3875 regulated operating points and lm338 heatsinking

Its been awhile since I posted over here! I built an LM3875 with BrianGT boards a few years ago. All sorts of fancy parts too. Last spring I sold that amp and took a troubled little built-for-ebay LM3875 amp in partial trade. The thing is a mess with junky old surplus caps, a big ground loop and funky resistor values. But it works, and I'd like to make it a subwoofer amp.

The transformer the guy choose has way too much voltage for a 4 ohm load. It puts out about 36-37V rectified and under load. Rather than replace it, I decided to do something I haven't played with before. I'm going to build a regulated supply with an LM338. I know that is just over their rating of 35V, but I am going to take a chance.

My local store had a pair of versions of the TO3 package LM338s in stock, so I took them despite the price. They should dissipate some heat on their own, and I have some little AL angle scrapes (maybe 1/2 on each side by 3"?) that would fit nicely in the small chassis (which has an perforated top) to mount them. So, for the 4 ohm subwoofers (they'll be pretty efficient), I'd think of running the chips at 24-25V. But, given my lack of proper heatsinking, should I use higher rails? How high can I go and have it happily drive 4 ohm subs?

Paul
Wild Burro Audio Labs - DIY Full Range Speakers
 
Because this one has 5% more voltage than I need? 5% is pretty darn close in electronics. Many transformers with 35V secondaries would probably give 37V anyway, as so many are rated for 115V lines. The transformer isn't a problem. No matter what I use, I've got to decide what the best operating point is.

Paul
Wild Burro Audio Labs - DIY Full Range Speakers
 
Last edited:

multisync

Member
2008-01-31 10:34 pm
Carp
an old trick /technicque to drop a volt or 2 is to put a diode or2 in series with the supply line. ie put 2 bridges one afte the other your +/- 37 becomes +/- 35.

The Lm 338 has a maximum voltage differential, input to output of 40 volts.. So if your input is 37 and your out is 25 the difference is 12. Well within range. If you play your music loud the Lm338 will get hot and may shut down if the heat sink is too small. Remember 25 volts/4ohm is 6.25 amps peak( with no losses). You will need 2 regulators for stereo. I have run, just for fun mind you Lm3386 at near their max rating +/- 42 vdc into 2.5 ohms. There were 3 of them in parallel. Yes they did get very hot. But it did work. Music is not sine waves. It is possible to run chips/power devices near the max rating with small heat sinks provided you don't do it for very long. If you are not selling the amp and take it easy with the volume control the only thing that may happen is that the chip will shut down due to over heat, over voltage, over current. Then again it may blow up too.

A long while ago I took some ics that were rated for +/-18 volts max and ran them up to near +/- 28volts before they failed. I also tested a number of small signal transistors and found that collector to emitter breakdown would quite often be 100% or more ie double the published rating.

there are not too many semiconductors that fail when you exceed the max by 10%.

Not too long ago nearly every resistor in a circuit was 20%, a few 10% critical areas used 5% very critical, usually instruments were 1%. Electrolitic values were only a sugestion. ie -20+50%. Now everybody wants 1% or better
 
I know that the voltage differential is well within the spec of the 338, I just want to keep heat dissipation to a minimum while achieving the best op point for the chip. I'm not to worried about blowing the 3875's either, I have more. The thing is, I want to maximize the fun. This amp will mainly function from 20-40hz, and while I won't need a ton of power (to drive little tapped horns), more wouldn't hurt. What I want to avoid is losing the peaks to the chip's protection function. It appears to me, based on my understanding of the data sheet, that you get next-to-no power into 4 ohms when you cross 27V or so. Yet plenty of folks seem to advocate running them a little (to a lot) hotter, which will also keep my regulators cooler. I'm tempted to stay down around 25-26V. When the movie goes "boom" I don't want them to run into protection mode and deny me the fun.

I know music is transients, but thermal concerns about the regulator are different, right? Won't it be dissipating more heat when the amp isn't working hard? (which will be most of the time).

So, rather than getting mired in what I need to do to the transformer, what experience or opinions do folks have about rail voltages for the 3875? Am I shooting for 25V, or closer to 30V? That question applies whether or not I am regulating and whether or not I buy a new transformer.

Paul
Wild Burro Audio Labs - DIY Full Range Speakers
 

multisync

Member
2008-01-31 10:34 pm
Carp
if you have lots of amp chips run them in parallel and then you won't have to use regulators and you will get close to 100watts @4 ohms. each chip will dissapate 50 watts. Like I said I tested 3 chips in parallel with +/- 42 volt rails. no problem and the chips did not go into protection. Ran hot yes. Needed a bigger heat sink. Like you said the peaks are only momentary, most of the time.
 
Many transformers with 35V secondaries would probably give 37V anyway, as so many are rated for 115V lines.
rubbish!
A transformer rated at 115/230:35Vac will give out 35Vac when fed with 115Vac and fully loaded, subject to the manufacturer's tolerance.
I would expect the tolerance to be less than +-1%.
 
Last edited:
Pacificblue, that makes more sense. I'm now less worried about heat sinking.

multisync, I would if I could. No room in the chassis, and if I lose the chassis, I may as well toss the whole thing. The point is trying to make is useable.

Andrew, the problem is that many folks in this country can only feed a transformer 115V with a variac. Some folks' line crosses 120V with regularity. I'm more up in the teens (as is common, AFAIK). (in the statement you quoted, I assumed most would know this, but why would you?) Sometimes the extra voltage helps (it certainly does with the Antek transformers in my tube amp), but it is always good to consider prior to ordering a specific voltage. So, if I order a transformer with 35V secondaries, for example, I'll get darn near exactly what I've got now, as would many others.

Paul
Wild Burro Audio Labs - DIY Full Range Speakers
 
PJ,
you are confusing your readers by combining rated voltages with supply voltage variation and transformer regulation variations.

Treat each part separately and then combine them to find what your transformer on your power supply will provide as a nominal DC voltage and also the worst case low voltage and worst case high voltage. The project must be designed not to be damaged by any of these "in tolerance" voltages.
 
I'm sorry if I confused you. To return to my original point though, a tolerance of 5% over the 338's rated voltage doesn't scare me much. As evidence for that tolerance's acceptability to me, I used rated/ vs. line voltage to illustrate that many of us are nearly always running 5% over voltage. (Unless we thought of it before hand, as in the case of my tube amp transformers. Though to use that amp as a further example, I started off having trouble getting within 20% of the HV I was after! Being only 2V off seems to me like a reason to celebrate, not replace parts. But, I really didn't intend to ever discuss 37 vs. 35V, let alone belabor it.)

What I'm really after are folks real world experience with operating voltages on the LM3875 and 4 ohm loads. I'm still thinking +/- 25V, especially as I'm now less worried about the heatsinking on the LM338.

Paul
Wild Burro Audio Labs - DIY Full Range Speakers
 
If you can't put more chips in the chassis and have no room for large heatsinks. You will have to lower your voltage. 35-0-35 ac will give you +/- 50vdc say good bye to your chips. I pushed my chip amps with 30-0-30 ac transformer ( +/- 42vdc but I only paid less than $20 for the Bp150 boards. I used the transformer cause I had lots of them lying around. If your load is 4 ohms stick to +/-25 vdc that's around 18-0-18 VAC transformer.

AndrewT knows his stuff and the chips even better. He has given tons of advice to chip amp builders, some don't take the advice and reget it later.

Pjand1 Yes all unregulated power supplies sag. This can be solved a bit by using lots of caps and a high current transformer
 
I was never contemplating running this transformer unregulated (though that is how the amp came to me), thats the whole point of the thread! Since I've got the regs, the voltage is easy enough to adjust. I'm thinking 25V, but I've seen several folks recommend higher, even with 4 ohms. I'm trying to determine if there is any good reason for that, or if I am better off staying safe at +/- 25V.

Paul
Wild Burro Audio Labs - DIY Full Range Speakers
 
I should add that I understand that the little amp would undoubtedly work more reliably with a new transformer. It would also work better with new signal circuitry. And a new case (more size, better AC connection possibilities). The whole point of this is to try and reuse as much of this piece of junk as possible, mostly because I think it is a fun exercise. The easy way to get a working stereo sub amp would be to buy this: 2*100W TK2050 Tripath T-AMP Upgraded + MW 27V 350W PSU - eBay (item 350407855780 end time Nov-25-10 17:09:04 PST) and call it a day. After all, I've already spent a good chunk of that cost on regulators.

Paul
Wild Burro Audio Labs - DIY Full Range Speakers
 
I'm thinking 25V, but I've seen several folks recommend higher, even with 4 ohms. I'm trying to determine if there is any good reason for that, or if I am better off staying safe at +/- 25V.

The reason is that we don't know how you use the amp. So we have to assume the worst case. And then, if you do the math you find that a single IC will not work reliably with higher voltages into 4 Ohm loads.

It is however possible that higher voltages work for you, if
- the amp has big heatsinks.
- the ambient temperature is low.
- you listen at civilized levels.
- the music you listen to only has short transient with high power demands, but only low power demands on average.
- you speakers are an easy load.
- any combination of the above.

You see that is a complex matter and difficult to assess. If you look at commercial amps, you will find most of them have heatsinks much smaller than theory predicts. In fact most hi-end monsters (or car amps) with kW of output power would be impossible to heatsink on worst-case conditions. The Overture Design Guide also calculates heatsinks much smaller than the worst-case demands. However anything below worst-case heatsinking is something you have to find out on your own by trial-&-error.
 
This amp will mainly function from 20-40hz, and while I won't need a ton of power (to drive little tapped horns), more wouldn't hurt. What I want to avoid is losing the peaks to the chip's protection function. . . . . When the movie goes "boom" I don't want them to run into protection mode and deny me the fun.

Does that help? I don't think heatsinking is the issue, as we are talking mainly about transients. Also, in response to godfrey's "try it an see", an approach that makes sense to me in general, I think it would be somewhat difficult to objectively assess the amp's performance in this application (low frequency transients). When folks run them at higher voltages into low loads, do they just rely on the fact that regular program material rarely puts more than a couple of watts where speakers tend to have their impedance minimum?

Paul
 
Does that help? I don't think heatsinking is the issue, as we are talking mainly about transients. Also, in response to godfrey's "try it an see", an approach that makes sense to me in general, I think it would be somewhat difficult to objectively assess the amp's performance in this application (low frequency transients). When folks run them at higher voltages into low loads, do they just rely on the fact that regular program material rarely puts more than a couple of watts where speakers tend to have their impedance minimum?

Paul
The LM38xx datasheets have lots of information about load, voltage and output power. There's also a calculator somewhere that lets you work out the maximum output from input voltage and load (and ambient temperature and so on) but for the life of me I can't find it at the moment... grrrr...

You're very unlikely to burn the chip with reasonably sane experiments, it has over-voltage and thermal protection.

On thermal protection, if it's clipping transients due to SPIKE protection, it's too hot. It comes down to that. Depending on circumstance, it may be impossible to cool the chip enough to avoid protection distortion at reasonable volume levels.
 
Last edited: