Measuring Distortions with the Scope:What you see is not what you really have..

[blackdog]

Hi,

In order to get a good understanding about the THD of different function generators,
I started up my measurement computer to do some measurements with a very good Soundcard, this is the: M-Audio Delta AP 192.

The yellow dashed line in the FFT window, indicates the bandwidth within which the THD is measured.

The Sinus of the scope picture does not look very clean, this is partly due to the RDC session I use to operate the measuring computer, so these are image artifacts and not a problem with the test signal.
The orange box shows the RMS level of he measurement signal.
For each generator this is set to 2V RMS, there are some exceptions, which I will indicate later.

Also a good sound card has a "sweet spot" this is only important when measuring the generator distortion of the Audio Precision measurement set.
All my function generators are well above the distortion and noise of the sound card.

The settings of this soundcard are these:
192Khz sampling
24-Bit
1-Channel
FFT = 1048576 points
Window = Kaiser-7
Averaging = 4x
Measuring software = Multi-Instruments pro 3.9

======================================
======================================


Let's start with the Siglent SDG1032x just like the previous measurements, here the measurement signal is 10KHz Sine at 2V RMS.



======================================
======================================


The bigger brother Siglent SDG2042x just like the previous measurements, here the measurement signal is 10KHz Sine at 2V RMS.



======================================
======================================


This is the HMF2525, measurement signal is 10KHz Sine at 2V RMS.



======================================
======================================


This is the Rigol DG4162 Chanel-1, measurement signal is 10KHz Sine at 2V RMS.



======================================
======================================


This is the Rigol DG4162 Chanel-2, measurement signal is 10KHz Sine at 2V RMS, the distortion of this channel is lower...



======================================
======================================


This is the generator output of a OWON SDS272s, measurement signal is 10KHz Sine at 855mV RMS, the max. level.



======================================
======================================


This is the generator output of a Silent SDS2100x plus, measurement signal is 10KHz Sine at 1.05V RMS, the max. level.



======================================
======================================


And now to show the limits of the measurement system two measurements of the distortion of the generator in the Audio Precision measurement set, the first measurement is again 10KHz and then performed now with a measurement signal that is the "Sweet Spot" of the total system, the level is just below 3V RMS.



======================================
======================================


And de last one, now at 1KHz, Sweet Spot is 2.5V RMS



======================================
======================================


Good, now it's time to start working on my business administration again, which I like a lot less, but it has to be done. 

Kind regards.
Bram

PS
Thanks mawyatt for finding the error, yes SDG1042X should be  SDG2042X!, my bad 

[mawyatt]

Bram,

Nice, you have a few AWGs

Is that a SDG1042X or should it be a SDG2042X ?

Best

[mawyatt]

Agree, the resistive combiner should produce the best overall IMD from the signal various sources. However, these AWGs under discussion have a built-in combiner, whether it's digital or analog don't know, but produces a Two Tone Signal without any additional components or cables, just one cable, and produces very respectable results (better than we expected) as shown by the various previous plots.

The question about the combining method has been dealt with multiple times already, see for example reply #159 here:

https://www.eevblog.com/forum/testgear/two-tone-test-with-scope-and-sa/msg4250476/#msg4250476

I am amazed about your stellar results. Maybe Siglent have been able to achieve some improvement for this function since the last time I've checked it...

Yeah, kinda figured there was a previous thread(s) on this, but memory is fading 

No complaints here about these SDG2000X & SDG6000X AWGs, altho one might complain about the UI, but we're more concerned/interested in the actual output and why we purchased such. As has been shown by many, these AWGs are pretty good in that respect 

Best,

[Martin72]

So..
I´ve made tests with both gens on my neutrik A1 analyzer(First time use since buying..).
First: Don´t laugh about the pics, it´s display didn´t got a backlight function, so I must use a pocket torch...
Something I have to upgrade on the A1, no doubt, also I should make the software running on a PC of today..
Pics are showing THD+N at 10Khz.
Interesting: Using an isolated transformer decreases the THD on the 1062X, but not on the SDS2122X.

[blackdog]

Hi Martin72

Interesting: Using an isolated transformer decreases the THD on the 1062X, but not on the SDS2122X

This is where learning how to make reliable measurements begins, commonmode problems ar a Bitch! 

I have several isolation transformers and various battery powered measuring instruments here, this is to minimize common mode interference, for reliable measurements always double check is my advice.

Kind regards,
Bram

PS
I have tuned a lot of tape recorders and other equipment with a Neutrik A2.

[Martin72]

for reliable measurements always double check is my advice.

Neutrik mentioned it in the manual ( using earth-free connection) so I thought ok give it a try.
As written, nothing changed on the 2122X, but 1062X.
Neutrik also explaining in the manual, how THD will be measured in the A1.
A notchfilter filters out the fundamental frequency.

[rf-loop]

Martin, you can try one small thing with SDG1000X

Load 50ohm.
1kHz
Set next SDG displayed output levels exactly
These are only some examples.

111.8 mVrms* look what THD you get.
Voltage level band change here (you can hear relay)
111.9 mVrms* look THD
353.4 mVrms* look THD
Voltage level band change here (you can hear relay)
353.5 mVrms* look THD
1.118 Vrms* look THD
Voltage level band change here (you can hear relay)
1.119 Vrms*
up to 3.536Vrms*
* Listen relay! for make sure V band border. I do not have any idea why it change some times example 353.4 or .5 and 353.5 or .6  and so on (in other V band change same, some times just tiny amount different)

Example just with my Keithley 2015 THD what is not so stellar but better than nothing. Yes I have many tables done long time ago but they are only on paper notebook.

(Setup for this Kethley 2015 THD: Mode Just plain THD (not THD+N) , freq.set manually 1kHz (instead of auto tune), Harmonics: up to 9th, SFIL:NONE (available NONE, A, C, CCITT, CCIRARM),  Average 100rdgs repeating (front panel button Filter, Keithley input feed thru 50ohm.
Examples
353.4 Vrms* my Keithley display 0.0196%  (this voltage band bottom and lowest THD) one V band top value
353.5 Vrms* my Keithley display 0.0918%  (this voltage band top and highest THD ) next V band bottom value

This THD changing behavior repeats through all level bands (but naturally with bit different numbers).


So THD is not just some value... it is many different values depending many things.
If user know his generator he can perhaps some times select better signal quality. Just example if want use 4dBm. Use 3.97dBm and THD is better.  THD is better when use internal voltage bands top values. And most bad when use just internal voltage bands bottom values.

[blackdog]

Hi rf-loop,

Thank you very much for explaining about the difference in behavior due to the output level of a function generator.
I wanted to do this in the third part based on the aberrations that occur on the flanks of the square wave signals from the function generators I have here.

The variation in THD also occurs with analog sine wave generators when additional gain stages are turned on or off there.
But the differences are usually much smaller.

My setup for the many measurements I have chosen (2V RMS at 50 Ohm load) where possible so that you have a bit of certainty,
that 2V RMS you can then externally with a potentiometer or with 50 Ohm attenuators back to a lower level, but the "Sweet Spot" of every function generator can and wil be different.
Look at my Rigol DG4162 chanel 1 and 2 distortion levels, same level, frequency but 40% different THD, probably has at to do with the linearity of the DA converters used in the DG4162 function generator.
This can happen with all function generators with two or more outputs.

In the measurements I showed here I assumed the audio frequency range.
The distortion that occurs with the multiple function generators I have is fairly constant over the audio range.
This is because the design of most function generators covers a much larger frequency range than the audio band and the opamps used i n the amplification stages are very broadband.

Distortion meters that cannot show a residual always give uncertainty about the indicated value of THD.
All THD meters implemented with some form of NOTCH filter almost always show the THD and many other forms of residue that have no relation to the test signal.

Many analog THD meters therefore have a number of filters built in to provide more certainty about the displayed THD value.
Often it is also referred to as THD + N which stands for Total Harmonic Distortion and Noise.

I purposely showed the original signal in yellow in my measurements on the scope photos with the output of my Audio Precision THD function in green.
So that a reasonable estimate can be made of what the residue looks like.
I am aware that a scope picture of the residue can poorly show that e.g. a lot of 1/F noise is present.
That is what the measurements I did last with my measurement computer and the proper sound card are for.

The user of one function generator will have to look carefully at what the datasheet says in terms of distortion at a given frequency range.
Almost always the distortion data are better than specified.
And as rf-loop points out, by playing with the output level you can often get the distortion lower because then there are less amplifier stages/relays in the signal way and the amplifier stages are more optimally in their working range.

But be aware that if you generate e.g. signals around 10MHz, then the DA and the amplifier stages will have more errors and the distortion will be much higher because of e.g. the slew rate of the opamps used in the output section of the function generator.

As almost always, figure out what your good and not-so-good points are of your used equipment and think about those points when you start taking measurements.
Several times I have done measurements on amplifier stages driven by the HMF2525 function generator which is very clean for a 14Bit model, see computer FFT for this.
And I was expecting during the measurements I was doing what a THD around 0.002% or less and I was stuck at 0.02% THD.
I had forgotten several times that I was using the THD meter in the Audio Precision, but the HMF2525 as a generator, before I made the THD measurements,
I had tested my D.U.T with a square signal from the HMF2525, for measuring low THD, I had forgotten to switch generators....

Brief conclusion regarding audio measurements with a sine signal.
The Siglent SDG1032X is quite useful for tube equipment, this is because with most tube equipment the THD is quite high.

For significantly lower THD, it is better to choose a Siglent SDG2042X, because this generator has a 16Bit DAC.
But please note that I am talking about sine wave signals, the SDG1032X has a very good point and I will soon show it in measurements of pulse signals from the function generators.

This time i do not remove the text below, now you now why "some times" mijn Englis is that bad. 
Translated with www.DeepL.com/Translator (free version)

Kind regards,
Bram

[rf-loop]

Here is one example.
Level do not change nearly anything but SDG change internal "voltage band" (whole range have many voltage bands)

I take here just one example. Previous one level band ends (highest in this band) at 353.5mVrms and next step go to next band most low level what is 353.6mVrms. (both are very close 1Vpp and roughly 4dBm (bit under).

I do not have good and steep filters for audio frequencies so my SA noise level and max input level give some limits what I can do. If I have good filter I can go more down because of course there can find more harmonics and also some spurs.  Also Spectrum analyzer mixer level need keep enough low for keep its own harmonics down enough. So in this case -110dBm level is border.

After generator output there is external attenuator set for drop SA input level to -25dBm. And because SA internal Att is 0dB "mixer level" is -25dBm (this SA is specified for -30dBm mixer level but with my cross check this 5dB did not give any readable difference in harmonics levels in this case. It can use higher mixer levels but I want keep SA own generated things low)

It is fun to see how much harmonics levels change. Yes it is natural if look how SDG works and not go deeper inside now.

But fore some users this behavior is good to know.
And, this around same happen between every internal voltage band change. This is just one of them.

Note that these THD % are calculated using only 1st (fundamental), 2nd and 3th harmonic levels

[Martin72]

Martin, you can try one small thing with SDG1000X

Load 50ohm.
1kHz
Set next SDG displayed output levels exactly
These are only some examples.

111.8 mVrms* look what THD you get.
Voltage level band change here (you can hear relay)
111.9 mVrms* look THD
353.4 mVrms* look THD
Voltage level band change here (you can hear relay)
353.5 mVrms* look THD
1.118 Vrms* look THD
Voltage level band change here (you can hear relay)
1.119 Vrms*
up to 3.536Vrms*
* Listen relay! for make sure V band border. I do not have any idea why it change some times example 353.4 or .5 and 353.5 or .6  and so on (in other V band change same, some times just tiny amount different)

Setup: SDG1062X with 50 ohm load termination connected to the A1.
Values with * = relay switch point


111.7mVrms   0.044%
111.8mVrms* 0.117%
.
.
353.4mVrms   0.043%
353.5mVrms* 0.116%
.
.
1.116Vrms      0.042%
1.118Vrms*    0.114%
.
.
1.2Vrms         0.110%
1.4Vrms         0.100%
1.6Vrms         0.091%
1.8Vrms         0.083%
.
.
.
2.9Vrms       0.050%
3.0Vrms       0.048%
3.056Vrms*  0.047%
3.535Vrms    0.041%

On every "relais switch point" THD increases, then decreasing until the next switch point with one exception: The last switch point won´t be affected(or it is a different relais from a different circuit).
Interesting things, thank you rf-loop 

[rf-loop]

Martin, you can try one small thing with SDG1000X

Load 50ohm.
1kHz
Set next SDG displayed output levels exactly
These are only some examples.

111.8 mVrms* look what THD you get.
Voltage level band change here (you can hear relay)
111.9 mVrms* look THD
353.4 mVrms* look THD
Voltage level band change here (you can hear relay)
353.5 mVrms* look THD
1.118 Vrms* look THD
Voltage level band change here (you can hear relay)
1.119 Vrms*
up to 3.536Vrms*
* Listen relay! for make sure V band border. I do not have any idea why it change some times example 353.4 or .5 and 353.5 or .6  and so on (in other V band change same, some times just tiny amount different)

Setup: SDG1062X with 50 ohm load termination connected to the A1.
Values with * = relay switch point


111.7mVrms   0.044%
111.8mVrms* 0.117%
.
.
353.4mVrms   0.043%
353.5mVrms* 0.116%
.
.
1.116Vrms      0.042%
1.118Vrms*    0.114%
.
.
1.2Vrms         0.110%
1.4Vrms         0.100%
1.6Vrms         0.091%
1.8Vrms         0.083%
.
.
.
2.9Vrms       0.050%
3.0Vrms       0.048%
3.056Vrms*  0.047%
3.535Vrms    0.041%

On every "relais switch point" THD increases, then decreasing until the next switch point with one exception: The last switch point won´t be affected(or it is a different relais from a different circuit).
Interesting things, thank you rf-loop 

Are these THD (Not THD+N) values plain THD without filtering and how many harmonics included in counting.  (looks bit high values if plain THD and No filters.)

[Martin72]

Hi,

No filters were set (forget, there is a high-pass filter from 400Hz and a band-pass filter that lets through everything from 22Hz to 22Khz), the A1 has only THD+N (IMHO), bandwidth up to 130Khz.
The data sheet of the SDG1000X series gives a THD of 0.075% (10Hz...20Khz), the measured values are significantly lower.

[rf-loop]

Here my some older measurements and also some today data  (perhaps these need do again, data partially from very old paper notebook.)

Revised, supplemented and updated table.

Plain THD (Not THD+N) %
SFILT: NO
Harmonics up to 7th.
Signal 1kHz.
Measured only using 50 ohm load.
Meter is Keithley 2015 THD

Edit: image changed.  New revised image:

[markone]

Here my some older measurements and also some today data  (perhaps these need do again, data partially from very old paper notebook.)

Plain THD (Not THD+N) %
SFILT: NO
Harmonics up to 7th.
Signal 1kHz.
Measured only using 50 ohm load.
Meter is Keithley 2015 THD

I would say that those numbers are great for a generator that goes up to tens of megahertz @ few hundreds of euros, i personally own a SDG2042x "upgraded" and i never expected a spectral purity at low frequencies on par with fancy audio 24 bits DACs for the simple reason that it's not its job.

At the same time i do not expect that a 12 bits DSO is able to provide an actual dynamic range over 73 dB, regardless of any mathematical process in place, at this frequencies a good soundcard does a much better job.

[Martin72]

Here my some older measurements and also some today data  (perhaps these need do again, data partially from very old paper notebook.)

Plain THD (Not THD+N) %
SFILT: NO
Harmonics up to 7th.
Signal 1kHz.
Measured only using 50 ohm load.
Meter is Keithley 2015 THD

Hi,

The THD specification of the A1 says nothing about the number of measured harmonics, but the measuring range is given, 20hz to 40Khz.
Whether this also means that at a fundamental frequency of 1Khz it also measures up to the 40th harmonic, I don't know.
I will ask Neutrik.
"Tomorrow" I repeat the measurements, then with the 20Khz filter, possibly the values decrease.
Either way, both measurement series are far below the specifications mentioned - a very good result for such a "cheap" device.

[rf-loop]

The data sheet of the SDG1000X series gives a THD of 0.075% (10Hz...20Khz), the measured values are significantly lower.

SDG1000X_DataSheet_DS0201X_E01I:

Total Harmonic Distortion Max 0.15 % 0 dBm, 10 Hz ~ 20 kHz

Note that Min or Typical is not listed, only Max.
Example I have not at all tested all frequencies between 10Hz to 20kHz.

And then in datasheet introduction and advertising text:

With 0 dBm output, the THD (Total Harmonic
Distortion) is less than 0.15%. Harmonics and
spurs are less than -40 dBc throughout the entire
bandwidth.

0.15% is specified only up to 20kHz!


Some may read this text so that 0.15% is also for entire badwith, if not calculate what this -40dBc mean.
Example if look harmonics levels graph there can  find that 60MHz have THD roughly 0.75% if look (only) 2nd and 3th levels alone.

If look again this image and there 100kHz 2nd and 3th levels it give roughly ~0.05%

But then, there is also some source for mess when we talk about THD.

Example your A1 do not measure harmonics itself at all. THis is why it do not have THD mode at all, it have only THD+N bacause its working principle and this N also include noise but also all non harmonic spurs of course... everything in band except fundamental filtered away using notch filter,  if I understand its user manual right.
It looks like it just measure whole all in its bandwidth just with notch filter over fundamental. So there is all, noise, all spurs and all harmonics.
And example Keithley 2015 measure harmonics. Because THD. It is Total HarmonicsDistortion.  If there is 2kHz signal and some spur in 5kHz. It is NOT harmonic at all. It is other tone, not part of pure THD.

There can also be some filters, as example Flat aka NONE and then C, CCIR A, CCITT, CCIR ARM etc but these available in Keithley.

With every setup I get different result. Also depending what all harmonics are counted. (Meters like A1 can not select harmonics at all. It can not measure harmonics, lt looks like it measure only just whole all in band. (explained in its manual)

Then, example Keithley have level ranges 100mV, 1V, 10V etc. Result also depends if test signal is near used range full scale or near bottom before lower range. (bit same as with SDG voltage bands behavior relative to THD) 
Keithley specification say 0.004% residual distortion for THD and individual harmonics. THD+N it say 0.056% (-65dBc) residual .

When someone tell just "THD" we need ask "what THD?", how measured and including what - if want avoid mess.

Btw, if I set Keithley for max harmonics up to 20kHz for 1kHz test signal and I select THD+N  and filter CCIT ARM  I get around same "THD" values as you. (just fast looked only with two levels)

The THD specification of the A1 says nothing about the number of measured harmonics, but the measuring range is given, 20hz to 40Khz.

It is somehow told in its manual. It looks like it have nothing for measure individual harmonics. This is why it have only mode THD+N. It have different working principle. Roughly saying: It measure fundamental and measure what is total without fundamental in measuring bandwidth. THD+N, so it measure all harmonics distortion and added with (+N) all non harmonics and noise in band.

If I set signal 1kHz and level SDG1000X voltage band IV top level (3.978dBm)
Then Keithley for THD+N,  harmonics up to 20th  and No filter aka flat.  0.0355 % (100 rdgs average)
Exactly same setup but THD alone: 0.0208 %
Just for fun:
Then continue same THD alone with CCIR ARM weighting filter: 0.0518 %
Then same but THD+N with CCIR ARM weighting: 0.0565 %

[Martin72]

Hi,
The A1 is an all in one audio measurement system that can also measure tape decks, not a dedicated harmonic analyzer.
And as always when you have everything in one device, sometimes something may be omitted.
But what it does, it does well.

SDG1000X_DataSheet_DS0201X_E01I:

I got this here:

https://www.batronix.com/files/Siglent/Funktionsgeneratoren/SDG1000X/SDG1000X_Datasheet_EN.pdf

Where the 0.075% for 10Hz....20Khz is stated.

[blurpy]

I did some THD measurements using a 24 bit audio analyzer a while ago, of the SDG2000X:

https://www.eevblog.com/forum/testgear/the-siglent-sdg2042x-thread/msg3606484/?topicseen#msg3606484

It does quite well I think.

[rf-loop]

Hi,
The A1 is an all in one audio measurement system that can also measure tape decks, not a dedicated harmonic analyzer.
And as always when you have everything in one device, sometimes something may be omitted.
But what it does, it does well.

SDG1000X_DataSheet_DS0201X_E01I:

I got this here:

https://www.batronix.com/files/Siglent/Funktionsgeneratoren/SDG1000X/SDG1000X_Datasheet_EN.pdf

Where the 0.075% for 10Hz....20Khz is stated.

Yes A1 is nice "all in one" instrument for repair-tune-service etc audio things.

It is shame Batronix still display old obsolete data sheet from year 2017.
Even TrueArb feature is missing.

(but yes THD is there Max 0.075% for 0dBm for 10Hz to 20kHz)
I can go down to 20Hz and Keithley tell THD 0.0293 % (100 rdgs average)
And up to 16kHz if want include 3th harmonic. (bacause it have 50kHz band max for harmonics, so with 16kHz it can still calculate 2nd and 3th): Result 0.0136%

Now need note thgat my SDG1000X is old when this old datasheet have been ok.
Why Siglent have changed this THD specification to double.

What is your SDG1000X hardware version?

If some one have more new HW and have possibility to measure somehow reliable THD... plese.
I'm curious to see if something has changed that would explain the change in specs.
 

[Martin72]

Hi,

What is your SDG1000X hardware version?

02-01-00-24-00

Interesting that this specification has changed so "drastically" in the last 5 years.

[rf-loop]

Hi,

What is your SDG1000X hardware version?

02-01-00-24-00

Interesting that this specification has changed so "drastically" in the last 5 years.

We have same HW.

I also wonder what have lead to this change.
 
I hope that some of the users who have latest HW can do comparable tests so that we can see if there have really been significant changes in this matter, or if the specification has just been changed due to some borderline case so that no one can complain under any circumstances that it does not meet the specifications. .

[pdenisowski]

Example your A1 do not measure harmonics itself at all. THis is why it do not have THD mode at all, it have only THD+N bacause its working principle and this N also include noise but also all non harmonic spurs of course... everything in band except fundamental filtered away using notch filter,  if I understand its user manual right.
It looks like it just measure whole all in its bandwidth just with notch filter over fundamental. So there is all, noise, all spurs and all harmonics.
And example Keithley 2015 measure harmonics. Because THD. It is Total HarmonicsDistortion.  If there is 2kHz signal and some spur in 5kHz. It is NOT harmonic at all. It is other tone, not part of pure THD.

When someone tell just "THD" we need ask "what THD?", how measured and including what - if want avoid mess.

Agreed.  This is how I would describe it as well:  THD is only the power in the (sum of N) harmonics.  The power of each harmonic is individually measured with a sufficiently narrow resolution bandwidth to minimize power from noise, spurs, etc.  In this case you have to specify the fundamental frequency, the number of harmonics you want to include in the measurement, and the starting RBW (this increases in a predictable way for each harmonic, so only need to specify it for the fundamental).

For THD+N, you essentially measure all power (harmonics, spurs, noise, etc.) over a bandwidth with the fundamental notched out (again, as narrowly as possible).  THD+N requires that you specify the bandwidth of interest.  Traditionally, THD+N was an easier measurement to make than THD, and I often see people confusing THD and THD+N.

I recently did two videos on this.  They are primarily focused on spec ans and RF (not AF), but the basic concepts are still the same

Understanding Harmonic Distortion Measurements


Measuring Harmonic Distortion with the FSW (which talks about the role of RBW)

[Martin72]

For THD+N, you essentially measure all power (harmonics, spurs, noise, etc.)

To me it sounds more plausible to measure everything than "only" the harmonics - Or do I have a thinking error?
BTW, here is an excerpt from the service manual on how to perform the measurement(pic below).
(The Googledrive link with the docs for the A1 is still active - If it´s interesting I could post it)

[RoV]

Hi, I share also here results of a measurement I posted under the topic "Re: Siglent SDG1032X sine distortion at 1 kHz ?"

Made a test on my SDG2042X (ok, 2122X  ).
Used an E-MU 0202 connected to my PC, acquiring ~20 s at 48000 Hz sampling, 24 bit resolution. Generator at 1 kHz sine, 1 Vpp output. Acquisition gain set to have about 40% level on sin peaks.
Processed with Welch method, using a Blackman window of length 4800 to minimize sidelobes and 50% overlap between data chunks: result in figure, with fundamental normalized to 0 dB.
2nd harmonic is at -103.5 dB, 3rd at -105.5. Worst is 10th at -101.9 dB.
THD, computed up to the 10th harmonic, is -97.5 dBc, or 0.0013%.

[Martin72]

Looks too fantastic...
Seriously, good thing.

@rf-loop:

Concerning thd specs of sdg1000x: I´ve read the specs of the sdg2000x, in the old sheet (2017) and in the new, a thd of 0.075% is mentioned.
Same as in the specs of the sdg1000x in 2017...