I've been looking at smoothing capacitors for my upcoming (solid state) amp builds, which could be up to a maximum of 100w per channel into 8ohms from a PSU with +/-42v rails.
I'm thinking 63v caps to be safe, as 50v doesn't 'quite' allow as much headroom as I'd like for transformer regulation plus mains voltage variations. And I was thinking something like 10,000uf per rail (20,000uf per channel) would be a reasonable capacity. Money is not irrelevant, unfortunately though. Cap prices soon start to spiral with better life expectancies, ESRs and ripple currents, yet I've seen some terrible looking specs at the budget end that I would want to avoid.
Roughly what sort of lifetime and total ripple current (or esr) would people suggest is a sensible sweet spot to aim for, and what sort of pairings are likely to be the best way of achieving it (e.g. 1x 10,000uf, 2x 4,700uf etc).?
Many Thanks
Kev
I'm thinking 63v caps to be safe, as 50v doesn't 'quite' allow as much headroom as I'd like for transformer regulation plus mains voltage variations. And I was thinking something like 10,000uf per rail (20,000uf per channel) would be a reasonable capacity. Money is not irrelevant, unfortunately though. Cap prices soon start to spiral with better life expectancies, ESRs and ripple currents, yet I've seen some terrible looking specs at the budget end that I would want to avoid.
Roughly what sort of lifetime and total ripple current (or esr) would people suggest is a sensible sweet spot to aim for, and what sort of pairings are likely to be the best way of achieving it (e.g. 1x 10,000uf, 2x 4,700uf etc).?
Many Thanks
Kev
You may be under-estimating my limitations 🙂 I have had a go at working out how the ripple affects temperature and so lifetime etc but I'm not really sure what I'm doing, or how real-world it is. Or in fact what 'ripple' effect would stem from driving the speakers with typical music, assuming (perhaps wrongly) that discharging the caps heats them as much as charging..
TBH I was sort of hoping for some rules of thumb or general estimates from experienced builders, which are likely to be rather more trustworthy than my uncertain attempts. I found some for estimating capacitance, but not for ripple and heat/lifetime. Is that because its always very specific to a given setup, or can ball-park ranges be of use?
Thanks
Kev
Last edited:
Ripple current is hard to calculate exactly, although modelling should give it. For a wild first stab I would assume about 1.5-2 times the DC current draw. Lifetime depends on temperature and build quality. You can affect the temperature through the rest of your design. Build quality depends on the cap manufacturer.
You could try looking at the long thread on reservoir cap size, but recently it has wandered off the point.
You could try looking at the long thread on reservoir cap size, but recently it has wandered off the point.
You could also download PSUD here:
PSUD2
It wil not help you with heat/lifetime issues, but give you an idea about the capacitance needed for a given ripple and/or the ripple for a given capacitance.
The tools is pretty straightforward and the is also a thread here on DIYAudio as well as many tutorials on the web.
PSUD2
It wil not help you with heat/lifetime issues, but give you an idea about the capacitance needed for a given ripple and/or the ripple for a given capacitance.
The tools is pretty straightforward and the is also a thread here on DIYAudio as well as many tutorials on the web.
Thanks, thats very helpful; I realise its just a rough ball-park but it helps a lot for checking the sanity of my choices. It also doesn't seem unreachable at the sort of currents I'd be looking at, which is quite reassuring.
Yeah, I've noticed that the manufacturer's life expectancies can vary hugely, though theres also quite a bit of variation within a manufacturer depending on the range and size of the capacitor. I've seen anything from 1,000 to 200,000hrs quoted at the rated temperature.
I'll have a look at that long thread again - I assume you mean this one. I'd found some later pages in my searches but perhaps I need to start reading from earlier on if its wandering.
Thanks, I'll have a go when I'm on my home computer. I'd kind of assumed such things were too advanced for me, but if there are tutorials and threads on it I should be okay.
Thanks again,
Kev
TNT and ESP have sections on PSU.
You won't get 100W into 8r0 from +-42Vdc supply rails.
30+30Vac gives +-42Vdc. This suits ~75W into 8r0.
35+35Vac gives +-50Vdc. This suits ~100W into 8r0.
And using the UK's 240Vac into a 230Vac transformer gives you a slightly higher rail voltage than what all the European and New World Members tell you will happen.
You won't get 100W into 8r0 from +-42Vdc supply rails.
30+30Vac gives +-42Vdc. This suits ~75W into 8r0.
35+35Vac gives +-50Vdc. This suits ~100W into 8r0.
And using the UK's 240Vac into a 230Vac transformer gives you a slightly higher rail voltage than what all the European and New World Members tell you will happen.
Well spotted; I should have checked the wattage myself - I just took it from the ESP site's P3a description, which I was thinking of using at the time of posting: "and with the transistors specified the amp will provide 100W into 8 ohms, at a maximum supply voltage of ±42V. This supply is easily obtained from a 30-0-30V transformer". I see that a tad lower down the page though there's a spec table which quotes it as 90w for the same supply voltage which is different but still above your estimate, so perhaps he's talking about some kind of peak as the driver impedance changes.
Thanks for pointing out my error, a 25% drop in expectation is significant enough to make a difference if I were to choose less efficient drivers; I was already considering the p101 mosfet project instead so maybe thats further support for it - though higher voltage capacitors would be needed.
Cheers
Kev
Thanks for pointing out my error, a 25% drop in expectation is significant enough to make a difference if I were to choose less efficient drivers; I was already considering the p101 mosfet project instead so maybe thats further support for it - though higher voltage capacitors would be needed.
Cheers
Kev
Consider a bank of smaller 63V radial caps. Even good ones don't cost that much, especially if you hit a quantity price break point, and the total will likely be less than for a big screw terminal can. You'll also get better dissipation factor and HF performance.
I think you're right. I've just been doing some price comparisons on a reputable dealer's website. Looking at just the standard type caps with 2000-3000hr@85c lifetimes, it looks like the best value sizes are between about 2,200uf and 4,700uf or so. Building multiples to my 10,000uf target (give or take) I end up with ESR totals of around 0.021-0.035ohms and total ripple currents of 8A to nearly 13A as standard. By comparison a single 10,000uf cap with suitable ripple current is 2-3 times the price as making it up from a few/several smaller ones.
Its slightly different for caps with more endurance. 105c caps are less economical in the small sizes, so starting around 4,700uf seems better value - but it still adds an extra 50% cost over the 85c ones; 3000hrs (rather than the standard 2000) at 105c is similar but more or less doubles the price of the basic 85c types. If I wanted say 12,000hrs at 85c then its 2-3 times the price of the basic ones.
(though that's just at this one retailer so may or may not be universal)
Looking at a few examples of temperature specifications, keeping an 85c capacitor at 40c seems like it can extend life by 25-30x its rating at 85c. Keeping a 105c capacitor at 40c can be more like 100x. The ripple causes heating though, so its a matter of keeping that down as well as just the ambient conditions. I guess with the internal heating from ripple and ambient temperature inside an enclosure case, 40c for the caps core temperature may be harder than it seems though, even with more smaller caps with more surface area.. perhaps standard 105c caps would be a better bet?
Cheers
kev
Last edited:
I recommend reading Bob Cordells book on Designing Audio Power amplifiers and take notice of page 351. Here he talks about the split reservoir capacitor supply where in the use of a small .1 to .22 ohm resistor is used to severely reduce the ripple and noise without raising the supply impedance much.
I built and used this arrangement with my LM60 amp modules and took the split reservoir idea much further. I have 400 VA Antek toroid and 35 amp bridge rect. feeding a set, + and -, 27,000 Uf caps then feeding .1 ohm resistors . This then feeds another set of 27,000 Uf caps and another .1 ohm resistors. This then feeds the main set of 100,000 Uf caps.
At full power output of 60 watts there is 10 millivolts of ripple on the main caps. Now of course the power supply cost about 5 times the amp board cost and is a true monobloc as everything is doubled for the stereo chassis. per rail.
Total capacitance for each amp board is 308,000 Uf's or 154,000 Uf's per rail.
But the theory is the use of cheap caps and 20 watt sand resistors to clean up the filtered DC rails. Then feed clean DC to the main caps of your choice.
I built and used this arrangement with my LM60 amp modules and took the split reservoir idea much further. I have 400 VA Antek toroid and 35 amp bridge rect. feeding a set, + and -, 27,000 Uf caps then feeding .1 ohm resistors . This then feeds another set of 27,000 Uf caps and another .1 ohm resistors. This then feeds the main set of 100,000 Uf caps.
At full power output of 60 watts there is 10 millivolts of ripple on the main caps. Now of course the power supply cost about 5 times the amp board cost and is a true monobloc as everything is doubled for the stereo chassis. per rail.
Total capacitance for each amp board is 308,000 Uf's or 154,000 Uf's per rail.
But the theory is the use of cheap caps and 20 watt sand resistors to clean up the filtered DC rails. Then feed clean DC to the main caps of your choice.
Ah interesting; happily I bought that book only a couple of weeks ago - haven't reached p351 yet though so I'll have to jump ahead when I get home tonight!
I've heard of the 0.1ohm resistor in a couple of other posts but didn't grasp the significance; thanks for the explanation, now I do (I think) see what its for - it partially decouples the two stages of capacitor banks allowing the second (some) ability not to respond to ripple affecting the first?
Cheers
kev
I've heard of the 0.1ohm resistor in a couple of other posts but didn't grasp the significance; thanks for the explanation, now I do (I think) see what its for - it partially decouples the two stages of capacitor banks allowing the second (some) ability not to respond to ripple affecting the first?
Cheers
kev
The thickness of the wiring connecting the transformer through the rectifier and on to the smoothing caps before reaching the amplifier power terminals does affect the RESISTANCE that you are inserting into the various RC and LC filters in your supply.
Thin wiring that is twisted on EVERY Flow and Return Pair will improve all these filters and cost you nothing extra, in return for reduced supply impedance and reduced interference.
Thin wiring that is twisted on EVERY Flow and Return Pair will improve all these filters and cost you nothing extra, in return for reduced supply impedance and reduced interference.
Yes that makes sense, I think its why I initially failed to grasp the use of the 0.1ohm resistor, because it seems counterintuitive to increase the resistance in the power circuit; I guess they feel the subsequent capacitors compensate for it.
I shouldn't have a problem using heavy, twisted-pair cabling - not so sure about the amp PCB end of things though. I was thinking of building on some more experienced person's PCBs as there's undoubtedly much that I don't know yet about power amp design. Perhaps I should reconsider, and be able to design in other things like the decoupling caps I'd like too.
I'm not very used to designs where such small resistances have such an appreciable effect. But if things like the cabling resistance and 0.1ohm resistors do contribute noticably to the the R/C and L/C aspects of the circuit, then going back to my OP, maybe I should care more about the internal capacitor resistances anmd impedences too. Many 'seem' to have tiny values, but these are actually still quite high relative to power cabling and that 0.1ohm resistor.
Thanks
Kev
I shouldn't have a problem using heavy, twisted-pair cabling - not so sure about the amp PCB end of things though. I was thinking of building on some more experienced person's PCBs as there's undoubtedly much that I don't know yet about power amp design. Perhaps I should reconsider, and be able to design in other things like the decoupling caps I'd like too.
I'm not very used to designs where such small resistances have such an appreciable effect. But if things like the cabling resistance and 0.1ohm resistors do contribute noticably to the the R/C and L/C aspects of the circuit, then going back to my OP, maybe I should care more about the internal capacitor resistances anmd impedences too. Many 'seem' to have tiny values, but these are actually still quite high relative to power cabling and that 0.1ohm resistor.
Thanks
Kev
Why use thick PSU cables?
Cordell very recently came to suggest that the added resistance of thinner cables may be a significant advantage. I replied that I have been recommending just that for a while.
Cordell very recently came to suggest that the added resistance of thinner cables may be a significant advantage. I replied that I have been recommending just that for a while.
hmm, more counterintuitiveness! Not sure my mental capabilities are up to this..
I've now had chance to read Bob Cordells paragraphs on the small resistor separating the two banks of capacitors; if I understand correctly, so that it becomes an R/C interaction with a low enough frequency to reduce the higher frequency ripple. (I guess something similar could apply with the resistance of supply leads and decoupling capacitors, though with very different R and C values). But the chapter itself covers quite a few more real-world aspects of the power supply and grounding, so I think I'll skip ahead and read it through tonight.
I'm beginning to feel that my 'learning as I go' (or at least as I plan) approach may not be ideal; probably I should slow down and do a lot more reading first. I've entered into this in the belief that I could just buy an expertly designed PCB with construction details and bolt it to almost any basic transformer/rectifier/capacitor setup... I'm sure that it would work, but its also becoming clear that even with much hard work already done for me, theres still considerable scope for me to then compromise the performance through making less than optimum component choices and implementation decisions.
Cheers
kev
I've now had chance to read Bob Cordells paragraphs on the small resistor separating the two banks of capacitors; if I understand correctly, so that it becomes an R/C interaction with a low enough frequency to reduce the higher frequency ripple. (I guess something similar could apply with the resistance of supply leads and decoupling capacitors, though with very different R and C values). But the chapter itself covers quite a few more real-world aspects of the power supply and grounding, so I think I'll skip ahead and read it through tonight.
I'm beginning to feel that my 'learning as I go' (or at least as I plan) approach may not be ideal; probably I should slow down and do a lot more reading first. I've entered into this in the belief that I could just buy an expertly designed PCB with construction details and bolt it to almost any basic transformer/rectifier/capacitor setup... I'm sure that it would work, but its also becoming clear that even with much hard work already done for me, theres still considerable scope for me to then compromise the performance through making less than optimum component choices and implementation decisions.
Cheers
kev
Last edited:
Very good, Kev. You understand that correctly. Now lets talk about noisy grounds vs. quiet grounds. Noisy grounds or returns appear around the first cap bank and the rectifier path. Most current so most noise.
In my example the quietest ground is at the third set of caps which are the big ones. So here is where the speaker return, amp board return and any bypass returns are connected. Virtually no noise at all.
The toroid center tap is connected to the center of the first cap bank so its far away from the quiet ground. The amp board input return is only at the board so no ground loops develop. In my amplifier chassis the power supply and the amp boards are not connected to the grounded chassis to avoid ground loops.
The amp boards are actually grounded through the RCA cables to the pre amp which is grounded properly. The amp chassis IS connected to AC ground through the line cord so its safe.
Hunting and killing ground loops took me on a journey that brought me to Bob's book and the final configuration.
In my example the quietest ground is at the third set of caps which are the big ones. So here is where the speaker return, amp board return and any bypass returns are connected. Virtually no noise at all.
The toroid center tap is connected to the center of the first cap bank so its far away from the quiet ground. The amp board input return is only at the board so no ground loops develop. In my amplifier chassis the power supply and the amp boards are not connected to the grounded chassis to avoid ground loops.
The amp boards are actually grounded through the RCA cables to the pre amp which is grounded properly. The amp chassis IS connected to AC ground through the line cord so its safe.
Hunting and killing ground loops took me on a journey that brought me to Bob's book and the final configuration.
Many thanks for the encouragement, and even more for the explanation! You explained the grounding so clearly that I think I actually understand it.
I'm reasonably aware of single-point/star grounding but hadn't got a grasp on where best to place this in the power-supply, or where any ground connection to AC earth would be (or even if one should be made at all). So thats prevented me making some potentially important mistakes.
Its also reassuring to know that earthing the chassis (independently of the internal circuits) to AC ground is still a possibility too, as I was starting to get a bit paranoid about such things, but would prefer not to trust 'only' to my skills at double-insulating.
Thanks again,
Kev
I'm reasonably aware of single-point/star grounding but hadn't got a grasp on where best to place this in the power-supply, or where any ground connection to AC earth would be (or even if one should be made at all). So thats prevented me making some potentially important mistakes.
Its also reassuring to know that earthing the chassis (independently of the internal circuits) to AC ground is still a possibility too, as I was starting to get a bit paranoid about such things, but would prefer not to trust 'only' to my skills at double-insulating.
Thanks again,
Kev
@Kev06: If you have not been through this before, I highly recommend starting with a single, conventional star ground, and perhaps some small isolation from an earthed chassis through a diode bridge, etc., and I mean regardless of CRC filters, "quiet" and "noisy" grounds, etc.
This means you start with something simple, and safe, and know 100% that it will work. May not be ideal, but ground arrangements are easy to change, wire is cheap... You can always make it more complex, but you would be starting with something everyone pretty much agrees does work.
Aside from that, one of the more productive members here, LazyCat, has just started selling his VSSA amp modules, which are 90% built (you just have to solder the caps and three transistors). The modules are specified for 100W into 8 ohms, with 45V DC rails.
If you want to roll your own, check out the related PeeCeeBee thread, where Shaan has reduced the same design to something that can be assembled by just about anyone with simple tools and available components, and still sounds very nice with a modest power supply (I am using 3x4700uF per rail, and depending on desired power level, others are using from 2x4700 to 4x4700, which means very inexpensive caps).
Both threads are very active, lots of people you can ask questions.
This means you start with something simple, and safe, and know 100% that it will work. May not be ideal, but ground arrangements are easy to change, wire is cheap... You can always make it more complex, but you would be starting with something everyone pretty much agrees does work.
Aside from that, one of the more productive members here, LazyCat, has just started selling his VSSA amp modules, which are 90% built (you just have to solder the caps and three transistors). The modules are specified for 100W into 8 ohms, with 45V DC rails.
If you want to roll your own, check out the related PeeCeeBee thread, where Shaan has reduced the same design to something that can be assembled by just about anyone with simple tools and available components, and still sounds very nice with a modest power supply (I am using 3x4700uF per rail, and depending on desired power level, others are using from 2x4700 to 4x4700, which means very inexpensive caps).
Both threads are very active, lots of people you can ask questions.
Thanks for the suggestions, I'll definitely look into those!
I shouldn't have too much trouble practically building the type of circuits that have been talked about so far (I've been an electronics hobbyist for some years). For me its more about actually understanding the design differences between a basic/generic power supply and a 'very good' one suitable for high performance with power amps, which is beyond my current knowledge.
Cheers
Kev
I shouldn't have too much trouble practically building the type of circuits that have been talked about so far (I've been an electronics hobbyist for some years). For me its more about actually understanding the design differences between a basic/generic power supply and a 'very good' one suitable for high performance with power amps, which is beyond my current knowledge.
Cheers
Kev
Last edited:
- Status
- Not open for further replies.