In some recent correspondence with a class-D amp enthusiast the topic of power supply capacitance came up. This is specifically for the combination of a linear power supply (transformer, bridge rectifier, caps) and a class-D half-bridge topology amp.
I come from the class AB amp world and am only now starting to use class-D stuff. For class AB, if I was going to build a power supply for 200W into a 4 ohm load, I might think of using 40,000uF or more per rail of capacitance. But the class-D enthusiast felt that for this amount of power or more I would only need 6800uF of capacitance per rail. I'm trying to understand how this would be the case. Even with class-D' s high efficiency, ya still gotta supply lots of current in between the charging interval at the peak of the mains waveform, and this current can only come from the caps. Current draw means a drop in the voltage in the cap bank, which causes ripple. Most class-D amps have relatively poor ripple rejection compared to most class-AB amps that I am familiar with, so this is all a bit concerning. Perhaps my class-D enthusiast friend is thinking of an SMPS which is recharging at a much, much more rapid rate and therefore does not need a large capacitance to store charge in between charging pulses?
So, what amount of capacitance might be appropriate?
I come from the class AB amp world and am only now starting to use class-D stuff. For class AB, if I was going to build a power supply for 200W into a 4 ohm load, I might think of using 40,000uF or more per rail of capacitance. But the class-D enthusiast felt that for this amount of power or more I would only need 6800uF of capacitance per rail. I'm trying to understand how this would be the case. Even with class-D' s high efficiency, ya still gotta supply lots of current in between the charging interval at the peak of the mains waveform, and this current can only come from the caps. Current draw means a drop in the voltage in the cap bank, which causes ripple. Most class-D amps have relatively poor ripple rejection compared to most class-AB amps that I am familiar with, so this is all a bit concerning. Perhaps my class-D enthusiast friend is thinking of an SMPS which is recharging at a much, much more rapid rate and therefore does not need a large capacitance to store charge in between charging pulses?
So, what amount of capacitance might be appropriate?
A class D amp that does not use feedback is very susceptible to power supply ripple.
With feedback...not so much of an issue.
Myself and some others working on TPA3116 types use a laptop power supply and this works very well. (on board uses 2 X 470uF)
One of my other projects is a higher power amp (TDA8954) and am planning initially to use a simple unregulated linear supply.
I will (someday) have this done and will have more information on this subject.
With feedback...not so much of an issue.
Myself and some others working on TPA3116 types use a laptop power supply and this works very well. (on board uses 2 X 470uF)
One of my other projects is a higher power amp (TDA8954) and am planning initially to use a simple unregulated linear supply.
I will (someday) have this done and will have more information on this subject.
What supply voltages? What output power into what load impedance?
The chip's manufacturer does warn quite stringently that a minimum supply voltage level must be maintained. They mention something about the warranty being void if you don't use their minimum required regulated supplies.
From reading other sources on the net, apparently "buss pumping" can be a big issue if the amp is current-starved. And apparently you can't get anywhere near the published spec without a solid regulated PSU.
The chip's manufacturer does warn quite stringently that a minimum supply voltage level must be maintained. They mention something about the warranty being void if you don't use their minimum required regulated supplies.
From reading other sources on the net, apparently "buss pumping" can be a big issue if the amp is current-starved. And apparently you can't get anywhere near the published spec without a solid regulated PSU.
The output stage is running at around +/-85Vdc. There are other housekeeping supplies that are lower voltage to run the driver and input stages of the amp. Rated output is 500W RMS into 4 ohms, single channel although the amp can do more at clipping.
When you mention "the chip's manufacturer" are you talking about International Rectifier and the IRS2092. Is that info from the datasheet? Can you point me to where you read it?
I am aware of the buss pumping phenomenon. More capacitance should help this, or at least that seems reasonable in my mind. I am more worried that the cap voltage will be dropping significantly unless there are enough F to support the amp's output...
The "warranties will be voided if you do not observe the following warnings" stuff is in section 2.5 of the pdf at http://www.parts-express.com/pedocs/brochures/320-313-parts-express-brochure.pdf . Sorry. I guess I was remembering wrongly when I wrote that it was said by the chip's manufacturer. It's actually by a particular board's manufacturer, Sure Electronics.
Anyway, I'm with you on the reservoir capacitance issue. There is no way around it, as far as I can see. But of course, if most people never approach the rated maximum output power, and aren't powering subwoofers, they might not ever see a problem when using too little capacitance.
But, running the numbers for 500 Watts, 4 Ohms, +/-85V, I do arrive at a reservoir capacitance of between 6500 and 7200 uF per rail per channel (i.e. double that if running two channels from one supply), for the onset of clipping to occur at 500 Watts. (I was using my spreadsheet so that was just guessing that the amplifier's clipping voltage was between 1V and 3V, for the voltage space it occupies, i.e. between the supply rail and the output. It might need to be higher, which would raise the capacitance needed.). And that was assuming the output signal was worst case: constant DC at the max peak values (or, say, a low-rate square wave, or a very low bass sine wave).
Of course, that amount of reservoir capacitance gives over 20V p-p of load-induced ripple (but only when operating at the rated max power). However (and I am still almost totally-unfamiliar with these types of amplifiers, and this chip), will that voltage matter to the chip? Or is it mainly the housekeeping supplies that are critical? I guess that you really don't want to trigger any under-voltage or over-voltage circuitry in the chip, or on the board. And the PSRR is apparently not all that great.
If you wanted 5% ripple (4.25 V p-p) instead of 24%, when producing the max rated output power, the capacitance would need to be more like 39000 uF per rail per channel. But hey, then you could also rate your amp at 755.7 Watts RMS max, assuming enough current was also available (13.74 A RMS, max).
I would probably go for the higher capacitance. But remember to use multiple smaller caps in parallel, mainly so the ripple current ratings can be reasonable. At 500 Watts, the output current would be about 11.2 A RMS. So the rectifier diodes and caps would be seeing something like 3X to 5X that, as a peak value, I'm guessing.
EDIT: Spreadsheet is posted at: http://www.diyaudio.com/forums/chip...fi-lm1875-amplifier-board-36.html#post3705483
Cheers,
Tom
Anyway, I'm with you on the reservoir capacitance issue. There is no way around it, as far as I can see. But of course, if most people never approach the rated maximum output power, and aren't powering subwoofers, they might not ever see a problem when using too little capacitance.
But, running the numbers for 500 Watts, 4 Ohms, +/-85V, I do arrive at a reservoir capacitance of between 6500 and 7200 uF per rail per channel (i.e. double that if running two channels from one supply), for the onset of clipping to occur at 500 Watts. (I was using my spreadsheet so that was just guessing that the amplifier's clipping voltage was between 1V and 3V, for the voltage space it occupies, i.e. between the supply rail and the output. It might need to be higher, which would raise the capacitance needed.). And that was assuming the output signal was worst case: constant DC at the max peak values (or, say, a low-rate square wave, or a very low bass sine wave).
Of course, that amount of reservoir capacitance gives over 20V p-p of load-induced ripple (but only when operating at the rated max power). However (and I am still almost totally-unfamiliar with these types of amplifiers, and this chip), will that voltage matter to the chip? Or is it mainly the housekeeping supplies that are critical? I guess that you really don't want to trigger any under-voltage or over-voltage circuitry in the chip, or on the board. And the PSRR is apparently not all that great.
If you wanted 5% ripple (4.25 V p-p) instead of 24%, when producing the max rated output power, the capacitance would need to be more like 39000 uF per rail per channel. But hey, then you could also rate your amp at 755.7 Watts RMS max, assuming enough current was also available (13.74 A RMS, max).
I would probably go for the higher capacitance. But remember to use multiple smaller caps in parallel, mainly so the ripple current ratings can be reasonable. At 500 Watts, the output current would be about 11.2 A RMS. So the rectifier diodes and caps would be seeing something like 3X to 5X that, as a peak value, I'm guessing.
EDIT: Spreadsheet is posted at: http://www.diyaudio.com/forums/chip...fi-lm1875-amplifier-board-36.html#post3705483
Cheers,
Tom
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.