The spreadsheet is, of course, fully documenting all of the math, if you look at the cell formulas and the VBA code.
In LT-Spice, I made the transformer model MUCH more useful by making it a component, with a symbol, and by making its parameters settable from the main circuit's schematic, where it was used.
To do that, I put something like
Vsb={Vrms_setting} Ssb={VA_setting} fb={mains_freq_set} Vpb={mains_rms_set}
in the PARAMS line in the box that pops up when I right-click on the symbol I created for the transformer.
Then I can set the params (that appear in the { }s, above) in the main circuit's ".param" spice directives, and can use them in sweeps!
It's been QUITE useful!
And if you have a variac, it's quick and easy to measure what's needed in order to make the model for a different transformer.
You can download a .asc file for a slightly-older version of the transformer model, from somewhere in this thread. And there should also be a zip file here that even includes the symbol for the transformer, etc.
In LT-Spice, I made the transformer model MUCH more useful by making it a component, with a symbol, and by making its parameters settable from the main circuit's schematic, where it was used.
To do that, I put something like
Vsb={Vrms_setting} Ssb={VA_setting} fb={mains_freq_set} Vpb={mains_rms_set}
in the PARAMS line in the box that pops up when I right-click on the symbol I created for the transformer.
Then I can set the params (that appear in the { }s, above) in the main circuit's ".param" spice directives, and can use them in sweeps!
It's been QUITE useful!
And if you have a variac, it's quick and easy to measure what's needed in order to make the model for a different transformer.
You can download a .asc file for a slightly-older version of the transformer model, from somewhere in this thread. And there should also be a zip file here that even includes the symbol for the transformer, etc.
Last edited:
Bigger isn't always best.
As you say, amplifier "tuning" or the sonic characteristics of an amp can be altered by its PSU design.
Again this is where objectivism and subjectivism might not agree.
A) The amp is better technically with the low ripple massive PSU.
B) The amp sounds better with a smaller "faster" PSU.
So which do you build ? A or B
Some good general useful info here... but it doesn't tell you how godd an amp will sound with a given PSU.
The Signal Transfer Company: Power Output
B) The amp sounds better with a smaller "faster" PSU
the consensus is general on this statement ?
What is the problem with too much capacitance in the PS ?
Actually the best amps around have huge capacitance
Some manufacturers even offer a capacitance upgrade for their amps
Thanks for any explanation.
Kind regards,
gino
Last edited:
Maybe there could be a problem if the transformer was too small. It needs to be able to recharge all of the caps and maybe also supply part of the music signal, during the short times when the rectifiers turn on. I can't think of any reason that more capacitance would be bad, especially if it had low-inductance connections. It should give a lower impedance, which should make its response faster, not slower. The signal is current, not voltage. The larger RC time constant slows the voltage droop but that's a good thing.
The larger the reservoir capacitance the greater the cost and complexity, both in and of itself and in the addition of components such as soft start circuits. Using large cap banks as a workaround for poor PSRR by minimizing ripple in a linear supply also reduces the diode conduction angle in the bridge, which tends to degrade both reliability and power factor. For DIY these aren't often significant concerns, but it is a bit silly to stack up 100,000uF per rail when you can get the same job done with a 50 cent regulator and 10,000uF per rail.
From a sound quality perspective, the narrower the conduction angle the broader the bandwidth of the ripple. So, in addition to there being less ripple, throwing cap at the problem results in spreading the charging energy over a larger number of mains harmonics, reducing their audibility. In an amp with insufficient PSRR to be transparent to the supplies (which would be, um, a lot of them---typical designs are in the 70dB range whereas 100+dB tends to be desirable) this means changing caps makes audible changes in the amplifer's tonal qualities. To answer Gino's question, some people like more supply dirt in their music, others like less.
From a sound quality perspective, the narrower the conduction angle the broader the bandwidth of the ripple. So, in addition to there being less ripple, throwing cap at the problem results in spreading the charging energy over a larger number of mains harmonics, reducing their audibility. In an amp with insufficient PSRR to be transparent to the supplies (which would be, um, a lot of them---typical designs are in the 70dB range whereas 100+dB tends to be desirable) this means changing caps makes audible changes in the amplifer's tonal qualities. To answer Gino's question, some people like more supply dirt in their music, others like less.
Like most things in audio there is not just "an answer". A playback system has to be engineered to work properly, throughout its entirety; and "easy fix" suggestions should be treated very carefully. The suggestion will have parts of the truth, but is not the truth in of itself.
If one thinks of creating a system like designing a bridge then one has a much better chance of getting it to work right. Every component is linked to everything else in some fashion, and just using some "fashionable" technique, or "impressive" idea that someone reckons is the go, is not going to cut it.
Which is a long-winded way of saying that a supply should be both low ripple and "fast". To do the two at the same time, correctly, is just an engineering issue ...
Frank
+1 for having the regulator. I experimented with raw capacitance and reached well over 100,000uF in paralleled 2,200uFs. The LF noise on the supply was lower with a regulator and 'only' 11,000uF but still adding the caps improved the noise above around 1kHz. This could be because I have a low value inductor after the reg (an LM350T) as when I first experimented without an inductor on the reg, the sound was distinctly flakey. I have yet to determine the optimum value of the inductor (its currently about 300nH, air cored).
To return to gino's question, I suspect the problems arising from higher cap values only apply to caps fed from mains freq rectifiers - the charging pulses become shorter and richer in harmonics and are more likely to couple into other parts of the circuit. At these higher freqs, the impedance of the cap becomes ESR limited. The solution to that is paralleling many smaller ones and adding some series L along the way. I rather suspect then that there's no upper limit on improvement by adding capacitance, just increasingly smaller returns for the outlay.
That's what tomchr and I found when measuring the load regulation of LM78xx and LM79xx parts, though with LDOs it depends on the loop configuration.
I think you probably meant to type ESL at high frequencies rather than ESR, or does high mean kHz in this content? Not sure why one would want to add ESL; most regulators land in the uH range on their own.
'High' meant 'above 1kHz' in this context - I agree at even higher freqs - say above audio then ESL becomes dominant. I added the explicit inductor because I suspected (though didn't measure) that the regulator had stability issues without. I have rather a lot of very low ESR ceramics on my supply too... 
Yes the rising output impedance of the reg looks like an inductor - back of the envelope of the order a few hundred nH.
<edit> Just worth mentioning - some incredibly large (hundreds to thousands of Farads) caps are available nowadays. They're not useful because their ESR is too high. They're very limited voltage (2.5V typically) so you have to put quite a few in series by which time the ESR kills them as audio band decouplers.
Yes the rising output impedance of the reg looks like an inductor - back of the envelope of the order a few hundred nH.<edit> Just worth mentioning - some incredibly large (hundreds to thousands of Farads) caps are available nowadays. They're not useful because their ESR is too high. They're very limited voltage (2.5V typically) so you have to put quite a few in series by which time the ESR kills them as audio band decouplers.
Last edited:
I don't try to sim the trafo myself, rather I use a sine source and feed that through a diode bridge. That works well enough to get the waveform into the filter looking how it looks on my scope. If you want to get closer to a trafo then add some inductance and resistance between the sine source and the bridge, depending on the properties of your particular trafo.
One big weakness of LTSpice is that it doesn't sim inductors' HF properties because it knows nothing about skin/proximity effect. Fortunately these two effects both tend to work in our favour, giving us better HF rejection in practice than we get from the sim.
One big weakness of LTSpice is that it doesn't sim inductors' HF properties because it knows nothing about skin/proximity effect. Fortunately these two effects both tend to work in our favour, giving us better HF rejection in practice than we get from the sim.
Last edited:
Yeah, a decent reservoir cap will be resistive to 100kHz, give or take. I've not seen stability problems with MLCCs into either standard dropout regulators (78xx, 79xx, 3x7, and similar) or LDOs. But for the latter I've never found a need to push the datasheet stability requirements. Similar considerations come up with polymer electrolytics; it's not uncommon for the cheapest ones I can get from the distributors which are convenient for me to be Nichicon R5s. Those are lower ESR than most 0805 and smaller MLCCs and are available in significantly larger sizes but they still don't transition from capacative to inductive until well above the regulation bandwidth of most linear regulators. Even a fast audio regulator like a TPS7A4700 is pretty well going to see them as capacitive but I haven't done a build with a 4700 yet.
Haven't worked with an LM350 but I'd be surpised if it behaved much differently from the other standard droput LM regulators. The 338 and 350 are similar enough to the 317 I'm inclined to assume they're pretty much the same part just set to different current limits.
And which solution would that be? Note the date on Michael Maida's app note and the relevant threads here on DIYA. If that was in reference tube and OPT emulation, most of the direct discussion and implementation is on sites other than DIYA but there's not exactly a shortage of unintentional tube emulation here---single ended class A MOSFET builds are a good place to start within DIYA.
Thank you. I have the same sensation actually
I experimented a little.
My impression has been that every parameter improves with more capacitance. And the more the better.
But as you say there could be problems with a undersized transformer
It could be the case where the transformer is not able to recharge quickly the caps. So if the transformer is little better not increase to much the capacitance ?
A reasonable start may be 2 x 20.000 uF x channel and listen
Thanks again. Very interesting .
Kind regards,
gino
Thank you very much for the very interesting reply
PSRR of the circuit is also a very important parameter
and also much more complicated
I just see at the datasheet of caps
For a given working voltage and a same brand and series if you go up with the uF there is a general improvement in all the parameters: lower ESR, higher ripple current, lower impedance
Everything is better, apart cost and size of course
These are huge problems when trying to upgrade caps in commercial units.
I think that this is also a choice by design because a big caps upgrade could lead to problems in other part of the unit (diodes, transformer, ecc.).
But clearly for me 2 x 4700uF x channel capacitance is too little
4 times this value could be a good start.
who are they ?
Weird ... very weird attitude
Not me for sure ... i like crystal clean sound.
Thanks a lot
Regards
gino
Hello ! i am everything but an expert but I understand the issues involved in modifying a circuit
My question about too much capacitance is an old doubt of mine
I read some reviews of cost no object audio equipment with huge amount of uF in the PS
Equipment that went on working for almost a minute just with the power stored in the caps (?!)
As usual the more expensive solution is the best (unfortunately)
Thanks and regards,
gino
P.S. by the way I have a problem with paralleling many little caps
i do not like the solution at all ... putting thins in parallel rightly is not an easy task. I prefer a couple of big caps .. maybe with a little by pass
A minute! I did up to 5 minutes with a power supply of my own creation for a chip amp - not deliberately to be fancy, but to enable it to have very "stiff" voltages - made nice sound too!
Well, you may not like doing the many thins, but it's the right technique, electrically! It reduces the ESR and ESL, allows you to use far higher quality caps, closer to where they're needed. The couple of fat boys might be easy to do, but it will mean a power supply that is, ooohhh, to pull a figure out of the air, 10 times worse in real terms -- poor value for money there ...
Frank
time to recharge the smoothing capacitance is not the problem. It is not even a problem.Thank you. I have the same sensation actually
I experimented a little.
My impression has been that every parameter improves with more capacitance. And the more the better.
But as you say there could be problems with a undersized transformer
It could be the case where the transformer is not able to recharge quickly the caps. So if the transformer is little better not increase to much the capacitance ?
A reasonable start may be 2 x 20.000 uF x channel and listen
Thanks again. Very interesting .
Kind regards,
gino
Let's take a 100W into 8ohms capable amplifier.
select two very different transformers to feed the +-10mF smoothing.
A 400VA, 35+35Vac, with 5% regulation will hold ~+-50.5Vdc when the amplifier is quiescent and drop to ~+-49.5Vdc when the amplifier is delivering 100W into 8r0.
A 80VA, 35+35Vac, with 15% regulation will hold ~+-51Vdc when the amplifier is quiescent and drop to ~+-48Vdc when the amplifier is delivering 80W into 8r0. The 100W amp can't achieve 100W with the lower supply rail voltage, it has become a 80W amplifier. However if one applies the IHF peak power test you will measure MORE IHF peak power from the 80VA transformer. This is because the supply rails have a higher average voltage while the IHF peak power is being delivered (by the capacitors).
The capacitance does not see a time to recharge problem. The transformer does not see a time to send charge problem.
The average voltage at the PSU differs to take account of the "shape" of the charging pulses in the charging circuit.
Now listen to those two assemblies. They will sound almost identical at very low sound levels.
But when one turns up the volume to normal or above normal sound levels the two will sound different.
I prefer a "stiff" PSU. To me it sounds better over a wider range of listening levels. I achieve "stiff" by using +-20mF for each 8ohms speaker and 1.5times to 2 times the output Wattage as a target VA for the transformer.
Well, you may not like doing the many thins, but it's the right technique, electrically! It reduces the ESR and ESL, allows you to use far higher quality caps, closer to where they're needed. The couple of fat boys might be easy to do, but it will mean a power supply that is, ooohhh, to pull a figure out of the air, 10 times worse in real terms -- poor value for money there ...
I agree - not liking to do something is no argument at all against it being the right solution
When I did a calculation on how many uF I could buy for my money, I get the most in smaller sized caps - 2,200uF/16V is about the optimum as I paid 0.25rmb each (roughly $0.04) for them, a bag of 250 cost $10 at my local electronics market. Buying bigger caps (I looked at 100,000uF/16V at $5 a pop) costs more, per uF. Sure it takes time to solder them all up, but this is DIY Audio is it not?This is extremely interesting and answer to my main doubt: if just a capacitor upgrade could be enough for some improvements
If i understand well the answer is positive
Of course a "stiff" PSU is what usually great amps have
It is difficult to find a great sounding amp with a weak power supply
And as someone has said, normal speakers can be a tough load indeed
Thanks for the very helpful advice
Kind regards,
gino
Joshua now has his own thread here,
http://www.diyaudio.com/forums/power-supplies/232289-modelling-l-c-r-power-supply.html
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.