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

[pdenisowski]

To me it sounds more plausible to measure everything than "only" the harmonics - Or do I have a thinking error?

No, that's absolutely correct:  measuring "everything" over a bandwidth (minus the fundamental), i.e. THD+N, is relatively easy:  just notch the fundamental out and integrate over the bandwidth, so to speak.

THD (just the harmonics, not noise, spurs, etc.) was a lot harder to measure before modern spec ans with automated measurement personalities.  You first have to measure the power of the fundamental - this is usually done in zero span mode with a RBW just slightly wider than the signal.  Then you have to change frequency to each harmonic (calculated based on the frequency of the fundamental) and measure those in zero span.  But since the width of the harmonics grows with harmonic order, you have to scale up RBW at each harmonic to make sure you're measuring the entire (wider) harmonic signal.

But again, modern spec ans with an automated harmonic distortion measurment function can do all of this very easily -- measuring THD "by hand" can be very time-consuming and error-prone.  I would say that THD is actually an easier measurement than THD+N (at least with a spec an).  THD+N requires a tunable notch filter, whereas THD with a modern spec an doesn't require any additional hardware.

It might also be worth mentioning that THD+N is essentially the inverse of SINAD, which is a measurement often supported on audio and some radio testers, but not common on RF spectrum analyzers.

[Performa01]

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?

I guess so…

Of course it used to be a lot easier to measure THD+N, as it just required an RMS voltmeter and a notch filter. This is how the traditional distortion meters work. No way to measure (and calculate!) THD without a spectrum analyzer.

But just because something was much easier to accomplish in the past, it is not automatically more plausible as well. You might think so as an end user ("I don't care what it is, it just doesn't belong there!"), but even as an end user you are affected differently by the different unwanted signals.

Consider the most popular area where distortion is a big thing: Audio. So you have, say, -50 dBc THD+N and you think this is fairly bad and it does not matter where all the unwanted signals stem from. But:

•   If it's pure noise, then your signal is essentially distortion-free but the signal to noise ratio is just 50 dB. Some folks won't even notice that noise, at least not with low dynamic pop/rock music…
•   If it's pure harmonics, then you might notice it because of a slightly altered sound. If the harmonics happen to be predominately even numbered, then the audiophools will rave about the "warm and fuzzy" sound  as they are used to it from their single ended tube amplifiers - using a (highly linear) light bulb as a pullup resistor 
•   If it's only spurs, then it will sound rather disturbing and folks who wouldn't have taken notice in one of the two previous cases might start to complain all of a sudden.

So even the end user would prefer separate specifications for THD, S/N and spurs.

For the designer of such gear it's of course all the more important to know where the unwanted signals come from. Needless to say that the investigation as well as the final countermeasures look very different for the three different types of unwanted signals mentioned before.

[rf-loop]

THD is Total Harmonic Distortion. It is just it, nothing else. It do not include any other things but harmonics (naturally there is never absolutely ideal meters).
When we measure other things, example harmonics and non harmonics and noise this is not anymore THD. It is THD+N where N include everything (if not more specified)  so why we even talk then about THD (if not just for make things messy), it is more like TD (total distortion   ).

Problem is that if we have measured THD+N and result is example 0.1 %  and then we tell "THD" is 0.1%. It is then just basic lie.
Also there is other things. One important is band width we have used for measurement or example what all harmonics we have included to measurement. Also some other things what make this whole "THD" thing quite messy - if we do not use right name and include other parameters. Some may example use also some weighting filter what may affect quite lot.
Many times peoples talk just about THD. I will repeat this. After someone tell to you that some equipment, example audio amplifier or signal generator have THD x%.
If not told anything more than "THD", then many times may need ask this question: "What  THD or is it something else?"

[Performa01]

THD is Total Harmonic Distortion. It is just it, nothing else. It do not include any other things but harmonics (naturally there is never absolutely ideal meters).
When we measure other things, example harmonics and non harmonics and noise this is not anymore THD. It is THD+N where N include everything (if not more specified)  so why we even talk then about THD (if not just for make things messy), it is more like TD (total distortion   ).

I like this term (TD)! Very appropriate!

Also there is other things. One important is band width we have used for measurement or example what all harmonics we have included to measurement.

Yes. If a more sophisticated spectrum analyzer calculates THD for us, it takes he number of the highest harmonic as a parameter. The traditional distortion analyzers usually have a fixed bandwidth, around 80 kHz as far as I remember. But with the many alternative solutions today, like soundcards plus software, we get rather dubious conditions. A soundcard that can provide 192 kSa/s could maintain an 80 kHz analysis bandwidth, but we don't know what analog (AA) filters are there in place to reduce the bandwidth to something more suitable for audio. Ultimately, there are many soundcard solutions where the attempt of measuring THD at 20 kHz would just be a joke because of insufficient bandwidth. Yet serious designers of audio gear want to know the distortion at 20 kHz, because they have to specify THD over the whole audio range – and at 20 kHz distortion figures usually don't look that good anymore (if measured correctly)…

Also some other things what make this whole "THD" thing quite messy - if we do not use right name and include other parameters. Some may example use also some weighting filter what may affect quite lot.

I hate weighting filters in general. No wonder, they are supposed to make measurements representative for the user experience, but are confusing and even misleading for the designer. They'll just get abused by the manufacturers of audio gear in order to make the banner specs look good.

Back in the mid-seventies of last century I was looking for a decent tape recorder. There have been many brands back then and I compared the specs as I wanted the highest possible S/N performance. Of course I had a favorite as well, and this machine specified about 55 dB S/N at 19 cm/s. Of course this was a weighted value after some standard of the DIN 45500 family.

One year later, in the newest catalog, they suddenly specified 62 dB under the same conditions. Only in the fine print you could find that it was now measured after IEC-A or something –I don't recall exactly anymore. In any case, I was a young teenager back then, overlooked the fine print and thought "wow, they have improved their machine by quite a bit!" – and you guess it, the new numbers went into the comparison table and made this machine the winner.

[Martin72]

Thankyou guys, interesting stuff as always..
Seen through the test field glasses, the matter would be relatively simple:
Target : Max. 0.15% THD
Actual : approx. 0.04% THD(+N)
Approx. 0.02% THD only
Test passed...

But it is of course very great to look behind the scenes at what is behind the measurements.
But the thing with the 0.15% is still suspect.
Sure, rather too much than too little - But that the measured values are so drastically below....

[Martin72]

Just for curiosity I´ve measured our rigol DG1022Z with the Lecroy HDO6034A.
Using it´s spectrum option.
THD according to the spec 0.075% (seems to be a magic number... ) 10Hz...20Khz.
What I really like on the HDO is the performance, it´s soo fast in everything...
Nevertheless, pics below.

[markone]

-snip
What I really like on the HDO is the performance, it´s soo fast in everything...
Nevertheless, pics below.

Just out of curiosity, how much money did your company spend for this piece of jewelry ?

[Martin72]

If I remember it right, a little bit more than 13000€.
Including the poweranalyzer option, spectrum analyzer option and severeal serial decoding options.
It makes always sense to buy by lecroy directly, together with the waverunner 9054 we got 42% discount and payed appx 26k for both together.
And the performance in general is really good by both.
They´re pc-based systems (Intel core i5 with 4x3.2Ghz, 16GB RAM, windows10), multigrid display, etc., tons of upgrades avaible (SSD, more RAM, more memory, triggers, math, decoders and so on).
And the UI is superb, having the both everyday as "references", siglent did the job with the 2000x+ and HD very well.
The HDO is my personal favorite at work.

[markone]

If I remember it right, a little bit more than 13000€.
Including the poweranalyzer option, spectrum analyzer option and severeal serial decoding options.
It makes always sense to buy by lecroy directly, together with the waverunner 9054 we got 42% discount and payed appx 26k for both together.
And the performance in general is really good by both.
They´re pc-based systems (Intel core i5 with 4x3.2Ghz, 16GB RAM, windows10), multigrid display, etc., tons of upgrades avaible (SSD, more RAM, more memory, triggers, math, decoders and so on).
And the UI is superb, having the both everyday as "references", siglent did the job with the 2000x+ and HD very well.
The HDO is my personal favorite at work.

I remember a smooth behavior also with an old 64Xs around 2008, it has been my favorite for 4 years (together a CP150 and ADP305) until i changed job, now it's time that i do not see/touch one and judging how things are getting here I think it will be unlikely that will happen again ...

[Martin72]

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.

Answer:

The A1 does not work with the THD+N function via a frequency analysis (FFT) or the impulse response like a modern analyzer, rather it works with a notch filter to mask out the fundamental frequency. Ultimately, the measurement results with and without the fundamental frequency are then offset against each other.
The procedure with the notch filter does not allow to look specifically between the harmonics, therefore no measurement 'without N' is possible.

The number of detected harmonics depends on the generator frequency and the bandwidth of the analyzer. The analysis range for THD+N is 130 kHz for the A1, but only generator signals up to 40 kHz are possible. With a generator signal of 1 kHz, you therefore have the harmonics up to k130 in the result.

[rf-loop]

My bad – In my initial comments I simply forgot that the linearity of the average scope frontend (including SDS2000X HD) is not up to the task of characterizing the harmonic distortion of an SDG1000X or 2000X.

One of the findings one could derive from this would be of a fatalistic nature, namely that one can forget the FFT function with oscilloscopes.
Is this really the case, or do you have to try a little harder to get a more credible result?
And if so, in which direction one would have to go for it....

If you look up the datasheet of PGAs that can be used for a scope like this, such as LMH6518, you will find the harmonic distortions specified somewhere between -44 and -50 dBc. Considering this, the SDS2000X HD still does a fairly decent job.

The FFT is still useful for many tasks, including distortion measurements within the usual range of interest, which hardly ever exceeds -60 dBc except for high end audio. There is a reason, why specialized audio analyzers exist.

You just have to take into account that the SFDR of a general purpose scope frontend might be only about 60 dB in practice – many (if not most) of the more affordable old SA boat anchors haven't been much better btw.

For instance, there is no restriction for single tone narrowband measurements, like phase noise or modulation spectra, where especially the SDS2000X HD will give you an exceptional dynamic range that clearly exceeds the ~72 dB that you could expect from the 12 bits.

The THD of a general purpose oscilloscope frontend is usually not specified in the datasheet. On the other hand, the Picoscope 4262 has a guaranteed linearity of 16 bits – but that is certainly not a general purpose instrument, with its 16 bit converters and limited bandwidth of just 5 MHz. Yet this is the way to go if you need to measure down to -96 dBc and don't want to be restricted to the audio frequency range like with soundcards.

Referring to the previous one.

It is really good to keep in mind the mixing results generated by the oscilloscope itself, especially when working with FFT settings where the processing amplify is high and we are looking at weak signal levels, such as, for example, rather low harmonics and nonharmonics spurs of the signal source being examined.

Here is just small part of observations.
In the test, I looked at the dual tone signal.

Throughout the test, the signal from the signal generator was kept completely unchanged. . The only variable was the CH1 V/div setting on the oscilloscope. I did a test for all possible values between 40mV/div to 100mV/div. Both signals had a level of about -12dBm. Signals freq. 101.3kHz 158mVpp and 103.3kHz 158mVpp.  At this level, dual tone peak-peak total is close to full scale (at 40mV/div).
I have full set of results but here just few of them. (there is really lot of changes when go through all possible fine steps, just one fine step and change may be over 10dB.. 15dB in 3rd order IM products)


img 71 just 100mV/div and img 21 just 50mV/div.

In images 17 and 35 there is perhaps most high and most low 3th order im levels (64dBc and 84dBc, 20dB change is quite big)

image 40 some kind of one just random medium example.

Then
Last image without number is just how this dual tone looks in time domain


ETA: Added image. With spectrum analyzer exactly same signal what was used with SDS2504X HD images.
This somehow proof that dual tone from generator is least not worse than in this SA image. (for avoid SA own generated distortion Attenuator was set for 23dB so that mixer level stay enough low for keep SA's 3rd order IMD as low as possible. SA's performance is not enough for analyze SDG 1000X in this matter.

[blackdog]

Hi rf-loop

Your measurements are a good example of what I often say on forums, know your instruments!

In one of my posts I mentioned that I used 2V RMS as the output signal and then terminated with a 50 Ohm termination resistor.
This is because with most generators, THD is fairly favorable and then you can then attenuate externally as needed.
This level may of course just not good for the particular generator used, therefore know your instrument!
I have tested it on five of my function generators where possible at the 2V RMS.

And then you get the second step that you will have to perform as a user, this as you did with your last measurements, which area of the input attenuator from here the osciloscope is the most linear.

There are different parts in the osciloscope where the gain can be controlled, these include these, relay input attenuator, " Voltage Control Amplifier" and digital.
So then you can also have a fairly large difference between the steps when you look at the THD and as in your last measurements at the third order products.

It's quite complex to take all those variables into account....
Technicians who do a lot of measurements with a Spectrum Analyser are used to using e.g. a passive attenuator to attenuate their generator signal in steps to see if they do not overload the input of their Spectrum Analyser.

I think most users of modern oscilosscopes are not always used to that when doing FFT measurements.
I think it's good to bring this to people's attention here.

Thanks for the nice measurements.

Kind regards,
Bram

[pdenisowski]

Technicians who do a lot of measurements with a Spectrum Analyser are used to using e.g. a passive attenuator to attenuate their generator signal in steps to see if they do not overload the input of their Spectrum Analyser.

I think most users of modern oscilosscopes are not always used to that when doing FFT measurements.

Yes.  I worked as a field applications engineer in test and measurement for > 20 years and this was my experience as well.

It bears repeating that whenever making any type of "distortion" measurement (IP3, THD, etc.) it's very important to ensure that the measured distortion is not being created within the measuring instrument itself.

[Performa01]

I have full set of results but here just few of them. (there is really lot of changes when go through all possible fine steps, just one fine step and change may be over 10dB.. 15dB in 3rd order IM products)

Yes, that was my observation as well. Sometimes I have cheated a little and taken advantage of this when presenting test results in the following thread - especially when I was talking about "sweet spots"

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

[mawyatt]

Since the DSO generally isn't designed as a Spectrum Analyzer, think what rf-loop has shown with the Linearity dependance on actual Input Signal Chain Gain/Attenuation Distribution is an important point in our DSO usage as a SA (FFT). These DSOs are really Data Acquisition Systems, and a SA is just input signal data processing, same as with a 'Scope usage. However, a Scope doesn't really need much Display DR or Linearity beyond 1% (~40dB) because we can't really "see" the effects on the display since this is a linear display, however when utilized as a SA in FFT Mode the display is now in dB and thus much higher on screen overall signal amplitude dynamic range, and subtle signal artifacts begin to show as do measurement instrument non-linear channel effects.

The quality modern DSO must possess a higher Signal Channel DR for use with the FFT Mode, and this imposes a higher initial design and build cost.

Overall Signal Chain "Sweet Spot" is an interesting topic. There was a unique class of amplifiers we employed long ago that had very interesting properties, one of which was the linearity "Sweet Spot" with a theoretical infinite dynamic range. We utilized very fast SiGe 400GHz Ft devices and later planned on >600GHz Ft InP devices for implementing special purpose custom ASICs. One of these had another unique property with gain expansion rather than compression with increasing input signal levels, all of these prior efforts were to try and "squeeze" as much DR from a given supply rail as possible and thus were operated mostly in current domain. So we were trying to place the Input signal within the "Sweet Spot" of the Signal Channel Dynamics to allow further signal processing, similar to what has been shown by changing the DSO Gain/Attenuation Distribution.

Best,

[elecdonia]

If you have a harmonic distortion analyzer (or tuneable notch filter), then it is incredibly useful to use the oscilloscope to view the output signal from the THD analyzer, which contains the distortion and noise components alone, without the fundamental. Today many digital 'scopes can perform FFT which will clearly display the amplitude of each harmonic. The first time I saw an FFT of THD analyzer output (in 1978!) I thought "this explains why some amplifiers sound better than others". But at that time FFT analyzers were strictly unaffordable for ordinary people like me. Times have changed!

In terms of perceived audio quality: The best sounding amplifiers have mostly 3rd harmonic (3x fundamental frequency), with equal or lower 2nd harmonic. Higher order harmonics should drop off rapidly. Ideally the levels of 4th, 5th, and 6th order harmonics are -60dB (or lower) compared to the fundamental. The best amplifier circuits have unmeasurably low levels of 7th and higher order harmonics, at least -80dB below the fundamental. There should never be any high order harmonics with levels above -20dB compared to amplitude of 2nd or 3rd harmonic.

Another "THD analyzer with oscilloscope" view which I have used for many years is to put the oscilloscope into X-Y mode.
This is extremely useful even when using only a modest quality THD analyzer (Examples: HP 331A, 332A, 333A, 334A, Heathkit IM-5258).
Apply the output signal from the amplifier under test to X axis.
Apply output signal from THD analyzer to Y axis.

Ideally the X-Y display will be a flat line until the amplifier under test begins to clip.
The onset of clipping will immediately raise large vertical components at the extreme left and right edges of the X-Y display.
More important, "crossover notch distortion" will appear as a single vertical line at the middle (0V) of the X axis. Crossover notch distortion is extremely audible, even in tiny amounts. Many early transistor amplifiers had plenty of crossover notch distortion. Especially Crown and Phase Linear. But improved solid-state circuit designs largely eliminated this issue after 1980. In contrast, tube (valve) amplifiers rarely have significant crossover notch distortion. In my opinion this is a big reason why people have been saying (for many years) that tube amps sound better than solid-state amps.

Rough diagram of X-Y displays for different conditions:

No significant distortion:               ---------------------

Clipping:                                     |--------------------|

Crossover notch:                          ----------|----------

I'll post images from actual measurements if anyone here is interested.

[elecdonia]

Why does this topic have two names?
     1) Re: Measuring Distortions with the Scope:What you see is not what you really have
     2) Re: Comparison between Siglent SDG1000X and 2000X

[tautech]

Why does this topic have two names?
     1) Re: Measuring Distortions with the Scope:What you see is not what you really have
     2) Re: Comparison between Siglent SDG1000X and 2000X

See reply #63, the OP chose to change it to better reflect the content.

[elecdonia]

Several posts in this topic display the FFT harmonics from the output signal of a "function generator." Typically the ratio of fundamental/harmonics shown is no better than -50dB for function generators. This isn't suitable for testing audio gear. It corresponds to THD of .5% to 1%. This is far higher than the THD of any decent audio amplifier circuit. In order to make useful measurements of distortion, the "residual" distortion of the signal source (oscillator or generator) should be at least 10 times lower than the expected THD of the amplifier under test.

Function generators always produce plenty of harmonics, especially higher order harmonics. The reason for this is that function generators internally start out by generating either a ramp or triangle wave shape. This is then filtered and/or processed with non-linear diode clipping circuits to "approximate" a sine wave. The best function generators have harmonics of -40dB to -60dB below the level of their fundamental. Function generators also produce extensive high-order harmonics up to and above 10th order. Adding up all these harmonics corresponds to about 1% THD, with the very best function generators approaching .25% THD.

In contrast, "sine wave audio oscillator" circuits actually generate a sine wave as their starting point. Some generators may offer a square-wave output also, but that is produced by running the pure sine signal through a comparator. The best sine-wave oscillators have THD as low as .0001% (after background noise is filtered out).
Pure digital generation of sine waves by high-grade 24-bit D-A converters can also achieve very low THD.

I recommend that future FFT data shown in this topic should use an "ultra-low distortion" sine wave oscillator (or 24-bit D-A) for the signal source.

[elecdonia]

Theoretically a very high-quality ultra-low-distortion sine wave oscillator could be phase-locked to the output signal from a function generator. But this isn't how most function generators are made.

Function generator:
     Available output signals: Sine, square, ramp, triangle, pulses of controllable width
     Output frequency range can be as large as .01Hz to 10MHz
     Output signal amplitude is tightly defined (changing the frequency doesn't alter output amplitude)
     Distortion for sine-wave output: .25% to 2% THD
     Some models offer voltage-control of output frequency

Low-distortion sine wave oscillator:
     Available output signals: Sine (and optionally square)
     Typical output frequency range is 20Hz to 100kHz or sometimes 1MHz
     Output signal amplitude may vary by +/- 1-3dB as frequency changes
     Distortion for sine-wave output: .0001% to .1% except older tube (valve) sine wave oscillators may have distortion spec of .5-1% THD
     Some sine wave oscillator circuits provide only a single fixed output frequency (typically 1kHz), or a limited number of fixed output frequencies

[mawyatt]

Think the context here where "Function Generator" is mentioned more than likely infers a AWG, which can have a respectable THD+N metric as shown by a few.

Best,

[Performa01]

Several posts in this topic display the FFT harmonics from the output signal of a "function generator." Typically the ratio of fundamental/harmonics shown is no better than -50dB for function generators. ...

Thank you so much for giving us valuable insight in the obscure field of audio engineering.  We learn something new every day!

In case you did not understand Tautechs response: This thread originally was (and somehow still is) about the distortion in AWGs (Arbitrary Waveform Generators), which have been shown to provide decent distortion levels down to -80 dBc.

In case you did not read the posts within this thread: unsurprisingly, it has been found that the linearity of the frontend of an average general purpose DSO cannot keep up with the modern AWGs, hence we get much worse distortion figures than what the generators are actually capable of.

Hint: the latter fact was the reason why the OP decided to change the thread title.

[Martin72]

Hint: the latter fact was the reason why the OP decided to change the thread title.

Exactly.
It actually started "harmlessly", with an FFT comparison between SDG1000X and 2000X.
rf-loop noticed (in a neighboring thread) that something could not be right and so it began...
What I noticed (later), the FFT of the two looked pretty much the same, which possibly already indicates an "error".
I find that very interesting what happens there and must read me again the pages here to create perhaps a summary of why you can not trust your eyes (apart from that you know the actual values according to the data sheet).
Also, we have not yet illuminated whether it gets "better" when you feed in a signal that has a not so low THD.
I had briefly addressed this at work or discussed with a senior developer and he said flatly, that comes from when you measure "digital" with "digital"....
Oh well, the display of my Neutrik A1 has an illumination, it is only totally weak:
EL foil, over 20 years old...
Replacement bought, should come next week and then I need no more flashlight.

[elecdonia]

Several posts in this topic display the FFT harmonics from the output signal of a "function generator." Typically the ratio of fundamental/harmonics shown is no better than -50dB for function generators. ...

In case you did not understand Tautechs response: This thread originally was (and somehow still is) about the distortion in AWGs (Arbitrary Waveform Generators), which have been shown to provide decent distortion levels down to -80 dBc.

Somehow I missed the word "arbitrary" when I first read this topic. I must re-read all 5 pages more carefully.
"Arbitrary waveform" implies a digital system with D-A converter at the output. Today 16 and 24-bit DACs with excellent linearity and very low noise are widely available and cheap. I'm not surprised that -80dB distortion ( .01% THD ) is feasible for an AWG. Perhaps I'm showing my (advanced) age because my concept of "function generator" applies to analog devices limited to producing sine, square, ramp, & triangle waveforms?
AWG systems are a whole different animal.

In case you did not read the posts within this thread: unsurprisingly, it has been found that the linearity of the front end of an average general purpose DSO cannot keep up with the modern AWGs, hence we get much worse distortion figures than what the generators are actually capable of.

I fully agree. This is one reason why it is so useful to connect the oscilloscope to the output signal from a THD analyzer (or notch filter) so that the oscilloscope isn't required to process the fundamental.
In any case one must first measure the signal source (AWG) with the analysis device (DSO) to determine the baseline distortion/noise level before putting a device-under-test into the signal chain.

[Performa01]

Now we are on the same page!