If there is no signal with high enough frequency component, you don't need in any way the PSU to have low impedance at "high enough" frequencies (OVBIOUS!!)
Of course, this is an oversimplification, because ambient EMI and cross-over glitches of class B and AB circuits will always cause current to be drawn from the power supply at frequencies at least an order of magnitude higher than the fundamental signal.
Indeed, I first become aware of multiple capacitor resonance the day that I saw the supply rails of one of my class AB amplifier designs ringing at a few Mhz during each output zero-current transition. I solved it by leaving the 100uF electrolytic and removing the ceramic.
Of course, we can't expect everybody to own a storage oscilloscope capable of easily showing these events. Indeed we can't even expect the people that have such equipment to worry about what is happening at low level inside their circuits. They enjoy watching the capacitors together and imagining the rest according to what they have been told.
The best advice that I can give on supply deocupling is to take an oscilloscope and find out. The truth is in the circuits, and some capacitor mixtures will effectively increase supply impedance to 10 or even 100 ohms in the 1Mhz to 10Mhz range instead of reducing it. Any symptom of ringing means increased supply impedance.
Of course, this is an oversimplification, because ambient EMI and cross-over glitches of class B and AB circuits will always cause current to be drawn from the power supply at frequencies at least an order of magnitude higher than the fundamental signal.
Indeed, I first become aware of multiple capacitor resonance the day that I saw the supply rails of one of my class AB amplifier designs ringing at a few Mhz during each output zero-current transition. I solved it by leaving the 100uF electrolytic and removing the ceramic.
Of course, we can't expect everybody to own a storage oscilloscope capable of easily showing these events. Indeed we can't even expect the people that have such equipment to worry about what is happening at low level inside their circuits. They enjoy watching the capacitors together and imagining the rest according to what they have been told.
The best advice that I can give on supply deocupling is to take an oscilloscope and find out. The truth is in the circuits, and some capacitor mixtures will effectively increase supply impedance to 10 or even 100 ohms in the 1Mhz to 10Mhz range instead of reducing it. Any symptom of ringing means increased supply impedance.
Cortez said:
I have good expiriences with bypassing elcos with PIO's. At high freqs oil caps are quite good- I have some infos about ESL of russian oil caps- medium size have ESL 30-60nH, but some types have non-inductive construction (they were designed for filters and pulsing currents circuits) and they have ESL 6-11nH.
I was surprised about the harm the paralleled small MKT cap did, does the same apply to larger Polyester caps in the uF range (1uF ... 4.7uF) ?
I'd be interested in the results of a polyester film or tantalum cap with a 1R resistor in series to raise it's ESR, this in parallel to a large electrolytic cap. If I only had an oscilloscope :/
I'd be interested in the results of a polyester film or tantalum cap with a 1R resistor in series to raise it's ESR, this in parallel to a large electrolytic cap. If I only had an oscilloscope :/
If I remember properly, 1uF MKT caused very very slight ringing below 1Mhz with both the small 22uF electrolytic and the 2200uF low-ESR one. Bigger MKT values would probably be fine. However, with a 1 ohm damping resistor in series you can parallel almost anything, altough output impedance improvement won't probably be worth the effort.
Intuitively, you can imagine a damped system as one that has much more resistance and/or capacitance than inductance (I know that it's like comparing apples to oranges, but it's an intuitive explanation). That's why very small capacitances or very high inductances, or very low resistances usually result in systems that resonate.
Intuitively, you can imagine a damped system as one that has much more resistance and/or capacitance than inductance (I know that it's like comparing apples to oranges, but it's an intuitive explanation). That's why very small capacitances or very high inductances, or very low resistances usually result in systems that resonate.
Hi Steinchen,
try adding a smaller r (100mohm to 500mohm) in series with the bigger cap to beef up it's ESR to damp the system. Keep the low ESR of the film cap as is to maintain best hi frequency performance.
To minimise the disturbance that causes the ringing, reduce the glitch at source by using bypass ceramics at 1. all high current users and 2. lower current users with step change in current needs.
try adding a smaller r (100mohm to 500mohm) in series with the bigger cap to beef up it's ESR to damp the system. Keep the low ESR of the film cap as is to maintain best hi frequency performance.
To minimise the disturbance that causes the ringing, reduce the glitch at source by using bypass ceramics at 1. all high current users and 2. lower current users with step change in current needs.
thanks to all for the helpful answers. I remembered an interested document I read a while ago and gave it a second read: http://www.calex.com/pdf/3power_impedance.pdf
actually they recommend a series resistor with the paralleled film / tantalum cap of the same resistance as the ESR of the cap, e.g. 0.2R
A larger film cap of at least 1uF probably does no harm, but does it any good ? Lower impedance at high frequencies than a low-ESR electrolytic cap (e.g. Panasonic FM) ? Faster transient response ?
Besides that I wonder how it come the myth of 0.1uF film caps in parallel to electrolytics widespread that far. Many people strongly believe to it and even reputated labs/shops like Welborne-Labs or THEL incorporate small film bypasses to their "audiophile" PSUs.
actually they recommend a series resistor with the paralleled film / tantalum cap of the same resistance as the ESR of the cap, e.g. 0.2R
A larger film cap of at least 1uF probably does no harm, but does it any good ? Lower impedance at high frequencies than a low-ESR electrolytic cap (e.g. Panasonic FM) ? Faster transient response ?
Besides that I wonder how it come the myth of 0.1uF film caps in parallel to electrolytics widespread that far. Many people strongly believe to it and even reputated labs/shops like Welborne-Labs or THEL incorporate small film bypasses to their "audiophile" PSUs.
The 0.1uF ceramic cap has proven to work quite well with low-cost high-ESR small electrolytics, as long as they are mounted and wired very closely in the PCB.
However, audio people with little electronics understanding roughly copied that practice and started to put dozens of 0.1uF ceramics and other types everywhere, not necessarily together or associated to high-ESR electrolytics. The funny side about decoupling and layout practices is that they put in evidence the rather high amount of pure "cook-book" electronics designers that we have currently in active, people that just copies what others did before without further thinking or testing.
One of the worst things that can be done (and I see frequently) is to have the 100nF capacitors near the op amps and the electrolytics on the other side of the PCB, with long inductive supply tracks going between them. In these circumstances, the supply impedance as seen from the op-amps is actually reduced by removing all 100nF ceramics due to the big resonance peak produced otherwise.
However, audio people with little electronics understanding roughly copied that practice and started to put dozens of 0.1uF ceramics and other types everywhere, not necessarily together or associated to high-ESR electrolytics. The funny side about decoupling and layout practices is that they put in evidence the rather high amount of pure "cook-book" electronics designers that we have currently in active, people that just copies what others did before without further thinking or testing.
One of the worst things that can be done (and I see frequently) is to have the 100nF capacitors near the op amps and the electrolytics on the other side of the PCB, with long inductive supply tracks going between them. In these circumstances, the supply impedance as seen from the op-amps is actually reduced by removing all 100nF ceramics due to the big resonance peak produced otherwise.
So low ESR isnt an obvious advantage in PS.
Where should we use low ESR and where high ESR caps ?
And what about that case, when the main PS is far away (20-30cm)
and there are naturally smaller caps on the amp's PCB, what
kind of caps and combinations would be the best solution ?
Its also important to put some cmaller bypass caps on the main
filter caps, and to the smaller local caps on the PCB right at the
output stage ?
For example:
If there is a 10.000uF main cap and a 100 or 1000uF on the PCB
how should I bypass them after all, if the wire is ~30cm long ?
Where should we use low ESR and where high ESR caps ?
And what about that case, when the main PS is far away (20-30cm)
and there are naturally smaller caps on the amp's PCB, what
kind of caps and combinations would be the best solution ?
Its also important to put some cmaller bypass caps on the main
filter caps, and to the smaller local caps on the PCB right at the
output stage ?
For example:
If there is a 10.000uF main cap and a 100 or 1000uF on the PCB
how should I bypass them after all, if the wire is ~30cm long ?
I would probably use small standard electrolytics with or without ceramics in a local basis, and a plain bigger electrolytic in a global basis (maybe low-ESR depending on filtering ripple requirements). Also, since PCB tracks are inductive anyway, you can just add high-ESR axial filter inductors to get a quiet RF-decoupled supply for each section of the circuit as long as the resulting CLC system is damped. However, there are by far too many factors involved so it's just not possible to give any cook-book style guidelines. You should get an oscilloscope and various flavours of capacitors, resistors and inductors, prepare some nice NE555 spike generator and experiment. There is not any other way to do it.
I think that one of the best rules of electronics design is "Never expect anybody else to do your homework" (or face the poor results otherwise).
I think that one of the best rules of electronics design is "Never expect anybody else to do your homework" (or face the poor results otherwise).
Looks like it is missing a bypass cap
I would have thought a 0.1 ufd cap would be connected to the negative rail and going to ground here. This method might be quieter.
This thread has been very educational. I have already built a regulated supply board that leaves out ultra low ESR electro's and the bypasses. Using small ultra lo ESR types right on the power pins. The large electros are damped with small film using a series 0.4 ohm resistor. The large caps are 10 cm away. Works great.
OT, I do not have a scope, or really a place to store one. Can those 100.00 cards and software for PC provide what is needed to monitor and measure these to Mhz range? I hope to look at actual power supply performance.
George
I would have thought a 0.1 ufd cap would be connected to the negative rail and going to ground here. This method might be quieter.
This thread has been very educational. I have already built a regulated supply board that leaves out ultra low ESR electro's and the bypasses. Using small ultra lo ESR types right on the power pins. The large electros are damped with small film using a series 0.4 ohm resistor. The large caps are 10 cm away. Works great.
OT, I do not have a scope, or really a place to store one. Can those 100.00 cards and software for PC provide what is needed to monitor and measure these to Mhz range? I hope to look at actual power supply performance.
George
Hi Panelhead,
are the 0r4 damping resistors in series with the electros or film caps?
That AD layout is fairly common;- 100n across supply pins and one 100n from one supply pin to ground. It has to do with the internal architecture of the opamp.
Would it do any harm to connect three 100nF to every fast opamp, i.e. across supply pins and from both pins to ground when the manufacturer gives less detailed advice?
are the 0r4 damping resistors in series with the electros or film caps?
That AD layout is fairly common;- 100n across supply pins and one 100n from one supply pin to ground. It has to do with the internal architecture of the opamp.
Would it do any harm to connect three 100nF to every fast opamp, i.e. across supply pins and from both pins to ground when the manufacturer gives less detailed advice?
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.