Hi thanks for reading.
I built a simple psu for my amplifier consisting of a 12v 15 amp transformer, followed by a bridge rectifier, then a 47,000 uf cap.
There was a lot of noise coming through the speaker so i inserted another 47,000uf and the noise went down. I kept inserting caps, the noise kept lowering, until i reached 135,000 uf and the noise disappeared completly.
I figured 135,000 uf was the perfect value so i took out the six caps and replaced with a single 150,000uf cap which is the closest value i could find, but some of the noise came back.
Why is that. Is it because the first cap smoothed the ac from the bridge rectifier, the second cap smoothed a little more, until it reached the sixth cap and was further smoothed to clean out the ripple totally, something a single cap cannot do.
Or is it the six caps being paralled connected had lower esr esl values which aided the smoothing process.
From this is it possible to build a power supply with just one smoothing cap or is this going too far.
Thanks.
I built a simple psu for my amplifier consisting of a 12v 15 amp transformer, followed by a bridge rectifier, then a 47,000 uf cap.
There was a lot of noise coming through the speaker so i inserted another 47,000uf and the noise went down. I kept inserting caps, the noise kept lowering, until i reached 135,000 uf and the noise disappeared completly.
I figured 135,000 uf was the perfect value so i took out the six caps and replaced with a single 150,000uf cap which is the closest value i could find, but some of the noise came back.
Why is that. Is it because the first cap smoothed the ac from the bridge rectifier, the second cap smoothed a little more, until it reached the sixth cap and was further smoothed to clean out the ripple totally, something a single cap cannot do.
Or is it the six caps being paralled connected had lower esr esl values which aided the smoothing process.
From this is it possible to build a power supply with just one smoothing cap or is this going too far.
Thanks.
Is it important that the number of capacitors should be equal to 1? If so, you can probably make it by consider the large capacitor to have higher impedance than the parallelled smaller. Use brute force and try an even larger. If not you can use the common CRC technique and begin with 47 mF - R 47 - mF. The value of R depends on the current throug the amp. Calculate for a voltage loss under 1 V. There is a easy tool for PSU simulation where you can measure the noise, voltage losses and many other things: PSUD2 at duncanamps.com/psud2.
Caps and PSUs have many different purposes.
1. You can calculate the cap required for a certain current draw to give a certain ripple.
2. Having a resistance and then capacitance to ground is a low pass filter - for example a 5.1ohm followed by 6000uF of capacitance (450V) for a valve amp is a 0.5Hz low pass filter.. 6000uF is hideously large for a valve amp.. for 450V+ caps it's also hideously large physically.
3. in parallel reduces ESR and increases available ripple current capability. perhaps less important for linear PSUs but for SMPS low ESR is something that needs consideration in the design. In a SMPS valve amp booster 48V->300V+ I had 8 low ESR electrolytic 100V 330uF caps in parallel. Compare that to the output caps at 0.03ohm ESR DC link non-electrolytic (on 80uF and one 20uF). Sizewise the film caps are massive compared to the equivalent electrolytic. It all depends on the purpose.
4. The point that mooly made about sag and current draw - you'll want to understand the peak current draw otherwise in overload the PSU will probably generate more noise, the voltage may sag and you will then find the amp design (things like the operating points of the amp chips/valves) then move causing additional distortion.
5. Adding more caps means more inrush stress without inrush control - typically they're designed for a specific energy rating (joules) so they can limit just enough. Without causes stresses on the power supply on startup.
Edit: I see it's linear
More coffee needed!
1. You can calculate the cap required for a certain current draw to give a certain ripple.
2. Having a resistance and then capacitance to ground is a low pass filter - for example a 5.1ohm followed by 6000uF of capacitance (450V) for a valve amp is a 0.5Hz low pass filter.. 6000uF is hideously large for a valve amp.. for 450V+ caps it's also hideously large physically.
3. in parallel reduces ESR and increases available ripple current capability. perhaps less important for linear PSUs but for SMPS low ESR is something that needs consideration in the design. In a SMPS valve amp booster 48V->300V+ I had 8 low ESR electrolytic 100V 330uF caps in parallel. Compare that to the output caps at 0.03ohm ESR DC link non-electrolytic (on 80uF and one 20uF). Sizewise the film caps are massive compared to the equivalent electrolytic. It all depends on the purpose.
4. The point that mooly made about sag and current draw - you'll want to understand the peak current draw otherwise in overload the PSU will probably generate more noise, the voltage may sag and you will then find the amp design (things like the operating points of the amp chips/valves) then move causing additional distortion.
5. Adding more caps means more inrush stress without inrush control - typically they're designed for a specific energy rating (joules) so they can limit just enough. Without causes stresses on the power supply on startup.
Edit: I see it's linear
More coffee needed!
Last edited:
It's a small diy class-a amp that i put together myself. There is no sag issue. The psu can deliver 15 amps.
This might sound weird but the more caps i put into the psu the more the sound changes, becomes inaccurate when compared against a single cap.
Caps add a certain colouration of their own, the more caps the more the colouration, the more inaccurate the music sounds.
Searching around i've read a ton of threads wherein people say things like elna caps sound different / better than nichicon, nichicon sound different to panasonic, panasonic sound different to mundorf. There too much evidence to think i am making a mistake or imaging it.
Just am disliking the sound of multiple caps and all the different colourations messed up the experience for me. Sorry for being so picky
For the six paralled caps I used two fischer caps, two elna made for audio, one nichicon, one panasonic.
What should i do next, try a bigger cap. What about small bypass caps across the transformer, bridge rectfier and large electrolytic.
Many thanks.
This might sound weird but the more caps i put into the psu the more the sound changes, becomes inaccurate when compared against a single cap.
Caps add a certain colouration of their own, the more caps the more the colouration, the more inaccurate the music sounds.
Searching around i've read a ton of threads wherein people say things like elna caps sound different / better than nichicon, nichicon sound different to panasonic, panasonic sound different to mundorf. There too much evidence to think i am making a mistake or imaging it.
Just am disliking the sound of multiple caps and all the different colourations messed up the experience for me. Sorry for being so picky
For the six paralled caps I used two fischer caps, two elna made for audio, one nichicon, one panasonic.
What should i do next, try a bigger cap. What about small bypass caps across the transformer, bridge rectfier and large electrolytic.
Many thanks.
You have a number of things to consider.
Firstly you should have a look with a scope and see what the actual power supply voltage looks like in terms of ripple content.
Bigger caps cause problems of their own, the transformer and bridge will see massive current demands but only for a short time period each cycle. I suspect the noise is actually some other issue being masked/modified by you altering the PSU.
You have to do measurements in the first instance to see what is happening, what the supply looks like and what this 'noise' looks like at the amp output.
Firstly you should have a look with a scope and see what the actual power supply voltage looks like in terms of ripple content.
Bigger caps cause problems of their own, the transformer and bridge will see massive current demands but only for a short time period each cycle. I suspect the noise is actually some other issue being masked/modified by you altering the PSU.
You have to do measurements in the first instance to see what is happening, what the supply looks like and what this 'noise' looks like at the amp output.
ripple voltage is directly proportional to load current and inversely to C or filter capacitance....the law is kept...
goodguys, you can look at this too.....have fun.....Capacitance Multiplier Power Supply Filter
What is the PSRR vs frequency of the amplifier?
Did you put a snubber on the rectifiers?
I found I got rid of the last hum by making a pi filter. 470 uf, 2 ohms, then 470 uf. The one or 2 ohms between sections performs an important function. 470 because it was an op amp mixer, low current.
2nd the advice on the snubber on the rectifiers. Even .01 uf across the output of the bridge can help a lot. This hum transmits by radio, has high frequency RF occuring at twice AC frequency. .01 uf series 5 to 10 ohms even better, but difficult to solder directly across the bridge. If you put the snubber away from the bridge rectifier, the wires from bridge to snubber act as transmit antennas.
2nd the advice on the snubber on the rectifiers. Even .01 uf across the output of the bridge can help a lot. This hum transmits by radio, has high frequency RF occuring at twice AC frequency. .01 uf series 5 to 10 ohms even better, but difficult to solder directly across the bridge. If you put the snubber away from the bridge rectifier, the wires from bridge to snubber act as transmit antennas.
Adding caps in 'bypass' with each other ends up with each cap's frequency impedance being added to the mix, in the end you end up with a filter with multiple poles - almost like a EQ. That also varies with temperature, model range (capacitance) and the model/make build etc depending on the cap.
The bypass caps across recifiers are snubbers - essentially they're there to quieten the diode switching noise.
I would agree with mooly - if you put an scope on and then work backwards you should find out what components are changing the signal in what way. It's difficult to improve something without knowing (a) the starting point (b) the 'error' and (c) the target to measure how close you are.
All that 50Hz/100Hz noise should show up prior to the caps..
Also do you have an EMI filter on the primary side of the PSU? It may be that mains noise is coming in and then the additional caps are removing the noise..
Because current cannot flow without voltage drop, a ground conductor is not equipotential if it has current flowing through it. The ground wires to the capacitors will have a small ripple voltage on them induced by the ripple current. How these grounds are connected to the rest of the circuit can have a huge impact on hum and modulation noise in the output. I wonder whether the wiring layout to the bank of 6 capacitors was having a nulling effect. Is the amplifier laid out with a star ground scheme to prevent ground noise being coupled into ground reference parts of the circuit?
I think it is connected with wiring (and only partially - with ESR).
Last edited:
Six x capacitors in parallel is quite likely to have a lower ESR at ripple frequencies than a single capacitor of the same value, which may also explain the lower noise. However from a ripple perspective, 1 x 47,000uF would easilly be enough in a properly laid out circuit to prevent hum, which points to an issue with circulating currents as alluded to in my previous post.
I think two photo of the wiring have to be enough - first photo '6-caps' wiring and second photo - one cap wiring.
Just wanted to ask about this statement, i am getting a distinct loss of treble, is there any way the cap's frequency impedance can be changed ie by changing the cap's make or capacitance value to avoid this treble eq'ing.
Last edited by a moderator:
In general electrolytics have higher impedance compared to plastic foil capacitors. Partly because the electrolytic have higher resistance than aluminium, partly because they are connected in just one end of the winded alufoil/electrolytic and the distance from one connection to the other through the dielectricum can be tens of centimeters and so the inductance may affect at higher frequencies. Plastic foil caps are connected over the whole edge of the alufoil so the distance between connections can be as short as a few millimeters.
Anyway, it seems unlikely that your problems comes from an ordinary PSU. Many good sounding amplifiers have quite simple power supplies, often made up from just electrolytics, no plastic at all. Modern electrolytic capacitors give good response up to many tens of kHz so methink it's something else that cause the loss of treble.
If you want low impedance in the MHz area, go nuts and add a >1uF polypropylene on the amplifier board. Don't use ceramics, the have too low impedance and may cause oscillation together with electrolytics.
Have you get rid of the noise?
Anyway, it seems unlikely that your problems comes from an ordinary PSU. Many good sounding amplifiers have quite simple power supplies, often made up from just electrolytics, no plastic at all. Modern electrolytic capacitors give good response up to many tens of kHz so methink it's something else that cause the loss of treble.
If you want low impedance in the MHz area, go nuts and add a >1uF polypropylene on the amplifier board. Don't use ceramics, the have too low impedance and may cause oscillation together with electrolytics.
Have you get rid of the noise?
