Need feedback on my powersupply design

[MTron]

I am in the process of building a LM3875 chip amplifier. Now due to the unique characteristics of the LM3875 in terms of supply voltage vs speaker impedance (see datasheet for more details) i wanted a power supply that was adjustable so i could dial in my supply voltage to suit different speakers.

I already have the transformer, a 200va Toroidal with dual 25volt 5a secondaries (not 25-0-25 center tap, 2 separate secondaries), so the power supply was built around this to give me a Vout of +/- 25-32 volts 5Amps

Ok, quick design criteria, Symmetrical power supply, with Vout range of +/- 25 to 32 volts. Max draw on the system 5amps as a safety rating, the amp should not draw more then MAX 4 amps even at max power at the lowest voltage. A really well filtered supply was the design goal, as this is for a HiFi audio application.

The +/- Vout rails lead to each chipamp channel, where another pair 2200uF electrolytic filtering cap is in close proximity to the V+ and V- pins on the LM3875

The LM338's will be connected to a generous heatsink.

Please look over my schematic to let me know if there are any problems...of if the design will even work!!



Matthew

[NiHaoMike]

It should work fine. The 2.2k resistors are not necessary.

[MTron]

Thanks, the 2.2k resistors were incorporated from another powersupply design i borrowed from, i was not entirely sure of their function in the global scale of things, so i left them.

Well, i'm going to go ahead and start buying the components, unless someone else has any input???

[Zero999]

Yes, there's no point in having the 2k2 resistors, they just waste energy.

[alm]

I think they were included to make sure the fairly large electrolytics are discharged immediately after it's disconnected from mains, even without a load connected. What's the point of the 1 ohm resistors is series with a 100nF cap? Series resistors are sometimes used to increase ESR/decrease inrush current or dampen resonance, but I don't see the point with lots of capacitance in parallel (including identical 100nF caps).

[Zero999]

Bleeder resistors are not needed for non-hazardous voltages.

The LM338 will discharge the capacitors when the power is disconnected.

The 1R resistors in series with the 100nF capacitors are pointless and actually degrade the performance so should be removed.

How much current are you intending to draw and at what voltage?

You do realise that the LM338 can't output 5A when the input-output voltage is higher than about 10V?

[tecman]

A couple of notes.  If you dial down the output voltage too low, along with a high line input you could exceed the maximum Vin to Vout rating.  I would also add a couple of ceramic or small tantalum caps to the output.  You want to be sure to have low impedance to the amp at high frequencies as well, so some caps will help.

Paul

[MTron]

Thanks for all the input. Ill just start with noting that i have no formal training in electronics, i jumbled this together from a few diff designs to suit my needs

I'll address the questions as they appeared

The 2.2K resistors: As stated, i don't know WHY these are there, they were just in the design, if i had to guess, i would say its to put a bit of a load on the transformer so the idle voltage isn't too high.

What's the point of the 1 ohm resistors is series with a 100nF cap?

No idea, again part of someone else's design, it came off another power-supply for a different gain-clone chip

The 1R resistors in series with the 100nF capacitors are pointless and actually degrade the performance so should be removed.

Will do, thanks.

How much current are you intending to draw and at what voltage?

You do realise that the LM338 can't output 5A when the input-output voltage is higher than about 10V?

  According to the datasheet, at a 10 volt drop the MINIMUM available current is 5, typical is 8. With a 15 volt drop (and ideal temp, granted,) the typical is still around that level. That being said, i am only going for about a 10v drop. The transformer is 25 volts, so rectified it should be about 35. I can also use a few series diodes to drop it another volt or so if the transformer outputs too high. The minimum voltage i would need to run is 25 volts, and at that voltage the max draw i should ever see is 4 amps.

Remember this is an audio amplifier, the max power output at 25 volts is 47 watts per channel, so about 94 watts, which comes to just under 4 amp draw at 25 watts...now this would be at max volume...and even then, audio signals never draw full power all the time.

Given the speakers i already have (6 ohm) i will likely be running a voltage of about 30v, with a power of about 55, this drawing slightly less current....again, full draw would be at full volume, with max load...unlikely with an audio amp.

I would also add a couple of ceramic or small tantalum caps to the output.  You want to be sure to have low impedance to the amp at high frequencies as well, so some caps will help.

ill take the tantalum caps into consideration, but as for the high frequency, isn't the whole point to eliminate any frequencies from the DC supply going to the chipamp?


Ok, all that being said, clearly there are some pitfalls to keep a watch out for, i suppose a more practical solution would be a voltage regulator that can handle more current, with similar output voltages. Again i really only need Vout in the range of 25-32, with a Vin of (hopefully) 35

Again, thanks for all the input!

[DJPhil]

I would also add a couple of ceramic or small tantalum caps to the output.  You want to be sure to have low impedance to the amp at high frequencies as well, so some caps will help.

ill take the tantalum caps into consideration, but as for the high frequency, isn't the whole point to eliminate any frequencies from the DC supply going to the chipamp?

I think I see the confusion here.

The short version is: yes, you do want to prevent noise from getting through, and the regulator will help with that a great deal. However, the regulator isn't a perfect voltage source, so when the current demand from the amp rises (think the beginning of a big bass note) the voltage drops and the regulator has to respond. As the frequency rises the regulator has trouble sourcing current at the speed the signal demands, and this causes fluctuation in the power rail. Think of the extra capacitors as small reservoirs of charge that can respond to that need more quickly than the regulator can. The 100uF cap on the output helps in this regard for lower frequencies, and adding a 1uF or 470nF (or both) after it can help with the higher frequencies.

I found a document that covers this much better than I can.

Hopefully someone will straighten me out if I've gone too far astray.

[MTron]

Hrmm, that document is actually very helpful!

I have no problem adding more caps, my digikey order is full of spare parts as they put me in the next price bracket, and i end up paying less for more!

Just as a question, where should these 1uf or 470nf caps be just after the voltage regulator or down by the chip amp. There will be some distance between the two, they will not be on the same board, because of the wire run from the psu to the chip, and per a design from someone else, there are 1500uF electros on each rail just before the chip, and i mean like a 1/4 inch away or less, should i put the 1uf/470nf caps before or after the electro?

also would 100nf metalized polypropylene work? i have a few extra of those already.

[DJPhil]

Just as a question, where should these 1uf or 470nf caps be just after the voltage regulator or down by the chip amp. There will be some distance between the two, they will not be on the same board, because of the wire run from the psu to the chip, and per a design from someone else, there are 1500uF electros on each rail just before the chip, and i mean like a 1/4 inch away or less, should i put the 1uf/470nf caps before or after the electro?

Note: I should have stated before, those values were just my guesstimates and aren't the result of any calculation.

Actually, I'm not sure, but I'd put them close to the (massive) electro near the chip. As long as the bypass caps are close to the chip I don't believe it matters what order they're in, though I could be wrong. Page 12 of the datasheet has details on power supply bypassing for the chip itself. If you have other circuitry (preamp, etc.) I'd suggest bypassing it locally, as this ensures a good bypass for each current sink.

Keep in mind, I'm at the edge of my understanding here. I'd feel a lot better if one of the pros would chime in if I'm over my head.

also would 100nf metalized polypropylene work? i have a few extra of those already.

They might not be optimal, but they could work. Ceramics are suggested for the lower values due to good HF performance and low cost, tantalum electros are recommended for the higher (1uF +) values to to low series inductance. Metalized poly caps only really suffer from high cost, as far as I know. As for the value, it's better to go with the datasheet than my weird guesses, so 100nF should be perfect.

After re-reading the above it seems as clear as mud, so I'll summarize so it may be subjected to peer review.

Add your 100nF in parallel to your giant 1500uF cap, close to the chip, on whichever side of the big one is easier to layout. Just for the heck of it, if you feel like it, add one in parallel to your 100uF on your power supply board. It probably won't help, but it shouldn't hurt either. The first should be sufficient, and that should cover you. For any other circuitry, try to bypass it as well with >10uF and ~100nF.

Hope that helps, and I hope I got it right.

[Time]

You need feedback on your powersupply design? 

HAHA, get it?


Constructive comments:

Maybe use a dual ganged pot for the varying resistances...


Why not use the LM337 for the negative rail instead?

Are there any drawbacks to using the positive regulator wired in this manner as opposed to the sister part which provides a negative voltage in the first place?  Just seems more... symmetrical.  I don't know.  This question is directed at everyone.

[DJPhil]

Maybe use a dual ganged pot for the varying resistances...

If it were a bench supply that would be ideal, but given that the load is an amplifier the voltages won't be frequently adjusted. The tracking on dual gang pots is usually poor unless you're spending a good deal of money, so separate adjustment allows each rail to be trimmed independently. It might be wise to split the resistance between a large fixed resistor and a low value trim pot for better accuracy.

Why not use the LM337 for the negative rail instead?

The 337 is rated at 1.5A max. Adjustable negative regulators top out at about 3A (LM333 I think), after which you either use separate windings (as was done here) or a pass transistor circuit. To properly implement current limiting and protection with a pass transistor system means adding several components and losing out on some of the convenience of a plug and go three terminal regulator. At some point it's just easier to design your own discrete regulation scheme.

Are there any drawbacks to using the positive regulator wired in this manner as opposed to the sister part which provides a negative voltage in the first place?  Just seems more... symmetrical.  I don't know.  This question is directed at everyone.

I know what you mean, I asked the same questions when looking to build an adjustable linear bench supply. The above pretty much explains the necessity, but I don't know of any specific drawbacks other than the necessity of using two separate windings.

Hope that helps.

[Zero999]

According to the datasheet, at a 10 volt drop the MINIMUM available current is 5, typical is 8. With a 15 volt drop (and ideal temp, granted,) the typical is still around that level. That being said, i am only going for about a 10v drop. The transformer is 25 volts, so rectified it should be about 35. I can also use a few series diodes to drop it another volt or so if the transformer outputs too high. The minimum voltage i would need to run is 25 volts, and at that voltage the max draw i should ever see is 4 amps.

Yes, I made a mistake, safe area protection doesn't kick in until at least 15V.

You also need to consider what is practical.

The thermal resistance from die to heat sink is 1°C/W, the maximum operating temperature is 125°C. If the heat sink has a thermal resistance of 1°C then with an ambient temperature of 25°C the power dissipation is then limited to 50W which is a 10V drop and a current of 5A.

Remember this is an audio amplifier, the max power output at 25 volts is 47 watts per channel, so about 94 watts, which comes to just under 4 amp draw at 25 watts...now this would be at max volume...and even then, audio signals never draw full power all the time.

Given the speakers i already have (6 ohm) i will likely be running a voltage of about 30v, with a power of about 55, this drawing slightly less current....again, full draw would be at full volume, with max load...unlikely with an audio amp.

You shouldn't need a regulated power supply for an audio amplifier which, if properly designed, will have a high power supply rejection ratio, meaning that the ripple on the power supply will not generate hum.

I think 47W per channel is a little optimistic with a 25V supply. Don't forget that the output transistors will probably need  to drop at least 3V to avoid clipping, a peak voltage of 22V will give 40.3W into 6?.

The peak current will also be much higher than what you've calculated, with 40.3W out, the peak current will be 3.67A per channel, which is 7.33A total, not a problem for the LM338 though.

[MTron]

You shouldn't need a regulated power supply for an audio amplifier which, if properly designed, will have a high power supply rejection ratio, meaning that the ripple on the power supply will not generate hum.

I think 47W per channel is a little optimistic with a 25V supply. Don't forget that the output transistors will probably need  to drop at least 3V to avoid clipping, a peak voltage of 22V will give 40.3W into 6Ω.

The peak current will also be much higher than what you've calculated, with 40.3W out, the peak current will be 3.67A per channel, which is 7.33A total, not a problem for the LM338 though.

Yea, my bad on the calculations.

But as for the regulated supply, yea normally i wouldn't bother, but the LM3875 has a rather quirky trait in that the speaker impedance it is able to work with is dependent on the voltage input. With a Vin of +/- 30v or more it would be unable to drive a 4ohm speaker properly (see page 10 of the datasheet, bottom left graph). To drive a 4ohm load a Vin of +/- 25 is what i have determined as a safe level, 30v for a 6ohm and 32+ for 8 ohm. For the highest power into 8ohm i would bypass the regulator to get the full 35available volts, but i doubt i will ever go there.

All this touble is just a result of already having this 25v Toroidal transformer and wanting the amp to work with any speaker impedance

[Mechatrommer]

if i'm designing the PSU, for the +ve rail (same with -ve), i'll remove R1,R2,C3. i'll move C2 and add similar to C1 on the right of LM338 (well, my right when looking at this picture).
and... i'll tap somewhere from left or right of the LM for a DC fan to cool the transformer down, so it will last longer Cheers!

[MTron]

if i'm designing the PSU, for the +ve rail (same with -ve), i'll remove R1,R2,C3. i'll move C2 and add similar to C1 on the right of LM338 (well, my right when looking at this picture).
and... i'll tap somewhere from left or right of the LM for a DC fan to cool the transformer down, so it will last longer Cheers!

Yea, removing R1,R2 and C3 is the general consensus, i will move C2 as you suggested, that just seems logical....as for doubling up C1...that would increse costs substantially....and may not be needed considering there is an additional 3000 uF (2x 1500 on each rail)  before each chip amp (stereo)

[Zero999]

But as for the regulated supply, yea normally i wouldn't bother, but the LM3875 has a rather quirky trait in that the speaker impedance it is able to work with is dependent on the voltage input. With a Vin of +/- 30v or more it would be unable to drive a 4ohm speaker properly (see page 10 of the datasheet, bottom left graph). To drive a 4ohm load a Vin of +/- 25 is what i have determined as a safe level, 30v for a 6ohm and 32+ for 8 ohm. For the highest power into 8ohm i would bypass the regulator to get the full 35available volts, but i doubt i will ever go there.

I don't think there's anything quirky about that. It looks like the IC has some sort of protection that limits the current when depending on the input voltage.

You could try adding a 2R power in series with the output but before the feedback loop. Negative feedback should keep the effective output impedance low enough to give a good damping factor but as far as the protection circuit is concerned, the impedance will be 8R. You'll  have to accept that the maximum power output will be 3/4 of what it says on the graph for 8R, so for 35V you're looking at 45W, although I think 37.5W is more realistic due to the voltage drop on the rectifiers and ripple on the capacitor. The resistor will have to be rated for 1/4 of the power output, although it's possible to de-rate as there's no chance that it'll be running at full power continuously.

[kmossman]

I would not tie the "negative" rail to the positive ground. I would use a 'proper' negative voltage regulator.

I suspect the 2k2 resistors are to 'bleed' off the 10,000uF caps to make it safer.

Personally I would put say 4,700 uF caps in for C1, and C6 and make C5 and C10 low ESR 10,000uF.
That would mean the regulator is charging C5 and C10, and so the amps are drawing off the caps, and not the regulator.

Turn-on thump might be a problem - you might want to design it for a slow turn-on - perhaps change the VRs into a resistor and VR, and put caps around the resistor to slow turn-on.