John Curl's Blowtorch preamplifier

[john curl]

Keep up the good work, PMA.

[syn08]

Here's the 600V/uS line preamp schematic. It looks rather complex (certainly much more complex than the 350V/uS pure opamp line preamp) but, conceptually, nothing really outstanding. What is missing in the schematic is the servo, currently implemented using a OPA132 FET input opamp in SOT-8.

For C3, C4, C5, C6, C8 and C9 it doesn't make much sense to specify the values, as they are barely lumped values only, but a combination of fixed caps and parasitics. C5 and C6 are in the low 1's pF range, C3, C4 and C9 in the low 10's of pF and C8 in the 100's of pF. They are also very strongly dependent on the opamp model. For example changing to THS4631 requires completely different values. Using OPA657, the open loop Fu was measured at around 45MHz and the OL gain at 60KHz was measured as around 90dB. Take these numbers with caution, I am not 100% confident on a precision better than +/-25%, I still have to work on better attenuators and shieldings for my OL measurement setup.

Don't bother to simulate this thing in the time domain, it won't work properly (in terms of accuracy) even if your opamp model has provisions for the supply current, so as to reflect the load current. AC simulation can be done (and some lumped values for the phase correction networks can be calculated) but this won't help much in laying out the real thing on a PCB.

PCB Gerber files (unfortunately not something that you could etch at home) and other details will be posted ASAP on my web site.

[Edmond Stuart]

Originally posted by G.Kleinschmidt
That's a good question. ................................
Cheers,
Glen

Hi Glen,

Now I've simmed PIM by means of FFTs, one of the input signal and one of the output signal. (Yes I know, this is brute force, could be done more efficiently. But who cares, my CPU is water cooled ).

Anyhow, I compared the phase of the two fundamentals (20kHz) and, expectedly, I got a complete different result (wrt the ZC mehod):
The phase modulation (at 20kHz) was only 59 nano degrees or expressed in time: (59/360)E-9 * 50E-6 = 8.2E-15s, ie 8.2 femptoseconds, far, far less than 11 picoseconds, obtained by only looking at the zero-crossings.

Surprised? Not me!

Cheers,
Edmond.

edit: Pavel, maybe I've missed something, but what the heck has PIM to do with a dominant pole compensation?

[Bob Cordell]

Every so often this PIM thing in relation to NFB keeps rearing its head. For those interested, read the PIM paper on my web site at www.cordellaudio.com.

I'm sure that PIM is difficult to accurately simulate, and my hat goes off to those here who have tried. I have never tried to simulate PIM in an amplifier.

During the PIM debates in 1983 I actually built a coherent IM distortion analyzer to measure PIM. I was able to show with those measurements that small open loop bandwidth does not increase PIM as long as the unity-gain frequency is held constant. The analyzer used phase locked loop techniques and coherent detection to isolate AM intermodulation distortion and phase intermodulation distortion. It basically relied on a coherent phase detector to measure PIM.

I also discussed and measured PIM in connection with my MOSFET power amplifier paper. That amplifier has lots of negative feedback. Its actual PIM measured only 0.05 ns.

I am not saying that PIM does not exist. Indeed, it exists even in amplifiers that employ no negative feedback. Nor am I saying that there is not a mechanism whereby PIM can be created by negative feedback. I discussed that mechanism in my PIM paper in 1983 and provided an explicit formula for it. Barrie Gilbert's work was little different, and his results did not conflict with mine. Nor did his results show that increased negative feedback and decreased open loop bandwidth exacerbate PIM.

In most cases, the application of negative feedback will actually reduce the total amount of PIM produced by an amplifier.

As was the case with TIM, Otala was completely wrong in suggesting that lots of negative feedback with small open loop bandwidth caused or exacerbated PIM. What matters is the unity gain bandwidth, not the open loop bandwidth.

Finally, it is very difficult for a circuit to create PIM without creating readily measurable IM and THD. The same nonlinearities are usually at work. If one deliberately put an all-pass filter in their amplifier which used a nonlinear junction capacitance for its tuning, then perhaps PIM could be created without accompanying IM.

Cheers,
Bob

[john curl]

Some people look to understanding why we hear differences, and other people would rather bury the differences, and their measurements, between components.

[john curl]

Keep it up, PMA, listen to nobody else but yourself.

[syn08]

Originally posted by john curl
Keep it up, PMA, listen to nobody else but yourself.

Sad, very sad... This is well beyond engineering, it's already a religious war.

Q.E.D

[john curl]

No, it is a 'sonic quality' war. We want to know WHY different designs sound different. It is important to us. Is this too much to ask of the 'hear (and measure) no difference' fraternity?

[dimitri]

I agree that life is boring without challenges. It seems though we choose different challenges.

Ovidiu, you advocate solid engineering approach to audio design. Baxandall and later Cordell showed that the rate of change of audio signal is very limited (taking into account frequency distribution). What was the goal in 600V/uS 4MHz line preamp design?

[hitsware]

Originally posted by G.Kleinschmidt



What a powerfull argument.

Thank You !
It seems to me that the ' blowtorch preamplifier '
is following the design philosopy of Microsoft.
Added performance ? .... yes
Added cost ? ..................yes
Inelegance ? ..................yes
If you don't even need feedback to reach
a ' good ( hearable and measureable ) ' performance .......
Then why even slightly confuse the issue
by using it ?