Bob Cordell said:
Hello Bob,
My concern with very high power supply reservoir caps is that the duty cycle of the diodes will get shorter and shorter. The charge current pulses will have quite a lot of higher harmonics. These can lead to measurable high-order line harmonics in the amp. Not all amps have very good PSRR.
For me there really isn't any advantage above 20kuF or so, and the disadvantage (and space and cost!) start to come into play.
Jan Didden
janneman said:
I think you will find that the duty cycle of the rectifiers (the conduction angle) tends to be more dominated by the impedance looking back into the transformer, including things like primary and secondary copper resistance, leakage inductance, and even ac line resistance. Given all of these factors, the conduction angle may not get much smaller with 100,000 uF than with 20,000 uF.
But, nevertheless, your concern about RF noise due to small rectifier conduction angle is legit. Keep in mind that you can always increase the conduction angle by adding some resistance or inductance in series ahead of the reservoir capacitor.
Bob
Mr. Cordell,
How important the rectifier dioda ripple to amplifier's sonic? I've seen many schematics put C or R-C accross rectifier diode.
How important the rectifier dioda ripple to amplifier's sonic? I've seen many schematics put C or R-C accross rectifier diode.
AndrewT said:
transformer -- 500VA for each rail, 30V secondary for each rail
voltage rail -- around 40V with rectification
power output -- probably 80W
speaker -- 8 ohm single driver
I know PSUD2 is not perfect; unfortunately I don't have other tools to rely on and I think I want to rely on some tools than my silverplated ears knowing that mine are not golden.
lumanauw said:
The C or RC snubber across the rectifier diodes can help a lot with the RF impulses created by the diodes switching on and off.
How much sonic difference it makes will depend a lot on the particular design and layout of the power amplifiers. Some are much better at rejecting both conducted and radiated RF noise than others.
But it goes without saying that it can only help to reduce it by using snubbers in the first place. You would also prefer not to conduct or radiate the RF diode switching noise to other components nearby, either.
Cheers,
Bob
Bob, I've heard though that the boutique ultra-fast, "soft recovery" diodes should not necessarily have snubbers added. Something about the way they operate already mitigating a lot of the radiated stuff.
This seems to be one of those "trial and error" areas. If you
are blessed (or cursed) with having to pass high frequency
noise emission requirements, then you will try all manner of
permutations to see what works best. Whatever gets the
harmonics lower than the limits gets the nod....
😎
are blessed (or cursed) with having to pass high frequency
noise emission requirements, then you will try all manner of
permutations to see what works best. Whatever gets the
harmonics lower than the limits gets the nod....
😎
Recently I tried to make selfoscilating classD amp. It drives me bananas 😀
It appears that the PCB is the most important component of these high-frequency designs (target=400khz selfoscilating), not the transistor, not the capacitor.
Then I look for reference. TI's SLOD006, chapter 17 with it's 2 references papers (by Bruce Carter) seems talking about this.
The first page of this page (17-1):
"Prototype, Prototype, PROTOTYPE".
Like what NP said, it seems in this area it is "trial and error" type of work. 😀. In the general summary, it says "Prototype the circuit", not saying "make this = this way, make that = that way"
In page 17-14 I found more surprise. Above 100khz, PCB trace behaves more like inductor, not resistive. Ground plane, antenna effect, input and return path, so many things to be aware of. If we use ground plane, PCB track is actually making capacitor with ground.
Well it makes me learn alot....I'm sure the experience of making selfoscilating classD PCB teaches many things that is perfectly implementable to classAB PCB designs.
Attachments
agent.5 said:
Of course it will. That's what capacitance is all about. But it will also slow the "fall time" if you will. The net result is that the power supply with more capacitance will provide a steadier voltage under load, as any given cycle of the same magnitude will use a smaller fraction of the available capacity. I'm no expert on electronics, but this issue is a simple matter of balancing both sides of an equation.
A stepped load in PSUD is there to look at ringing behavior in the power supply. You have to be careful making conclusions about what it means for music signals.
Sheldon
Edit: Your figures are exactly what would be expected. If you drain the power supply to a given point, the more capacity it has, relative to it's impedance on the front end, the longer it will take to fill up - or drain. A very slow power supply, by the way you've defined it and presented it here is, in fact, a more constant voltage source. Take it to the other extreme. By this logic, a very fast power supply would swing right along with the signal - not what you want, I think.
Final Edit. I promise: Look at it it this way. By adopting your definition of speed, a very large battery would be a very slow supply.
Well, my conclusion was that more capacitance is NOT always better. I did not try to go to either extreme of no capacitance or so much capacitance that only NASA can afford to experiment with such setup.
Looking at the 3 sims above, it appeared that they all have pretty constant voltage and similar ripples, and the lowest capacitance setup is the fastest.
Maybe I should ask, Sheldon, which one of the three will you pick based on the graphs?
Looking at the 3 sims above, it appeared that they all have pretty constant voltage and similar ripples, and the lowest capacitance setup is the fastest.
Maybe I should ask, Sheldon, which one of the three will you pick based on the graphs?
I did not say that more capacitance is always better. Who knows what might sound best to any particular person. What I did say is that more capacitance means that you will have less ripple on the rails, either from the rectifier input or from the effects of the amp drawing power during music signals. If that's your definition of slow, then yes, slower would generally be considered better. If by "faster" you mean a "stiffer" power supply (which I would think is the usual definition, or at least a definition based on the physics), then your analysis is flawed.
Sheldon
Edit: I pick the one behind door 1.
Sheldon
Edit: I pick the one behind door 1.
Now let me ask you a question. Do you want the power supply voltage to drop rapidly when you begin drawing more current, or would you like it to stay where it was?
Sheldon
Sheldon
Sheldon said:
Concerning the next major musical passages that will draw massive amount of current, I probably want it to recover the fastest. If this means it has to drop the fastest, then so be it.
So you want it to sag the most and recover the fastest. Well, you've got yourself a nice guitar amp then.
Sheldon
Sheldon
Sheldon said:
again, it is all within a small range. Looking at my 3 sims, they don't drop all that much to begin with.
I agree, the analysis is flawed.Sheldon said:
this sounds like "do you want high dynamic overhead" or good bass?Sheldon said:
I think you have asked PSUD2 the wrong question.
Your supply with 0r6series resistance is superb at eliminating mains hum, but less good at maintaining rail voltage under load, particularly the varying load of a normal ClassAB amplifier.
Put a scope on the supply rails and measure what they are doing.
Try reducing the series resistance significantly (a much thicker wire in your choke) then your recharge will be quicker and by keeping the biggish caps you keep stiff rails under load.
BTW, I would normally recommend +-15mF to +-20mF/ch for your amp and speakers. The lower 15mF based on rail ripple and the higher 20mF based on PSU RC>=160mS. You are way above this. Is this a very high bias ClassAB amp? A low bias amp that suffers from hum with +-20mF of smoothing is faulty/badly designed.
- Status
- Not open for further replies.