Provided there is adequate PSRR and feedback, and provided that there remains sufficient supply voltage to cover the signal voltage then no. With inadequate PSRR, and insufficient feedback, then power supply variations can end up in the load but this shows poor design. As I said, the concept of signal path is not always helpful as it seems to lead people astray.
Member
Joined 2009
Paid Member
Member
Joined 2009
Paid Member
you have misunderstood me - your post no.24 accurately points out that good PSRR and feedback (often the good PSRR comes from use of nfb) removes a lot of the influence of the power supply caps from the current that flows through the load. All I am saying is that with simple Class A you often have very poor PSRR and often no feedback (except degeneration perhaps) and therefore, the power supply caps may be more relevant.
Last edited:
It's on the data sheet, quiescent device current, IDD max at 25°C.
Typical values are listed as 0.01µA. Furthermore, the
IDD spec is for all six inverters in parallel (see fig.15).
At best we have a class AB amplifier with a bias of 0.01µA, so if he is really drawing no output current (less than 0.01µA) then it might be class A.
Typical values are listed as 0.01µA. Furthermore, the
IDD spec is for all six inverters in parallel (see fig.15).
At best we have a class AB amplifier with a bias of 0.01µA, so if he is really drawing no output current (less than 0.01µA) then it might be class A.
Sorry, I was allowing my prejudice against zero feedback SE Class A to show through in my remarks. I regard such a circuit as OK for a 50-year old wireless set, but not for my hi-fi. When designing for music reproduction one of the aims should be to ensure that PSU caps are not in the signal path, although I realise that not everyone sees things that way.
They sort of "materialize" the virtual (+) input of the operator used as an amplifier.
Since they belong to the same chip, they share the same threshold voltages and can be used as reference/virtual ground.
Here, it is necessary, since the OP doesn't want a coupling capacitor.
They also provide a resonably low impedance reference, again without using bypass capacitors.
@ CMOS amp schematic and quiescent current:
When reading datasheets, ALWAYS check the conditions the parameters were measured at. All CMOS gates have a certain level of cross-conduction when the inputs are biassed at Vcc/2, one could even argue it is necessary. Some have enought that prolonged exposure to such conditions can even lead to overheating. Others can get to the same conditions with marginal power supplies that result in P and N channel MOSFET's treshold voltages overlapping.
In the case of the particular device shown, measurement of supply current vs input voltage shows an orders of magnitude increase compared to normal logic levels. This information is, however, rarely given in datasheets.
@Power supply caps in the signal path
I would agree with DF96 that the idea of 'signal path' is one of the more counter-productive ones when it comes to proper understanding of how circuits work. It is evident in this argument as well - the capacitor is not there on it's own, it's a part of the power supply, therefore if anything is in the 'signal path' then it would be the whole power supply, simplifying it down to merely the cap is oversimplifying, even if most of the time the amp supply current flows through it. But, whithout the other part where the cap is charged, the whole device would not have it's intended function, so it simply cannot be disregarded. For one thing PSRR would be meaningless as there would be nothing to reject without the ripple current.
That being said, caps do matter. Even in an amp with lots of NFB making them largely invisible. Why? Well:
0) Think in dB. This gives a better idea of how you would hear things, rather than see them on a scope.
1) Think parasitic components, some of which are non-linear, and others make up resonant circuits. Since we have nonlinear elements that alo may exxagerate certain frequencies while attenuating others, through which two currents flow that should ideally have no common paths, we have a source for frequency and amplitude dependant intermodulation.
2) Think constants are not constant. PSRR gained from filtering is not constant because filters have turnover frequencies and components are not ideal. PSRR gained from applying NFB is not constant because open loop gain is not constant (neither in the frequency nor in the amplitude domain!) so PSRR is not constant as well.
3) Finally, think intermodulation. Both the audio signal and the ripple current are rather complex waveforms in reality and contain lots of harmonics, most of which may not be audible. However, intermodulation creates all possible sums and DIFFERENCES of all present frequencies so non-harmonically related but audible intermodulation products can appear in the output of amps (by various mechanisms, mostly related to improper understanding of the concept of a ground, but also through the power supply lines), giving 'the capacitors' a sonic signature. Note quotes - it's actually the power supply that has a sonic signature in this case, the capacitors are just one part of it, so yes, in most cases there will be differences in the signature if the caps are changed, but that's really not even 30% of the story. If you think caps are complicated, try transformers and the interaction between the two through rectifiers.
All that being said, a properly designed 'simple' cap filter PSU can be made that imparts very little of it's own 'sound' to be truly audible (other conditions will influence the sound more), but it's sometimes done in unexpected ways. In particular once grounds, current loop induction and transformer stray fields are managed, it often becomes obvious that the mechanisms of 'PSU capacitor sound' are not where one expects to find them, in the capacitors themselves. It should be noted that large caps indeed do have their own set of problems, at large sizes and currents most of these become related to mechanics (ever heard music from your PSU caps when testing an amp on a dummy load?). It should also be noted that the best possible PSU design may not be the one employing components as close as possible to ideal ones. In a lot of cases a well designed PSU is well designed because it works well with standard caps, and premium ones may well actually degrade it's performance. It's all down to carefully managing the non-ideal components so that their inadequacies work for you rather than against you.
When reading datasheets, ALWAYS check the conditions the parameters were measured at. All CMOS gates have a certain level of cross-conduction when the inputs are biassed at Vcc/2, one could even argue it is necessary. Some have enought that prolonged exposure to such conditions can even lead to overheating. Others can get to the same conditions with marginal power supplies that result in P and N channel MOSFET's treshold voltages overlapping.
In the case of the particular device shown, measurement of supply current vs input voltage shows an orders of magnitude increase compared to normal logic levels. This information is, however, rarely given in datasheets.
@Power supply caps in the signal path
I would agree with DF96 that the idea of 'signal path' is one of the more counter-productive ones when it comes to proper understanding of how circuits work. It is evident in this argument as well - the capacitor is not there on it's own, it's a part of the power supply, therefore if anything is in the 'signal path' then it would be the whole power supply, simplifying it down to merely the cap is oversimplifying, even if most of the time the amp supply current flows through it. But, whithout the other part where the cap is charged, the whole device would not have it's intended function, so it simply cannot be disregarded. For one thing PSRR would be meaningless as there would be nothing to reject without the ripple current.
That being said, caps do matter. Even in an amp with lots of NFB making them largely invisible. Why? Well:
0) Think in dB. This gives a better idea of how you would hear things, rather than see them on a scope.
1) Think parasitic components, some of which are non-linear, and others make up resonant circuits. Since we have nonlinear elements that alo may exxagerate certain frequencies while attenuating others, through which two currents flow that should ideally have no common paths, we have a source for frequency and amplitude dependant intermodulation.
2) Think constants are not constant. PSRR gained from filtering is not constant because filters have turnover frequencies and components are not ideal. PSRR gained from applying NFB is not constant because open loop gain is not constant (neither in the frequency nor in the amplitude domain!) so PSRR is not constant as well.
3) Finally, think intermodulation. Both the audio signal and the ripple current are rather complex waveforms in reality and contain lots of harmonics, most of which may not be audible. However, intermodulation creates all possible sums and DIFFERENCES of all present frequencies so non-harmonically related but audible intermodulation products can appear in the output of amps (by various mechanisms, mostly related to improper understanding of the concept of a ground, but also through the power supply lines), giving 'the capacitors' a sonic signature. Note quotes - it's actually the power supply that has a sonic signature in this case, the capacitors are just one part of it, so yes, in most cases there will be differences in the signature if the caps are changed, but that's really not even 30% of the story. If you think caps are complicated, try transformers and the interaction between the two through rectifiers.
All that being said, a properly designed 'simple' cap filter PSU can be made that imparts very little of it's own 'sound' to be truly audible (other conditions will influence the sound more), but it's sometimes done in unexpected ways. In particular once grounds, current loop induction and transformer stray fields are managed, it often becomes obvious that the mechanisms of 'PSU capacitor sound' are not where one expects to find them, in the capacitors themselves. It should be noted that large caps indeed do have their own set of problems, at large sizes and currents most of these become related to mechanics (ever heard music from your PSU caps when testing an amp on a dummy load?). It should also be noted that the best possible PSU design may not be the one employing components as close as possible to ideal ones. In a lot of cases a well designed PSU is well designed because it works well with standard caps, and premium ones may well actually degrade it's performance. It's all down to carefully managing the non-ideal components so that their inadequacies work for you rather than against you.
On the topic of PSU caps... and the load return current flowing "through" transformer secondaries and/or bridge rectifiers and PSU caps etc.
I feel this point needs pressing home
Consider the following text book example. A 741 driving a 390 ohm resistor.
The input is adjusted to give around 4 volts pk (8 volts pk-pk) over the 390 ohm. The input is either a squarewave or a DC voltage input. That allows approximately 10 ma to flow in the load.
A scope shows a perfect output squarewave as expected or a steady DC potential depending whether a squarewave or DC voltage is used as an input.
The wires to the PSU are now suddenly and unceremoniously snatched from the circuit leaving the 4700uF caps to briefly power the amplifier.
The output shows no change at all until such time as the rails collapse to far to allow the circuit to operate. That applies to either the squarewave output or a steady DC output supplying current to the load.
My definition of this is that the load current does not pass through the reservoir caps, it certainly can not pass via "the transformer and bridge" because they are not in circuit.
I hope that makes it clearer to those that still think that in some way the load current is in series with the PSU, its caps, the transformer etc etc. It is not.
I feel this point needs pressing home
Consider the following text book example. A 741 driving a 390 ohm resistor.
The input is adjusted to give around 4 volts pk (8 volts pk-pk) over the 390 ohm. The input is either a squarewave or a DC voltage input. That allows approximately 10 ma to flow in the load.
A scope shows a perfect output squarewave as expected or a steady DC potential depending whether a squarewave or DC voltage is used as an input.
The wires to the PSU are now suddenly and unceremoniously snatched from the circuit leaving the 4700uF caps to briefly power the amplifier.
The output shows no change at all until such time as the rails collapse to far to allow the circuit to operate. That applies to either the squarewave output or a steady DC output supplying current to the load.
My definition of this is that the load current does not pass through the reservoir caps, it certainly can not pass via "the transformer and bridge" because they are not in circuit.
I hope that makes it clearer to those that still think that in some way the load current is in series with the PSU, its caps, the transformer etc etc. It is not.
Attachments
your point is that because the system can continue as long as there is sufficient charge/voltage in the caps, then the caps are not in the signal path...
Phyy....on one hand you can argue and even prove that...on the other hand that is what creates engineering solutions...as you the can calculate a sufficient size and ESR requirement....
But practice shows that the power supply may just be the single most important part of an amplifier .. the engine room so to speak...and there is huge difference in performance with different caps and transformers...they all show their signature...regardless of the amplifier module attached to it...Kind of hard to explain then...
Phyy....on one hand you can argue and even prove that...on the other hand that is what creates engineering solutions...as you the can calculate a sufficient size and ESR requirement....
But practice shows that the power supply may just be the single most important part of an amplifier .. the engine room so to speak...and there is huge difference in performance with different caps and transformers...they all show their signature...regardless of the amplifier module attached to it...Kind of hard to explain then...
Member
Joined 2009
Paid Member
Mooly,
It's a good thought experiment - Einstein was a big proponent of them so you're in good company! But I think your conclusion is incomplete. Actually, being able to remove the rest of the power supply except the caps and yet still play music, for me, results in a confirmation that the caps must very much be in the current path - otherwise the darn thing would stop playing immediately on the creation of an open circuit condition for the signal. Think about it - the signal current through the load must flow in a circuit, current is conserved around a loop - the psu caps complete that loop. The rectifier diodes are conducting only briefly even when you leave them in the circuit so the transformer rarely 'sees' the signal current - it's the caps that provide a good ac path for the signal current as it traverses the power supply.
Not to make things more confusing, I find that people use different meaning for the words 'signal path'. For some it means tracing out the Voltage signal (wrt 'ground') from the input connector to the output connector. This is very different from considering the path of Current, which flows in loops, often many separate loops throughout the chain of amplification. In the former definition the psu caps are out of the picture, but in the 2nd definition they are in the picture.
It's a good thought experiment - Einstein was a big proponent of them so you're in good company! But I think your conclusion is incomplete. Actually, being able to remove the rest of the power supply except the caps and yet still play music, for me, results in a confirmation that the caps must very much be in the current path - otherwise the darn thing would stop playing immediately on the creation of an open circuit condition for the signal. Think about it - the signal current through the load must flow in a circuit, current is conserved around a loop - the psu caps complete that loop. The rectifier diodes are conducting only briefly even when you leave them in the circuit so the transformer rarely 'sees' the signal current - it's the caps that provide a good ac path for the signal current as it traverses the power supply.
Not to make things more confusing, I find that people use different meaning for the words 'signal path'. For some it means tracing out the Voltage signal (wrt 'ground') from the input connector to the output connector. This is very different from considering the path of Current, which flows in loops, often many separate loops throughout the chain of amplification. In the former definition the psu caps are out of the picture, but in the 2nd definition they are in the picture.
Last edited:
On the topic of PSU caps... and the load return current flowing "through" transformer secondaries and/or bridge rectifiers and PSU caps etc.
I feel this point needs pressing home
Consider the following text book example. A 741 driving a 390 ohm resistor.
The input is adjusted to give around 4 volts pk (8 volts pk-pk) over the 390 ohm. The input is either a squarewave or a DC voltage input. That allows approximately 10 ma to flow in the load.
A scope shows a perfect output squarewave as expected or a steady DC potential depending whether a squarewave or DC voltage is used as an input.
The wires to the PSU are now suddenly and unceremoniously snatched from the circuit leaving the 4700uF caps to briefly power the amplifier.
The output shows no change at all until such time as the rails collapse to far to allow the circuit to operate. That applies to either the squarewave output or a steady DC output supplying current to the load.
My definition of this is that the load current does not pass through the reservoir caps, it certainly can not pass via "the transformer and bridge" because they are not in circuit.
I hope that makes it clearer to those that still think that in some way the load current is in series with the PSU, its caps, the transformer etc etc. It is not.
Actually, it is, but not all of the time, i.e. ALL of the load current does not pass through the caps. If we were talking class A, then all of the AC load current would be split differentially and pass through the caps. So, the caps are not strictly in series with the load in this example, just the same as they are not strictly in series with the rectifier and transformer - i.e. not all of the time. In the case of the rectifier side, they never are always in series, except maybe if you used a full-wave 3-phase rectifier and heavily loaded the power supply. However, this 'intermittent' connection and rather unfavorable combination of high demands from a component that still has to be relatively cheap, are in fact what creates 99% of the problems (and let me make it clear again, these problems can be lessened to a very large degree without resorting to exotic components).
The example shown however leaves a lot to be desired, if it is to model real-world conditions in a power amp driving a typical speaker. First of all, a speaker is not a resistor. It's reactances are capable of returning current back into the power supply, For reasons I will not go into here, this is quite unusual for a regular power amp but class D amps are often designed as a bridge circuit just to mittigate this problem, otherwise you get a power supply 'pumping' problem, i.e. the load charges the caps up over the normal PSU voltage. In a typical transformer + rectifier PSU no portion of load current ever goes backwards into the transformer as the rectifier will not alow this. However, since the actual power comes from the transformer, the current in the transformer windings is clearly modulated by the amp's output current (and in fact you can often hear the music coming from the transformer at high powers into a dummy load).
Secondly, in a power amp, PSU ripple current and load current are usually the same order of magnitude, even for relatively low output power of a couple of W. The thermodynamic version of Murphy's law states that things only get worse under pressure, and in our case, they get worse with more current and therefore increased ripple voltage. Milliamps are rarely a problem, amps usually are a problem.
Thirdly, the class of amplification makes quite a lot of difference, and in fact heavily influences the design of a PSU, including cap selection. Class AB roughly puts half-wave rectified load current through the PSU caps, while there is no charging going on from the rectifier side (if there is, things get sligltly more complex, ans in the moments of comutation, either from the amp or the rectifier side, they get really complicated). The complex current and voltage waveforms indeed are distorted from their theoretical values and shapes due to the action of various parasitic components in the cap, which at the usual currents present, cannot be disregarded any more.
However, the experiment above tends to show some interesting things if applied to a typical, class AB amp, especially if it has very large reservoir caps. It does take some doing as a regular power amp will usually only work for several (tens of) seconds once the plug is pulled, but I've been present at some experiments where there was an audible improvement when the plug was pulled! i.e. the charging pulses from the rectifier were removed from the 'mix'. Not surprisingly, I have yet to hear any such thing from a preamp where currents are a few orders of magnitude smaller, yet caps are not proportionally smaller (I am assuming well designed PSUs on all equipment). Although it is difficult to measure with widely available equipment, it is sometimes possible to hear improvement when a small value resistor (up to a few ohms) is added in series with the rectifier, which often drops the ripple current quite significantly, especially with toroidal transformers - with no change of the PSU caps. Granted, many mechanisms are at play here, but that's really the point - the caps cannot be considered in isolation for most normal power supply implementations. In situations where they CAN be considered as a lumped component in series with the signal, their influence, while being present, is far less insidious. The problem is really that they tend to be considered as 'the' component in a PSU where the story can be much more complex, sometimes just on the account of a few details, such as for example using fast diodes just because someone said it will sound better
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.