into the world. I’ve even had an user apologize for that he crashed my app.  But bringing them to light is important and taking responsibility and taking care of them is part of the job.

[HendriXML]

The test I did with the SAG1021, I repeated with a SDG 1032:

I wanted to see how useful the method of doing multi window measurements would be using the SAG1021.

From -2.0 to 2.0 V, I did 20 partial and overlapping measurements (batches). Each batch climbing in voltage (DSO offset and AGW offset).

The AWG was loaded with a stairway ramp of 50 different value blocks evenly spread with an amplitude of 400 mV and thus a difference of 8 mV between them.  These value blocks where mapped to (multi segment) DSO acquisitions and then averaged. (Multiple frames and the block itself). For each batch a data file was written, containing the set/expected wave voltage, the measured one (DSO), and the difference between them.

These are shown in the graph against each other. Ideally it should be a straight line.

On a secondary axis the difference between the values are shown. This shows whether there's an offset, amplitude error etc for each batch.

The VDiv on the scope was 100 mV/div resulting in a 4 mV resolution.

The way the SDG1032X acts, is the way I would like it to perform in this test. The largest error, is about the size of the scopes granularity. No amplitude stuff, and no strange offsets. (I almost sound like Trump)

Will repeat more tests, but I'm happy with this outcome. (Maybe at -1.5V something is not completely ok, but that can also be the DSO )

[Performa01]

I'm glad you think this upgrade was worth it. As already discussed before, it's not only signal quality, but also things like solid system ground, high output power, dual channel and a lot of directly accessible features.

And I am glad to learn that Siglent have paid more attention to detail on this nice low cost benchtop AWG than on the SAG1021 USB brick

Still bear in mind that an AWG is just meant to deliver reasonably distortion-free waveforms at reasonably accurate amplitudes and frequencies. And the SAG1021 still does this, despite the various shortcomings in its HW and FW. The average user, who don't try to use it as a precision source will have little to complain about the signal quality.

But then it's an obvious idea. Many folks look for a precision power supply to do similar things as you do, but I've always held the view that an AWG is the much more appropriate tool for this - and less dangerous for the DUT, as it cannot deliver high currents from the outset. But then again, you need to keep the high putput impedance (50 ohms) in mind. For low frequencies, we could easily build an OpAmp output buffer though - this way we can get both, zero output impedance and superfast and reliable current limit at the same time.

[HendriXML]

After many hours of measuring and averaging, the value (dac) transitions of the two awg channels can be shown.

They're better than the first SAG1021, but not in a completely different league then the second.

I kind of hoped for better. Maybe the 16 bit SDG2000X performs better in this regard.

However if I would like to, I could put 2 channels in serie and let them cancel each others error out.

As can seen in the error graphs, the absolute errors are in the 1% range, the amplitude was 3.6V.

The relative stuff (steps) is measured properly, however in the absolute stuff, there's also the accuracy of the scope's offset dac in play.

When I started using an AWG for measurements, I used a simple linear ramp. This made it possible to do linear interpolation and made it also possible to do vertical averaging.

But this also means that glitches (non linearity) can disturb the reliability of the averaging.  Using stairway ramps and not doing any interpolation might give better results, especially when using multiple "windows"/offset voltages. The DSO offset dac then needs to be trustworthy, so I need a way to test/ensure that.

[HendriXML]

For instance, you could use the dual channels to build up an even higher resolution source by e.g. combining Ch1 + Ch2/256, using a stable precision signal attenuator + combiner. You could then make an automated calibration routine that gathers correction values for each output value of Ch1 and use this as an offset for Ch2 (using some very special arbitrary waveforms, or DC mode alternatively), where Ch2 performs the fine steps and needs no additional correction, because the absolute INL error there is already devided down quite nicely.

While doing measurements, I saw behavior that needed explaining.
It seems that the combine function is software/digital and not analog.
So no small stepping possibilities....

https://www.siglentamerica.com/operating-tip/generating-complex-waveforms-using-siglents-combine-function-x-series-dual-channel-generators/

[Performa01]

The Combine function of all the Siglent SDG devices is analog - it is a very nice function that can replace an external power combiner in most applications. The internal combiner has low port to port isolation though, thus generating some third order intermodulation distortions.

EDIT: It turns out that the channel combine function has to be a digital operation indeed.

However, this is not what I have been referring to. With the internal Combine function there is no way to precisely adjust the relative weighting of the channels. You could achieve this by using external attenuators and an external power combiner, but the comercially available components cost some money and accuracy would be limited.

So my suggestion is a voltage divider network, homemade with just a few metal film resistors, operating the SDG in High-Z mode and certainly good enough for frequencies up to a couple of MHz:


2-Way Combiner CD

V1 and V2 are the two output channels (V2 = coarse = 1/3, V1 = fine = 1/3000).

Be aware that R1, R2 just represent the output impedance of the AWG channels.

So the idea is simple: When set to High-Z reading, a V2 output set to 3V will provide 1V at the divider output. If V1 is set to 3V, it will result in a voltage rise by 1mV. So you can adjust the output voltage from 0 to 1V if you set V2 from 0 to 3V and fine adjust it by 0 to 1mV by setting V1 from 0 to 3V, see the graphs below:


2-Way Combiner V1


2-Way Combiner V2

[HendriXML]

The Combine function of all the Siglent SDG devices is analog - it is a very nice function that can replace an external power combiner in most applications. The internal combiner has low port to port isolation though, thus generating some third order intermodulation distortions.

Thanks for your effort, will dive into it. However in my case I need an output that can source some more current. Will take the output resistors in consideration and adjust for them.

But are you really sure the combining is done with lets say voltages? Instead of calculations before the dac’s. I’ve read that this feature has been put in some models afterwards using a firmware update. That implies it is not implemented in HW. The measurements I did have would also have  some  unexplained behavior if its HW. That is off course possible, and I can dive into it if you’re certain. For me it is a blackbox, and I have no clue how thing could be wired without the use of relays. (I cannot hear them).

My measurements will be slow, so I think its could still be usable regarding distortions.

[Performa01]

However in my case I need an output that can source some more current. Will take the output resistors in consideration and adjust for them.

The network as I've suggested it has exactly 50 ohms output impedance, just like the generator outputs itself.

You can still use an OpAmp configured as a voltage follower to act as an output buffer. I would recommend an OPA134 for this.

But are you really sure the combining is done with lets say voltages? Instead of calculations before the dac’s. I’ve read that this feature has been put in some models afterwards using a firmware update. That implies it is not implemented in HW. The measurements I did have would also have  some  unexplained behavior if its HW. That is off course possible, and I can dive into it if you’re certain. For me it is a blackbox, and I have no clue how thing could be wired without the use of relays. (I cannot hear them).

You're correct, I' cannot be sure about that. For some reason I was convinced it works that way, just because of the intermodulation distortions, but when thinking about it, they would be generated just as well even when it is just a digital signal mixing (through the non-linearities of the output amplifier). On the other hand, the fact that the max. output amplitude remains unaffected would rule out a traditional power combiner (50 ohms) and point towards a digital side solution indeed.
 

[HendriXML]

The combine function is presented in a way that indicates there’s is some analog switching. So its quite understandable to assume that. If its just calculating a new waveform, then I even find it a bit misleading.

Just like using the math function on a scope, doesn’t give a true differential probe when using two probes together. So it is rightfully called math on a scope, and not “differential probing”. It should also be called math on the awg if it is just calculating a sum of 2 waves + parameters.