To get 1000 Watts RMS into one Ohm, continuously, not just on peaks, the key is to use higher-voltage rails, so that the reservoir capacitance doesn't need to be so large.
(This is a good use for my spreadsheet, using Excel's "What If" feature to get the power ratings or capacitances that I want to see, by having it change the ratio of theoretical to max power.)
With 60 Hz mains:
For example, with 55-Volt rails, you could do it with only 37430 uF (or 53000 uF if you wanted to be able to push constant DC at that power level). But with 50-Volt rails, you would need to up the reservoir capacitance to 116100 uF (or 164200 uF if constant-DC-capable).
66-Volt rails might be a good assumption, since there are big 48V and 50V RMS transformers. Then it would only require 15027 uF per rail to rate the amp at 1000 Watts into 1 Ohm, average, continuous (and 225 W into 8 Ohms, and about 410 W into 4 Ohms).
But then you could just up the capacitance to 40600 uF per rail and the amp could be rated for 1500 W into 1 Ohm.
To get the ripple down below one volt with an 8-Ohm load, you might as well up the capacitance to 66000 uF per rail, which would give 0.73 V p-p max ripple at max rated continuous average power of 243 Watts into 8 Ohms. That would also give you a max of 475W into 4 Ohms with 1.44 V p-p max ripple, and 1665W continuous max rating into 1 Ohm, with 5.4 V p-p max ripple.
At 1665 Watts into one Ohm, a sine wave would be 40.8 Volts RMS. So the peak voltage would be 57.7 Volts. So the peak power would be 3330 Watts.
You would also want decoupling caps. So if you added another 27000 uF per rail near the output devices, it would be even be able to push constant DC at the maximum peak level (or up to whatever the transformer would give, if it's not rated for that).
Since your amp might not be more than 67% efficient (if it's class AB), you would want to use a 2500 VA transformer, for each channel.
Mouser.com has this one, for $298:
http://www.mouser.com/ProductDetail...EpiMZZMvwUzoUXIIvyTw34uqMHR%2bueH%2bNT7ETNDM=
That one is 50 V RMS output, so the rail voltage might be closer to 69 Volts and the max power ratings would be a little higher than I calculated above.
Remember, though, that all of the above is for max continuous rated output power. Chances are that you could use a power transformer with a significantly-lower VA rating and still be OK.
I should also mention that I assumed a Vclip voltage across the output stage (rail to output) of 2.9 Volts. So "your mileage may vary".
Cheers,
Tom
(This is a good use for my spreadsheet, using Excel's "What If" feature to get the power ratings or capacitances that I want to see, by having it change the ratio of theoretical to max power.)
With 60 Hz mains:
For example, with 55-Volt rails, you could do it with only 37430 uF (or 53000 uF if you wanted to be able to push constant DC at that power level). But with 50-Volt rails, you would need to up the reservoir capacitance to 116100 uF (or 164200 uF if constant-DC-capable).
66-Volt rails might be a good assumption, since there are big 48V and 50V RMS transformers. Then it would only require 15027 uF per rail to rate the amp at 1000 Watts into 1 Ohm, average, continuous (and 225 W into 8 Ohms, and about 410 W into 4 Ohms).
But then you could just up the capacitance to 40600 uF per rail and the amp could be rated for 1500 W into 1 Ohm.
To get the ripple down below one volt with an 8-Ohm load, you might as well up the capacitance to 66000 uF per rail, which would give 0.73 V p-p max ripple at max rated continuous average power of 243 Watts into 8 Ohms. That would also give you a max of 475W into 4 Ohms with 1.44 V p-p max ripple, and 1665W continuous max rating into 1 Ohm, with 5.4 V p-p max ripple.
At 1665 Watts into one Ohm, a sine wave would be 40.8 Volts RMS. So the peak voltage would be 57.7 Volts. So the peak power would be 3330 Watts.
You would also want decoupling caps. So if you added another 27000 uF per rail near the output devices, it would be even be able to push constant DC at the maximum peak level (or up to whatever the transformer would give, if it's not rated for that).
Since your amp might not be more than 67% efficient (if it's class AB), you would want to use a 2500 VA transformer, for each channel.
Mouser.com has this one, for $298:
http://www.mouser.com/ProductDetail...EpiMZZMvwUzoUXIIvyTw34uqMHR%2bueH%2bNT7ETNDM=
That one is 50 V RMS output, so the rail voltage might be closer to 69 Volts and the max power ratings would be a little higher than I calculated above.
Remember, though, that all of the above is for max continuous rated output power. Chances are that you could use a power transformer with a significantly-lower VA rating and still be OK.
I should also mention that I assumed a Vclip voltage across the output stage (rail to output) of 2.9 Volts. So "your mileage may vary".
Cheers,
Tom
Last edited:
According to my other spreadsheet, which plots the power supply behavior, the capacitance I gave for 1665 Watts into 1 Ohm would not quite be sufficient for using 93000 uF while pushing constant DC at the max peak output level). It will, however, do 1540 Watts into 1 Ohm, in that simulation, without clipping.
But that might be a lesser worry, because it shows that the repetitive rectifier current peaks would be 229 Amps! (222 A @ 1540 W)
However, the diode model I used is not known to be valid above 20 or 30 amps. So the 229 amps might be rubbish. And the transformer model is a scaled version of a 500 VA transformer model. So it might be crapping out, too.
EDIT: Oh drat. I see why. My spreadsheet tests with the output power at 1.414x the rated RMS power. The transformer is 2500VA, 50V @ 50A, and the output would need to be 57.7 Amps DC at 57.7 Volts DC, to run at the PEAK output power level into 1 Ohm, which is 3330 Watts. I would need to tell it 1250 Watts, so it would test with DC at the max peak, i.e. 50 Volts/50A. Or, to test the 1665W level, I could tell it 1177 Watts.
But that might be a lesser worry, because it shows that the repetitive rectifier current peaks would be 229 Amps! (222 A @ 1540 W)
However, the diode model I used is not known to be valid above 20 or 30 amps. So the 229 amps might be rubbish. And the transformer model is a scaled version of a 500 VA transformer model. So it might be crapping out, too.
EDIT: Oh drat. I see why. My spreadsheet tests with the output power at 1.414x the rated RMS power. The transformer is 2500VA, 50V @ 50A, and the output would need to be 57.7 Amps DC at 57.7 Volts DC, to run at the PEAK output power level into 1 Ohm, which is 3330 Watts. I would need to tell it 1250 Watts, so it would test with DC at the max peak, i.e. 50 Volts/50A. Or, to test the 1665W level, I could tell it 1177 Watts.
Last edited:
With that kind of voltage , it would require you to have a massive output stage , not to mention 240volt mains ...
Not that I'm against such ....
50 volt rails is more appropriate for me , as is 220K of capacitance perside, 160k in the main PSU and the rest on the board as close to outputs as possible ...
Yeah baby ...
Nige, while I, admittedly purely intuitively, agree that 10,000 uF is an almost "ideal" cap size, there are times when you have to go bigger. I had that when restoring my Marantz 170 DC power amp. There was just barely enough space inside for two cans, period. Originally it came with a single dual concentric 12,000 uF cap, but since the case was generic, there was another place inside for another cap.
I just barely managed to squeeze in two BC Components 22,000 uF caps, but that was it.
As expected, this produced a very deep and authoritative bass, and anyway, the original was exactly 35 years old, so I guess simply exchanging old for new would have produced positive results all by itself. Making it almost twice as large only made the good even better.
However, given a choice, I would have preferred 2 10,000 uF in parallel.
I just barely managed to squeeze in two BC Components 22,000 uF caps, but that was it.
As expected, this produced a very deep and authoritative bass, and anyway, the original was exactly 35 years old, so I guess simply exchanging old for new would have produced positive results all by itself. Making it almost twice as large only made the good even better.
However, given a choice, I would have preferred 2 10,000 uF in parallel.
With that kind of voltage , it would require you to have a massive output stage , not to mention 240volt mains ...
Not that I'm against such ....
50 volt rails is more appropriate for me , as is 220K of capacitance perside, 160k in the main PSU and the rest on the board as close to outputs as possible ...
Yeah baby ...
If you kept it to 1500 Watts or less, it could run on 115VAC, theoretically, especially if you did a little power factor correction to get the current peaks down, and used a soft start circuit.
But OK, let's see what it would really take with 50V rails:
It turns out that even a 0.031 Ohm secondary resistance is significant, when you're talking about an average of almost 45 Amps of current. It drops about 1.39 Volts. And, at these high currents, the rectifier diodes will drop more than usual. I'll assume that it's a peak of about 4.3 Volts total, for two diodes. So that's already 5.69 Volts, skimmed right off the top, which means that the transformer output peak voltage would need to be at least 55.69 Volts, to get a 50-Volt rail. I would say at least 57 Volts, since there's probably another 30 milliohms of resistance lurking around. So, 57 V peak equates to 40.3 VAC RMS. Let's just say 40 V RMS for the transformer output, but assume a 50-Volt DC rail peak voltage. You could use a transformer with more than 40 V RMS output, but not less, unless you didn't mind having less than the 1000 Watt max rated power. If you want to also account for a possible low mains condition, you can do that.
So we'll say 50 Volts for the rails. And we'll use 63V-rated capacitors, although 100V ones would have lower ESR and should then also dissipate less heat from the ripple current. I would use 100V. But calculating with 63V will be a worse case. But the reservoir capacitance for 63V caps is only about 2000 uF higher than with 100V caps, in this case.
You would need 115000 uF per rail to get 2.38 Volts maximum peak-to-peak ripple voltage, to be able to push 1000 Watts RMS continuously into 1 Ohm, which would be a sine wave with a peak voltage of 44.72 Volts (31.62 Volts RMS), and a current with the same peak and RMS values. The 44.72 Volts maximum peak level would dissipate almost exactly 2000 Watts, max peak.
You have to be careful about the rail voltage. If something unaccounted-for would make it end up at 49 Volts instead of 50 Volts, you would need almost 100000 uF of additional capacitance to maintain the 1000 Watt rating for max output power.
With a 4 Ohm load, the same 115000uF would have a rated maximum output power of 270 Watts RMS, with a max p-p ripple voltage of 0.62 V. And into 8 Ohms, you'd have a max rated output power of 137 Watts, with a max ripple of 0.36 Volts.
If you bumped the capacitance up to 163000 uF per rail, then instead of just a sine wave you could handle any type of signal, up to and including constant DC at the peak level, without clipping, which would boost the rated maximum power into 1 Ohm from 1000 Watts RMS to 1414 Watts RMS, and a peak output of 2828 Watts.
Edit: The numbers for the 4 and 8 Ohm loads are conservative, because the lower current would drop less voltage across both the secondary winding resistance and the diodes.
Last edited:
Isn't designing for CONTINUOUS 1 kW/1 Ohm a bit ridiculous?
I mean, how many people do you know who sit in their rooms and listen to full power sine waves?
In peak, I understand, and in Wayne's case, knowing he has extremely low impedance speakers, it even makes sense, but in continuous mode, I think about one half that, or 500W, should be more than enough, assuming it can peak at 1 kW.
I mean, people are throwing watts around like they were confetti. Anyone who has driven his own speakers at just 2W continuously should know just how unpleasantly loud that is even with moderately efficient speakers (say, 90 dB/2.83V/1m).
I mean, how many people do you know who sit in their rooms and listen to full power sine waves?
In peak, I understand, and in Wayne's case, knowing he has extremely low impedance speakers, it even makes sense, but in continuous mode, I think about one half that, or 500W, should be more than enough, assuming it can peak at 1 kW.
I mean, people are throwing watts around like they were confetti. Anyone who has driven his own speakers at just 2W continuously should know just how unpleasantly loud that is even with moderately efficient speakers (say, 90 dB/2.83V/1m).
Good points . My focus was based on an accident of a repair I did . I repaired a Naim NAP 250 . 100 V caps were on spacial at Farnell . I was surprised how good it sounded ( same series as Naim use ) . Doing my searches yesterday I noticed ripple currents seldom top 12 A . 3 x 10 A might just do 1000 watts 1 R transient ( 3 x 10000 ) . Looking at a voice coil we never could sustain 1000 watts even if a Bose 901 type arrangement .
Douglas Self states that there will be no advantage to exceed the ripple requirements in a quest for tighter bass and dynamics . Equally most have noticed excess capacitance does alter the sound . I suspect an output tripple ( Darlington with driver ) having three outputs can do the transient current for 1000 W 1 R . It might only need 30 mA of standing current .Thus if sitting at +/- 90 V the heat output is 6 watts . If we only have +/- 80 V to the drivers backwards the heat output might be slightly higher with music compared with not having excess voltage . The speed of the power supply will be good and cost will be less . Compare now with a switch mode PSU of the required spec .The capacitor ( s ) will be tiny . I hate switch mode although sometimes there is no option . Commonsense says they are not made to last .
Amps that have load of caps and a flat sound I call muscle bound . The oversize decouplers should solve that .
An amp with +/- 90 V will need the cascode VAS BC 550/560C MPSA92/42 BF 720/721 ( BF469/470 as was ) . Input stage 2SA 970 , 1085 . Outputs ( and compliment )
http://docs-europe.electrocomponents.com/webdocs/0f92/0900766b80f92435.pdf
For an amp this big I might consider the output device as a driver ( device No 2 ) . The FT is good . The old ways were often better .
BTW . By sacrificing +/- 10 V the clipping will be very nice relatively speaking .
Thanks Tom,
Planned on using 220k /uf /ch , 100v caps , 50volts after rectification quasi protection of output stage (only 7pr/ch) another spread sheet gave me .55v ripple with the above, suprising to see it much higher by your calculations..
100watts rms when running 93db +, the xtra is for dynamic headroom necessary for realism ...
Man, do you need a good amplifier, or what ....
!!?Yes, you've got to throw everything at getting the power supply right ...
Mr Wayne .
You seem a candidate for a magnetic amplifier . It is said some are under the oceans and could out live the pyramids with zero servicing .
From what I know 200 000 watts is as far as they went . Distortion was low enough to be OK . They are a transformer held at or near saturation . The input signal alters that and results in amplification . Some resemblance to a CVT transformer .
Most data is US Navy and now declassified .
LTP . My rule of thumb is make Re + re about 50 R . That is if tail is at 2 ma re = 26 R ( 1 mA ) . Re then needs to be 24 R if so . Any better rule ?
You seem a candidate for a magnetic amplifier . It is said some are under the oceans and could out live the pyramids with zero servicing .
From what I know 200 000 watts is as far as they went . Distortion was low enough to be OK . They are a transformer held at or near saturation . The input signal alters that and results in amplification . Some resemblance to a CVT transformer .
Most data is US Navy and now declassified .
LTP . My rule of thumb is make Re + re about 50 R . That is if tail is at 2 ma re = 26 R ( 1 mA ) . Re then needs to be 24 R if so . Any better rule ?
http://www.themeasuringsystemofthegods.com/magnetic amplifiers.pdf
If you go 3/4 thought this there is more ( P361 , the pages ran through issues , not 500 pages per issue ) . Some interesting stuff . Components about the same price as now in many cases . One page says get into computers and how .
http://ia600805.us.archive.org/25/items/PracticalWireless1973August/PracticalWireless1973Aug.pdf
Quicker to open and interesting .
http://www.rfcafe.com/references/po...c-amplifiers-jul-1960-popular-electronics.htm
If you go 3/4 thought this there is more ( P361 , the pages ran through issues , not 500 pages per issue ) . Some interesting stuff . Components about the same price as now in many cases . One page says get into computers and how .
http://ia600805.us.archive.org/25/items/PracticalWireless1973August/PracticalWireless1973Aug.pdf
Quicker to open and interesting .
http://www.rfcafe.com/references/po...c-amplifiers-jul-1960-popular-electronics.htm
Last edited:
I work with this stuff all the time . Most if they are honest wouldn't have a clue where to start . My guy worked with the senior engineer of Partridge when being trained . It is fairly safe to say if Graham says it is a lost art he will be at the top of the tree in knowledge . His best shot was if I found one he would clone it . I wanted this for an industrial application . The distortion had to be below 1% for it to be worth doing . Graham said often magnetic shunts are required and special cuts in the core . Graham would have no difficulty making the transformer for the Ongaku . Magnetostiction would be better than a mains transformer as it is in harmony with the music . Potting would help ( vacuum impregnation and correct curing ) . If anyone gets further have a surplus of power as the dynamic behaviour is more like a zero feedback tube amp ) .
Oneac might know ( Chloride ) . CVT's are similar . Oneac's designs are very good . 1% is typical of their range . 3 % is what most offer .
Here is a more typical application . Same principles .
http://www.google.co.uk/url?sa=t&rc...=rWLhYkzfWY4vcZv0qXGMcA&bvm=bv.49405654,d.bGE
Last edited:
- Status
- Not open for further replies.