Author Topic: "Words of Wisdom", by John Curl  (Read 10028 times)

Gen. Dreedle

  • Commander, Alabama wing, Confederate Air Farce
  • Global Moderator
  • Hero Member
  • *****
    • View Profile
"Words of Wisdom", by John Curl
« on: May 26, 2010, 06:40:02 AM »
Some will claim it contains no wisdom.

Others will complain that this work was culled from other forums, without the permission of some folks who are mentioned in it.

Others will complain about if, for some other reason.

Fine.

I did not create this. (It was done by one of the folks who tried to make DIYHiFi.org work. It was an attempt, by that person, to help grow the forum. Like everything else we tried, it really didn't work.) Discounting all of that, it might make for good reading. Even if you don't believe any of it, you may find it sparks your imagination. That alone makes it worth the download.

So, here it is:

/data/johncurl-v.0.1.pdf

Speaking of people who will complain...............

Shortly after this was made public on DIYHiFi, one person was outraged by it, and quit immediately. Well, nothing I could do about it. He left, and that was that. But funny thing...............

Quite some time after this, one of the members sent me an e-mail, that went something like this:

"I don't know what you did to run off *** *****, in the early days of the forum, but.............

Thank you! Thank you! Thank you!"

See what I mean about not being able to please people?

Jocko
"Major Danby, sir."
"Danby. D-A-N-B-Y."
"Take him out and shoot him."
"Sir?"
"I said take him out and shoot him. Can't you hear?"

seagreen

  • Hero Member
  • *****
    • View Profile
Re: "Words of Wisdom", by John Curl
« Reply #1 on: May 27, 2010, 09:16:44 AM »
How much effect do you think that matched JFETs have in the input stage?

How about matching for the MOSFETs used in the folded cascode?

Just how critical is this?
No double blind testing was used in the making of this post.

Gen. Dreedle

  • Commander, Alabama wing, Confederate Air Farce
  • Global Moderator
  • Hero Member
  • *****
    • View Profile
Re: "Words of Wisdom", by John Curl
« Reply #2 on: May 27, 2010, 09:18:08 AM »
Sounds like the kind of question a newbie would ask!

Jocko
"Major Danby, sir."
"Danby. D-A-N-B-Y."
"Take him out and shoot him."
"Sir?"
"I said take him out and shoot him. Can't you hear?"

seagreen

  • Hero Member
  • *****
    • View Profile
Re: "Words of Wisdom", by John Curl
« Reply #3 on: May 27, 2010, 09:23:58 AM »
Can't help it.  We are what we are.

You answered pretty quickly; do you just sit there and follow your new forum pages?   ;D

Back to the questions...

I can understand why the input devices have to be matched, but not the cascode devices.  Except for DC offset purposes.
No double blind testing was used in the making of this post.

Gen. Dreedle

  • Commander, Alabama wing, Confederate Air Farce
  • Global Moderator
  • Hero Member
  • *****
    • View Profile
Re: "Words of Wisdom", by John Curl
« Reply #4 on: May 27, 2010, 09:32:03 AM »
I just happened to be on-line.

I imagine if there was a big difference in gate threshold, that would mess it up. Haven't played with enough MOSFETs to have a feel for how much their parameters move around. That 1/f thing kinda sorta worries me.

Although, I have used them in some low-noise regulators. Open-collector style. Allows the error amp to operate away from Vcc.

Of course, all speculation, since no one has actually seen a Blowtorch preamp schematic. Save JC, of course. I imagine some Ayre stuff is close.

Jocko
"Major Danby, sir."
"Danby. D-A-N-B-Y."
"Take him out and shoot him."
"Sir?"
"I said take him out and shoot him. Can't you hear?"

seagreen

  • Hero Member
  • *****
    • View Profile
Re: "Words of Wisdom", by John Curl
« Reply #5 on: May 27, 2010, 09:49:03 AM »
Wait - I saw some schematics on a couple different web pages.  I have to believe that they're close to the real thing.  I figured some guy got to borrow one and take it apart.  Or even that the designer gave out the design since I hear that these preamps aren't made any longer.

Why does Vgs matter for the cascode parts?  Aren't the input JFETs changing their own voltages with input signal?

Don't you wonder how much better this preamp would sound if it had some feedback applied?
No double blind testing was used in the making of this post.

Gen. Dreedle

  • Commander, Alabama wing, Confederate Air Farce
  • Global Moderator
  • Hero Member
  • *****
    • View Profile
Re: "Words of Wisdom", by John Curl
« Reply #6 on: May 27, 2010, 09:55:37 AM »
I was talking about the MOSFETs. If one turns on at 2 V, and the other side at 3 V, that would probably muck it up. (You get the idea.)

Jocko
"Major Danby, sir."
"Danby. D-A-N-B-Y."
"Take him out and shoot him."
"Sir?"
"I said take him out and shoot him. Can't you hear?"

seagreen

  • Hero Member
  • *****
    • View Profile
Re: "Words of Wisdom", by John Curl
« Reply #7 on: May 27, 2010, 10:01:00 AM »
Wouldn't that just change the drain voltages on the input JFETs?  But that changes when Vgs changes, too, because the gate voltage on the cascode parts is fixed.

This is kind of a challenging problem.  In some ways the JFETs work better when Vds is held constant, but in other ways not.
No double blind testing was used in the making of this post.

Gen. Dreedle

  • Commander, Alabama wing, Confederate Air Farce
  • Global Moderator
  • Hero Member
  • *****
    • View Profile
Re: "Words of Wisdom", by John Curl
« Reply #8 on: May 27, 2010, 10:01:28 AM »
Yeah...............feedback..............yeah, go for it!

The guys at The Pub think feedback is great, because there is less gm variation in the input stage. That, and lower voltage swing means less change in input non-linear capacitance. So, less modulation of the corner frequencies.

Something about global feedback doesn't sound quite right. Always liked the results I get with no feedback.

Maybe transient intermodulation distortion isn't the answer!!!!!!!!!!!

Or PIM, or SID, or all the other bizarre distortion mechanisms. Real, or imagined.

The horror, the horror.

Jocko
"Major Danby, sir."
"Danby. D-A-N-B-Y."
"Take him out and shoot him."
"Sir?"
"I said take him out and shoot him. Can't you hear?"

Gen. Dreedle

  • Commander, Alabama wing, Confederate Air Farce
  • Global Moderator
  • Hero Member
  • *****
    • View Profile
Re: "Words of Wisdom", by John Curl
« Reply #9 on: May 27, 2010, 10:06:11 AM »
Wouldn't that just change the drain voltages on the input JFETs?  But that changes when Vgs changes, too, because the gate voltage on the cascode parts is fixed.

This is kind of a challenging problem.  In some ways the JFETs work better when Vds is held constant, but in other ways not.

The drains, I would imagine, should be the same, on both sides.

Oh, wait...........we are talking about a Blowtorch, something we have not seen, just guessing...................uh....................ok, trying to imagine what I think the Blowtorch looks like.............

Yeah, the drains would be at different points. The non-linear capacitance would be different. Some of the distortion components might not cancel out. Sounds like a job for Simulation Man. Maybe we should try to recruit him? He is a malcontent.

Jocko
"Major Danby, sir."
"Danby. D-A-N-B-Y."
"Take him out and shoot him."
"Sir?"
"I said take him out and shoot him. Can't you hear?"

seagreen

  • Hero Member
  • *****
    • View Profile
Re: "Words of Wisdom", by John Curl
« Reply #10 on: May 27, 2010, 10:46:51 AM »
Feedback is a strange topic to think about.  For example, just what *does* happen to the energy of the distortion products?

Here's an interesting discussion of feedback and its effect on music reproduction.


http://www.its.caltech.edu/~musiclab/feedback-paper-acrobat.pdf

No double blind testing was used in the making of this post.

seagreen

  • Hero Member
  • *****
    • View Profile
Re: "Words of Wisdom", by John Curl
« Reply #11 on: May 27, 2010, 12:35:08 PM »
One more thought - my weekly allocation.  Forget the earlier feeble attempts at joking - those weren't thoughts.

I wonder how much of the performance of a preamp or amp or whatever is based on the topology and how much on the execution.  Probably hard to put a number on, but I almost think that the execution and little details almost matter more than the topology in a lot of ways. 

I don't just mean the use of the right components that a lot of people claim don't matter.  (And to be fair to them, for their design goals it might not.)  More the layout, the "grounding", power supply noise, coupling, yada yada yada.

You can't make a crappy topology sound brilliant just by using a great layout scheme, but you can make a modest topology sound pretty good through clean circuit living.  It might sound better than a great topology put together badly.

It certainly works that way at frequencies higher than audio. 
No double blind testing was used in the making of this post.

Gen. Dreedle

  • Commander, Alabama wing, Confederate Air Farce
  • Global Moderator
  • Hero Member
  • *****
    • View Profile
Re: "Words of Wisdom", by John Curl
« Reply #12 on: May 28, 2010, 06:54:48 AM »
One place execution shows up is in cascodes. Stick the wrong capacitor on the bias pin, and the square wave response shows strange things. Maybe Mr. Simulation could predict some of that. Might be time to try to fetch him.

Jocko
"Major Danby, sir."
"Danby. D-A-N-B-Y."
"Take him out and shoot him."
"Sir?"
"I said take him out and shoot him. Can't you hear?"

tommy_t

  • Newbie
  • *
    • View Profile
Re: "Words of Wisdom", by John Curl
« Reply #13 on: May 29, 2010, 09:54:21 AM »
Well, I'm kind of puzzled about the question because I don't know the specific circuit being talked about.  Jocko asked me whether SPICE could predict some of the strange effects on a folded cascode caused by different types of capacitor used to bypass the cascode "bias pin".  The answer is... "maybe".

If it's parasitic oscillation of a MOSFET at over 100 MHz, probably not.  But the fix for that is likely just a gate stopper.  OTOH, I've done a simulation of a bootstrapped cascode input stage that showed peaking in the frequency response from input voltage to output current to be over 10 dB at around 20 MHz.  The solution in the simulation was to add a stopper.  Shorting out this stopper in the real hardware on the bench caused oscillation.

SPICE is a "garbage in, garbage out" system.  The weakness is the models.  BJT models can be quite good if the model parameters were extracted correctly (a big "if" most of the time).  Standard SPICE has no model for vertical MOSFETs at all, so to get a good one, it's necessary to use a subcircuit having a lateral MOSFET implementing the DC V-I characteristic, a nonlinear capacitor to implement the variable Cgd, a diode whose nonlinear capacitance implements Cds (and which also implements the DC characteristics of the body diode) and a fixed capacitor for Cgs.  It's a lot of work to create these models, and quite time-consuming.

So let's assume you've gone to all this work such that all your semiconductor models are very good, and in addition, you've accurately characterized capacitors for ESL and ESR.  My rules of thumb are this:

1) Below 1 MHz you should get very good results.
2) From 1-10 MHz, accuracy begins to degrade
3) From 10-100 MHz, accuracy degrades further, but if the simulation shows anomalies like parasitic oscillation, peaking in the frequency response or ringing in the transient response, better fix them or you're likely to have problems in the real circuit.
4) Above 100 MHz, forget about it

Gen. Dreedle

  • Commander, Alabama wing, Confederate Air Farce
  • Global Moderator
  • Hero Member
  • *****
    • View Profile
Re: "Words of Wisdom", by John Curl
« Reply #14 on: May 30, 2010, 10:10:41 AM »
The thing I have noticed is an odd anomaly. (OK, I haven't done much of this in the last 10-20 years, but not much has changed, I'm sure.)

In the folded cascodes I have built, there is an odd glitch, in the waveform. On the positive half of a square wave, right before it goes negative, there is a small, positive-going glitch. Not very big, but certainly noticeable on a >100 MHz 'scope. Futzing with the cap, or whatever, is hung on the base (gate) of the cascode device, has a lot to do with that. While some things are obviously wrong (and noticeable), minor things can have a big effect.

That gets back to the question of how much execution plays in the final result. In this case, I would say it plays a very important role. While that anomaly may not seem like much, I can assure you that it can be the difference between "sounds good", and "something doesn't sound right".

Our hobbyist brethren, with neither the simulation skills (ok, applies to me as well), or measuring abilities, will make something, and it will exhibit these sort of problems. In turn, they fill countless pages on DIY forums, complaining, because of their ill-fated attempts, that all cascodes sound bad.

So, even if they had the schematic to a Blowtorch preamp, and the parts (almost as unlikely), the odds they will get the same results are pretty slim. But, won't stop them from Googling for it. Some will stumble upon this. The perceptive ones may learn something. The rest will continue to kvetch. The end result being guys like us leave, and go somewhere off on our own. Where we won't have to deal with the likes of the DIYers.

Jocko
"Major Danby, sir."
"Danby. D-A-N-B-Y."
"Take him out and shoot him."
"Sir?"
"I said take him out and shoot him. Can't you hear?"