36V symmetric power supply

[TheAmmoniacal]

So I'm working on a project that is designed for a 36V symmetric power supply, from -36V to 36V across two nodes on the circuit. I've been looking around a bit to find a good schematic for such a supply, so I could 'bodge' one up. I haven't been able to find one with the full documentation a noob like me need, but I did find a 24V symmetric supply that looked good. The documentation for it states that the max input voltage is 24 VAC (2x), with a max heat dissipation of 20W. The circuit is added as an attachment, and my question is which changes/modifications would allow me to connect it to 36VAC? And thereby dissipating almost 40W?

[Richard Crowley]

How much current does your power supply need to source? 
What is the load?  Does the supply voltage really need to be regulated?
The LM317 has a max output voltage of 37V which is rather cutting it close for a 36V supply.
You may want to start back a step or two and analyze what you need vs. simply copying something not quite right.

[TheAmmoniacal]

How much current does your power supply need to source? 
What is the load?  Does the supply voltage really need to be regulated?
The LM317 has a max output voltage of 37V which is rather cutting it close for a 36V supply.
You may want to start back a step or two and analyze what you need vs. simply copying something not quite right.

At this moment I am just copying an audio amplifier circuit that is rated to supply 100W to a 4 ohm speaker, so I would assume the power supply has to supply ~1.4 A (100/72), or 0.7 A per secondary winding. At the moment I'm just powering the circuit from a computer power supply, which has a -12V and a 12V pin, which is working fine, just not that much power - I guess 24V would be sufficient though. I'm just not sure how well the LMs would operate at their limits, if I can get it up to 36V or if I have to look at different designs.

[Richard Crowley]

Why do you even need a regulated supply at all for a power amplifier?
We typically don't bother regulating the power for power amplifiers because of negative return on investment.

[dannyf]

We typically don't bother regulating the power for power amplifiers because of negative return on investment.

2nd that.

[madires]

Yep, just build a solid power supply with a recitifier and caps. For 100W audio (class AB?) the power supply should able to deliver roughly 150W. I'd suggest 4700µF for the caps at minimum.

[fcb]

It's all about the current...

Regulating power-amplifiers does happen and it does deliver results. As for negative return on investment....? I hope this doesn't turn into another audiophile bashing thread.

I would suggest you look-up the schematics of the Naim NAP135 or NAP250, they both have regulated main-rails and there is a lot of cloning going on.

[mariush]

A computer power supply can't provide a lot of power on -12v, the -12v is usually rated only for 0.5A or less.

You don't need regulated power supply for power amplfiers, just get a transformer with a center tap or two independent secondary windings, a bridge rectifier and a bunch of big capacitors to smooth out everything as much as possible.  A  48-56Vac transformer with center tap will do, you'll have 24-28Vac on each winding, so 33-38v peak DC... a lot of input capacitance (8000-16000uF on each side) will make the output hover around peak dc - 1-2v.

[Richard Crowley]

A power amplifier that benefits from regulated power rails is, ipso-facto, inadequately designed and uses regulated power rails as a crutch to make up for faulty design.  Cloning poor designs is the stock-in-trade of audiphools.

[dannyf]

Cloning poor designs is the stock-in-trade of audiphools.

To be fair, the particular distortion from poor designs may well be what the audiophiles are going after.

[TheAmmoniacal]

So you think a simple rectifier circuit, as attached, will do? I'm not an "audiophile" by any means, and have no idea how well a regulated supply would compare with it. For some reason my mind just locked onto the idea of a regulated supply, but as mentioned it's probably not worth it. Thanks for getting that clear.

[mariush]

Yes, but the capacitance could be better (more).  A formula approximates capacitance in Farads =  Current / (2 x AC Frequency x Vripple) where Vripple is how much you are willing to have the output voltage go down from peak voltage.  For 1A and about 2v ripple and 50Hz mains frequency, you'll have C = 1/2x50x2 = 1/200 = 0.005 F = ~ 5000 uF

[madires]

For that current please go for 10000µF caps or multiple caps in parallel. You could add 100nF caps in parallel to the rectifier diodes to reduce EMI if you are using standard Si diodes.

[NiHaoMike]

A power amplifier that benefits from regulated power rails is, ipso-facto, inadequately designed and uses regulated power rails as a crutch to make up for faulty design.  Cloning poor designs is the stock-in-trade of audiphools.

Some digital amplifiers scale the supply voltage to the volume setting in order to increase efficiency and get better resolution at low volume settings. In these, the main supply rails are not just regulated, but variable.

[calexanian]

I agree.. A properly designed, and internally grounded amp running at a proper quiescent current should have no perceivable hum, and if you are having supply voltage sag appearing on your output, you are invariably in clipping output anyways in which case nothing can help you. 

[David Hess]

The LM317 has a max output voltage of 37V which is rather cutting it close for a 36V supply.

The LM317 is a floating regulator so the maximum 37 volt specification is maximum input to output voltage.  It will quite happily regulate higher voltages as long as provisions are made to limit the maximum input to output voltage under startup and fault conditions.

[kodi]

Sorry for getting this old post revived, but I'm just doing a project of PSU for power amplifier (originally there was a switching power supply inside) with +-36V rails and about 1-1.1A output, and I was about to build something VERY similar to the design in post one of this thread. As I have to build a total of 12 power supplies like that, cost is one of important factors. And putting battery of total 10-20mF 50V caps is total opposite to "cost effective". So can anyone tell if the schematics from the first post is valid for +-36V output?
Input voltage will be around +-38-39V and I know 317/337 are limited to drop down voltage of 37V, so in theory there is no problem here, but I was wondering if during initial power-up there will be too much stress on the chip.

[Kleinstein]

For about 1 A of output current you should get away with something like 3300-4700 µF if more ripple is tolerated. This also helps not to have to much peak current for the transformer. There is no fixed limit, its just that less capacitance means more ripple and less peak current, thus less load to the transformer. 

The fused belong between the transformer and the rectifier, as the rectifiers might fail short (the typical mode of failure of a rectifier) and the resistance of the fuses this way helps a little reducing current peaks.

At high voltages a short of overload is critical for the 317. So turn on might be a problem with a large capacitive load. Under normal operation the 317 will not even see the high voltage, but only the difference from input to output. Also sufficient cooling gets important as thermal shutdown under high voltage could lead to disaster.

[Halvmand]

Sorry for getting this old post revived, but I'm just doing a project of PSU for power amplifier (originally there was a switching power supply inside) with +-36V rails and about 1-1.1A output, and I was about to build something VERY similar to the design in post one of this thread. As I have to build a total of 12 power supplies like that, cost is one of important factors. And putting battery of total 10-20mF 50V caps is total opposite to "cost effective". So can anyone tell if the schematics from the first post is valid for +-36V output?
Input voltage will be around +-38-39V and I know 317/337 are limited to drop down voltage of 37V, so in theory there is no problem here, but I was wondering if during initial power-up there will be too much stress on the chip.

At 1 A you will drive a lm317 almost to it's limits.
Peak current is higher and it might limit the current internaly.
I would go unregulated,  but there is bigger 3 pin regulators out there if you insist.

[kodi]

Sorry for getting this old post revived, but I'm just doing a project of PSU for power amplifier (originally there was a switching power supply inside) with +-36V rails and about 1-1.1A output, and I was about to build something VERY similar to the design in post one of this thread. As I have to build a total of 12 power supplies like that, cost is one of important factors. And putting battery of total 10-20mF 50V caps is total opposite to "cost effective". So can anyone tell if the schematics from the first post is valid for +-36V output?
Input voltage will be around +-38-39V and I know 317/337 are limited to drop down voltage of 37V, so in theory there is no problem here, but I was wondering if during initial power-up there will be too much stress on the chip.

At 1 A you will drive a lm317 almost to it's limits.
Peak current is higher and it might limit the current internaly.
I would go unregulated,  but there is bigger 3 pin regulators out there if you insist.

Power supply I want to replace was rated at 1A and the highest peak I measured was 1.087A (at turning on).  LT317/337 datasheet states that max output current is 1.8A @ 150 degree and 2.2A at 25 degrees, so I think I will be safe here.

For about 1 A of output current you should get away with something like 3300-4700 µF if more ripple is tolerated. This also helps not to have to much peak current for the transformer. There is no fixed limit, its just that less capacitance means more ripple and less peak current, thus less load to the transformer.

Are you talking about unregulated solution? 

The fused belong between the transformer and the rectifier, as the rectifiers might fail short (the typical mode of failure of a rectifier) and the resistance of the fuses this way helps a little reducing current peaks.

At high voltages a short of overload is critical for the 317. So turn on might be a problem with a large capacitive load. Under normal operation the 317 will not even see the high voltage, but only the difference from input to output. Also sufficient cooling gets important as thermal shutdown under high voltage could lead to disaster.

Good thinking about fuse.
And there is enough room for ample heatsink

Well I ordered some transformers so the best method is just to do an experiment LTSpice shows that there should be no problem, but.. we will see

[Halvmand]

Well in most cases you probably will be fine.
But don't blame me when the party stops just because you skimped on the power supply in your amp.
How did you measure the peak current? I'm not sure your multimeter catches those fast transients. Your power supply should handle all what the amp is requiring. that is including the peaks.
You've said you had something similar to the topic starters amp which is 100 w. I don' see that happening with a LM317.

That is the beauty of the transformer / rectifier. As long you don't continuously overload it, it'll happily absorb the peaks.
The lm 317 won't. I don't know how it behaves when overloaded in that manner, but I suspects it will be audible at best.

Those numbers from the LM317 data sheet are the absolute maximum values. Don't get to close to them.

And the last thing. Why do you fell the need to piss away heat in a linear regulator, ad complexity and cost, especially when the last was one of your concerns?

[Halvmand]

Just thought of an other thing to further discourage you.

The extra power dissipated in the regulators, you will have to factor in when buying transformers. Now you again will have to spend extra (bigger transformer). I'd say that's a minimum of 6 w per channel at best.

Lets say a 10 VA step up in transformer is appropriate.

50 VA transformer price = 27 $: http://www.digikey.com/product-detail/en/triad-magnetics/VPT36-1390/237-1334-ND/2090072
Price pr VA = 27 / 50 = 0,54$
0,54VA * 10 = 5,4$

5,4 Dollars extra in transformer cost, just because of that regulation.

[Richard Crowley]

@kodi seems to have fallen into the same trap as @TheAmmoniacal (who apparently gave up and never returned).
Neither of them can explain why they even THINK they need a "regulated" power supply for a power amplifier?
Until that is explained adequately, this discussion seems like a waste of time.

[kodi]

Ok. Background story. I have 6 Alesis M1 Active Mk2 monitors. The configuration of it is Bi-Amp.

Original PSU is rated
+/- 36V (2x 1A)
+/- 18V (2x 1A)

Problem with them is that PSU after few years gives up. I have two of them still working, so I want to first revive the dead ones and then

Regarding costs: Transformers will be custom made, so there is no difference in price.
Good caps 10mF rated 50-63V  are about 10-12Euro (depending if single bigger or mutliple smaller ones - I prefer the second one of course). So we are talking minimum 20 Euro per +-36V filtering plus around 6-10 for +- 18V. Thats 26 Euro times 6 = 156 Euro (not much for some, but it matters to me). OTOH I have  14 2200-2700uF @50V + some @ 35V caps at my disposal at home + all LT317 and LT337 I would need. I have almost all needed to do a regulated power supplies.

@kodi seems to have fallen into the same trap as @TheAmmoniacal (who apparently gave up and never returned).
Neither of them can explain why they even THINK they need a "regulated" power supply for a power amplifier?
Until that is explained adequately, this discussion seems like a waste of time.

I am sorry I am wasting your time, I wasn't intending to.
There are no proper service manuals for those monitors and yes, I was looking over the Internet, and no - they not fit, even when the description says that they are for this particular model, all I know that on those rails voltages are pretty stable under different loads (and yes, I know how to measure peaks, yes my Brymen can do this, and yes I verified that on the scope).
Someone went a pretty long way to stabilize voltage on those rails (please see attachment how the original PSU looks like).

[NiHaoMike]

A lot of the time, it's the caps that fail. Check them first and it could be as simple as replacing them to have a working PSU again.