Stu,
I have been working on the issue of power supply noise. Surprisingly small amounts of noise at higher than expected frequencies (above 1 MHz.) raise the audio band noise floor out of proportion to the RF noise level.
This is very device and topology dependent.
In a differential input, common emitter voltage amplifier stage with emitter follower output designs, the gain stage acts as a common base amplifier to the rail noise.
I am still at the very early stages of this. I have shown power supply noise versus basic design issues on this thread a bit earlier.
I may go for a patent on some of the more interesting things I have found and so have not mentioned those areas.
It should be possible to simulated the amplification of power supply noise over a wide bandwidth into a common base amplifier followed by a much slower emitter follower for some insight into the issue.
As to an actual number, I would aim for -190 db up to the low RF range!
ES
Is this supported by any independent measurement that you could show or link? The results published in Stereophile do not prove that for example JC-1 would be the lowest noise design in power amplifiers.
Why not -200?
In any case, I'd like to know from John what he thinks is reasonable power supply noise for his market and that he generally achieves, expressed either as a noise voltage density (preferably with some notion of the noise corner frequency) or total noise over the audio bandwidth. Me, I design circuits that have very little power supply criticality and I don't have to have a good story to tell, so my life is easy.Why not -200?
But have you tried a -190db re rail voltage power supply? In a double blind test?
When you use tubes the problems go way down. There are no complementary forms to provide the common noise path. Also the bandwidth doesn't vary greatly by stage.
On the other hand you can't quite go as quiet as a cooler solid state gizmo.
But have you tried a -190db re rail voltage power supply? In a double blind test?
Why would I need to? I already can't hear any noise from an inch away from my speakers. And I don't listen to them from an inch away!
It is not just the noise you can hear. It is the noise modulating a desired signal followed by a low pass filter that can also be an issue.
Of course sticking your ear an inch away from some of the loudspeakers I play with is a very dangerous thing to do. A 30% efficient loudspeaker powered by a kilowatt (real power) amplifier can cause instant hearing loss if there is a glitch at that time!
But you have an interesting point, it is your ears that are the judge of your design. Numbers are nice for rough comparison or more importantly internal refinement.
Of course we don't let John off as easily!
SY said:
I could use two. How does that help?
Only thing I can think of is that it would make the current sources less sensitive to temperature change of the transistors, with the larger degeneration resistance. I only used an LED in a current source for the shunt part, not in the series current source. I could get rid of that LED too, using a different type of current source.
Even with very low THD there is nonlinearity. No circuit with amplification is perfectly linear with zero THD.
I'm sorry my tin/lead ears have a hearing threshold...so I don't need a zero THD circuit. But why do audiophiles obviously prefer high THD circuits with consequently more modulation than less?
Okay, I don't see what you mean. For the LED the DC voltage doubles, and the dynamic impedance also doubles, right? That would be bad in my circuit.
I'm sure, I'm not understanding something, but I don't have time to post my circuit right now, and I'm not entirely anxious to have it needlessly raped on the message board.
Okay, I will try to post it later.
- Status
- Not open for further replies.