From my "THD Challenge", one-third of DIY'ers could hear a 7th harmonic at 0.01%. I designed the test to make the "distortion" easily audible. State-of-the-art amplifiers have 1-2 orders of magnitude lower total distortion.
Ed
Ed
Your circuit is not so good, if you want better results use 3 or 4 step emitter follower and cascode vas. Faster matched power transistor will be improved the result as well.
I suggest you to try to measure the distortion in different temperatures and how the bias has changed.
I suggest you to try to measure the distortion in different temperatures and how the bias has changed.
You'd be surprised at what some can hear under informal conditions. Yet you can have in excess of 1% distortion that isn't immediately identified. We are talking normal music played as they typically would. They eventually get that something sounds off. By the same token, over time some extremely minor distortion can be noticed.
It really depends on what distortion there is and how it affects the sound. But looking at a needle will not get you there.
It really depends on what distortion there is and how it affects the sound. But looking at a needle will not get you there.
This is an example of transfer function of a MOSFET power amp with error correction. Again, gain as a function of input voltage and idle current. The higher the idle current, the more linear the gain is, though there is no optimum. Just compromises.
Hi Ed,
Nor do I. As my historical posting might suggest ...
Nor do I. As my historical posting might suggest ...
Wasn't this what you just said? I was responding to this.
Right, usually speakers are the most THD contributor in the chain, and a good pair of speakers is not cheap.
And another BJT class AB, now high power amplifier. Please note loss of gain with amplitude.
"Listening" - avoid the ditch. It may be audible. I am missing John Curl in the debate.
"Listening" - avoid the ditch. It may be audible. I am missing John Curl in the debate.
lol!
Right now I am working on a JC amplifier. Another going on the bench next.
Beta droop Pavel? Newer devices perform much better in that regard. 2 pairs Pavel, for 250 watts output? That is daring.
Right now I am working on a JC amplifier. Another going on the bench next.
Beta droop Pavel? Newer devices perform much better in that regard. 2 pairs Pavel, for 250 watts output? That is daring.
Your point that the spectrum matters applies only when the harmonics are high enough to be audible. State-of-the-art amplifiers have the total distortion far below audibility.
Ed
... and this is where the whole 26mV started with.
Jan
Jan
Exactly Jan!
Hi Ed,
Distortion spectra is always important. I just serviced a large tube power amp, now I have a solid state one on the bench. Both very good designs and brand names. Some distortion is audible no matter their frequency, although in the peak of your ear sensitivity is the worst for obvious reasons. How the distortion products relate to the stimulus that created them is important too.
We can safely say, the less distortion in all it's forms, the better. That is what I strive for and clients have agreed with me for decades.
Hi Ed,
Distortion spectra is always important. I just serviced a large tube power amp, now I have a solid state one on the bench. Both very good designs and brand names. Some distortion is audible no matter their frequency, although in the peak of your ear sensitivity is the worst for obvious reasons. How the distortion products relate to the stimulus that created them is important too.
We can safely say, the less distortion in all it's forms, the better. That is what I strive for and clients have agreed with me for decades.
An interesting plot might be distortion v. part cost v. energy consumption for various topologies. I suspect CFP output stages win here due to better thermal performance (lower energy for typical use) and lower PSU voltage requirements (smaller PSU parts). Class A obviously loses.
Since the bias current runs through both resistors, one only one bias can be adjusted to 'optimal', the other then isn't.
But apparently, equalising the gain/local feedback for the two halves lowers the distortion more than the increase from non-optimal bias on one side.
Is the 0.2 ohms on the NPN?
Jan
But apparently, equalising the gain/local feedback for the two halves lowers the distortion more than the increase from non-optimal bias on one side.
Is the 0.2 ohms on the NPN?
Jan
Dear Pavel,
I very much like your claim to question "recipes" and not just measurement alone.
Creating a bird's wing diagram is a very simple task with MC12. But regardless of gm-doubling, double cross and Nicks class ABBA, looking at the static transmission behavior of the OPS in the OL & CL case is fantastic.
I am very curious to see how this information thread will develop.
You are looking at the whole thing and not just the ordinary push-pull output stage.
HBt.
I very much like your claim to question "recipes" and not just measurement alone.
Creating a bird's wing diagram is a very simple task with MC12. But regardless of gm-doubling, double cross and Nicks class ABBA, looking at the static transmission behavior of the OPS in the OL & CL case is fantastic.
I am very curious to see how this information thread will develop.
You are looking at the whole thing and not just the ordinary push-pull output stage.
HBt.
Please give more examples of “yes” (type A) and “no” (type B). It seems that the topic of optimal V_BE is dependent on the overall topology (and perhaps also on the dimensioning), so sometimes yes and sometimes no.
Once again as a reminder:
everything revolves around the transfer (cross over) distortion of a trivial push-pull output stage. And how the snot affects the global nfb. Or to put it the other way round: what is a feedback system ... and so on?
The whole subject of “bias or quiescent current” and "switch on" and "switch off" is honestly a hackneyed, endless topic.
But to look at the scenario with a little more serenity and, above all, to investigate it again and again in practice using measurement technology ...
wow!
Douglas Self's recommendations for the “blameless topology and design rules” apply without restriction. Point taken.
kindly,
HBt.
Last edited:
Let me reference Nelson Pass' document. https://www.firstwatt.com/wp-content/uploads/2024/01/art_the_square_law.pdf
He mentioned putting degenerating resistor could help flat out the output resistance of the bjt push pull output stage, but he did not say how big the resistor or how much the bias is.
That also confirms that there would be an optimal degenerating resistance by given the bias current, although we don't know how to calculate the value.
PS: From the chart, I can see the degenerated resistance is about twice of the no degeneration. Can I say the added resistor is about Vt/Ibias?
He mentioned putting degenerating resistor could help flat out the output resistance of the bjt push pull output stage, but he did not say how big the resistor or how much the bias is.
That also confirms that there would be an optimal degenerating resistance by given the bias current, although we don't know how to calculate the value.
PS: From the chart, I can see the degenerated resistance is about twice of the no degeneration. Can I say the added resistor is about Vt/Ibias?
The output impedance of the pair with degeneration is the same as one emitter resistor.
Ed
Ed
Simulation of BJT devices at low currents is usually far from accurate. My main concern about the low theoretical optimum current is "thermal memory" causing temporary under bias. A short term over bias if a device has got hot is going to cause far less distortion rise. Under bias would happen if the Vbe multiplier compensation got warmed and then loud music stopped, allowing the outputs to cool.
I suspect the CFP configuration is less vulnerable to this.
I suspect the CFP configuration is less vulnerable to this.