I concur with AndrewT on acceptable voltage drop range.
We should firstly filter noise off a basic power supply to reduce voltage fluctuations from HF noise/nuisance charge. And, with decent filtering (filters on transformer primary, secondary and diodes) the voltage fluctuations are reduced to fit his 1v to 5v range.
Noise is a weak current source that charges caps extraneously higher until load pulls down the extra, and thus noise IS the primary cause of extraneous voltage fluctuations. So, first we filter that. Later, after filtering off HF noise, we can consider series filters, such as CRC. I've constrained my CRC to 1.4v drop (at most), and the results still fit within AndrewT's guidelines.
P.S.
There are possible exceptions:
It is possible that a Class D power amplifier has insufficient idle load to pull down noise/nuisance charge, and in this case the voltage fluctuation could be a bit higher (so of course one should filter harder--no free lunch). Unrealistic non-music testing conditions applied to audio amplifiers (inapplicable test), over-current management (retail current dumper), undersize transformer (bad design), and diodes that list a more aggressive voltage drop at current in the datasheet graphs (abused rectifier) are other possible causes of discrepancy that may exceed the 1v-5v range. As you can see, most of what looks like exceptions is really just noise. Meanwhile, back to filtering off the noise.
We should firstly filter noise off a basic power supply to reduce voltage fluctuations from HF noise/nuisance charge. And, with decent filtering (filters on transformer primary, secondary and diodes) the voltage fluctuations are reduced to fit his 1v to 5v range.
Noise is a weak current source that charges caps extraneously higher until load pulls down the extra, and thus noise IS the primary cause of extraneous voltage fluctuations. So, first we filter that. Later, after filtering off HF noise, we can consider series filters, such as CRC. I've constrained my CRC to 1.4v drop (at most), and the results still fit within AndrewT's guidelines.
P.S.
There are possible exceptions:
It is possible that a Class D power amplifier has insufficient idle load to pull down noise/nuisance charge, and in this case the voltage fluctuation could be a bit higher (so of course one should filter harder--no free lunch). Unrealistic non-music testing conditions applied to audio amplifiers (inapplicable test), over-current management (retail current dumper), undersize transformer (bad design), and diodes that list a more aggressive voltage drop at current in the datasheet graphs (abused rectifier) are other possible causes of discrepancy that may exceed the 1v-5v range. As you can see, most of what looks like exceptions is really just noise. Meanwhile, back to filtering off the noise.
Last edited:
you should look at the Pioneer SPEC2 amp psu, standby rails of 91 volts drops to 75volts at full power.....these are actual measured data....not something imagined.....
link.....Pioneer Spec-2 | Owners Manual, Service Manual, Schematics, Free Download | HiFi Engine
link.....Pioneer Spec-2 | Owners Manual, Service Manual, Schematics, Free Download | HiFi Engine
Long thin cable from transformer to board and other creative current management (current dumpers) is standard for most retail builds. It is extra cheap protection. Necessarily, decreased return/defect and lower cost is very important for retail mass market amplifiers. The disadvantage is that mass market style current dump protection makes a given transformer act like a bit smaller transformer.you should look at the Pioneer SPEC2 amp psu, standby rails of 91 volts drops to 75volts at full power.....these are actual measured data....not something imagined.....
Suggestion: build your own hi-fi instead.
Alternative suggestion: it should be possible to alter a current dumper retail supply into a CRC supply without losing protection--just relocate the loss, more usefully, into the CRC resistors; however, this guess needs double-checked for safety. And I would also use the Fairchild Stealth 1200v diodes to make a bridge rectifier with an attractive little bit more current dumping protection. Anyway, I think we can do better than a retail current dumper cable. But doing better means the costs are not suitable for mass production--and that is something to remember when you are shopping retail audio.
EDIT:
Also a 250 watt per channel (total 500w) amplifier can be expected to have a larger voltage drop.
Last edited:
what's with the 1 to 5 volt drop anyway? how was it arrived at?
i know that the same amp circuit using different sizes power transformers will have different voltage sags.....
for full power sine wave testing, for as long as the power transformer can deliver required voltage and currents at the same time, then that should be a valid design....i have been building power transformers longer than amplifiers, since 1970, i think i have something to say...
maybe the more knowledgeable members here can investigate further....
to make this amp drop rails in the 1 to 5 volt range i estimate that i need to build a transformer at least 3x as heavy as what i have in my super leach amp....
i know that the same amp circuit using different sizes power transformers will have different voltage sags.....
for full power sine wave testing, for as long as the power transformer can deliver required voltage and currents at the same time, then that should be a valid design....i have been building power transformers longer than amplifiers, since 1970, i think i have something to say...
maybe the more knowledgeable members here can investigate further....
to make this amp drop rails in the 1 to 5 volt range i estimate that i need to build a transformer at least 3x as heavy as what i have in my super leach amp....
I believe that the 1 to 5 volt drop is "average home amplifiers" and not referring to big PA amplifiers. And I guess that the 1 to 5 volt drop is using excellent filtering (snubbing the transformer primary, secondary and rectifier) to minimize HF nuisance charge voltage fluctuation. It just seems to be a quality control check.
Tony,
I understand what you are saying when you say that your amplifier is not a PA amplifier. The discussion has been generally revolving around amplifiers of lower output in the 50 to 100 watt range and like you I feel that in some applications this is not realistic. With high efficiency loudspeakers then those power levels are fine, but with low efficiency speakers of say 84 to 86db per watt output then you are looking at much more power to drive them to any kind of high level of output. I think that we can all agree that with low efficiency speakers you will be doubling up the power output for each 3db of increase in sound level at a fast clip. But at the same time I think that the conclusions drawn from the lower powered amps should still follow the same rules and you will have to scale the application to your power requirements. The filtering and smoothing of the transformer will not change and the component layouts will still follow the same pattern, just much more current and or voltage will be needed to get there.
Steven
I understand what you are saying when you say that your amplifier is not a PA amplifier. The discussion has been generally revolving around amplifiers of lower output in the 50 to 100 watt range and like you I feel that in some applications this is not realistic. With high efficiency loudspeakers then those power levels are fine, but with low efficiency speakers of say 84 to 86db per watt output then you are looking at much more power to drive them to any kind of high level of output. I think that we can all agree that with low efficiency speakers you will be doubling up the power output for each 3db of increase in sound level at a fast clip. But at the same time I think that the conclusions drawn from the lower powered amps should still follow the same rules and you will have to scale the application to your power requirements. The filtering and smoothing of the transformer will not change and the component layouts will still follow the same pattern, just much more current and or voltage will be needed to get there.
Steven
Very interesting thread, I've only read the first 15 pages so far in one seating, will need to do more later after work.
Now, my question is.
What is the reason behind the sound being more "musical?" with smaller caps vs bigger caps? Shouldn't the sound be the same with the big caps, vs the smaller ones not being able to handle the demands for lower end?
thanks!
if we put instrument's in subjective position, i think is end of the world.
Instrument measure it very well,maybe...not ear.
I agree that this thread contain many good info and good job of gotee,as others have post formulas,idea, etc. but i feel that at end, with new measures (with real instrument's) and comparition, we return at begin of questions.
We all know that multi small caps are better, instead one big. this is old (but can not reduce drop voltage at 1-5V). (or i not have indagate well..
Is my opinion that: 50-100w without crossover in loudspeaker, can work well with good job on psu.
----------------------------
With 200w and crossover (this mean peak 750w) medium level at home. please, if someone have (as me) mosfet amp 200w8R and 3 way B&W or other loudspeaker, measure it while listen a good Jazz.
Obviously I do not want to listen to all musical instruments, soft, with the dynamics far from the reality.
in a few days, I'll put the measures, and the photo of psu + module mosfet amp.
we can use as the comparative with others.
...while I wait to see the measure on psu 1-5V
regards
Instrument measure it very well,maybe...not ear.
I agree that this thread contain many good info and good job of gotee,as others have post formulas,idea, etc. but i feel that at end, with new measures (with real instrument's) and comparition, we return at begin of questions.
We all know that multi small caps are better, instead one big. this is old (but can not reduce drop voltage at 1-5V). (or i not have indagate well..
Is my opinion that: 50-100w without crossover in loudspeaker, can work well with good job on psu.
----------------------------
With 200w and crossover (this mean peak 750w) medium level at home. please, if someone have (as me) mosfet amp 200w8R and 3 way B&W or other loudspeaker, measure it while listen a good Jazz.
Obviously I do not want to listen to all musical instruments, soft, with the dynamics far from the reality.
in a few days, I'll put the measures, and the photo of psu + module mosfet amp.
we can use as the comparative with others.
...while I wait to see the measure on psu 1-5V
regards
not NICO, but i believe there is no explanation that can satisfy everyone....Nico's opinion is true for himself only....YMMV...
(I'm back from travel.)
AndrewT,
The calculation you were looking at was for calculating the ABSOLUTE MINIMUM capacitance. You reacted as if you thought it was giving a recommended design value!
The maximum supply voltage drop at full output power can be as small as one wants it to be, obviously. But that's basically the whole point of this thread! HOW SMALL IS "small-enough"? But mainly, WHY? And, AFTER that, how do we easily predict the reservoir capacaitance needed, for a particular amp and specs?
As you should know, I had been focusing, FIRST (in order to try to find an easy way to set a lower bound, just as the FIRST STEP), on finding ways to predict the ABSOLUTE MINIMUM reservoir capacitance that would keep the charging pulses from causing gross distortion of the output signal.
And with DF96's correct insight, I was able to see that, with the amplifier load in the spice circuit I've been using, the MINIMUM usable capacitance is the one that just barely doesn't allow the voltage rail (at the bottom of the ripple voltage waveform) to force the amplifier to have any less voltage across it than is physically possible.
The calculation you were looking at was for calculating the ABSOLUTE MINIMUM capacitance value.
You reacted as if you thought it was giving a recommended design value!
Changing the subject, momentarily, to what an acceptable lowest-excursion of the power rail voltage might be:
I have no good rule, yet, except that the power rail voltage had better not crash into the absolute minimum, i.e. vload + vamp .
But, what would be so wrong with a worst-case, at MAXIMUM rated output power, that got "kind of close" to that but was guaranteed to never reach it?
No harm no foul, as I see it, in one sense, but I'll also bet that the output distortion figures will be lower when it DOESN'T approach the absolute minimum, as closely.
However, if the distortion difference was very small, between the "huge dips" case and the "1-5V dips" case, then we could possibly have to say that it's fine to have huge dips in the rail voltage.
I posted such distortion figures (distortion vs reservoir capacitance, for C from below the absolute minimum to very large), for several cases, but the distortion figures were not well-calibrated, back then. I do have the calibration method very well-refined, now, for the simulation setup I'm using. I will start posting some plots as I get groups of them finished.
I realize that square wave distortion and sine THD are not necessarily good measures of an amplifier's sound quality. But we do also have reports that lower reservoir capacitance (and therefore larger dips in the rail voltage) "sounds better". Other differences in sound quality seem likely to be due to differences in the transient response (since distortion pretty-much covers the quality of a steady-state response). The transient response will also get looked at, here. Personally, I usually tend to think that the transient response is more important than the steady state response (or, at least, more important to think about).
I realize, too, that I am moving rather slowly, toward the goal of finding a way to pick the value of the reservoir capacitance. Investigating how to at least put a lower bound on it was interesting and maybe it should have been more-obvious to me. But I learned a lot, the way we went about it, and actually verified and validated some simple "well known" ideas.
The next step, as I see it, will be to see if there's a simple rule for calculating the reservoir capacitance that puts the steady-state distortion just beyond the "knee" of the distortion vs C curves, which always start out almost vertical, then curve quickly at the knee and become almost horizontal. (Maybe we just found most of that rule, based on the ripple voltage excursion etc, but we need some easier way to express it, and first need to see how that point relates to the actual distortion numbers, and how their "diminishing returns" behave, vs increasing C.)
After that "knee" point, the distortion doesn't improve much at all and we might be able to assume that there would be no good-enough reason for adding any more capacitance, after that (unless someone can give a convincing argument for not having downward excursions of vrail beyond a certin percentage of nominal, for example).
So I THINK that we'll be looking for a rule of thumb that will relate that distortion knee to the load impedance, output power, and transformer Vrms and VA ratings. I'm hoping that the the distortion plot knee will relate to something that the standard equations already use as a constraint, such as a maximum ripple amplitude (lowest Vrail excursion), which would then be applicable, more or less, to all cases, although it might again relate to the minimum possible voltage across (or impedance of) the amplifier, or whatever is sitting between the rail the the top of the load.
Maybe it's already obvious to some others, here. But it would be nice to show it working out as expected, for at least several disparate example configurations.
By the way, what's a good/reasonable LOW LIMIT for frequency? I have been using 25 Hz. Is that low-enough?
Cheers,
Tom
Last edited:
In relation to how 25Hz has been used to calculate, the caps are likely to get charged twice as music plays, worst case is when the charge and discharge are 90 degrees out of phase. Use 50Hz or 60Hz. I showed something interesting earlier http://www.diyaudio.com/forums/power-supplies/216409-power-supply-resevoir-size-7.html#post3098100 ,http://www.diyaudio.com/forums/power-supplies/216409-power-supply-resevoir-size-8.html#post3098155
http://www.diyaudio.com/forums/power-supplies/216409-power-supply-resevoir-size-7.html#post3098100 ,http://www.diyaudio.com/forums/power-supplies/216409-power-supply-resevoir-size-8.html#post3098155
Last edited:
my own personal non proven theoretic fact is that the last cap should be big enough to 'make it on its own 99% of the time
the rest are 'just' backup and smoothing, if any other needed
then we can still argue about 'big enough'
but very big caps are simply slow in my opinion
as in many other things in life, going big introduces lots of problems
the rest are 'just' backup and smoothing, if any other needed
then we can still argue about 'big enough'
but very big caps are simply slow in my opinion
as in many other things in life, going big introduces lots of problems
80,000uF timer can be like driving with the parking brake on.Nico Ras said:
When a charging cap is an open door to noise, and when the power board has such large capacitance to form a slowdown timer, we have to interrupt this action with capmulti, or regulator, or a series schottky/fast diode prior to cabling the power to the amp board, because we don't want the amp board caps to charge with the same slowness as the big power board reservoir.
Materials list:
Schottky Left V+ at the output of the power board
Schottky Right V+ at the output of the power board
Schottky Left V- at the output of the power board
Schottky Right V- at the output of the power board
(Amplifier board power cables attach to the schottky.)
Of course one could use regs or capmulti, which is more orthodox, but schottky or fast silicon diodes are cheaper and easier.
Credit to diyaudio.com member "theprof" for researching this inexpensive repair.
P.S.
Some applications may be partially effective with an ordinary cable instead. My suggestion is greatly applicable to the "all in one" amp board power supply made all together as a single unit. But, my suggestion is less applicable to other layouts that have an interlink cable.
P.P.S.
Other fixes include smaller size low esr caps for amp board (improve quality, reduce size--low esr caps usually win the fight for charge), and/or (sounds just like) adjust your compensations to sharpen up your square wave test for more ear perking fun than rounded.
Last edited:
But, but, but, a larger total capacitance (whether all in one big cap or in several paralleled) reacts FASTER, in the sense that it takes a smaller (or slower) drop in the "ambient voltage" to make any given amount of current surge out of them.
Or do you mean the rise-time is slower? i.e. the time it takes to go from zero to the discharge level, for a particular lowering of the "ambient voltage".
Or are you thinking about the RC time constant, and the longer, slower decay time of the "e to the minus t over RC" equation for the voltage across the cap? In this case, the longer the better, since that's what keeps the ripple amplitude small, and enables the caps to power the output all by themselves, between charging pulses, i.e. it enables them to last until the next one.
Maybe you are referring to ideas derived from digital circuits, where the decoupling caps are said to need to be "small", to be fast-enough. In that case, they are talking about having low-enough inductance to be fast-enough to provide the fast transient currents. And that does depend on the physical smallness of the capacitors (and the shortness of their connections). They even go so far as to turn the surface-mount caps so that the smaller of the two dimensions is across the decoupling points.
BUT STILL, you would want to use the LARGEST capacitance you could get, in the small package size, for those digital decoupling applications.
So I'm not sure what would be considered to be "slow" about larger electrolytics, UNLESS we are talking about transient response. THEN I TOTALLY agree (because of inductance due to larger dimensions and connection distances). HOWEVER, that job, the "fast" part of it at least, will be handled by the DECOUPLING CAPS, which HAVE to be very near the points of load (to be fast-enough), not by the reservoir caps.
Last edited:
60 cycles per second in the USA (from an amp it sounds like 120hz distortion), plus a lot of higher frequency bonus noise from every appliance that you and all of your neighbors who are on the same transformer have either running or switching. The bonus noise is why I've been promoting filters installed at both the primary and secondary windings of transformers. Even if an audio amp has excellent advertised power rejection, it is still better to avoid inserting noise into the audio amp. Filtering off the HF trash is the first step towards realistic digital replay and, surprisingly, that filtering need also includes the audio amp's transformer. Results do vary--some are dramatic and others are not.Hi Daniel what is the switching frequency of mains
Last edited:
Daniel,
The RC time constant of the reservoir caps should not be a problem, when charging the decoupling caps on the amp board. You're probably thinking of the slower exponential decay of a larger capacitance's voltage, as it discharges current through a resistance. But look at the very beginning of that curve. There is no slow ramp UP to it. The current starts flowing blindingly quickly, and at its highest rate.
The power rails themselves do have resistance and self-inductance that could slow down the charging of the decoupling caps (not by a lot, though). But larger reservoir caps should only help make it faster and better for the decoupling caps, as far as I can see.
As far as what Nico experienced, there were far too many variables to know why he heard what he heard. What were the rail inductances and resistances (i.e. the physical lengths)? What were the ripple amplitudes on the rails, and how were they changing? What was the average voltage on the supply rail and how much was it changing, due to the music signal? What decoupling capacitances were used at the points of load? What were the transformer Vrms and VA ratings? Etc.
As far as bass is concerned, there IS a minimum reservoir+decoupling capacitance that must be present, for the system to be able to handle the demands for current at bass frequencies, since low-frequency bass waves have periods (1/f), or half-periods if only one polarity is considered at a time, that could make them need current for the entire time between charging pulses. Can capacitors charge and discharge at the same time? I think they must. But I wonder if it would help to offset the voltages at which several parallel caps would each start charging (and discharging).
Cheers,
Tom
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.