why don't we use predistortion more in the audio world?

[nigel pearson]

I can't help but think that frequency-dependent EQ is a waste of time and is bound to sound unnatural no matter how the settings are derived. The alternative is to go with an impulse response correction that aims to give you the headphone sound at your listening position - but we know that doesn't work if you move your head by a millimetre. Seems to me that the best compromise will be what most people say: cut down the reflections with acoustic treatments and only apply time- and frequency-diminishing impulse response correction at the lowest frequencies.[/QUOTE]


As John Deans a friend says you can't EQ time domain problems . However it helps to try .

[CopperTop]

Originally Posted by nigel pearson

As John Deans a friend says you can't EQ time domain problems.

And I suppose it depends what you think of as a problem. A bit of 'live-ness' in a listening room does not sound wrong, even if the steady state frequency response looks truly terrible. Listening to stereo in an anechoic chamber might be theoretically correct, but I bet it would feel horrible.

Anyway, although EQ is still pre-distortion by certain definitions, it's off topic for this thread! (But I feel better now I've got it off my chest).

[nigel pearson]

Oxford town hall is where I did many recordings . I could get a sound infinitely better than the actual sound . It took me years to ignore the sound of the building . One day I noticed some unoccupied more expensive seats where I knew it would be better . I have to say the sound difference when forced to take my proper seat was a more dramatic change than a new set of speakers . I fell out with the safety man at the town hall who considered my microphones to be a safety risk ( idiot ) . I decided to call it a day . On returning some years later I found I could still process the sound in my head and enjoy it . That surprised me .

Our local Littlewoods store used a string quartet to celebrate the enlargement of the store . They were playing in the section that sells coats . It was completely anechoic . Hayden sounded dreadful . I asked if they had any Ravel . They were very snooty and said only a piece of 20 bars . It was total magic . It seems Ravel somehow built echoic qualities into his music . It is better played by slightly detached professionals as it needs no schmaltz . They were perfection .

[nigel pearson]

Pre-Distortion Techniques

[DF96]

That predistortion will have the same effect as NFB, in that it will sharpen the eventual clipping - there is no free lunch! However, it is a good example of applying predistortion where NFB cannot be applied (at least, not easily). In that case NFB would require somehow detecting the response of the tape head or even the tape. The same issue does not arise with amplifiers, although it could perhaps arise with loudspeakers.

[nigel pearson]

How about feed-forward error correction ? That seems more of a free lunch . As someone said , no free lunches , cheap lunch ? Perhaps . Bootstrapping seems the best cheap lunch .

[jcx]

Black's 1st pass idea was feedforward error correction, famous audio implementation by Quad "Current Dumping" Amp

Originally Posted by jcx

the "non-interacting" power output combiner is a difficulty

the L,R output combiner could be seen as a XO, the low frequency branch thru the output series L is driven by the main power amp

the fast correction amp path to the output through the resistor would be a 1st order highpass

but since the correction needs to add up to unity at the output you need a 1st order roll off for the correction amp's signal - hence the integrator

people like to draw the Quad Dumper circuit as a AC 4-arm bridge, the parallel circuit shown has a similar correction amp signal frequency shaping


I'm not yet really sure of the utility Black's Feedforward Error Correction scheme for audio power amps - except as a "band-aid" for being stuck with a poor performing main power amp
since you still have to accurately measure the error it seems like negative feedback around a better power amp would just be better
In the example of the BobAmp circuit I think doubling up ouput Q to avoid Beta droop and using 5x faster RET would let you get similar distortion improvement with 2-pole compensation

the Vanderkooy paper did suggest Black's Feedforward Error Correction as a way to help out Class D main power stages

[CopperTop]

Originally Posted by DF96

there is no free lunch!

If you took a conventional amplifier with NFB and 0.005% THD or whatever, are you saying that there is nothing you could do to 'null out' the remaining THD (and IMD of course) with just a smidgeon of the right pre-distortion?

[jcx]

practically I'd expect the distortion residual of a high feedback amp to be highly variable with operating point, thermal history – and very, very small

which would make it hard to accurately model, predict to maintain the cancellation with a predistoriton approach

to predict the Q die temp in the amp is a multi-physics problem = thermal radiation, convection, conduction
how much instrumentation are you going to afford - air, heatsink, transistor tab temperature sensors?, output current, power supply V and Re deltaV to get electrical power dropped in the output Q? - what about the driver Q now?

To be motivated to try predistortion in audio power amps over really good implementations of high negative feedback in audio amplifiers I'd like to see better evidence of “the problem” with negative feedback

sighted, uncontrolled listening impressions by the crew that failed the Carver Challenge aren't doing it for me


with their own source, speakers, in their own listening room Stereophile's "Golden Eared" reviewers couldn't tell Bob Carver's $600 SS amp from their own choice of "SOTA" tube amp after Bob tweaked the SS amp frequency response, output impedance for a deep null with the tube amp

[CopperTop]

Originally Posted by jcx

practically I'd expect the distortion residual of a high feedback amp to be highly variable with operating point, thermal history – and very, very small

which would make it hard to accurately model, predict to maintain the cancellation with a predistoriton approachto predict the Q die temp in the amp is a multi-physics problem = thermal radiation, convection, conduction
how much instrumentation are you going to afford - air, heatsink, transistor tab temperature sensors?, output current, power supply V and Re deltaV to get electrical power dropped in the output Q? - what about the driver Q now?

If I was minded to try this experiment, I would try to maintain the amp's temperature at a fixed value in order to reduce that dimension at least, but I could feed the output of various sensors into the processor, too. The question would be mainly one of repeatability: I imagine I could learn to pre-distort optimally using previous and future samples as inputs to my neural-style network, but would it work the same two weeks later?

how much instrumentation are you going to afford

I don't think you are really into this audiophile thing! A few temperature sensors, extra heating transistors, even air conditioning for the whole room are as nothing compared to what these people spend on a mains cable. I want to do this for the audiophiles..!