Whatever you may need or not, unless there is zero THD there is nonlinearity, which enables modulating HF noise with the audio signal.
Who said they do?
If some of them do, sometimes, what is the relevance of it to the issue of HF noise modulating the audio signal?
That's right, but it isn't necessarily bad. If you're using the LED as a reference, the voltage doubles, but the noise voltage adds as power, so the noise voltage increases by 3dB because the noise between the two devices is uncorrelated. So you gain a 3dB advantage in signal to noise ratio by stacking two of them. For example, a common red LED has a Vf of 1.6V or so. The noise voltage is roughly 2uV (20Hz-20kHz). So the s/n of a single device is 20 log (1.6/2x10e-06) = 118dB. Stack two of them and the noise voltage is sqrt([2x10e-6]^2 + [2x10e-6]^2) = 2.83uV, but the reference voltage is now 3.2V, for a signal to noise of 20 log (3.2/2.83x10e-06) = 121dB.
There is no 'the reference' in SY's statement. He's saying to use many. Here's a thought experiment -
Say you need a 10V reference and you wanna use TL431s. I can see two ways to go. First use resistors in 3:1 ratio to multiply up the 2.5V reference to 10V, using a single TL431. Second, use 4 TL431s in series with no resistors. The DC voltage is the same but the noise will be 6dB better in the second case. In fact the DC voltage will be more accurate on average in the second case because of lack of resistor tolerance and the power of statistics.
Its possible to mitigate the noise in the first case by bypassing the upper resistor with a suitable cap to reduce the noise gain.
Because HIGH THD is your personal number. "They" prefer LOW ENOUGH THD instead of sacrifice of other parameters they value in order to get EVEN LOWER THD ON 100 HZ ON ONSET OT CLIPPING.
LOUD ENOUGH FOR TIN/LEAD EARS?
You should know him by now, that he never gives hard facts.
Doing so would in this case reveal the complete absence of any PSRR in his input stage.
After all, his business model is based on "Tales Of Mystery And Imagination"
regards, Gerhard
(with due respect for Alan Parsons)
Last edited:
Really, just about any cascode stage. And that's OK, it's a difference in design philosophy- some like to make high PSR stages to simplify power supply requirements, others don't bother about PSR and design very low noise power supplies. It depends on where you prefer your complication. So I hope John will not feel embarrassed about coughing up a number.
I like to think in terms of nV/rt Hz. Of course, what is important depends on the operating level and the power supply rejection of the circuit.
YES, it is true that the resistor fed folded cascode circuit (my favorite) has minimal power supply rejection, so I have to build quiet power supply follower stages (after first using a 317/337 combination) of about 5nV/rt Hz or less. My latest design is perhaps 1nV/rt Hz, but it is VERY expensive and exotic. Not for amateurs.
A real 'breakthrough' is the new Bybee power supply filter. It will get you down to about 20nV/rt Hz, which is pretty good, certainly better than my HP power supplies.
By comparison, perhaps a LM317 might be 100nV/rt Hz or more, because it does not have a cap bypassed voltage reference. I can't be sure of this, because I can't find a noise rating that I can use directly from the data sheet.
YES, it is true that the resistor fed folded cascode circuit (my favorite) has minimal power supply rejection, so I have to build quiet power supply follower stages (after first using a 317/337 combination) of about 5nV/rt Hz or less. My latest design is perhaps 1nV/rt Hz, but it is VERY expensive and exotic. Not for amateurs.
A real 'breakthrough' is the new Bybee power supply filter. It will get you down to about 20nV/rt Hz, which is pretty good, certainly better than my HP power supplies.
By comparison, perhaps a LM317 might be 100nV/rt Hz or more, because it does not have a cap bypassed voltage reference. I can't be sure of this, because I can't find a noise rating that I can use directly from the data sheet.
We are talking about noise in the low RF range causing modulation of a stage inside of a complete amplifier. The only way to increase linearity is by device selection or gain degeneration (semi-intrinsic feedback.) Lower stage gain is the actual result and lower loop gain and thus an increase overall THD. So there may actually be a trade-off between lower THD and lower noise IM!
So reducing power supply noise is a better solution for some designs than increasing the stage's power supply noise rejection.
A bit more respect for JC would be in order.
What a denigrating comment. So disrespectful of John and his achievements. I'm shocked about this tone and behaviour.
What does PSRR have to do with the marketplace or the cutomers John working? If he's circuits need low noise supplies, and he knows the hows and whys, so what? He is a really sincere character. Hope you have as much success with your creations as he has.
- Status
- Not open for further replies.