Distortion Cancellation Experiment

[RayDunzl]

Using REW, send a 450Hz tone at -20dB to the system (JBL LSR 308 speakers), take an RTA with UMIK-1.

Note high level of 3rd harmonic distortion in the speaker output.

It is clearly audible.

Using the Distortion controls, add third harmonic to the signal at -40 degree phase angle.

 

[amirm]

Interesting. What did it sound like after?

 

[Blumlein 88]

Your second example had a lot more low frequency noise. Was that disturbing results? I see of course noise at the frequency of interest was about the same.

 

[RayDunzl]

Interesting. What did it sound like after?

The 450Hz tone with third harmonic sounds like the tone with a hint of another tone (the third, at 1350Hz).

The tone with the added anti-third-harmonic sounds like the 450Hz tone.

In this case, it sounds like it looks.

---

While reading up on the Benchmark DAC3, it mentions its own internal harmonic distortion cancelator (a feature of the ESS chip), so I thought I'd investigate a little.

DAC3 vs. DAC2 The DAC3 builds upon Benchmark’s highly successful DAC2 product family. The DAC3 maintains the familiar DAC2 form factor, but adds the higher performance available from the new ES9028PRO D/A converter. The DAC3 offers the following improvements over the DAC2:

Active 2nd Harmonic Compensation
Active 3rd Harmonic Compensation
Lower THD  Lower passband ripple
Improved frequency response
Increased Dynamic Range
Faster PLL lock times
Faster switching between inputs​

... Harmonic Compensation The ES9028PRO has two distortion compensation systems that independently remove most of the 2nd and 3rd harmonic distortion in the D/A converter. Benchmark's ultra-clean analog output stages allow these systems to be fully leveraged in the DAC3.

 

[RayDunzl]

Your second example had a lot more low frequency noise. Was that disturbing results? I see of course noise at the frequency of interest was about the same.

Disturbing the results - no... The area above 450Hz isn't much different.

I took the measurements moments apart, played around with some different frequencies, then sometime later, I noticed I didn't like the capture of one, and repeated it.

18Hz is related to the HVAC, other low low stuff I can't identify, but have noted possible Interstate Truck noise (2.67 miles away), it's summer so remnants of distant Thunder rumbles, subsonics of helicopter rotors, MacDill AFB (12 miles, 4 of it over water), and of course, the ever present ???.

 

[Cosmik]

While reading up on the Benchmark DAC3, it mentions its own internal harmonic distortion cancelator (a feature of the ESS chip), so I thought I'd investigate a little.

DAC3 vs. DAC2 The DAC3 builds upon Benchmark’s highly successful DAC2 product family. The DAC3 maintains the familiar DAC2 form factor, but adds the higher performance available from the new ES9028PRO D/A converter. The DAC3 offers the following improvements over the DAC2:

Active 2nd Harmonic Compensation
Active 3rd Harmonic Compensation
Lower THD  Lower passband ripple
Improved frequency response
Increased Dynamic Range
Faster PLL lock times
Faster switching between inputs​

... Harmonic Compensation The ES9028PRO has two distortion compensation systems that independently remove most of the 2nd and 3rd harmonic distortion in the D/A converter. Benchmark's ultra-clean analog output stages allow these systems to be fully leveraged in the DAC3.

It all seems a bit odd. If a circuit has a bent transfer function (e.g. instead of a straight, sloping line it is a bit of a 'S' shape), this will give you harmonic distortion when you feed in a sine wave and intermodulation distortion when you feed in something more complex. The relative level will vary depending on amplitude e.g. if low enough, the sine wave might be entirely in the linear region but begins to distort as it gets louder.

Going into the frequency domain is a very sensitive way to detect the distortion, but does not directly tell you what mechanism is causing it. By testing with numerous signals at various amplitudes it might be possible to deduce what the distortion characteristic is.

The way to fix a bent transfer function would be to pre-distort the signal with the mirror image of the curve (e.g. 'S' shape) - no need to go into the frequency domain at all. Considering the circuit to be literally a harmonic generator on the basis of tests using a few sine waves and attempting to fix it on that basis might not be meaningful.

With a DAC, what might cause it to generate harmonics? Well, in the case of Sch**t, one of their DACs cobbled together from based on non-audio DAC chips reputedly has glitches at the zero crossing. It might be possible to paper over the glitch somehow, but is the glitch stable with temperature?

 

[RayDunzl]

Noise, 3:40am

AC running



AC off



Maybe some fan noise spikes, still working on repacement PC, haven't replaced the noise shield behind the couch.

 

[RayDunzl]

It all seems a bit odd.

I'm not qualified to argue anything else you said.

ESS 9028 blurb sheet - http://www.esstech.com/files/9814/5193/1789/ES9028PRO__ES9026PRO_Product_Brief_121615.pdf

"Programmable THD compensation to minimize THD caused by external components"

 

[Cosmik]

I'm not qualified to argue anything else you said.

ESS 9028 blurb sheet - http://www.esstech.com/files/9814/5193/1789/ES9028PRO__ES9026PRO_Product_Brief_121615.pdf

"Programmable THD compensation to minimize THD caused by external components"

Oh, I see. So it's not to compensate for the DAC's own shortcomings, but allows you to compensate for external distortion mechanisms. I bet it will effectively be a transfer function bending thing. Use with care, however: do you know that your external distortion is stable and repeatable?

 

[Cosmik]

Ray, your experiment is fascinating, by the way. It shows that if you pre-distort the waveform, you can fix distortion - something many of us may have thought about but never actually done. The question is, however, whether the same fix works at different amplitudes and with multiple sine waves, etc. Distortion in speakers is also unlikely to be purely a 'bent transfer function', unless that transfer function can be envisaged in multiple dimensions. A comprehensive fix for speaker distortion would probably be mapped in a complex way to the signal's content and its history (for a few hundred milliseconds, say?), so a fix that 'worked' for all steady state sine waves wouldn't necessarily work for complex music - but if we only tested using steady state sine waves we would never know... I presume the distortion would also vary with ageing, temperature and, for all I know, humidity.

 

[RayDunzl]

The question is, however, whether the same fix works at different amplitudes and with multiple sine waves, etc.

In my simple case, no. The amplitude and (apparently) the phase of the speaker distortion in-room varies with frequency and amplitude.

The 450Hz example worked well, so I posted it.

Correcting an electronic circuit might be a little more straightforward than correcting an electromechanical device, though.

Either would likely be an interesting coding challenge for a DSP filter.

 

[Brad]

Getting the correct Q (or filter steepness) for the distortion correction would be difficult - would need to properly characterise the distortion and its response with amplitude as it would be nonlinear.
On a related note, Uli on the acourate forum recently did a simple demonstration how adding 2nd order distortion (eg simulating tube amp) causes more than 2nd order distortion, as it also adds intermodulation products

 

[March Audio]

Pre distortion is a technique used to improve linearity in RF amps

http://www.keysight.com/en/pd-17617...ortion-builder?nid=-34261.919304&cc=AU&lc=eng

http://literature.cdn.keysight.com/litweb/pdf/5990-6428EN.pdf?id=1909191

 

[Don Hills]

... Correcting an electronic circuit might be a little more straightforward than correcting an electromechanical device, though. ...

[cough]Dynagroove[/cough]
From the Dynagroove page on Wikipedia:

... the Dynagroove process used tracing compensation, which pre-distorted the record groove to cancel out the distortion created by playback with a conical-shaped phonograph stylus, which could not track the record groove accurately in the high frequencies, especially in the inner grooves of an LP. ...

 

[fas42]

The other factor is time, and previous usage - unfortunately, distortion is a moving target with just about everything ... I've had some experiences hearing obvious speaker distortion, and seeing what happens when the moving parts are given some "exercise" - literally, one can condition away an obvious anomaly by giving the speakers a workout; may take a half hour, say, but is certainly doable.

The lovely comfortable world of taking some reading, and that being the 'absolute truth', irrespective, is a long way from how the world actually works ...

 

[RayDunzl]

I've had some experiences hearing obvious speaker distortion, and seeing what happens when the moving parts are given some "exercise" - literally, one can condition away an obvious anomaly by giving the speakers a workout; may take a half hour, say, but is certainly doable.

These have been exercised for 9 months, and the exorcism hasn't yet occurred.

The lovely comfortable world of taking some reading, and that being the 'absolute truth', irrespective, is a long way from how the world actually works ...

Who does that?

 

[March Audio]

The other factor is time, and previous usage - unfortunately, distortion is a moving target with just about everything ... I've had some experiences hearing obvious speaker distortion, and seeing what happens when the moving parts are given some "exercise" - literally, one can condition away an obvious anomaly by giving the speakers a workout; may take a half hour, say, but is certainly doable.

The lovely comfortable world of taking some reading, and that being the 'absolute truth', irrespective, is a long way from how the world actually works ...

If the speakers had high enough distortion to be obviously audible then they were faulty. Bin them, you wont fix them with exercise.

Franks comfortable world, being the absolute truth, is a very long way from how the world actually works ....

 

[RayDunzl]

If the speakers had high enough distortion to be obviously audible then they were faulty.

These do distort a little earlier than the MartinLogans on test tones.

Music masks the distortion harmonics with more complex harmonics, I suppose.

Played loudly (they are only 56W) they lose to the bigger system.

Played reasonably, they're my Daily Drivers now for casual music and TV.

So, I don't rate them as faulty, they just have their own Operating Envelope*.

* Limited range of parameters in which operations will result in safe and acceptable equipment performance.

 

[March Audio]

Sorry Ray I was referring to Fas' comments. If something is significantly changing in the distortion with exercise then something is wrong with the speaker suspension/coil.

Regarding your results, yes music is going to mask the harmonics to a large extent. Mind you at 35 dB Im sort of surprised that 3rd is easily audible. You prompted me to have a play and see what I can hear.

For 75 dB fundamental level, still seems a bit high, was the measurement taken at the listening position or close to the drivers? i sort of wonder if something else was "rattling" in the room

 

[RayDunzl]

Sorry Ray I was referring to Fas' comments.

I know.

Mind you at 35 dB Im sort of surprised that 3rd is easily audible.

Well, easily audible when switching between having and not having it.

Fire up a newish version of REW and play with the distortion tool - it's part of the Sine Generator.

As for measurements, my mic lives on the back of the couch. If I move it that would be the exception and definitely be specifically noted. Default, and usually noted, is "listening position".

I'm not building speakers, so, essentially, I'm only interested in what comes to the listener.