Power Supply Resevoir Size

[gootee]

Originally Posted by danielwritesbac

How to calculate transformer RMS AC Output Voltage and change it into common specs used with transformer purchases, such as their AC, Current, and VA figures?

Same question more simply: Can the chart show real transformers?

I think that everything you listed is already given in the spreadsheet, except for VA. The AC RMS output voltage is given, as is the maximum RMS current for a given maximum output power. The transformer model is not ideal. So the values should be somewhat close to the real transformer you would need.

It seems like one of the most important things is to use a high-enough transformer output voltage. Find the output power you want, with a relatively-low minimum required capacitance, and use the corresponding transformer output voltage. Then use triple the Cmin, or more. You want the absolute-minimum C to be low, for the VAC RMS and Pmax combo that you choose, even though you'll probably use much more C than that. It just means that you have sufficient headroom and will have (or at least could have) low ripple amplitude.

But you also need some excess A (and VA) capability. I'm still working on trying to quantify that.

[djk]

IMO, the optimum transformer size is 2VA for every 1W you expect out of the amplifier into the minimum impedance you expect it to drive.

This makes the assumption that you are not using stupidly large capacitors that give you a really bad power factor.

An example might be a 100W x2 amplifier with a 400VA transformer.

Yes, you can use a much smaller transformer. (but) The bigger the transformer is, the better it sounds. The point of rapidly diminishing return seems to be at 2VA for every 1W of output.

YMMV

[Tarasque]

What I normally do is use determine the voltage by the highest impedance that the amplifier needs to drive with the coresponding wattage and do the same for the current for the lowest impedance.
The result is typically a bulky powersupply with lots of headroom. The result is also an amplifier that typically sounds well and has umpf.

Looking at the famous amps, they all have plenty of power supply on board and they range in xformer from 1.5 to 4 times the power that the amp can deliver at 8 ohm's.

[gootee]

It makes a lot of sense to assume that the max RMS output power is actually double the max RMS sinusoidal output power, since a square wave WOULD literally double it, and we have seen music that approximates a square wave.

So is that enough or should we then double that? I tend to want to say double it again. e.g. For "50 Watts RMS per channel" spec, assume 100 Watts per channel might be needed. Then 200 VA is really 1X and we want 2X so we would want 400 VA.

It also kind of falls in line better with the plot in post 1359, at Power Supply Resevoir Size .

http://www.diyaudio.com/forums/attac...00va-500va.jpg

Cheers,

Tom

[gootee]

Originally Posted by gootee

It makes a lot of sense to assume that the max RMS output power is actually double the max RMS sinusoidal output power, since a square wave WOULD literally double it, and we have seen music that approximates a square wave.

So is that enough or should we then double that? I tend to want to say double it again. e.g. For "50 Watts RMS per channel" spec, assume 100 Watts per channel might be needed. Then 200 VA is really 1X and we want 2X so we would want 400 VA.

It also kind of falls in line better with the plot in post 1359, at Power Supply Resevoir Size .

http://www.diyaudio.com/forums/attac...00va-500va.jpg

Cheers,

Tom

If a high-enough Vrms transformer output voltage is used, such that the absolute-minimum capacitance is "low" (maybe 5000 uF or less?), and 2X or 3X the absolute-minimum C (or more) is actually used, then there should be no reason to use much more VA than about 2X the maximum RMS sinusoidal output power spec. (But I might still want to add 20% to that.)

[gootee]

If anyone is interested, here is the version of the spreadsheet that also calculates the average rail value and the ripple amplitude, for the cases with the absolute minimum C and 3X as much C.

[danielwritesbac]

Originally Posted by gootee

If anyone is interested, here is the version of the spreadsheet that also calculates the average rail value and the ripple amplitude, for the cases with the absolute minimum C and 3X as much C.

Is it possible to list transformer VA on the spreadsheet?

[DF96]

Transformer VA cannot be derived from winding resistance alone. You would need to add a thermal calculation, together with knowledge of the exact behaviour of the insulation with temperature. That is why a 'x VA for y W' rule of thumb can never be more than an approximate rule of thumb, because it uses VA as a proxy for winding resistance but it is a weak proxy.

[gootee]

I decided that the spreadsheet needed to be changed, after noticing that while the calculated Cmin values agreed well with the values found during exhaustive simulations with LT-Spice, for cases where Cmin was less than about 4000uF-5000uF, the agreement deteriorated significantly for cases where the calculated Cmin was higher, with the spreadsheet predicting Cmin values that were significantly too low, compared to the worst-case square wave simulations.

The calculated values would probably be fine, most of the time, if only sine waves were considered, as is customary. But my intent was to try to provide Cmin values that would not allow the power supply to grossly distort the output, in any case at all. (But I suppose that in that case, the spreadsheet should just double the maximum RMS power entry. I might actually end up doing something similar to that.)

The new version that is attached should be consistent with the simulations whenever the calculated Cmin values are less than or equal to slightly less than 10000uF. Calculated Cmin values of 10000uF or higher should be considered to be at least 50% too low, which makes sense if they are for sinusoidal rather than square wave inputs. It also makes sense that the diference only becomes prominent when the Pmax is approaching the practical limit for each transformer output voltage.

Changes made, compared to previous version:

- Increased V_amp_min from 3.00 to 3.35 Volts, per reality.

- Changed "rectifier and rail drop" number, in calculation of Vc_pk, from 1.7 to 1.7+(0.35*1.414*iload_rms_max)

- Added column that calculates "Transformer Worst-Case Volt-Amps Per Rail", which just multiplies the peak output current by the peak output voltage. This would be doubled for a transformer that would power two stereo channels.

- Added columns for p-p ripple voltage and average rail voltage, for each of the two capacitances given

Cheers,

Tom

[danielwritesbac]

The affordable and low ripple selection is of most interest to me.
Question: Is this applicable to a split rail amplifier, like 80W, 20,000u per rail, 28,0,28vac 4.3a transformer?
Question2: For example, my 28,0,28vac 6a stereo build could support up to 70 watts per channel (140 watts)?
Question3: If not applicable to a regular split rail amplifier driving an ordinary 8 ohm speaker, then how would the following need to be different?

  0.5W   1,400u    4.1VA
  1.0W   2,200u    5.4VA
  1.3W   3,300u    6.3VA
  2.5W   3,300u   11.9VA
  3.1W   3,300u   13.2VA
  9.1W   6,800u   28.6VA
 15.0W   8,200u   44.7VA
 25.0W  10,000u   71.3VA
 30.0W  15,000u   77.9VA
 32.0W  20,000u   80.4VA
 39.2W  20,000u   97.5VA
 55.0W  18,000u  135.7VA
 65.0W  20,000u  158.5VA
 98.0W  20,000u  234.9VA
100.0W  20,000u  237.3VA
125.0W  20,000u  295.7VA
155.0W  20,000u  363.2VA
180.0W  30,000u  395.4VA