John Curl's Blowtorch preamplifier

[1audio]

Were the pot distortion measurements done with the wiper feeding an extremely hi-Z load, like a JFET input, or were they done with the wiper feeding a rather finite resistive load to ground. I would guess the former case, in which there is virtually no AC or DC current flowing in the wiper, would exhibit much less distortion. If any distortion is exhibited in that case, what is the distortion mechanism thought to be?

Thanks,
Bob

I was measuring through the wiper with a Radiometer CLT-1 at around 3V I think. This was a while ago and just a quick effort to weed out unsuitable vendors. I think I may be able to set up two CLT-1's if I can find the bench space and do a better evaluation, closer to the actual circuit.

The distortion is probably from the connection across the wiper and the narrowing of the current path (or so the CLT app notes say, the physics are beyond me). There are dissimilar materials and a host of other issues that may affect the connection. And I don't have the patience to set a pot and leave it for a month to see what happens as the contact sits. I know from fixing a plethora of old test equipment the contacts will decay over time, but not always.

Even though these distortions are quite low they seem to be an indicator of audio problems. I would never claim that they are audible unto themselves.

 

[1audio]

I have the Alps "Black Beauty" in my preamp and it sounds very smooth.

That is the pot we used in the production Spectral DMC-10. It had the best sound of the practical options at the time. We used a 10K pot to keep the frequency effects at a minimum. Lower would be better but would induce (more) distortion in tube sources. We tried to keep the wiper load as high as possible to reduce distortions, errors and wiper noise to a minimum.

 

[1audio]

Toroids can have good high frequency isolation (another way of saying low interwinding capacitance) if they're made with a Faraday shield between primary and secondary, which most off-the-shelf toroids do not have.

Personally I've always liked the dual, split bobbin, UI power transformers.

Still the problem of high capacitance exists with these shielded xformers, since the shield must be connected to somewhere (earth, that is). If one wants true low coupling (low leakage/"GND" currents) there is no other way than true low capacitance xformer design like R-Core, UI and specialist toroids with sectoral (non-concentric) windings (Avel-Lindberg has some small wattage models).

- Klaus [/QUOTE]
If the goal is low line noise then isolation the secondary side from capacitance is part of the problem, the magnetic coupling is also a part of the problem. And the fields from the transformers are more insidious than one might think.

A true double shielded transformer (ultra-isolation) when implemented properly should reduce the line induced noise to a minimum, but its important to tie the shields to the right places or you won't get anywhere with one. Its best to model the whole system on paper identifying all of the leakage paths, current sources and current sinks. For perspective- 1 microVolt can be enough to cause a significant problem, and its 160 dB below powerline voltages in the US. And 1 microAmp of current across 1 Ohm will get you that 1 microVolt of bad stuff. 100 MegOhms is all the leakage necessary to get that microAmp, and it could be 27 pF of leakage capacitance, not a lot. (I think I have the arithmetic right, no guarantees)

I saw an instrument transformer years ago that had a big plastic former keeping the winding well away from the core. Not very efficient but good isolation. Probably for high voltages.

I had an "ultimate" power transformer made years ago. It was a laminated core, "O" shaped, 6" X 8" for about 15W load. The primary and the secondary were on opposite legs. They had a large spacer between the windings and the core. The fill was VERY low. The flux was also less than 1/2 the typical flux level for a transformer with that iron. The results were amazing, suggesting that all of the above was important.

 

[Bonsai]

Any comments on the integrated potentiometers?

I noticed that Audio Research uses these on one of their pre-amps . . . . and it has gotten very good reviews.

 

[h_a]

Thank you for your comments on transformers!

By the way, does anybody have an idea where one could get these Vishay bulk metal foil potentiometers like the 1240W? Mouser says 50$ for 25 pieces of a certain value. 2$ per pcs is quite inexpensive and would be interesting to try out.

Have fun, Hannes

 

[Bob Cordell]

I recall this well, but decided that it was necessary to freeze the
voltage across the device in order to isolate the effect. As you
know, varying Vds or Vce introduces a subtle gain variation on its own.

So I cascoded a 2SK370 for constant Vds at 10V and measured
harmonics for a constant 1V output into 1 Kohm at an 8 mA bias
with varying degrees of degeneration.

What I saw was all harmonics declining with increasing
degeneration, although not initially in exact proportion.

When I find myself with more time, I intend to repeat the test
with enough precision to enable graphs with circles and arrows,
with a paragraph on the back of each one explaining what it's
about, and I'll also repeat the loop test with cascoding.

😎

Hi Nelson,

The cascoding is a good point, and may affect the results somewhat, but I don't think many people use the "Hawksford-like" cascode (forgive me for using this convenient terminology) where Vce or Vds is held constant. There is no question that such cascoding will reduce the distortion somewhat in a BJT or JFET, but it is a little unclear how it will change the behavior as a function of varying amounts of applied emitter degeneration vs applied global feedback. I'll look forward to seeing those results.

Of course, even apart from cascoding, there will be some visible differences between such experiments between BJTs and JFETs.


Cheers,
Bob

 

[Bob Cordell]

In case of multiple local feedbacks you have by-products of modifications of each function by local feedback multiplied on each other, in case of a global feedback you have transfer functions of each stage multiplied then they are modified by a global signal. From mathematical point of view results will be very different. And they don't depend on who and how draw general conclusions from the observation.

Your "pretty much the same" probably don't include frequency dependence and transients... I can show you on fingers how and why, if you will be so kind to follow me...

There will certainly be more complex interactions when multiple loops are involved and there are combinations of local and global feedback involved. It is also true that at much higher frequencies there will be frequency-dependent effects introduced that may affact the behavior. Whether or not the result I plotted would be different at 20 kHz is a fair question, for example. Indeed, I was just keeping it simple.

Most of these interactions are sufficiently complex that it is a bit unreliable to speculate based on hypothetical arguments. One really needs to either make the real-world distortion measurements or do the simulations. I'm not sure what you mean by "showing you on fingers", but I suspect that translates to argument rather than measurement or simulation. You are right that things get more complex, but unless you measure or simulate you are not necessarily right about the actual behavior.

Do you have measurements or simulations that show what you are referring to?

Cheers,
Bob

 

[zinsula]

This is of particular interest to me and I have compared some
actual measurements to the simulations presented here, but
the results are ambiguous - specifically I was unable to confirm
the creation of higher orders with degeneration.

The subject does not seem to be popular, however.

You will find the discussion ranging approximately over posts
#2873 through about #3842.

Thank you Nelson.
Yes, a more profound discussion about this would be interesting.

Maybe the attenuator position I proposed isn't that bad 😎

Tino

 

[KBK]

Manufacturing bad spots are on many common, and even very high quality pots. First found in Levinson JC-2 in 1974. BIG PROBLEM Hi Z loading helps, but distortion can be modeled as lossy inverted side by side diodes. Have made the measurement comparisons, myself. Other pots can sometimes look mostly like 2'nd harmonic.

Treating a pot like a feedback loop when it comes to controlling the variables of layout and construction (and thus sensitivity to all aspects of path length, type, position in space, number of junctions, etc) helps keep the pot's variables contributing to the least degree possible, in my experience.

 

[john curl]

Unfortunately, KBK, pots put directly into the feedback loop impart their nonlinearities without being fixed by feedback. Look at the feedback loop as a template, that the amp tries to follow.

 

[Sigurd Ruschkow]

Less treble? How come? Did you measure the frequency response of that pot?

Regards,
Edmond.

Not less treble. A more 'muddy' treble.



Sigurd

 

[zinsula]

Degeneration vs Distortion

Simulated a bit...

R1 / R11 / R21 are degeneration resistors, simulated between 100 Ohms and 1k, input was always 1V peak.

Analyzed the current through R1 / 11 / 21 (= the voltage over it), so no need for source resistors and cascoding etc.

Transconductance of the FET Models seem to be around 28mS.

But look for yourself.

PS Another way would be to hold the current constant, i.e. changing input voltage.

Tino

 

[Nelson Pass]

Myself, I don't think I would trust sims at -150 dB.

😎

 

[john curl]

Thanks Tino. I will never have to learn Spice at this rate! 😀

 

[zinsula]

Hehe, I don't trust it either 😉

Here with constant peak current of 7.4mA (quite high), but with different input voltage.

Again, maybe not to trust, but as a tendency....higher degeneration seems better.

Anyone with appropriate gear to measure the same thing? Easy enough to build.....

Tino

 

[hitsware]

Doesn't 'degeneration' (by definition)
imply less distortion ?
More passive in proportion to active
circuit involvement. .... ?
(for a given function)

 

[1audio]

Unfortunately, KBK, pots put directly into the feedback loop impart their nonlinearities without being fixed by feedback. Look at the feedback loop as a template, that the amp tries to follow.

John:
I think KBK was making the point that the layout around a pot is critical (I agree) not to put it in the feed back loop, something that opens up a host of potential problems (bad pun).
-Demian

P.S. The Quan-Tech is ready.

 

[Lumba Ogir]

PMA,

To say that asymmetrical LTP input is same good as complementary-symmetrical is a common wrong point of view of many circuit designers (e.g. Self), who have never analyzed JFET complementary-differential input stages in deep, and who always kept CCS source biasing in their analysis.

From what point of view would it be less good?
Have you analyzed the complementary-symmetrical structure and the push-pull topology in general? In that case maybe not thoroughly enough or in right manner. Keep it going.

 

[Lumba Ogir]

Mr. Cordell,

Originally posted by Wavebourn
It is obvious that if to multiply a function by a function from itself the order goes up.
The problem "local vs global" almost leads to different results of frequency dependence and different behavior on transients.

Your first statement is correct

The second one also, unconditionally.

The analogous chart where the NFB is applied globally instead of by emitter degeneration looks pretty much the same.

How can you make a statement like that, Sir?

 

[Bob Cordell]

Mr. Cordell,

Originally posted by Wavebourn
It is obvious that if to multiply a function by a function from itself the order goes up.
The problem "local vs global" almost leads to different results of frequency dependence and different behavior on transients.

The second one also, unconditionally.

How can you make a statement like that, Sir?

Because I made the chart and looked at it.

Cheers,
Bob