John Curl's Blowtorch preamplifier part II

Status
Not open for further replies.
I am current aiming between -125db to -150db in spice. (edit - but I have no idea how this translates in real life )

mike

In case you speak about high order harmonics, it is not that difficult to achieve this, in real life.

Avoid too high global NFB and use distortion cancellation techniques, as John does. You will get under measurement capabilities of current test equipment then.
 
Avoid too high global NFB and use distortion cancellation techniques, as John does.

If high global feedback generates or increases the closed loop high order harmonic components, then the (open loop) amplifier design is flawed. There isn't anything like "too much feedback", it is the low feedback that (in certain circumstances) may create problems.

But then this is engineering 101 and has nothing to do with the audio nirvana that seems to be the main topic here, so I apologize for the intrusion.
 
Pavel,

Thanks for your reply. I guess we have to define "too high Global NFB" to make this understanding clear.

I like the relaxed sound of amps with low FB but I have noticed in spice that if I reduce global FB from 33db to 23db then all HD rises from H2 - H20.

So perhaps this idea of of more high order HD with more FB kicks in with higher levels of FB or perhaps with more complex designs ?

mike
 
The confusion lies in generalization from the particular (in a device with pure second order distortion, higher harmonics are generated). Increase or decrease in higher order harmonics in actual amplifiers (which are decidedly not pure second order) depends on their open loop level and the amount of feedback. High levels of feedback reduce high order distortion to well below the noise level. See Bruno Putzey's excellent article in Linear Audio for a thorough and accessible analysis.

(edit: x-post with Scott)
 
Mikelm, this is where the 'subjective' conflicts with the 'objective'. The general math and the simulations show that the more feedback you use, the better. Heck, I truly believed in lots of negative feedback back in 1969 when I worked in the Ampex Research Department.
I even came up with some exotic ideas that might of worked, IF I could have found a SUPER op amp, at that time. Later, when I tried to make high feedback op amps 'work' in reality, my clients ultimately rejected my efforts and went back to tubes.
I, too, find that my best circuits work with low to no global loop feedback. Of course, I have to be VERY careful about the higher order distortion products that are not going to be 'masked' by feedback, like so many IC op amps are, because of the finite feedback that I have to use. This is WHY I have to measure below -100dB or .001% harmonic distortion, AND I have to make sure that the higher order distortions are virtually not generated by the circuitry I design. That it the challenge.
Unfortunately, even IF I pass THIS test, other factors, not yet well measurable, tend to make an individual design, a GO or NO-GO with listeners, reviewers, and ultimately, sales. I have two examples of my own design that have 'failed' in the marketplace, even though they were all discrete, with lots of jfets, and measure pretty good. I just took one out of storage, this week. However, to test it again, even with my best efforts, will probably not turn anything up that will clue me as to its 'failure' as a design.
Other factors are probable, and their implementation is the reason for my persistence on this thread.
 
Classic feedback theory, that we learned in school, assumes that the amplifier forward path is linear. That's usually a pretty good assumption/ simplification, but in one common case it fails. Crossover distortion really messes with the simplification (which assumes monotonicity).

I wonder if a lot of the modern distrust of long-loop feedback comes from crossover distortion, which is more common in modern stuff than in old school stuff.

Thanks,
Chris
 
Also I think there are more opportunities for localized resonances in high FB & complex designs.

These will have no harmonic relation to the applied signal even though they may be excited by it, but I guess their demodulated derivatives could still show up in a fourier analysis . . . and ruin the sound of the amplifier.
 
Still, there must be SOMETHING that will measure the differences in two line stages (for example) and give me direct evidence of why one sounds better than another. I'm open to serious suggestions.

That was my point John, if a device does not register what is the use of the test one way or the other? In my opinion the Hirata test is just a special case of the multi-tone power ratio test where you can mathematically create any crest factor test signal you want (within physical limits).

We're still left with negative indicators, beloved amplifiers measure big numbers.
 
Also I think there are more opportunities for localized resonances in high FB & complex designs.

These will have no harmonic relation to the applied signal even though they may be excited by it, but I guess their demodulated derivatives could still show up in a fourier analysis . . . and ruin the sound of the amplifier.

You're on dangerous ground here, highly speculative and not based on much physical reality.
 
You're on dangerous ground here, highly speculative and not based on much physical reality.

There are two elements to my argument.

1) higher FB designs more prone to local resonances.

2) HF resonances on audio amplifiers somehow get demodulated and ruin the audible sound of the amp.

I admit that point 1) is intuitive and not proven and I also admit that there are plenty of opportunities for local resonances in low gain amps . . . but I think point 2 can be substantiated with simple logic.

If we imagine an audio circuit with a mosfet present that is in need of a gate stopper resistor but the gate stopper is either too low value or simply not present. The circuit will be prone to resonance or oscillation lets assume the former - prone to resonance but not actually oscillating.

Do we not all know from experience that the above description is an almost guaranteed formula for an amp that will sound awful ? I thought this was common knowledge. I know I have heard this effect.

Somehow that resonance, which will typically be in the 10 - 20Mhz range, manages to ruin the audible sound of the amp - if it isn't somehow getting demodulated into the audible range and thus appearing in the H2 - H20 spectrum can someone explain to me how this effect can be heard.

I would welcome a fuller understanding of this phenomena.

cheers

mike
 
Last edited:
OK

So do we think that this audible effect comes from the fact that when this resonance is excited by the music signal and the some oscillation begins the transistor has been transformed from a reasonably linear device to a non-linear device and this causes audible distortion ?

Is it that simple ?
 
Status
Not open for further replies.