You've conflated several things there (like power and current). John has (and I congratulate him for being very clear and succinct in his answer) explained that a watt in load A is not that same thing as a watt in load B- the actual load range being designed for, including load angle and magnitude range versus frequency, is critical in making these estimates, not just "microfarads per watt." All amps do not have the same power supply rejection. All amps aren't required to have the same noise. All amp topologies do not have the same power supply modulation or the same voltage rails. And on and on.
Design rules are nice, but what was pointed out is that a rule for "microfarads per watt" is a meaningless one without context. The analysis to determine minimum capacitance for a given design and performance target is straightforward, so why not do it?
Depends on the weight
technically higher rpm means more speed , gearbox is TQ converter, now amplifier , I want big stiff supply , err , Ahh , for bedroom system too...
Last edited:
I think you missed the "joules per watt at a given load" part. I don't think I ever said "microfarads per watt".
I'm ready to accept that there may not be a good answer, maybe amp designers don't think or power supplies in "joules per output watt" terms. I usually don't when building an amp - I just look at what others are doing and figure out what might work for me. If I wanted "One Stupid Trick for Great Power Supplies" I'd just stick as much capacitance in there as would fit. And buy a huge transformer.
But I was asking if perhaps someone had found a relation between energy stored in the power supply and sound quality. Something more than a simple "More is Better" or "Just enough to kill the ripple." Those are pretty easy to figure out by yourself. Looking at typical class AB pro amps, they tend to have fairly low capacitance and small transformers compared to their power ratings. Why? Size and cost are important for a touring amp. Weight is very important. But what about "High End" amps? (I mean that in a good way). They tend to have beefier power supplies for a rated power than their pro cousins. Low cost consumer amps are even lighter weight. Can we guess why?
Not all of us have the luxury of stuffing a huge power supply into the box. Where is the point of diminishing returns? Is there one? What might be the trade-offs of going smaller?
Amps have to be designed with some sort of load in mind. Is the only PSU design rule "Bigger is Better"?
I have just been doing a parametric search for op amps . Not for audio exactly , an active filtering application for measuring ( GBP 10 MHz ) . I think I will buy myself a big stock of MC33079 as they are good enough for these applications and have 4 op amps in the package . They look to be going obsolete .We are then left with MC33078 NE5532 (34) and LM833 . So far so good . Then absolutely nothing . OPA134 , 2134, 604 ,2604,2227 . It is not exactly an endless list . AD823 with OPA627 seems to head the list . I think it is time to design discrete op amps . If not we become stuck in a diminishing market . I have a friend who is designing a phono stage , if low noise the list is minute . To be frank 6nV/Hz is too noisy . It needs to be < 2 nV .
Others did- but joules per watt is just as nebulous. What's the rail voltage? What's the load? At what frequency? And all the other bits I mentioned.
Someone talked about using enough capacitance to keep the amp running for 5 seconds after turnoff. I'd rather have an amp that collapses faster than that so that in the event of a power outage, my speakers don't suffer a full power BANG from the preamp.
Rail sag? Get a bigger transformer- the caps shouldn't be a band-aid.
Uhhhhh..... isn't that why we use joules instead of just capacitance? Joules takes into account the voltage and capacitance. Low voltage amps need much more capacitance to store the same energy as high voltage amps.
I dunno. What's the amp rated for? A minimum 8 ohm load? 1 ohm?
Audio amps are generally supposed to be consistent from 30Hz to 15KHz or better. Or do you want a specific frequency?
Not sure what those are.
Indeed! But can they be an aid without being a band-aid? Is there any sonic advantage to having all that energy on tap tied directly to the output stage? If there is, how much?
Seriously Sy , power cuts,
Then have the pre stay up for 10 ... The caps are not for transformer rail sag , Isn't for maintaining voltage during diode switching, storage and controlling ripple.. I Guess best to keep adding caps until no detectable sonic advantages or measurable ripple ...
Ooooooohhhh, are they freshly harvested? Washed with morning dew? I want, I want, I want ...
Frankly, I don't understand what the whole argument is about. It's fairly clear that there is no, nor can there be any universal answer to how much capacitance you need, as it varies as per several variables.
However, if you compare John's answer for a nominal 100 WPC/8 Ohms amp with the little formula proposed by Motorola, I think you'll find a very big percentage of agreement.
I think assuming 1.5 Joules of energy per every dissipated 10W of power is nice, safe bet. If you know the load might be an evil one, go for 2 Joules per dissipated 10W. Just don't forget to clearly define what you expect from the amp into low loads, say 2 Ohms, or you might run out of steam. Work it out for both 1.5 and 2 Joules, and find your own niche somewhere in between.
I could be wrong, but I don't think it can get any more specific than that.
However, if you compare John's answer for a nominal 100 WPC/8 Ohms amp with the little formula proposed by Motorola, I think you'll find a very big percentage of agreement.
I think assuming 1.5 Joules of energy per every dissipated 10W of power is nice, safe bet. If you know the load might be an evil one, go for 2 Joules per dissipated 10W. Just don't forget to clearly define what you expect from the amp into low loads, say 2 Ohms, or you might run out of steam. Work it out for both 1.5 and 2 Joules, and find your own niche somewhere in between.
I could be wrong, but I don't think it can get any more specific than that.
And that's basically what I was looking for.
For some reason when you ask for a rule of thumb or general practice, people fly into a fit and want to hit you over the head about "no Magic Answers" or "You can't generalize" or some such. Sheesh..... All I wanted to know is if anyone had noticed a relation between energy stored/output power and sound quality. (Sound Quality vs Measurements). It appears that no one posting here thinks in those terms, fair enough. Tho we can see that someone at Motorola did.
FWIW, many deluxe amps have much more than 1.5 joules per 10W. Is it just overkill?
For some reason when you ask for a rule of thumb or general practice, people fly into a fit and want to hit you over the head about "no Magic Answers" or "You can't generalize" or some such. Sheesh..... All I wanted to know is if anyone had noticed a relation between energy stored/output power and sound quality. (Sound Quality vs Measurements). It appears that no one posting here thinks in those terms, fair enough. Tho we can see that someone at Motorola did.
FWIW, many deluxe amps have much more than 1.5 joules per 10W. Is it just overkill?
I just had to do this with a valve amp . 330 uF was not enough and was humming ( 450 V 140 mA ) . So for the loss of 5 V I used a capacitance multiplier ( Darlington and 10 uF , 4 K , 430K ) . It was bits from my useful box and worked a treat ( 10 to 1 reduction and nicer looking ripple ) . It needed 10 nF from collector to base . It even sounded nice ! With transistor amps my feeling is ripple current tells you all you need to know . Then add caps as you fancy . My goodness this simple thing has dragged on .
I beg to differ a little . Say whatever maximum current will be . Times it by 1.5 for safety . Job done . The manufacturer will give the ripple ratings . Then add caps as your ears require . I would admit that the currents that might be required baffle most of us . Would anyone like to say ? Lets talk daft speakers and stupidly high wattage's ( 1 ohms 600 W ) . One problem is that the big amps draw current at the crest of the sine wave . Sometimes the caps hardly get a look in . My friends in the PA world use 3 phase power as it helps replenish the caps more often . They find there comes a time when no matter how big the caps are 3 phase does it better .
Bigger is better.
See "le Monstre", so named because of its power supply. An exercise in "How far can you go?". On average Hiraga's power supplies were bigger than normal, it was an important part of the design. Of course, the school of massive power supplies has come into vogue recently, so the average seems to be creeping up.OK, as the ears require. That's a tough one. What I was hoping to find is that there may be an approximate range that many designers have used over the years that seems to work well. I'm sure the range will be fairly broad, but what is it? Just because most folks have never thought of it in those terms does not make it an invalid metric. And of course just because I thought of it does not make it valid, either.
I'm curious.Perhaps stored energy doesn't relate well to any sonic qualities. Maybe PSU impedance is a better metric?
Not necessarily. Motorola were mainly hard-headed engineers, so it is very likely that their rule of thumb was derived in the way I explained i.e. the criterion was not 'sound quality' but x% droop in 10 ms (or 8.33 ms in US). This could be regarded as a ripple calculation, but ripple calcs usually assume full power sine-wave; I assumed full power square wave.Pano said:
If your amp has good PSRR (including low power supply IM) then it may be the case that there is nothing more to be said: low enough ripple guarantees good enough sound. When people try to explain anything further they usually quickly fall back on myths and legends rather than genuine explanations. Now it may be that we don't know how much hum is low enough, or how much ripple IM is low enough, given that we can't have zero of either unless we use battery power. 'Too low to measure' is meaningless, as you can always measure lower if you try a bit harder.
- Status
- Not open for further replies.