Class D amp needs big power supply?

[eem2am]

Hello,

Please could you verify this?

I hate to say it but I (think) I have just proved that a 350W Class D guitar amplifier needs a power supply rated to supply 700W in order to properly supply the Class D amp without rail sag.

The class d simulation (in the free LTSpice) which prooves this is posted at the bottom of this post.


If you run the simulation, you will see that the average power dissipated in the 8R “speaker” load is 357W……this would mean a nominal 4.5A rated power supply. (Since the supply is 80V)
However, look at the current draw of the D amplifier during the peaks of the 82Hz sine wave……the current draw for a 3ms interval around the peak of the 82Hz sine wave is 8.5A (!!!!!!)
…..therefore the power supply must be rated to supply 80V at 8.5A…..that’s 680Watts (!!!!!)

So it is true….the simulator never lies.

If you want to power a D amplifier then you need a power suppy rated to supply TWICE the power of the amplifier.

Is this correct?


[82Hz was chosen as it’s the low E string on a guitar]


LTsice simulation file (.txt but just change to .asc)
http://www.2shared.com/document/pFvYW_Q7/CLASS_D__80V__82_Hz.html


Here are the two files which control the fet switching, to produce the 82Hz sine wave. (i couldnt attach them with the provided "attacher" because the files are too big)
(ensure they are in the same folder as the simulation)

http://www.2shared.com/document/IhulBpr2/that.html

http://www.2shared.com/document/rk5QRezk/that1.html

 

[Aucosticraft]

can we say that the powersupply rating needs to be approx, 2x of the rated capacity of amp(amp+losses). and why should this be true only with classD ?

also here the design of power supply comes into picture. say the supply is designed for 6A and peak current for 10ms is 10A.

So the power supply needs to be designed as per requirement.

 

[Saturnus]

If you want to power a D amplifier then you need a power suppy rated to supply TWICE the power of the amplifier.

Is this correct?

Well, yes and no. Peak output of any amp, class D or otherwise, is 2x the RMS output. That is hardly surprising.

 

[Viennatom_new]

ee2am, have you never heard that electrolytics can store energy, so that the PSU needs only be able to supply little more than average power input to class D stage.

 

[Richard Ellis]

Remember the charging peak....as the supply has very little time to "restore" the charge. Use some Toroids, they use up very little space inside your chassis.

____________________________________________________Rick........

 

[eem2am]

Vienna Tom

ee2am, have you never heard that electrolytics can store energy, so that the PSU needs only be able to supply little more than average power input to class D stage.

To provide enough capacitance to hold the rail up like that would take an enormous amount of capacitance......it'd be cheaper and take up less room to just build an SMPS that can supply twice the average power.

 

[gmarsh]

Vienna Tom

To provide enough capacitance to hold the rail up like that would take an enormous amount of capacitance......it'd be cheaper and take up less room to just build an SMPS that can supply twice the average power.

Enormous, eh?

Assume the power supply delivers 350W constant power into a capacitor, regardless of output voltage. And we're pulling a 82hz sine wave from it with a peak amplitude of 700W, giving 350W RMS.

Psupply(t) = 350, Pload(t) = 700*sin^2(2*pi*82*t)
Pcap(t) = Pload(t) - Psupply(t) = power leaving the cap.

Did the integration on pen/paper and won't copy it over here... but it works out that the capacitor has to supply 0.68 joules over the duration when Pload > Psupply.

Now lets assume our 350W amplifier has +-60V rails (suitable for a 4 ohm speaker) and we let those rails drop to +-55V in this case. Since E=1/2CV^2... 0.68J = 1/2 * C * (60^2 - 55^2), and C = 2365uF.

So 2700uF per rail. You can get 3300uF/80V in a 22mmx30mm case. Absolutely freakin' massive, right?

 

[eem2am]

you cannot let a class d's rails sag by that amount........dropping down 5v like that is not acceptable for class d supply.....class d has very poor PSRR.

 

[Saturnus]

you cannot let a class d's rails sag by that amount........dropping down 5v like that is not acceptable for class d supply.....class d has very poor PSRR.

What on earth does PSRR have to do with voltage sag?

 

[nigelwright7557]

class d has very poor PSRR.

Only class d with no feedback has poor PSRR.
Class d amps with feedback compensate for the ripple by feeding forward its antiphase.

I bulit a discrete class d amp with no feedback and it hummed like whatnot.
My irs2092 amps are almost silent.

 

[Pafi]

gmarsh!

Pcap equals to Pload-Psupply only in full bridge. With split supply you omitted supply pumping.

And why 82 Hz, why not 25 Hz? And why 5 V, why not 1 V? There are bad PSRR ClassD amps too (most full digital).

 

[B-Red]

.....
Is this correct?

No.
Class D amplifier require a very stable power supply, that is not necessary a synonymous of a very powerful supply.

In a class D amp the current consummation is very variable in particular during the peaks, so we need for a reactive and stable supply.

 

[gmarsh]

gmarsh!

Pcap equals to Pload-Psupply only in full bridge. With split supply you omitted supply pumping.

And why 82 Hz, why not 25 Hz? And why 5 V, why not 1 V? There are bad PSRR ClassD amps too (most full digital).

I assumed in that analysis that the power supply puts the entire 350W into the same rail that the amplifier is instantaneously drawing current from. Full bridge means you need the same capacitance value, but just one of them instead of two.

If the supply uses synchronous rectification and a coupled inductor, it can move voltage from one rail capacitor to the other and you can halve the capacitance.

When you're designing a SMPS for audio use, you should design it to handle peak current without limiting or blowing up - set the current limit thresholds, saturation current of the magnetics appropriately. But winding resistances, Rds(on) ratings, heatsinking, etc. can be rated for RMS current draw.

 

[eem2am]

yes feedback with class d helps but the feedback does not completely get rid of the problems.....its better to just have a power supply which gives solid, rails which dont vary more than +-1%.

Feedback with an smps is easy....because you have a known demand value (Vout)....with a d amp the required output is the current and the required value of current is constantly changing as the music changes and thats a very very hard feedback job.......best you just use a decent power supply.

Class D amplifiers with good feedback loops are extremely expensive to build and take ages to design.......to have good feedback in a class d amplifier across the entire audible range is a very very hard task.........and to implement it is very expensive............as i said, its far better to just get a power supply rated to twice the average power of the D amplifier and do it that way.

for a feedback system you need a demanded value and you look at the error between demanded and actual......in a d amp the demanded value is constantly changing and you have a mega big control system issue on your hands.....i thus tell you its cheaper and easier to just use a big power supply.

 

[Pafi]

I assumed in that analysis that the power supply puts the entire 350W into the same rail that the amplifier is instantaneously drawing current from.

And ignored supply pumping.

(I don't know why you quoted me. Your post doesn't reflect to mine.)

If the supply uses synchronous rectification and a coupled inductor, it can move voltage from one rail capacitor to the other

This is contradictional to your assumption.

 

[gmarsh]

I propose we make a discussion out of this question: how does a class D amplifier differ from a class AB amplifier in power supply rejection - what makes it worse?

- class D will draw sin^2(t) power waveform, class AB will draw a sin(t) power waveform. Class AB will have a higher peak unless you're bring the output to the rails (not a good idea if you have rail pumping happening)
- Class AB amplifiers almost entirely use feedback, which is pretty much necessary for PSRR. Class D amplifiers sometimes don't (eg TI PurePath things) but I think it's fair to say that the better ones do.
- You can design a class AB amplifier with unlimited open-loop gain/bandwidth if you want, and design a feedback loop around it to provide virtually unlimited PSRR. Class D needs the feedback rolled off at a lower frequency to keep the switching frequency reasonable.

 

[gmarsh]

And ignored supply pumping.

I believe I calculated the capacitance which results in +-5V of supply pumping in a specific case?

This is contradictional to your assumption.

Don't see the contradiction. Assuming that the power supply switching frequency is far above the audio frequency, it's a valid statement - the power supply will transfer charge between and equalize voltage between the two rails, effectively doubling the capacitance on a single rail if you're only using one rail at that given moment.

edit: Never mind, my previous analysis assumed 3-state modulation...

 

[Pafi]

I believe I calculated the capacitance which results in +-5V of supply pumping in a specific case?

Is this a question for me?

Don't see the contradiction. Assuming that the power supply switching frequency is far above the audio frequency, it's a valid statement - the power supply will transfer charge between and equalize voltage between the two rails, effectively doubling the capacitance on a single rail if you're only using one rail at that given moment.

edit: Never mind, my previous analysis assumed 3-state modulation...

OK, I don't mind, but 3-state modulation can solve supply pumping only in case it is done by a true 3 level power stage (with switched GND), but this is almost without an example in audio application. (Motor drive is an other story.) In an audio amplifier half bridge 3 level modulation is achieved with eg. BCA topology, and there is no GND switch, what means supply pumping is still the same. (Speaker current have to flow back somewhere, doesn't it?)

But I think this is quite OFFtopic.

In the other aspect you are right, the PSU doesn't have to be "big". But I wouldn't dare to say "this is the way you have to do", or choose 1 particular (and rare) solution, and make general statements based on it. There are different strategies for providing the extra peak power, while average power requirement is unbelieveably low (for some people). Somebody decides to provide it with capacitor (optimal for idle loss), some other choose active (less space).

 

[eem2am]

gmarsh

If the supply uses synchronous rectification and a coupled inductor, it can move voltage from one rail capacitor to the other and you can halve the capacitance.

Are you sure coupled inductors act like that?......i know coupled innductors improve cross regulation between the two split output rails, but i dont think they are very very good at it......there is a unitrode design note on them, which says they are good, but they are not that good.....i will find it.

 

[AP2]

I honestly do not understand...
d class amplifier, after which the devices have been dimensioned well. The power supply decides the output power of the amplifier. therefore, if the absorption is 8Amp and rail voltage (measured at the instant of current) is 60V. the output power = (480W-% of inefficiency), assumed loses 3.5%, peak power is 465w. (just an example)
what does absorb twice?
Yes, i agree that a nice amp need regulated fast psu, specially if it is D Class.