Thank you very much When you say is not very big i immediately think to a bottleneck I have noticed the secondaries They are not thick as i would like
Unfortunately i cannot measure they impedance A bigger transformer same VAC out should be beneficial Maybe even more than bigger caps ?
If you have to choose where to put money would you buy a higher VA rating transformer of add capacitance ? i could do that in two steps
But which first ? trans or caps ? that is the question
Once you are at 20,000 uF per rail you don’t need any more. Get there first. If you are not there yet get there. Then go to bigger VA. Anything over about 400 VA and 6 amp diodes won’t be big enough anymore, though.
Upping the trafo VA wont change much at normal listening levels, but at high volume you will not notice clipping as much. So you WILL be tempted to play louder, at least some of the time. If you end up running it harder than you used to, it will generate more heat in the output transistors - just because of higher average power.
When one is used to using 800 to 1.5kVA transformers in his builds, anything else seems “small”. A 225 VA may seem “large“ to many. Matter of perspective.
Upping the trafo VA wont change much at normal listening levels, but at high volume you will not notice clipping as much. So you WILL be tempted to play louder, at least some of the time. If you end up running it harder than you used to, it will generate more heat in the output transistors - just because of higher average power.
When one is used to using 800 to 1.5kVA transformers in his builds, anything else seems “small”. A 225 VA may seem “large“ to many. Matter of perspective.
It is not theory but practice. There are small electrolytic capacitors and those small diodes withstand impulse currents from the secondary, Increase the capacitors significantly, or put modern low ESR capacitors, and it can be anything. I use PSUD2 free simulator for those simulations.
Hi thank you very much indeed This is a very valuable advice to me The two now present should be around 2x5000uF more or less
So a reasonable size should be 5000uF per channel per rail ? Good ! very important point to fix
About transformer VAs i guess that a rule can be establish. The present should be around 200VA Considering an efficiency of 50% that means that there are 100 VA available for the two channels and 50VA for one channel That would mean 50W on 8 ohm per channel ?
It seems to me that the transformer VAs in some way fix the limit of the power obtainable from an amp
And actually the first thing i look in an amp is the power consumption usually reported on the back panel
When i see something below 300VA i am not happy
You are right and i can say that the amps that i have listened to which sound more relaxed and in control had all very big transformers And possibly also of very good quality
They are one of the most expensive part in an amp and where the savings are done like big supply caps as well
On the basis of your words and considering that i am looking at high efficiency speakers the caps upgrade seems really the first step
However i will have to remove the board to do that and in the process i will put some more robust diodes as well
Thank you very much again
The voltage usually sets the power output limit in an 8 Ohm load, the current the output limit in 4 Ohm loads or lower, assuming the power transistors can handle the load. The heatsink needs to be considered to, for sinus output. This amp seems to have no (?) heatsink, only the case? So your continuous power is likely not going to increase when you up the VA's of the tranny. So, it is not just one variable, but they all have their own role in the mix.
Transformer VA simply sets the curve of output voltage vs. the current drawn. Manufacturers simply size the VA according to the current drawn under normal music conditions, assuming one is turning it up till you just hear a bit of clipping (distortion). Continue to turn it up, and the voltage continues to sag along that curve. Clips earlier, you hear more distortion. Bigger transformer will drop more slowly as current increases, so you won’t hear as much clipping. The point at which it just starts to clip doesn’t move, but you can drive it further into clipping before the sound gets too nasty to tolerate. It therefore SOUNDS louder. To put out full power sine wave indefinitely, it needs to be sized OVER 2X the output power at minimum impedance. No one ever does, unless it a cost-no-object design. It can be sized anywhere between the two extremes, depending on how much performance you are willing to pay for.
Heat sinking is the same way. That chassis probably provides ALL that is required when the transformer is operating at its VA rating, with the music turned up till it’s just clipping. It will not survive war volumes. If you want that, you up the heat sinking, use a fan, or both. If you want to calculate what you need ahead of time, you need to simulate your actual use condition with expected power output. Calculated based on the maximum possible dissipation will result in a very large impractical heat sink in many cases. A good compromise is to put it on something practical, run it high as you ever will for an hour or so (into a dummy load) and measure the temperature rise. If you are not comfortable with the result, you need more.
Capacitors are far easier to calculate. You simply want the IMPEDANCE of the power supply (aka capacitors) to be significantly lower than the speaker impedance at the lowest frequency you intend to reproduce. Look at the pole frequency (aka 3 dB cutoff) between the combined impedance of the speakers loading all channels and the cap value. Then decide if that Number is low enough for you. 20,000 uf and 2 ohms (both channels loaded in 4 ohms) is 4 Hz. That is the magic “five times away“ from 20 Hz. If you bass needs are modest or speakers high impedance the caps can obviously be reduced. Below the cutoff frequency, the output voltage of the supply drops FASTER with in increasing current draw than above it… by you guessed it… 3dB. “Five times away” from the 3 dB cutoff is 0.5 dB which is 10% down. It is simply filter theory at work. Ripple current ratings need to be respected too for long cap life, but following the 20,000 uF guideline is more than enough for up to 1000 watts. Above that it can get iffy. And if you design for reduced bass output your ripple current demands will tend to relax quite a bit.
When building a price point amplifier, you are going to make compromises in all 3 places. Well, there is a fourth too - the SOA of the output transistors, which dictates how low an impedance you can run before magic smoke appears. Lateral mosfets are actually very good in this regard.
Heat sinking is the same way. That chassis probably provides ALL that is required when the transformer is operating at its VA rating, with the music turned up till it’s just clipping. It will not survive war volumes. If you want that, you up the heat sinking, use a fan, or both. If you want to calculate what you need ahead of time, you need to simulate your actual use condition with expected power output. Calculated based on the maximum possible dissipation will result in a very large impractical heat sink in many cases. A good compromise is to put it on something practical, run it high as you ever will for an hour or so (into a dummy load) and measure the temperature rise. If you are not comfortable with the result, you need more.
Capacitors are far easier to calculate. You simply want the IMPEDANCE of the power supply (aka capacitors) to be significantly lower than the speaker impedance at the lowest frequency you intend to reproduce. Look at the pole frequency (aka 3 dB cutoff) between the combined impedance of the speakers loading all channels and the cap value. Then decide if that Number is low enough for you. 20,000 uf and 2 ohms (both channels loaded in 4 ohms) is 4 Hz. That is the magic “five times away“ from 20 Hz. If you bass needs are modest or speakers high impedance the caps can obviously be reduced. Below the cutoff frequency, the output voltage of the supply drops FASTER with in increasing current draw than above it… by you guessed it… 3dB. “Five times away” from the 3 dB cutoff is 0.5 dB which is 10% down. It is simply filter theory at work. Ripple current ratings need to be respected too for long cap life, but following the 20,000 uF guideline is more than enough for up to 1000 watts. Above that it can get iffy. And if you design for reduced bass output your ripple current demands will tend to relax quite a bit.
When building a price point amplifier, you are going to make compromises in all 3 places. Well, there is a fourth too - the SOA of the output transistors, which dictates how low an impedance you can run before magic smoke appears. Lateral mosfets are actually very good in this regard.
therefore if i get it right keeping a same VA rating if one does not need much power better to use low voltage rails ? very interesting
I have to measure the voltages in this unit
It is better to have it connected to a load ? do i have to short the inputs ?
absolutely Only the case
I have to measure it first What i do not understand is that if at idle i measure the voltages on the mosfets pins combined with the datasheets will not be possible to find replacements of any kind
This sounds surprising to me What other information are needed ?
Thanks a lot for the accelerated course in amp design In the past i did some rudimental tests replacing caps and transformers
Every time using more uF or VA changed the sound for the better
One amp kept on sounding at low volume for almost one minute after being switched off
There's a trivial way to check the adequacy of the power supply
To place a voltmeter across the main caps
If the needle stays always steady is fine
If it sags it's not good
Maybe using a heavy dance track at high volume on a dummy load
A very basic but extremely telling test imho
Every time using more uF or VA changed the sound for the better
One amp kept on sounding at low volume for almost one minute after being switched off
There's a trivial way to check the adequacy of the power supply
To place a voltmeter across the main caps
If the needle stays always steady is fine
If it sags it's not good
Maybe using a heavy dance track at high volume on a dummy load
A very basic but extremely telling test imho
Last edited:
Re: 20,000uF..
I suppose it has to hold most of its voltage for eons (e.g. 1 / 100th of a second) between charging pulses? And the amount of ripple depends on the audio, so even below any clipping, the rails will encounter cyclical sagging and top-up, which the amplifier has to work hard to filter it out.
In my mind an SMPS should be a better solution. Yes, it has HF ripple and noise, BUT with each cycle (less than 1 / 20,000th of a second) all the accumulated sag gets reset at a fast rate.
Probably the best load you could hope for, from a power supply perspective, is if the amplifier is single-ended with a resistor load and runs extremely hot with a high bias. That way the audio only changes the overall load by a small fraction of the constant load. And it's not highly distorted either, as would be the case with class-B, or AB, or push-pull class-A (in order of decreasing severity), where sine waves get rectified. So, if it absolutely must be a brute transformer, the amplifier should not be too aggressive in splitting the audio into half-cycles for high efficiency. So it's kind-of a Catch 22, all or nothing thing, where you ostensibly save some power with class B, but still trip the circuit breakers with a giant trafo you could almost use for welding.
I suppose it has to hold most of its voltage for eons (e.g. 1 / 100th of a second) between charging pulses? And the amount of ripple depends on the audio, so even below any clipping, the rails will encounter cyclical sagging and top-up, which the amplifier has to work hard to filter it out.
In my mind an SMPS should be a better solution. Yes, it has HF ripple and noise, BUT with each cycle (less than 1 / 20,000th of a second) all the accumulated sag gets reset at a fast rate.
Probably the best load you could hope for, from a power supply perspective, is if the amplifier is single-ended with a resistor load and runs extremely hot with a high bias. That way the audio only changes the overall load by a small fraction of the constant load. And it's not highly distorted either, as would be the case with class-B, or AB, or push-pull class-A (in order of decreasing severity), where sine waves get rectified. So, if it absolutely must be a brute transformer, the amplifier should not be too aggressive in splitting the audio into half-cycles for high efficiency. So it's kind-of a Catch 22, all or nothing thing, where you ostensibly save some power with class B, but still trip the circuit breakers with a giant trafo you could almost use for welding.
SMPS is a catch 22. In principle, it is just plain better. But they don’t take to overload gracefully, and are far easier to overload with an audio amp than one might think. That 20,000 uF acts as an integrator providing energy between charging pulses AND can deal with the fact that an audio amp itself draws a constantly varying current. Here’s the kicker. The current varies between zero and THREE TIMES the average. If you go through the math it’s all there. With a small output capacitor, the time that it can put out the required three times it’s rating is limited. Play any real bass through it at all, and most supplies will go into limiting and sound positively NASTY - far worse than a transformer supply sagging some 20%. Supplies that can deal with this loading profile gracefully cost real money - FAR more than the equivalent size transformer, 25A rectifier block, and two 20,000 uF caps. You can always just buy a good Meanwell or whatever, at 3X the capacity you calculate you need and of course that will work well as it will simply operate normally at peaks of current draw. Probably better than the linear supply. But for a large amplifier you are talking about a large, very expensive supply. Or two, to make +/- rails out of. Lighter weight for sure, but it will lighten your wallet too.
Some of the cheap pro amplifiers cheat - and design their supplies to allow those 3X peaks to pass un-limited. Documented example: Behringer iNukes. Run them at war volume for a year or two (nightly) and they blow up, un-repairably. Most of the cheap stuff out there just limits and sounds positively terrible at war volume - or has DSP that backs everything off and you might as well just have a 50 watt amp in it.
Some of the cheap pro amplifiers cheat - and design their supplies to allow those 3X peaks to pass un-limited. Documented example: Behringer iNukes. Run them at war volume for a year or two (nightly) and they blow up, un-repairably. Most of the cheap stuff out there just limits and sounds positively terrible at war volume - or has DSP that backs everything off and you might as well just have a 50 watt amp in it.
Last edited:
And as one can see wading through all this, the choice of BJT or mosfet outputs is way down the list of priorities when making a quality amplifier. Either will work well, when designed specifically for it. Some CIRCUITS lend themselves to easy retrofitting one way or the other. Others don’t. @ginetto61 ‘s circuit lacks a push pull driver stage, which would need to be added to go to BJTs. It’s also not a giant amplifier using six output pairs in parallel, requiring a driver to handle all that capacitance.
Last edited:
Thank you for the very interesting advice It would be interesting to try a switching mode dual power supply but they are not very common
I really do not know where to look for them
However i said something wrong To check for some voltage sagging the voltmeter should be placed across the output stage pair and see the needle during the musical peaks Ideally it should not move
If it moves there is some problem somewhere in the power supply That is clear even to me
Maybe even the main caps distance from the output devices play a role ?
The caps recharge 120 times per second I think that the parameter to look at is the caps ripple current ? the higher the better ?
this is a very important issue
I have seen an amp with a quite big transformer and small PS caps I am quite confused
But the voltmeter would provide evidence of some problems in the power supply for sure
Hi ! i have checked again the transformer It is a 40VAC-0-40VAC 😵
after rectification they should be about 55VDC+55VDC
why so high voltage for such low power i do not understand And maybe this high voltage can put under stress the components ? 🤔
I have seen amps of similar power supplied by a double 20VAC transformer
And they did sound very good anyway
I took the opportunity to place under the transformer a silicon disk used for hot cups
The bottom plate used as heatsink must become very hot during use
after rectification they should be about 55VDC+55VDC
why so high voltage for such low power i do not understand And maybe this high voltage can put under stress the components ? 🤔
I have seen amps of similar power supplied by a double 20VAC transformer
And they did sound very good anyway
I took the opportunity to place under the transformer a silicon disk used for hot cups
The bottom plate used as heatsink must become very hot during use