Discrete Opamp Open Design

[coluke]

What about this dirty trick (see 1st attachment) to trim AC symmetry and lower output THD _without_ affecting DC balance? Without it THD is about 0.00065% from 100 Hz up to 20 kHz @ 0 dBV into 30 ohm load (almost only 2nd, measured with a subtractive method); with the shown trick, provided the trimmer is a good quality 10-turns one, you can lower 2nd harmonic down to an impressive -130 dBV, while 3rd stably sits at about -125 dBV (@ 1 kHz, 0 dBV into 30 ohm), which means something like < 0.0001% THD (rising at ~ 0.0002-0.0003% @ 20 kHz) - it seems THD doesn't vary at all with time (on a couple of hours scale, at least). I must admit that lowering further and already quite low THD figure could be a mere technical exercise, nevertheless getting down to < 0.0001% @ 30 mW into 30 ohm using 10 devices isn't actually an everyday experience

L.

[coluke]

Originally Posted by bcarso

I may start a new thread for in-depth explorations of linuxguru's approach.

Thumbs up! I'm really interested

L.

[Joachim Gerhard]

This is the idea.

[bcarso]

Originally Posted by coluke

Thumbs up! I'm really interested

L.

I also just got word (thanks D!) that there is some prior art closely resembling the approach, if not subsuming it. I am looking at the claims soon --- not that any of us in here should be deterred from diy applications

From the descriptions it would appear as if the basic notion of the current sensing for compensation is part of a couple of patents.

[fas42]

Originally Posted by john curl

I don't think so, fas42. Reducing feedback also reduces potential distortions that are MORE exotic than we can easily measure.

Sorry to go a bit OT again, but just to extend: I agree that there are distortion mechanisms at work creating audible problems, that can certainly be classified as "exotic", as in that they are imposing themselves from outside of the world of the direct and generally accepted parasitic behaviours of the circuit parts. And these other mechanisms either interfere with the "textbook" functioning of the feedback, or lie beyond the area of the circuit that the FB looks at.

So, IMO, playing with the level of FB is not addressing the real problems of getting correct sound, I'm with Bruno in saying there ain't no such animal as too much, correctly functioning, feedback. Firstly, make sure that the area of the audio system that has designed in FB as part of its structure is working as "perfectly" as it can, and secondly address the problems that lie elsewhere ...

There's a simple issue at the heart of audio: test after test says that people begin to lose the ability to hear "distortion" at about 60dB down, at -80 it's completely invisible; and I agree with these findings -- yet people thrash around, fighting to get, and achieving -120dB performances, in parts and circuits, which other people then dismiss as having obvious, audible characteristics. So what's going on ... ??

Frank

[abraxalito]

Originally Posted by fas42

There's a simple issue at the heart of audio: test after test says that people begin to lose the ability to hear "distortion" at about 60dB down, at -80 it's completely invisible; and I agree with these findings -- yet people thrash around, fighting to get, and achieving -120dB performances, in parts and circuits, which other people then dismiss as having obvious, audible characteristics. So what's going on ... ??

One of the things I think is going on is there's almost an obsession with single tone THD measurements which tends to exclude other more promising avenues of measurement. Where are the multitone IMD measurements which Audio Precision made possible decades ago?

The reason single-tone testing doesn't cut it is its crest factor is very unlike music. Music (not including the modern highly compressed genres as 'music' here ) spends most of its time below -20dBFS whereas a full-scale sinewave spends half its time above -6dBFS. Hence traditional THD doesn't weight a circuit's linearity where it matters most - around the zero crossing.

[SY]

I think multitone IMD might go back a wee bit further than Audio Precision...

[abraxalito]

Of course it does - my point was that the AP made it more accessible to the average punter. Just as computers went back prior to the founding of Microsoft yet it was Bill Gates who made them widely accessible.

[Kindhornman]

abraxalito,
So what are you proposing as a new test? Do we chose a three tone sample with 20hz, 1khz and 20khz and test with burst waveforms simultaneously, or do a longer time frame test reaching the upper power limits of the device in question? In speaker testing we use log chirp testing to see what is going on and perhaps something like that would work for an amplifier at the correct levels to work the NFB circuits to see what is happening? Give us an idea what would be a better test than single frequency testing methods.

[RNMarsh]

The harmonics of a single tone produces some 'grass' on the FFT plot. Say .01% thd worth; Can't be heard. More simultanious tones, more grass growning and IM grass, too. After 20,000 tones have been applied, that .01% THD of each of thousands of tones adds up to enough to reach audible levels. The audible effect is an equivalent increase in the background noise level. Which is described as masking details or changing character of the sound depending on music freq content played. This might be why we need Really low thd numbers on a single tone test.... to prevent the accumulated harmonics of many simultaneous tones from reaching an audible threshold. -Thx RNM