I've done some some research for a high performance power supply recently but also keeping value in mind. I've settled for a linear supply, reasons not subject to this discussions.
as a result of simulations I've come to the conclusion that as far as price/performance goes I'm much better off with many cheap caps in parallel.
precisely, 8 x 10000uF paralleled Samwha's per rail give:
- larger total capacitance
- lower equivalent ESR
- lower ripple voltage
- lower ripple current per cap
- lower price
all this is compared to the largest (capacitance-wise) Epcos caps of the same price.
the single advantage that I can see for the expensive caps is longer life but if not pushed (and considering the large number of them connected in parallel they most likely won't be) the Samwhas should be more than ok. they will be used in a class D amp, powering moderately sensitive (86dB) speakers with well-behaved impedance. and after all, I've seen Samwha's (pictures available upon request) used in top-of-the-line Meridian players.
to get similar measured performance with say Epcos caps bought off Farnell I would have to spend a few times more. subjective performance aside (try to pretent for a second that this is not even an amp's PS that we're talking about) what should make me spend (considerably more) money on some 'big name" caps?
as a result of simulations I've come to the conclusion that as far as price/performance goes I'm much better off with many cheap caps in parallel.
precisely, 8 x 10000uF paralleled Samwha's per rail give:
- larger total capacitance
- lower equivalent ESR
- lower ripple voltage
- lower ripple current per cap
- lower price
all this is compared to the largest (capacitance-wise) Epcos caps of the same price.
the single advantage that I can see for the expensive caps is longer life but if not pushed (and considering the large number of them connected in parallel they most likely won't be) the Samwhas should be more than ok. they will be used in a class D amp, powering moderately sensitive (86dB) speakers with well-behaved impedance. and after all, I've seen Samwha's (pictures available upon request) used in top-of-the-line Meridian players.
to get similar measured performance with say Epcos caps bought off Farnell I would have to spend a few times more. subjective performance aside (try to pretent for a second that this is not even an amp's PS that we're talking about) what should make me spend (considerably more) money on some 'big name" caps?
Last edited:
my calculations are based on 2 channels load. so either one single supply with 2 x 80000u or two with 2 x 40000. what's wrong with that?
Last edited:
Lots of smaller caps, as long as they are not rubbish, will be better than one big cap.
E-Bay has several examples of fields of smaller electrolytics on PCBs to do exactly that.
You don't need to go mad though. A field of 1000uF Caps would probably be the best compromise.
E-Bay has several examples of fields of smaller electrolytics on PCBs to do exactly that.
You don't need to go mad though. A field of 1000uF Caps would probably be the best compromise.
80000u was the lowest value that gave a ripple current below maximum ratings for the Samwha under worst case scenario (maximum load). I realize that such extreme conditions are never to be found with real music and normal listening levels in a small room but I would like the reserve for the days when some very low sensitivity speakers with nasty impedances might land in my living room. it's not a fixed value, just an example.
anyway my question is more general: how does an expensive cap of well known brand justify its price as long as there don't seem to be any obvious benefits compared to many cheap caps in parallel, except for longer life? or what is it that I'm missing?
anyway my question is more general: how does an expensive cap of well known brand justify its price as long as there don't seem to be any obvious benefits compared to many cheap caps in parallel, except for longer life? or what is it that I'm missing?
Last edited:
The higher dissipation losses give rise to more self generated HF noise. You will need to also address that with smaller bypasses.
bypasses go without saying. still, the added cost is negligible.
are you referring to thermal noise?
are you referring to thermal noise?
Any loss can be modeled as a resistance. All resistors have both 1/F and "Thermal" noise. Good bypasses properly executed will address thermal noise. I do not have a handle on how important 1/F noise is as the capacitor that generates it also sinks it!
Also verify the esr of the smaller caps, then compare to the esr of the expensive big caps. The sonics will be different, from type and brand, you will have to decide which is better for you. Most of the amplifiers i have favored sonically in the past had big caps, with bypass, not multiples of small caps.
the ESR of the cheaper caps is much larger (55 mohms but the value is not very accurate as it isn't specified explicitly but rather derived from loss tangent) than that of the more expensive ones (varies based on capacitance and price), as indicated by much lower allowed ripple current but paralleling them should lower overall ESR.
I would like to keep subjective impressions aside (too many reasons to list), I'm only interested in objective data.
I would like to keep subjective impressions aside (too many reasons to list), I'm only interested in objective data.
I've often wondered if, instead of gigantic capacitors, some other technology e.g. li-ion or ni-cad batteries could be used instead, along with smaller capacitors? Might be possible to design a supply with smaller components that still had high peak capabilities?
@Mr push pull...
That's my point, while adding caps lower ESR, without knowing the total or actual esr , you might still be higher than the big caps, 8 might not be enuff...
That's my point, while adding caps lower ESR, without knowing the total or actual esr , you might still be higher than the big caps, 8 might not be enuff...
just looked again at the datasheet to check and the loss tangent is specified at 120Hz, so I unless I'm missing something the ESR can be derived from there.
don't worry about esr.
One excellent low esr cap will struggle to get down to 8// caps where the effective esr is ESR/8.
But take care to keep inductance low. Wiring and/or traces to 8// caps is likely to be long and open increasing inductance.
Don't put bypassing on the smoothing bank unless you can prove it is a benefit. Scope and switching test signal to investigate.
One excellent low esr cap will struggle to get down to 8// caps where the effective esr is ESR/8.
But take care to keep inductance low. Wiring and/or traces to 8// caps is likely to be long and open increasing inductance.
Don't put bypassing on the smoothing bank unless you can prove it is a benefit. Scope and switching test signal to investigate.
there's much controversy about the bypass caps. I remember Nelson Pass advising against them as they tend to form high frequency resonators combined with stray inductances.
- Status
- Not open for further replies.