As mentioned above, some of the cases investigated were showing a peak power that was limited by transformer VA instead of by rail voltage itself. I thought I would discuss how I came up with this factor.
As most people know, music has an average power level that is below, sometimes far below, the peaks in the signal. This can be captured in what is known as a "Crest Factor", which is the ratio of peak power to average power. So, your 100W amplifier might be on average producing about 1W of output power, however, brief peaks of 30W-60W might be common. Transformer heating should be based on the average power throughput. This is different than the peak power, which is determined by rail voltage.
In my spreadsheet, I calculate the power supply capability in two ways: one considering rail voltage alone, and another considering the average power that the supply must generate for a given crest factor (for music, this is typically 10-20dB). Given that the transformer should not over heat, I compare the rail-voltage-limited and continuous-power-limited peak output power, and then choose the smallest of the two as the limit.
Because the dissipation in the output stage of a class AB amplifier (at full power) is proportional to the rail voltage squared, higher rail voltages will result in higher power dissipation. The dissipated power must also be generated by the power supply, and gives rise to additional transformer heating. So, in the case of the 40V secondaries, above, the rail voltages are actually higher than the case with the 35V secondaries, so more dissipation.
This can be manipulated by the crest factor to some degree. For instance using a crest factor of 20dB changes things so that the transformer secondary voltage is the limiting factor instead of the VA rating. This just means your amplifier will not be able to work at a very high constant power level, or looking at it another way your 100W amplifier is really a 25W continuous amp with 6dB of headroom... it's just another design factor to consider.
-Charlie
As most people know, music has an average power level that is below, sometimes far below, the peaks in the signal. This can be captured in what is known as a "Crest Factor", which is the ratio of peak power to average power. So, your 100W amplifier might be on average producing about 1W of output power, however, brief peaks of 30W-60W might be common. Transformer heating should be based on the average power throughput. This is different than the peak power, which is determined by rail voltage.
In my spreadsheet, I calculate the power supply capability in two ways: one considering rail voltage alone, and another considering the average power that the supply must generate for a given crest factor (for music, this is typically 10-20dB). Given that the transformer should not over heat, I compare the rail-voltage-limited and continuous-power-limited peak output power, and then choose the smallest of the two as the limit.
Because the dissipation in the output stage of a class AB amplifier (at full power) is proportional to the rail voltage squared, higher rail voltages will result in higher power dissipation. The dissipated power must also be generated by the power supply, and gives rise to additional transformer heating. So, in the case of the 40V secondaries, above, the rail voltages are actually higher than the case with the 35V secondaries, so more dissipation.
This can be manipulated by the crest factor to some degree. For instance using a crest factor of 20dB changes things so that the transformer secondary voltage is the limiting factor instead of the VA rating. This just means your amplifier will not be able to work at a very high constant power level, or looking at it another way your 100W amplifier is really a 25W continuous amp with 6dB of headroom... it's just another design factor to consider.
-Charlie
I made a couple of additions, so I have posted a minor update, version 2.1.
The latest version is available here:
Transformer_capability_calculations_VER2.1.xls
-Charlie
- I fixed the plot in the transformer model worksheet that was blank.
- I added the rail voltage in the upper left plot on the "music program material" worksheet.
- I added a visual dashed vertical line in the plots on the "music program material" worksheet so that it was easy to see the different power limits for each.
The latest version is available here:
Transformer_capability_calculations_VER2.1.xls
-Charlie
I concur with AndrewT on this!
Practicality constrainments would make the spreadsheet a lot smaller, less misleading and easier to read.
I'd like to see a very small chart (with only low ripple selections), listing amplifier power, DC operating voltage, transformer, and power supply capacitance.
Here's some DC operating voltage (rails) selections:
8 + 8
13+13
15+15
18+18
22+22
26+26
30+30
33+33
36+36
40+40
44+44
47+47
51+51
57+57
What's missing is an affordable Low Ripple selection of transformer amperage, transformer voltage, amplifier watts (split rail monobloc) and capacitance, all on a highly practical chart. Estimation is okay. What I'm after is applicable data.
P.S.
If there's some data that even I can read, then new power supply designs can start from an orthodox basis for seriously improved performance.
Practicality constrainments would make the spreadsheet a lot smaller, less misleading and easier to read.
I'd like to see a very small chart (with only low ripple selections), listing amplifier power, DC operating voltage, transformer, and power supply capacitance.
Here's some DC operating voltage (rails) selections:
8 + 8
13+13
15+15
18+18
22+22
26+26
30+30
33+33
36+36
40+40
44+44
47+47
51+51
57+57
What's missing is an affordable Low Ripple selection of transformer amperage, transformer voltage, amplifier watts (split rail monobloc) and capacitance, all on a highly practical chart. Estimation is okay. What I'm after is applicable data.
P.S.
If there's some data that even I can read, then new power supply designs can start from an orthodox basis for seriously improved performance.
Hey Daniel,
I invite you to use my spreadsheet to get this data yourself...
1. Enter whatever +/- (split) rail voltage you want (minimum 12V, max 60V) and transformer VA on the "transformer model" worksheet.
2. on the "music program material" worksheet, choose:
- dB crest factor = peak amplitude / RMS amplitude. Typical values for music are 10-20dB. Enter 0 to show continuous power data.
- load resistance, per channel
- number of channels in amplifier
- Vod: output stage drop-out voltage = Vrail - Vout,peak (typically 3-4V)
- Mains Frequency [Hz] (50Hz or 60Hz)
- Volt, Maximum allowable ripple voltage
You can investigate lots of scenarios this way... and since your interest in "low ripple" will be contrary to others interest in "minimum capacitance", each user should choose what suits them, or can investigate how the parts they have lying around will perform.
I'm currently working up an "info and instructions" page that should be helpful.
-Charlie
You can weigh them. Typically about 2.2 lbs per 100 VA is about right. Do the transformers have dual secondaries or a center tapped secondary?
Charlie,
I'm still catching up on reading posts (truly-excellent progress, Charlie!) but I thought that I had better mention...
This thread actually isn't about the MINIMUM capacitance that can be used. It's actually about "How much capacitance should be used?" and also "Is there a simple 'rule of thumb', such as 'always use X uF per RMS amp of output', that we can develop?". (I believe that there is no simple rule for determining the capacitance, unless we add so many constraints that the rule would then be almost useless.)
I mentioned finding the minimum capacitance because back when we started it seemed like a logical first step for trying to converge on the capacitance that COULD be used, on the way to trying to narrow it down to what capacitance SHOULD be used. Anyway, that effort got a LOT more involved than I anticipated, and I am still not quite finished with just that part.
ALSO, Onaudio (Harrison) is still waiting VERY patiently for a similar effort to be applied to selecting the decoupling capacitors! <grin> That should be even more interesting than this is!
All,
Last night I did multiple simulations where I swept the VA ratings, and observed the large effects (which we already knew about) that I need to try to account for, in my spreadsheet. But since I have the "per-unitized" scalable transformer model equations, I should already have the answer and just need to throw it all into the spreadsheet.
I also started to look at differences between my worst-case type of analysis (square wave signal, which finds cases where clipping will occur, with sine signals, that cannot be found with sine-only peak or average-level types of calculations) and the usual standard sine-signal type of calculations, and believe I might have identified at least one case where I might have mixed the two, in the equations I derived for the spreadsheet. So I'm going to re-do the derivation, BOTH ways, and will probably include both results, somehow, maybe by providing one set of numbers that "will work almost all the time" (i.e. a somewhat-lower "soft minimum" capacitance value, for example) but also providing what my original intent had evolved into, which will be somewhat-larger absolute "hard minimum" capacitance values and absolute-maximum RMS output power ratings such that even a millivolt of clipping "CANNOT" occur, with a square wave input at either 0 or 180 degrees phase angle, which covers all combinations of possible signal peak positions/phase angles vs time, relative to all possible charging pulse peak timing, including when low frequencies make the capacitors "hold their breath" through multiple recharging opportunities. Maybe later I can allow for a "minimum signal frequency" input. For now I'll either stay with the 25 Hz square wave repetition rate or go slightly lower.
Cheers,
Tom
Last edited:
i have seen the traffo for the st120 it uses a regulated B+ using a single ac secondary winding.....
An externally hosted image should be here but it was not working when we last tested it.
I used monobloc setting with the ripple zeroed.
Output drop voltage is at 3v, db crest is at 15.
Speaker load set to 8 ohms, and mains at 60hz.
I wish there was a supply capacitance readout.
The "vct" is for EI core center tap transformers.
Code:
9.1W 6,800u 1.1a25vct 28.6VA 12.5+12.5vac
15.0W 8,200u 1.5a30vct 44.7VA 15+15vac
25.0W 10,000u 2.0a36vct 71.3VA 18+18vac
30.0W 15,000u 1.9a40vct 77.9VA 20+20vac
38.0W 20,000u 2.2a44vct 97.5VA 22+22vac
55.0W 18,000u 2.7a50vct 135.7VA 25+25vac
65.0W 20,000u 2.8a56vct 158.5VA 28+28vac
98.0W 20,000u 3.7a64vct 234.9VA 32+32vacDue to transformer prices, I opted for the low ripple selections of watts, capacitance and va from Tom's spreadsheet (that I may have misread), then matched up the figures to your spreadsheet, and deleting all mismatches to make the short list. On first glance, this seems to have worked.
But, could you double-check it for me?
Last edited:
72vac? Okay. 430VA apiece. Suited for a tube amp.
Need MosFet rail splitter if used for solid state amp.
After rail split there's about 52+52vdc rails, and 6a.
Capacitance minimum, 6800u//6800u caps, per rail.
Such high voltage then 6a is not suited for 4 ohms.
For 8 ohms you get about 120 Watt monobloc amp.
Last edited:
probably that was what they had plenty of at the time of production, and they got those very cheap........
Would that mean only 3 amps per transformer? I would think that a 8 pound transformer could put out more than 3 amps but I do not know. 120 watts was what was coming out of the amp to start with at 60 watts a channel.
We all listen to sinewaves. Even when simulated squarewaves and sawtooth waves are inserted into digitally created music, these are composed of sinewaves.
Hi Daniel,
Looks good. To start, a couple of comments:
A. You can't "zero out the ripple" in my spreadsheet. Zero ripple could only be achieved with infinite capacitance...
B. Tom's output power in Watts figures seem to be slightly higher than what I calculate (See below) in most cases, but his figures and mine are in pretty good agreement.
I've copied your table from above, and done the following:
1. I used the capacitance specified in your table. To do this I varied the ripple voltage in my spreadsheet until the capacitance was a close match. There is a new column "Vripple" listing the corresponding ripple voltage.
2. I updated the output power column to reflect the values that my spreadsheet predicts for output power given the transformer listed and the capacitor size (per rail).
Here is the result:
Code:
output cap per ripple secondary AC
power rail voltage transformer info voltage
8 W 6,800u 0.73V 1.1a25vct 28.6VA 12.5+12.5vac
13 W 8,200u 0.76V 1.5a30vct 44.7VA 15+15vac
22 W 10,000u 0.78V 2.0a36vct 71.3VA 18+18vac
29 W 15,000u 0.59V 1.9a40vct 77.9VA 20+20vac
37 W 20,000u 0.50V 2.2a44vct 97.5VA 22+22vac
50 W 18,000u 0.64V 2.7a50vct 135.7VA 25+25vac
65 W 20,000u 0.65V 2.8a56vct 158.5VA 28+28vac
90 W 20,000u 0.76V 3.7a64vct 234.9VA 32+32vac[/FONT]I'm not sure what you mean by the above statement. I believe that the output power is limited by the rail voltage for these cases, so the transformer VA is not going to play much of a role in output power. Reducing the capacitance will reduce the available output power, and increasing the capacitance might slightly increase power, but it depends on several things. You would have to run many scenarios... rather than making a huge table or tables, you can just run the numbers in my spreadsheet anytime and change any one of the variables to see the effect on the output power.
-Charlie
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.