Nothing to see here

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
And you were expecting normal?
 

Attachments

  • NothingToSeeHere.gif
    NothingToSeeHere.gif
    27.2 KB · Views: 656
  • NothingToSeeHere.asc.txt
    14.9 KB · Views: 77
Nevermind soft clipping at +/-22V, and +/-6.5A,
nor a non-switching AB crossover. None of that
stuff even slightly important. The secret entirely
about rounding sharp corners off the schematic!

No-one else has figured this by now? As nature
abhors a vacuum, so too does LTSpice abhor a
sharp corner. Place an oil slick there, and watch
the electrons spin wildly out of control into the
grandstands.

I've added 22uF in series with R5, and 680R+68pF
across R8. Defining -3dB rolloffs at 5Hz and 96KHz.
 
Last edited:
Intriguing circuit, especially the input stage.
But the performances do not seem to match the complexity however.

The input stage has a number of symetries, and by some aspects it reminds me of cross-quads, but stacked vertically.
Except there is a rather large ratio between the collector current of say Q3, Q19 and Q5, Q8 f.e.: this and the 470R coupling resistors ruin most of the Xquad effects.

What is this topology supposed to achieve, Ken? (Apart from the clipping control and Iq control, which are pretty obvious.)
 
Intriguing circuit, especially the input stage.
But the performances do not seem to match the complexity however.

The input stage has a number of symetries, and by some aspects it reminds me of cross-quads, but stacked vertically.
Except there is a rather large ratio between the collector current of say Q3, Q19 and Q5, Q8 f.e.: this and the 470R coupling resistors ruin most of the Xquad effects.

What is this topology supposed to achieve, Ken? (Apart from the clipping control and Iq control, which are pretty obvious.)

Try 2n3904 2n3906 for the low, 2n2222 2n2907 for the high?
Similar transistors scaled for Iq might have made more sense.
Pretty sure I've seen all the above surplus, and as quad DIPs.

I tried the whole sim with just 2n2222 and 2n2907, no big diff.
Also not worried @53Mhz, thats high enough to be quenchable
with a ferrite bead somewhere. Somewhere I also posted a
mod for 5Hz and 96KHz -3db rolloffs too, yeah #8 it was...

So, nobody noticed the virtual cascodes? That I did not allow
critical decision making transistors to ever work with more nor
less than 2 emitter drops tracking above each collector?

I thought the first Watt looked pretty clean considering this
bill of junkbin materials? Nothing fancy, and mostly all same.
Sim is nearly this clean all the way to the soft clipping knee.

Course you gotta play nice with LT's FFT to keep from getting
same crazy windowing that unrealistically splatters the results.
I'm no master of the FFT, I just mess with start stop timing.
This from like 5 to 15mS with .options plotwinsize=0
+/-0.125V in, +/-4V out into 8ohms = 1W.

I don't get where the 470's are ruining anything? The cross
coupling resistors are necessary to make a voltage diff into
a pair of equal and opposite current diffs. I had thought to
abuse 340's (a pair of 680 in parallel), just to keep the bill
of materials monotonous. Its not at all picky an exact value.
 

Attachments

  • 1W8ohm.gif
    1W8ohm.gif
    10.7 KB · Views: 246
Last edited:
Yes, X crossed quad of matched JFETs. Same thing, only completely different...
Excepting you would need at least 8, to boast a similar self cascode action.

I definitely got JFET beat on cheap matched pairs and quads.
Only caring threshold of VBE, which is very repeatable.

My input impedance sucks, but those 22Ks could look bigger with bootstraps.
I'm not sure an input bootstrap's effect on stability, not quite ready to endorse
that as a solution yet. Still, 7K333 ain't the worst input impedance ever...
 
Last edited:
The input stage still puzzles me. And I don't think it achieves its full potential.
As I understand, the 470R's set the transconductance, and it's pretty low.
The virtual cascodes may be a nice idea, but with such a level of degeneration they have little work to do.

This is confirmed by the OL gain, approx 1,000, which I find very low for such a topology.

What about bypassing the 470R's with 47R + 470µf?

It would easily bring 20dB more AOL (and perhaps some instabilities too, but that should remain manageable).
 
If you don't like 470, degenerates too much transconductance?
Just drop 3 diodes in series with 150R, no need for bypass caps.

Just got back from Tanner with two sandwich baggies fulla 2n4401
and 2n4403, and a hundred pack of 1/4W 12R carbon comps 5%.
Unfortunately, these are just a bit small to clamp two with a TO5
sink unless'n I were to wrap with some aluminum foil or whatever...

Een with some wallwarts and other crap I bought, less'n 20 bucks.
I got off cheap today. Maybe I'll build this one Manhattan style?

Already have MOSFETs on heatsinks, and 24V power supplies.
Damn, I forgot Schottkys....
 
Last edited:
I just simmed with 2 diodes and 220R in place of 470R, much better. Thanks Elvee.
Only 2 diodes for reason of too lazy to make more room (for three) in the drawing.
Not saying two is better than three. Will I waste transistors here to save another
shopping trip? You bet!

There was not much distortion in the original sim to tell any difference, but Cordell
models for IRF MOSFETs showed up some distortion. Now added transconductance
through "the 470's" reveals how much it helps.

I have some 10A Shottkys somewhere, just need to find them...
 
Last edited:
Maybe I went too far with buying the cheapest to prove a point?
The point that this topology makes holds even crummy transistors
at very stable operating points, thus evading all sorts of variables.

Just ran sim with 2n4401 + 2n4403 and realize I can exceed 40V
on two of them. Also the current limits are no longer completely
symmetrical. Well, whatever. What I got, probably what I'll use.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.