suppression of interference
It always makes sense to suppress interference at source rather than plaster over the cracks.
Consider the PSU alone.
It's job is to supply DC to a client.
Suppressing artefacts at any frequency should be done at the PSU.
Consider the Amplifier as a modulator of a DC supply into a reactive load.
Interference this time are the glitches that the amplifier places on the DC line when trying to drive near instantaneous current changes into that reactive load.
Decoupling right next the the devices that create the glitches is equivalent to suppressing interference at source.
We cannot and must not forget that both ends of the supply and client circuits needs best affordable performance to achieve great results.
We need PSU that are as close to DC as affordable.
We need amplifier that keep the DC line clean while trying to do their modulating of DC purpose.
We must not put all our eggs in one basket and disregard the consequences of omitting interference suppression at (some) source in any system.
It always makes sense to suppress interference at source rather than plaster over the cracks.
Consider the PSU alone.
It's job is to supply DC to a client.
Suppressing artefacts at any frequency should be done at the PSU.
Consider the Amplifier as a modulator of a DC supply into a reactive load.
Interference this time are the glitches that the amplifier places on the DC line when trying to drive near instantaneous current changes into that reactive load.
Decoupling right next the the devices that create the glitches is equivalent to suppressing interference at source.
We cannot and must not forget that both ends of the supply and client circuits needs best affordable performance to achieve great results.
We need PSU that are as close to DC as affordable.
We need amplifier that keep the DC line clean while trying to do their modulating of DC purpose.
We must not put all our eggs in one basket and disregard the consequences of omitting interference suppression at (some) source in any system.
Andrew,
I think it is more subtle than that. Local decoupling for higher freqs is indeed a good idea, if only to avoid the problems with the inductance of the wire going back to the supply!
But with for instance a drum sole, you need real capacitance and 10.000uF local decoupling isn't practical. And at any rate, those lf high current half-waves WILL modulate the DC, short of using a fully regulated high power source.
But I am not comfortable with your first statement:
If 'plastering over the cracks', whatever that means in practise, leads to cleaner amp output (less supply ripple breakthrough), we're done. There's always a tradeoff in supply complexity for clean DC and amp complexity for high supply rejection ratio. It's your job as a designer to select the best compromise in terms of cost, performance, complexity, reliability etc.
jd
local decoupling is mandatory to help attenuate the line variations effect that will happen and will affect adjacent stages if not done.
"Papering over" implies a non optimal correction that does not address the problem and it's physics. There must exist a more effective solution that is probably more economic if the physics is understood and the problem addressed directly.
My own preference for using +-20mF for an 8ohm amplifier is probably just such a case of papering over. It seems to give the result I want to hear. But is it optimal? Probably not and the reason I cannot make it more optimal is that I don't understand the physics that requires the change or optimisation.
I guess your design style is a little more "conventional" than my own. Each to his own, there's a lot to be learned from other methodologies. I'm no longer someone who gets any particular kick out of considering a component in isolation, I really enjoy exploring synergies between say an amp and its power supply, or an amp and the speaker it drives. That's one reason I'll never go back to using passive speakers.
Yes, but what of artifacts which occur enroute between the PSU and amplifier? Such as RF pickup on supply leads? Can these really be suppressed in the PSU? I just pick on this one example as a way of saying that from my perspective, the component by component approach to design leaves gaps to be filled.
OK, looks like a false alarm from me. Rev 3 seems to have the right
diode bridge connection in the negative part of the PSU.
There was an incorrect older version going around for a while.
I'd suggest it depends on the impedance of the lines.
Surely the supply lines are very low in impedance ref. the PSU ground and the chassis ground.
If the above is true then RF in suppressed first by the chassis, then by the low loop area of the lines, then by the low impedance, at RF, of the lines. Twisted lines must have some capacitance. I suspect that line capacitance will be effective to VHF, maybe UHF.
This to me is systems thinking.
I was using the "think alone" analogy to draw attention to the fact that each end is actually a source of interference and that source area becomes the first port of call to expend resources in attenuating the problem.
Yes, the first statement is true. The second one is only true at low frequencies - since we're talking RF here its basically false.
I can't quite follow what you're saying here. But I think you're not quite clear on how RF affects twisted lines. The RF won't really affect the differential mode signal but it will tend to mess around with the common mode signal. So capacitance between the twisted lines is not an issue - both lines will be wobbling up and down together in response to the RF field.
Hi,
I am not too good at this difference between common mode and differential mode on the PSU lines.
Could you explain the mechanism of the RF and common mode on the PSU lines and what happens when this common mode interference arrives at the amplifier's power input?
I am not too good at this difference between common mode and differential mode on the PSU lines.
Could you explain the mechanism of the RF and common mode on the PSU lines and what happens when this common mode interference arrives at the amplifier's power input?
Its similar to the difference between a common mode and a differential mode signal going into an opamp's + and - inputs. Differential mode means the + and - signals move in opposite directions, common mode means they both move in the same direction.
Happy to have a try, if it doesn't work out well, please get back to me.
PSU lines are generally the longest wires within any design. The longer a wire is, the more effective it is as an antenna over a broad range of frequencies. Longer wires receive down to lower frequencies.
With a twisted pair, the two wires are so tight together that except at the very highest frequencies, they look to be a single wire to RF. So there will be a common mode RF signal on them.
Suppose these two twisted wires are the + and - supply to the chip amp. The chip amp's PSRR falls at higher frequencies so it has no natural immunity to this interference. However I haven't found this low PSRR to be the source of an audible problem in itself. I have though noticed that RF causes problems when applied to the input stage of a bipolar opamp or chipamp (not so much with JFET input opamps). It can get to the input stage from the power supply wires via the decoupling caps, if those caps are from + to GND and from - to GND. The GND connection of course has some inductance so won't short out this RF, which can get through to either the + or - input pins via any passive components connected to the same GND as the decouplers.
The upshot of this is - if you want to decouple from the rails to GND, make sure that ground is a separate ground all the way back to the star point. If not, your polluted signal ground will induce RF into your ICs with negative impact on your sound. This is audible in the form of sibilance but that is just the easiest audible effect to spot - it tends to reduce dynamics and soundstage depth too.
A dual polarity supply is +ve and -ve supplies.
This is a three wire supply.
The three wires and the enclosing chassis will all have that common mode RF imposed on them.
Does this situation fit your model?
This is a three wire supply.
The three wires and the enclosing chassis will all have that common mode RF imposed on them.
Does this situation fit your model?
Yes. How much CMRF appears on the chassis depends on the quality of its grounding. With a long wire back to safety earth, which looks like an inductor, the grounding will be fairly ineffective at RF even though its good enough at mains frequencies for safety purposes. I believe this RF grounding is the main reason people say that different mains cables sound different.
I'd use a bit more copper if it were me and make the star more like a real star by thinning the tracks as they get closer together. Oh, also I have no idea what kind of loading this might get, but the 0.47R resistors look a little close to the caps - a bit more space between them would help cap lifetime.
Other than those minor nit-picks it looks fine.😀
Hi Abraxalito, Thanks for your valuable guidance.🙂🙂
0.47R mounted vertical to reduce space
Trace width increased
More prominent STAR GROUND
trust this time my efforts are in right direction.🙂
Actually not in the case of the star ground - it looks even less like a star now, more like a fill.😀 Did you read what I wrote earlier? Thin (means neck) the tracks as they get towards the centre of the star so they don't all join together before the central point.
Hi Abraxalito,🙂
There are 3 types 1st is half done for your guidance & rest are for your comment.
Thanks😀😀
There are 3 types 1st is half done for your guidance & rest are for your comment.
Thanks😀😀
I like the first one best as the star point is (potentially when finished) closest to a point. You just need to neck down (to say 1/3 - 1/4 their width) the four tracks uncompleted so they don't join up before the star.
- Status
- Not open for further replies.