I don't think it's quite that simple.
If we don't take John's idea of not losing sight where we're going literally but by essential meaning, as I understood him I guess the best power supply is the one which satisfies the requirement of a specific topology best. Meets all of its needs, with a healthy safety margin. Some trade-offs are inevitable.
But then, there are quite a few topologies around, when one takes into account the also available variations on any theme.
So how can there be one ideal power supply to fit all perfectly?
Stick your ear up next to my speakers with the volume turned up. It's a pretty simple test. Ask Jan what he heard when he did it.
Sorry if practical reality gets in the way sometimes when you want to obscure a simple issue for fun and profit.
It would be nice to see the 'basic math' on this.simon7000 said:
See Mains Quality for a plot which shows the advantage of a poor mains supply.
Do a lot of electronic and ear only tests, what else?
Sometimes, odd things happen which seem to defy logic. If you are persistent enough, you will get to the bottom of it, but may still be left with some appearently odd choices. Polypropilene is a good example: it generally works better than most with decoupling audio circuits, but on occasion, it may turn out to be the best choice for PSU lines as well. Especially if you use a wide bandwidth transistor, like MJE 15030/15031; ON Semi say they are 30 MHz, but just look at their data sheet and you'll see that for typical reg currents for line work, delivering 100-300 mA, they in fact don't drop below 45 MHz.
You are confusing nonlinear loading of the AC mains resulting in flat topping with the generation of harmonics at the point of load switch on.
I have previously posted the spectrums that result from the sudden change in current. You are welcome to do your own measurement and analysis of a step impulse due to diode switch on.
If you look carefully at your cite and calculate the voltage levels that would result from injecting 1 watt into an AC power line, (Source resistance around .1 ohms) which results in around .316 volts AC at the 120 volt primary and see that after stepping down through a bandwidth limited transformer and filtering there is still around 200 microvolts. This is a filtering effect on the voltage distortion of 60 db. Now for a good PSSR circuit that should be enough. But that is just the noise introduced by a single 20 watt power supply load. Add in all the DC power supplies on the mains distribution transformer and that 20 watts can actually be more than 1,000 watts. So even circuits with decent PSSR can still have problems.
Last edited:
Stick your ear up next to my speakers with the volume turned up. It's a pretty simple test. Ask Jan what he heard when he did it.
Sorry if practical reality gets in the way sometimes when you want to obscure a simple issue for fun and profit.
Now, that's a strong scientific point, to be sure.
If anyone else said it that way, he'd be promptly asked for measurements and scientific paper references.
You gotta love objective double standards!
Stick your ear up next to my speakers with the volume turned up. It's a pretty simple test. Ask Jan what he heard when he did it.
Sorry if practical reality gets in the way sometimes when you want to obscure a simple issue for fun and profit.
And again you miss the point. Power supply noise affects perception more than just a noise floor. Also what numbers are associated with the "volume turned up?"
What is the gain of your preamp, power amp and sensitivity of the loudspeakers?
I have some loudspeakers that do better than 109 db/w @ 1M, would they still be "dead quiet?"
Now as to obscuring issues, what you were bragging about was that your preamp using a load of a low noise high PSSR current source is less power supply critical than other designs. Perhaps there are other designs?
Last edited:
That's a remarkable claim. Any evidence?
My speakers are 87dB/2.83V/1M. Not terribly efficient or inefficient. Noise at the output of the power amp, 20kHz bandwidth, is -87dB referred to 2.83V. The limitation is not the power supply, it's the input stage.
It is the 1W assertion for which I would like to see the 'basic math'. Never mind, I have done a back-of-envelope calculation myself. 20W PSU, 120V RMS supply, 0.1R source resistance - I will use your figures.simon7000 said:
120V RMS gives a peak voltage about 170V. 20W DC output then means about 120mA DC. For a typical cap input supply the RMS current in the AC circuit is about 2-3 times the DC current output. Lets say 2.5 times, so we have 300mA RMS. Across the 0.1R source impedance this dissipates 9mW. This is the total power sent back into the incoming mains, so in any particular frequency range the power will be less.
Now from where did you get your 1W figure?
I am not an audio professional but maybe some insight could be gained by considering what an audio amp's power supply is supposed to do: The signal that gets passed to the load of the final active amplifying device (e.g. a transistor) is a current. So the first job of the PSU is to accurately supply current, on demand.
I realize that a stable power supply voltage for the active amplifying device (and other parts of the circuit) can also be important, especially if PSRR is not high-enough. But the voltage's stability is still secondary, since the accuracy of the output voltage depends on the ability of the PSU to supply a precisely-controlled current signal. Click on the link in my signature.
If PSRR is a problem, then... you have a problem. Capacitors will only release current if we make their voltage fall. The only obvious solutions (to me), with a linear supply, are: 1) larger capacitance values, and... 2) feedback.
After that is solved well-enough, then the fun can begin. The "real" signal path is the path of the current that goes directly from the PSU and decoupling caps to the load, through the active output devices, which are just current- or voltage-controlled current valves. So it seems that that signal path should be optimized for extremely-accurate supplying of current, for all relevant frequencies/waveforms. To me, that would mean cap selection and sizing, and low-inductance layout and routing, maybe with an emphasis on decoupling.
I realize that a stable power supply voltage for the active amplifying device (and other parts of the circuit) can also be important, especially if PSRR is not high-enough. But the voltage's stability is still secondary, since the accuracy of the output voltage depends on the ability of the PSU to supply a precisely-controlled current signal. Click on the link in my signature.
If PSRR is a problem, then... you have a problem. Capacitors will only release current if we make their voltage fall. The only obvious solutions (to me), with a linear supply, are: 1) larger capacitance values, and... 2) feedback.
After that is solved well-enough, then the fun can begin. The "real" signal path is the path of the current that goes directly from the PSU and decoupling caps to the load, through the active output devices, which are just current- or voltage-controlled current valves. So it seems that that signal path should be optimized for extremely-accurate supplying of current, for all relevant frequencies/waveforms. To me, that would mean cap selection and sizing, and low-inductance layout and routing, maybe with an emphasis on decoupling.
Last edited:
People seem to be getting pretty confused here ... noise in the power supply area doesn't translate into "noise" from the speakers - as in a hiss, or gurglings, mutterings or suchlike from the speakers. No, it results in good ol' distortion - as in, dirty, grey, unkempt sound, the sort of reproduction that bores one, makes one wonder why anyone bothers listening to it - you might call it, cigarette ash sound, ...
Working hard to get the PS to be on its best behaviour gets rid of that cigarette ash quality in the reproduction, IME ...
...Working hard to get the PS to be on its best behaviour gets rid of that cigarette ash quality in the reproduction, IME ...
Of course, by measurement and listening.
But first we have to define noise as any unwanted signal.
Now the simple case is where the 1/F noise moves the woofer cone and introduces Doppler distortion in the audio signals below the crossover frequency. (But you knew that, you just want to change the subject from your lack of data on dead quiet.)
The second common case is when RF range noise causes increased distortion in the Vas stage, probably by a shift in bias due to some rectification at the base-emitter junction.
One really nifty case is where the RF range noise increases the bias in an output stage. (Actually a problem in a football stadium I did. Only happened in a few amps.)
But then I already mentioned that increased harmonics in the midrange may not be heard by themselves but will mask other information. (The ear can pick out tones quite a bit below the noise floor.)
Now why don't you ask about double blind tests of these effects to complete the circle of you don't have to but everyone else does.
Of course, by measurement and listening...
Now why don't you ask about double blind tests of these effects to complete the circle of you don't have to but everyone else does.
Yes, I get that you don't want to acknowledge the difference between an ordinary and extraordinary claim. But still, you've made an extraordinary claim- "Power supply noise affects perception more than just a noise floor." Do you have any evidence for this?
A 120 Volt AC power supply at 20 watts draws an RMS current of 167 ma. The actual duty cycle of current draw starts when the secondary voltage after the rectifier becomes greater than the voltage stored on the filter capacitor, it stops when the capacitor is charged to the peak secondary voltage minus the rectifier drop. Now a low power transformer will sag a bit when first loaded and then rise in voltage as it is unloaded. This unloading extends the charging time. Typically the charging current for a 60Hz. line is 3 times the RMS current so the peak charging current will be .5 amps.
Now in your method you are confusing the energy into the load with the energy generated at the source. A 20 watt power supply with a load of .167 amps will indeed only create a voltage drop to add to the flat topping of .5 amps into the AC line source impedance (which is now the load!) of .1 ohms or a drop of 50 mV. Of that only a few millivolts will be in the energy band of concern. Your calculation of this being 9 mW would correspond to a launch energy of 3.6 watts. (You do correctly note that only part of this is midband.) Having measured the harmonics it really is about 5% of the supply power.
Now what I have also shown is that this energy is not just dissipated in line and source resistances but some is radiated into the surrounding chassis and can actually provide a few mV of potential across the chassis even at the very low resistances in a sheet metal chassis. Of course that is only part of the energy dissipation.
The cite you showed had a typical 5% line voltage flat topping. With a line resistance of .1 ohm and a 6 volt flat topping that would be a surge current of 60 amps. (However the .1 ohm figure is typical of AC line branch wiring where the line source resistance from generation to local power panel is lowered by the shunt resistance of the loads.)
Of course the conclusion is that most of the power supply noise injected back into an AC power line is filtered by the lines impedance itself and less by the actual power supplies filters. The big but is that there is so much noise from all the other sources that a power supply is require to filter out noise in addition to reducing ripple. (Of course a much larger amount of the local switching noise will have to be handled by the power supplies filter circuits.)
Last edited:
Gee and here I thought you were making extraordinary claims. The issues I cited were noticed and were sent out for repairs. I don't think you get a blinder version of a test than somebody walking up cold to a sound system and asking if it is broken.
BTY if your loudspeaker has 87 db sensitivity @ 1 M and the amplifier chain has 87 db s/n then noise should be heard as the midband threshold is around -6db for normal hearing and the ear by the loudspeaker should provide another 20 db of gain!!!! That is why I consider your claim extraordinary. ( I know Jan, he doesn't seem to have a 26 db hearing loss. So a reasonable conclusion would be your listening environment is noisier than 26 dba.)
Let me try three guesses, your room is not that quiet, your room is not that quiet, your room is not that quiet.
Have you ever measured the noise level in your listening room?
Last edited:
As has been stated, the more one investigates what really goes on in the process of extracting usable energy for your circuit from the mains the more complex the whole situation resolves to be. The simplicity of the typical "how to add a PS!" treatise conveniently ignores a whole swag of detail devils ...
Three points for anyone who can name all of the false assumptions here.
So I take it from your continued kicking up of dust that you have no evidence for your claim?
(Actually, the reason I brought up Jan is that when I was troubleshooting a software issue preventing some music from playing, he stuck his ear at the speaker and thought that the system was powered down and that was the cause. So since you cite that as an acceptable blind test, then yes, I did one)
There once was a guy that walked around the North End in Boston with a board covered in fantastical numerical coincidences offered as profound prophecies.
- Status
- Not open for further replies.