John Curl's Blowtorch preamplifier

[andy_c]

Dimitri, at first I wasn't sure what you were saying, but after thinking about it a bit, it hit me. Is there a "mirror image" effect on the other crossing?

 

[KSTR]

Agreed, where is transition from negative to positive for the same circuit?

In my example both transitions look similar, though not 100% identical. For the highest level, there is quite a bit of even distortion (~ 0.1% sum of even orders) which might account for this. Looking at the fundamental phases, there is a slight decrease, from -0.8deg to -0.92deg in the FFT output, and there are shifts in the normalized (to fundamental) component phases, the 3rd ie increases from 80deg to 90deg, and the 2nd jumps around at will. 25kHz was the test frequency.

- Klaus

 

[GK]

Hmmm.....so, what we basically need for this technique to work the best is an amplifier with as low an open loop bandwidth and as much global nfb as possible for minimal distortion.



http://www.diyaudio.com/forums/showthread.php?postid=1535888#post1535888

 

[scott wurcer]

Scott, if you don't care, why do you criticize the efforts of others?

I never criticised anyone's efforts. You have to divide invention from doing your job.

 

[syn08]

What about PMA simulations, which looks like late Otala PIM? .

Why couldn?t we discuss alternative to overall FB schemes? Life is boring without challenges.

Andy already addressed some concerns regarding the 5532 model. I took a look into a standard 5532 macromodel and its definitely not good for any accurate distortion simulation. Two more comments/questions: a) PIM was already analyzed to death. Why would it be specific to circuits using integrated opamps? b) if one would integrate Pavel's circuit (well, at least the part that is prone to) would it have more, less, or the same amount of PIM compared to the discrete implementation?

Otherwise, I don't see any problem in non nfb designs. What makes me a little nervous is the way the non fb designers are sneaking nfb in their designs, be it degeneration or other kind of local feedback, claiming that only gnfb is evil. But then every effort to linearize the circuit before applying gnfb is valuable, not for concerns regarding the amount of required gnfb to reach the target performance but for potential stability issues.

I agree that life is boring without challenges. It seems though we choose different challenges. Which is normal and healthy.

 

[KSTR]

The other concern I have is op-amp models. If the model is device-level, then that's fine. But most op-amp models are of the macro-model type, which aren't intended to model distortion at all. So they should be used with care in that application. I'm thinking specifically of the 5532 model that was used above.

I bet Pavel has addressed this issue, using some proper reverse engineered model from the known 5532 "equivalent schematic". BTW, both the Philips and the OnSemi datasheet of the 5534 still have the input pair drawn wrong, after all this years...

- Klaus

 

[john curl]

Sounds backward to me.

 

[john curl]

I need to know more, PMA, about the test that you are doing. Don't worry about the others, and their comments.

 

[scott wurcer]

Well, this isn't a text, but an app note that was taken from a 1974 article in the IEEE Journal of Solid State Circuits. It's The Monolithic Operational Amplifier: A Tutorial Study by Solomon. It's dated but very nicely done.

And this reminds me of a question I had intended to ask Scott. Does the output stage dissipation of modern IC op-amps still cause variations of open-loop gain due to thermal effects as described by Solomon in this paper? If not, what has changed?

People have become smart enough to eliminate the first order effects of this. We have amplifiers with 2mA bias and 400mA peak current that don't have this as a primary problem.

 

[KBK]

You didn't ask for help. Here's an example of a discrete op-amp that you can't integrate. Note JE990 style inductors in input, but here circa 1966. The use of an explicit ground pin is now discouraged showing that even Dick Burwen didn't quite get it. Still can't find an earlier use of folded cascode as in that Harris op-amp.

I like those inductors. I have my reasons. It has to do with the actual idea of conduction itself. You see, there has to be an 'alignment' along the entire conductive path, before conduction occurs. This means, that inductors, in their own special little way... act like capacitors at near lightspeed, just before conduction or current passing/transfer takes place. Thus, a properly applied inductor is an energy storage device, that surpasses any known capacitor in quality..but only for that incredibly short period of time.

The rub..is learning how to apply that point--- in design.

As for disallowing change..think if it as the lightspeed equivalent of a perfect 'non-newtonian fluid'. The ultimate in brick wall 'slam back' (controlled) stability. The quality of the inductors and matching would be critical in that application. Matching would be more critical than absolute value, from my understanding of the inductors themselves, as a consideration. Mass stabilization at it's finest.

Think of choke loaded plates..and why they work so well..and it begins to make sense.

 

[KBK]

I need to know more, PMA, about the test that you are doing. Don't worry about the others, and their comments.

I saw the graph.and it took me a sec to figure out what I was looking at..but I liked it much when I figured it out. A very important piece of information, it provides.

 

[john curl]

You should see the thermal graph of a ua741. Shocking!

 

[andy_c]

We have amplifiers with 2mA bias and 400mA peak current that don't have this as a primary problem.

Thanks Scott. Just trying to catch up with the rest of the world .

 

[john curl]

If I only had my op amp tester handy.

 

[hitsware]

What's wrong with "stinkin overall feedback" ?? Please explain ??

consider this circuit:
(your devices, values)

An externally hosted image should be here but it was not working when we last tested it.

No feedback related issues

 

[john curl]

It is primitive. Why did you put it up here?

 

[GK]

I bet Pavel has addressed this issue, using some proper reverse engineered model from the known 5532 "equivalent schematic".

If that was the case then I would expect the results to be not dissimilar to mine.

 

[hitsware]

>It is primitive. Why did you put it up here?

Into 8 Ohms I got ~ 0.06 % THD 20 to 20kHz
Quasi Monotonic
(THD constant with output 0 to 5 Watts)
I'm surmising that with more suited small signal
devices and the lighter load presented to a
' blowtorch preamplifier ' the same scheme could
generate good numbers compared to feedback
driven designs .... with no feedback issues <|:^))

 

[snoopy]

consider this circuit:
(your devices, values)

An externally hosted image should be here but it was not working when we last tested it.

No feedback related issues

Sir, you have avoided answering the question and your example contradicts yourself. The 1 Meg resistor used in the gate-drain circuit of the first FET is providing negative feedback !!

Somehow providing the same feedback over two stages instead of one is supposed to make it infinitely worse

 

[snoopy]

>It is primitive. Why did you put it up here?

Into 8 Ohms I got ~ 0.06 % THD 20 to 20kHz
Quasi Monotonic
(THD constant with output 0 to 5 Watts)
I'm surmising that with more suited small signal
devices and the lighter load presented to a
' blowtorch preamplifier ' the same scheme could
generate good numbers compared to feedback
driven designs .... with no feedback issues <|:^))

I could provide many examples of feedback amplifiers which perform even better so what are all of these issues with feedback amplifiers ?? Please explain ??