SDG1032X vs SDG2042X for Audio

[KungFuJosh]

Hi,

It was suggested to me that I use a sweeping function generator with my SDS1104X-E to look at distortion characteristics of audio equipment. I'd mostly be testing tube amplifiers and LD microphones. I'm also interested in seeing what things look like with Bode Plot II.

As the title says, I'm looking at the SDG1032X and the SDG2042X. The 1032 being 14bit, and the 2042 being 16bit, I assume the latter is better for audio, but given the bit depth of the scope, does it even matter? Thoughts?

Side note: the cost difference doesn't matter to me more than the usefulness of either unit.

Thanks,
Josh

[BillB]

Since you've already got the SDS1104X-E that can control a Siglent SDG for Bode Plots, and if cost doesn't matter to you, then why wouldn't you get the SDG2042X?

[Performa01]

You are right. An 8-bit DSO cannot measure distortions below some -48dB because of the limited linearity of the ADC.

Likewise, even though modern waveform generators have up to 16 bits, their fast DACs cannot be ultra-linear and this is most likely also true for the analog amplifiers inside the generator.

I don’t have a SDG1000X or 2000X, but I can show you what to expect from a 16bit system with my SDG6052X. It should perform pretty much the same as a SDG2000X at low frequencies.

First a 20kHz reference measurement with a 16bit DSO, which is optimized for low distortion and noise and can cleanly measure down to -96dB:


SDG6052X_THD_20kHz_Pico4262

An ideal 16bit generator would have a THD of no more than -96dB.

From the screenshot above we can see that the strongest harmonic is the 2nd and it is only ~82dB down.
THD up to 1MHz is measured at -78dBc, hence some 16dB higher than the theoretical ideal value.
Total Harmonic Distortion measured at 1MHz bandwidth is still low at 0.012%.


Now let’s try a similar measurement with the Siglent SDS1104X-E:


SDG6052X_THD_20kHz_SDS1104X-E

The Siglent doesn’t support automatic THD measurements, but we can still compare the strongest harmonic. For this we need to calculate the difference between the fundamental at 20kHz and the 2nd harmonic at 40kHz: +8.95dBV - -37.9dBV = 46.85dBV; if this were the only harmonic then THD would be measured as slightly less than 0.5%, hence much higher than it actually is.

Compare this to the 82dB we got with the 16bit DSO…

So it makes no difference for distortion measurements on an 8bit scope whether you get the SDG1000X or SDG2000X. For Bode Plots, a SDG2000X can go up to 120MHz, whereas the SDG1000X is limited to 60MHz max. Since you seem to be mostly interested in audio, this wouldn’t matter either.

[KungFuJosh]

Since you've already got the SDS1104X-E that can control a Siglent SDG for Bode Plots, and if cost doesn't matter to you, then why wouldn't you get the SDG2042X?

Cost is less important when it comes with a benefit. No benefit = no value = why waste almost $200?

The SDG1032X is a good bit cheaper and while 14 bit, less memory and no touch display otherwise offers the same functionality.

Should I care about having less memory for audio? I don't care about the touch display.

Thanks,
Josh

[Performa01]

Should I care about having less memory for audio? I don't care about the touch display.

Memory is only significant for arbitrary waveforms. Even though these might be useful for audio in some special situations, I don't think that you'd ever need long memory for this.

The touch screen on the SDG2000X and 6000X is just a gimmick. The user interface is still the same as for the SDG1000X, i.e. not at all optimized for touch use. You'll hardly miss it.

[BillB]

Since you've already got the SDS1104X-E that can control a Siglent SDG for Bode Plots, and if cost doesn't matter to you, then why wouldn't you get the SDG2042X?

Cost is less important when it comes with a benefit. No benefit = no value = why waste almost $200?

True.  Others have struggled with this decision, too, and short of a bandwidth requirement I would't talk someone up to a 2042X.  I also agree that the touchscreen isn't really worth it.  The 1032X now has EasyPulse and TruArb as well.

https://www.eevblog.com/forum/testgear/sdg-2042x-as-first-fg/msg2189009/#msg2189009

[KungFuJosh]

True.  Others have struggled with this decision, too, and short of a bandwidth requirement I would't talk someone up to a 2042X.  I also agree that the touchscreen isn't really worth it.  The 1032X now has EasyPulse and TruArb as well.

https://www.eevblog.com/forum/testgear/sdg-2042x-as-first-fg/msg2189009/#msg2189009

They are/were both hackable to higher bandwidth from what I've seen. But since I probably don't need higher than 40K, that doesn't make a difference.

Ha, I called Saelig to see if one was out of stock, and they're both in stock. So that doesn't help.

Here's a dumb question: should I get the 16bit 2042X in case I ever get a 16bit scope? Even though I like this excuse, it seems dumb because chances are that better AWGs will come out in that price range long before I get another scope.

Thanks,
Josh

[rf-loop]

<clip>
Now let’s try a similar measurement with the Siglent SDS1104X-E:


SDG6052X_THD_20kHz_SDS1104X-E

The Siglent doesn’t support automatic THD measurements, but we can still compare the strongest harmonic. For this we need to calculate the difference between the fundamental at 20kHz and the 2nd harmonic at 40kHz: +8.95dBV - -37.9dBV = 46.85dBV; if this were the only harmonic then THD would be measured as slightly less than 0.5%, hence much higher than it actually is.

Compare this to the 82dB we got with the 16bit DSO…

So it makes no difference for distortion measurements on an 8bit scope whether you get the SDG1000X or SDG2000X. For Bode Plots, a SDG2000X can go up to 120MHz, whereas the SDG1000X is limited to 60MHz max. Since you seem to be mostly interested in audio, this wouldn’t matter either.

Are you sure there is not somethng wrong with your SDS FFT setup (do not drive signal to nonlinear range. Set signal level or V/div to ~6div-pp and look harmonics level rel carrier. Then rise signal or adjust V/div so that signal p-p (YT display)  is higher and higher, Look where harmonics start rising much  more than fundamental     ).

Here with SDG1062X (SDG1032X)

Signal 1Vrms, 20kHz (checked my hand write test notebook, 1.09Vrms)

(I do not remember window)


FFT Window Flattop

[Performa01]

Josh, the question is, how much difference there is between a 14bit vs. 16bit AWG.

I cannot test a SDG1000X, but I have reviewed the SAG1021, which is also 14bit. It sure will not perform quite as good as a SDG1000X, because of the lower sample rate and low voltage output stage.

The SAG1021 shows -70.5dB THD with a measurement bandwidth of 100kHz. The bandwidth difference is insignificant for a low distortion sinewave like this, so we're talking about less than 8dB difference, whereas in theory the difference between 14 and 16 bits should be 12dB. That means the 14bit DAC is closer to the ideal performance than the 16bit.

As I said, I'd expect a SDG1000X to be better, so the difference would be even smaller. It's up to you to decide if this is enough to want the SDG2000X.

I think both have a good price/performance ratio and you're not going to get something significantly better anytime soon. As much as I wanted to, I cannot give you a technical argument why you should get the SDG2000X for your needs. So it's really up to you - looks like it has to be a gut decision

[BillB]

True.  Others have struggled with this decision, too, and short of a bandwidth requirement I would't talk someone up to a 2042X.  I also agree that the touchscreen isn't really worth it.  The 1032X now has EasyPulse and TruArb as well.

https://www.eevblog.com/forum/testgear/sdg-2042x-as-first-fg/msg2189009/#msg2189009

Here's a dumb question: should I get the 16bit 2042X in case I ever get a 16bit scope? Even though I like this excuse, it seems dumb because chances are that better AWGs will come out in that price range long before I get another scope.

Well that is a dumb question to ask a bunch of TEAs that hang out in a test equipment internet forum.   

Seriously, while having the better one is always nice, the fact that the 1032X is on sale makes it a pretty good deal for what you get.  If you got that one now, you could try it out and if it didn't meet your needs or you grow out of it, you could always sell it and make most of your money back.  (Or if you were a true addict, you'd keep it and buy the fancy 24-bit 1GHz model when the time comes.  )

In fact, I originally purchased the 1032X and was happy with it, but then I stumbled upon a cheap used 2042X.  So, I sold the 1032X to justify buying the 2042X.  I later on needed an additional AWG and immediately regretted selling the 1032X.  So, I bought another 1032X (luckily a cheap used one, but for more than I sold the first 1032X for!  )

[rf-loop]

Josh, the question is, how much difference there is between a 14bit vs. 16bit AWG.

I cannot test a SDG1000X, but I have reviewed the SAG1021, which is also 14bit. It sure will not perform quite as good as a SDG1000X, because of the lower sample rate and low voltage output stage.

The SAG1021 shows -70.5dB THD with a measurement bandwidth of 100kHz. The bandwidth difference is insignificant for a low distortion sinewave like this, so we're talking about less than 8dB difference, whereas in theory the difference between 14 and 16 bits should be 12dB. That means the 14bit DAC is closer to the ideal performance than the 16bit.

As I said, I'd expect a SDG1000X to be better, so the difference would be even smaller. It's up to you to decide if this is enough to want the SDG2000X.

I think both have a good price/performance ratio and you're not going to get something significantly better anytime soon. As much as I wanted to, I cannot give you a technical argument why you should get the SDG2000X fort your needs. So it's really up to you - looks like it has to be a gut decision

In my previous message

Are you sure there is not somethng wrong with your SDS FFT setup (do not drive signal to nonlinear range. Set signal level or V/div to ~6div-pp and look harmonics level rel carrier. Then rise signal or adjust V/div so that signal p-p (YT display)  is higher and higher, Look where harmonics start rising much  more than fundamental     ).

Are you sure your test with SDG2000X and SDS FFT is ok.  (These harmonics level looks terrible high)

[Performa01]

rf-loop, I've used a generator signal of 8Vpp = 8div pk-pk, which should be fine since the total input range is actually 10div pk-pk.

I've repeated the test with 7Vpp:


SDG6052X_THD_20kHz_SDS1104X-E_7Vpp

Now we get 7.82dBV - -39.7dBV = 47.52dB;

Not much of a difference here, but the 3rd harmonic has dropped indeed, which is somewhat unexpected.

Of course I was wondering, why you get so much better results. Is it my early pre-production sample of the SDS1104X-E that has a worse frontend?

So I looked for the differences and noticed that you probably have used a 50ohm inline terminator, whereas I just made a direct connection into 1Mohm, just like I did it with the Picoscope.

I've tried again, this time with through termination:


SDG6052X_THD_20kHz_SDS1104X-E_3.5Vpp_50

This is now  1.68dBV - -52.3dBV = 54dB;

Okay, this is a bit better now. Still not quite as good as the 58dB in your measurement, so the frontend might indeed be improved in current instruments.

[rf-loop]

@Performa01

Have you noted that over  6-7 div p-p  it is not very linear... over 8-9 div p-p it is far from ideal linearity. In normal YT scoping and signals in displayed area it still is quite ok for normal use. But sinewave is not anymore accurate sine and this can see in frequency domain. It is well known that YT dsisplay human eye can not detect small distortion but frequency domain... SA or scope FFT  we can see these very clearly.
Sidenote. My SDS1000X-E is also bit old but perhaps not as old as yours! (My is November -17)

And sidenote: SDG2kX and SDG1kX  Harmonics distortions are specified for 0dBm signal.
SDG2kX -65dBc  and 1kX -60dBc under 10MHz and 0dBm

[Performa01]

rf-loop, I've just tried 6Vpp and the 2nd harmonic dropped by one more dB.

Yes, the input doesn't seem to be very linear indeed, but up to 7 divisions total should be okay.

There's still the funny phenomenon that the distortion got so much better with 50 ohm termination. So I've tried unterminated connection again, but only 3Vpp this time - and lo and behold, the 2nd harmonic is 55dB down.

So the 1V/div range has a significantly worse distortion performance than the 500mV/div range.

It's not a problem with the generator level, as it has shown very low distortion even at 8Vpp into the Picoscope.

[rf-loop]

rf-loop, I've just tried 6Vpp and the 2nd harmonic dropped by one more dB.

Yes, the input doesn't seem to be very linear indeed, but up to 7 divisions total should be okay.

There's still the funny phenomenon that the distortion got so much better with 50 ohm termination. So I've tried unterminated connection again, but only 3Vpp this time - and lo and behold, the 2nd harmonic is 55dB down.

So the 1V/div range has a significantly worse distortion performance than the 500mV/div range.

It's not a problem with the generator level, as it has shown very low distortion even at 8Vpp into the Picoscope.

There is some differences between V/div ranges when look linearity.

Generators, SDG1kX and SDG2kX and so on are much better what can evaluate using example SDS1000X-E FFT.
I have looked these Generators using SSA3kX  (also SDG1k and 5k with HP8568B what I do not anymore have) anmd it is clear that SDS scopes FFT can not evaluate these generators performance. There come: The wagon pulls the horse - effect.

Time ago I made some further tests for evaluate SDS1kX-E  FFT performance but  after new  FW version 6.1.33  I throw all these tests to garbage collection.

Time ago I made some further tests for evaluate SDS1kX-E  FFT performance but  after new  FW version 6.1.33  I have throw all these obsolete tests to garbage box.  Some times frustrating. Later something more about SDS1kX-E  new 6.1.33 FFT. perhaps in its own FFT thread after lot of fiddly, time-consuming, tests are even somehow ready and cross checked for avoid garbage class tests.

I have one  rough general thumb rule recommendation for SDS1kX-E  FFT users: Price of this "rule" is part of dynamic range. Keep level of SUT (Signal Under Test) max  6 div p-p on the YT display with used V/div setting.  ( 6 division "rule" is a nice coincidence with the old Tektronix service manuals. )
(And sidenote - when use BodePlot II and Channel Gain Auto - it adjust automatically well)

Next pictures show how it go using just one tiny set of settings.
There can see how harmonics level drops lot of more than fundamental drop.
There is used harmonics measurement function. left table show 1st harmonic (fundamental/carrier) and other harmonics abs level dBm and right column harmonics level vs 1st harmonic (can read also as dBc)
There is 100mV/div  screen height is 8div (800mV)  and full vertical range is bit over 10 div / bit over 1V before it clip (1024) but due to fact analog part is not linear there is some compression before ADC limits.
YT mode measurements also change around 8div and example over 790mVpp it add > character before number. Inside 7-8 vertical div in YT mode this compression is not lot. But... it generate harmonics and when we are looking in vertical log scale...boom...

Note these 2. 3. 4. and 5. harmonics level related to fundamental (rightmost column in table).
Images are from input level +4dBm full scale and then input level 1dBm steps down to half voltage level.

[Performa01]

I have one  rough general thumb rule recommendation for SDS1kX-E  FFT users: Price of this "rule" is part of dynamic range. Keep level of SUT (Signal Under Test) max  6 div p-p on the YT display with used V/div setting.  ( 6 division "rule" is a nice coincidence with the old Tektronix service manuals. )

It turns out that the rules for an optimized setup for distortion measurements are actually a bit more complex than that.
Thankfully, the distortion performance can be amazingly good and there are just a few "bad" setups, that we should avoid.

See my reply #186 here:

https://www.eevblog.com/forum/testgear/siglent-sds1104x-e-in-depth-review/msg2502165/#msg2502165

[rf-loop]

It turns out that the rules for an optimized setup for distortion measurements are actually a bit more complex than that.

Of course all are always much more - and even more^2 - complex than extremely simplified "thumb rules". 

[KungFuJosh]

I ordered the 2042X. It came down to 16bit 2000X series vs. a supposedly better square wave generator in the 1000X, and obviously I went with the 16bit unit.

[n3mmr]

This is an old thread, but still:

If you want to work with high performance audio, and measure thd and IM and stuff att very low levels, Quantasylum is the way to go.

The QA401 is a usb audio interface with the addition of a very good analysis engine implemented in an FPGA + an SoC. It outperforms all other audio interfaces that I know of.
This unit comes very close to measuring near the limits of a 24bit system.

You need a ( very low perf required ) tiny pc + a good hi res screen to complete the system

All in all about the same price as a DSO and an SDG. And is actually good enough for audio work.