I'm aware of the problem with perfect matching in simulation, therfore I usually use slightly different models (FET vto and kp params with 10% variation) and also non perfect matching of resistors etc. Still I did get dominating 3rd with almost identical levels for the grounded and the Zinsula variations, while Patrick's cross-connection seems to be a tad more sensitive to mismatches.
- Klaus
- Klaus
Then I don't have an explanation for your results, or rather I imagine
that the 3rd harmonic is simply dominant in your sim circuit.
In actual measurement I have observed significant (large) quantities of
distortion looking at one output to ground which are not seen in the
differential output across the load. It is common mode content and
it can be shown that part of it is distortion and noise which originated
on the other half of the circuit and was communicated through the diff
pair to the other side.
This common mode distortion and noise is comparable, but not
identical, to the open loop figures for the circuit.
The opposite will occur for the non-inverting but shared feedback
arrangement, where distortion and noise will be reproduced out of
phase at the opposite output, reinforcing it across the load.
And of course for the simple bridged version which doesn't share
feedback, you have relative independence of the two sides.
that the 3rd harmonic is simply dominant in your sim circuit.
In actual measurement I have observed significant (large) quantities of
distortion looking at one output to ground which are not seen in the
differential output across the load. It is common mode content and
it can be shown that part of it is distortion and noise which originated
on the other half of the circuit and was communicated through the diff
pair to the other side.
This common mode distortion and noise is comparable, but not
identical, to the open loop figures for the circuit.
The opposite will occur for the non-inverting but shared feedback
arrangement, where distortion and noise will be reproduced out of
phase at the opposite output, reinforcing it across the load.
And of course for the simple bridged version which doesn't share
feedback, you have relative independence of the two sides.
Patrick-
I agree--if you can get the Toshibas. I wouldn't suggest that the Vishay/Siliconix devices are anywhere near as good as the Toshiba's, but you can still build a very good sounding amp with them if you can't get Toshiba jfets.
I very well may build a balanced F5. I had a bit of hum coming from my speakers the other day. The hum is gone now, but if it becomes a problem, a balanced F5 is the best solution I can think of.
I will try to post some pics of sqare wave output soon. I don't have a camera right now, but I may be able to borrow one from work.
JJ
Without any doubt reality overrules simulation, and I have to say that my study was in no regard extensive (Monte-Carlo methods are a bit cumbersome in LTspice).Nelson Pass said:
On the other hand, when I try to inject a sharp noise pulse of 2V in one of the MOSFET gates (using the perfect circuit) the spike seen at the load and many other points is virtually identical in X vs. non-X, which should not be the case?????
What I am missing here ????
- Klaus
I've few basics questions:
> This is a pretty incredible little amplifier at very low cost, so why not just build it and see for yourself.
I would not hesistate to second that, especially that I always like high bandwidth amps.
But the "very low cost" should, IMHO, be taken into context. While the face value of the circuit components would probably cost no more than US$50 per channel, that does not include power supply, heatsink, connectors of various sort, and other mechanical parts, which can easily be 3 to 4 times the circuit components.
In addition, most people do not have a collection of matched JFETs and MOSFETs at hand, and would have to pay a premium for matched devices, or buy 10x of what they would otherwise need nominally.
Of course if you only want a quick mock-up to test the circuit, and are not bothered with matched devices or minimum distortion, have laboratory power supplies and large heatsinks handy, then you can probably get one built for a few bucks.
I have to say I am surprised that the F5 is not receiving much attention, compared to say the F4 or the Zen amps. It is really a much better amp than most of those, and is not lacking in power either in the balanced (4 MOSFETs) version.
Patrick
I would not hesistate to second that, especially that I always like high bandwidth amps.
But the "very low cost" should, IMHO, be taken into context. While the face value of the circuit components would probably cost no more than US$50 per channel, that does not include power supply, heatsink, connectors of various sort, and other mechanical parts, which can easily be 3 to 4 times the circuit components.
In addition, most people do not have a collection of matched JFETs and MOSFETs at hand, and would have to pay a premium for matched devices, or buy 10x of what they would otherwise need nominally.
Of course if you only want a quick mock-up to test the circuit, and are not bothered with matched devices or minimum distortion, have laboratory power supplies and large heatsinks handy, then you can probably get one built for a few bucks.
I have to say I am surprised that the F5 is not receiving much attention, compared to say the F4 or the Zen amps. It is really a much better amp than most of those, and is not lacking in power either in the balanced (4 MOSFETs) version.
Patrick
neutrino_th said:
That's true
But I'm a noob with poor english vocabulary. I just try to understand
Your English is better than my French.
Please to give my warm regards to Paul Bocuse of your fair city. I love truffle soup.
Patrick -- the power supply, chassis and inter-connects are pretty much the same for any power supply you're going to build -- so whether you build a Leach or an LM4780 you still have the same "fixed cost burden".
I cost out the F5 parts at $26/channel.
jackinnj said:
I like it very much .
Distorsion figures are pretty good also not matchin transistors IMHO.
Its easy to find out that the amp sound much better after some hours of burn in .
Or its just my ears that get used to that different sound ?
Lots of work required F3 ( or Zen V9 ) in matching , following the goal of the lower distorsion or obtaining 2 identical channels .
Anyway F5 is very very good . The next one I will match transistor and source resistors as I can and will see .
I used some snippable female pin headers for the JFET source adjustment resistors. I haven't seen the technique written up yet, so I assume that you examine the residuals on your distortion analyzer and adjust until everything is hunky-dory.
Jacco -- last time I was in Lyon we stayed in some swanky hotel and it rained through the skylights -- I really liked the Musee des Beaux Arts.
An externally hosted image should be here but it was not working when we last tested it.
Jacco -- last time I was in Lyon we stayed in some swanky hotel and it rained through the skylights -- I really liked the Musee des Beaux Arts.