Want to build a SS PA

[lowell]

I'll check out the LM3886 info. I've replaced a few of those in my time, they are familiar. But yeah designing/building is a whole other story. I've built my share of tube amp kits and 5 tube amps of my own design all the same model. Solid state is a whole new story. I'll be back with more questions for sure... my first question will probaby be about a power transformer... we'll see.

[R.G.]

my first question will probaby be about a power transformer... we'll see.

You may not need the help, but I'll jump to one of my personal conclusions.

The ideal power transformer for one of these is very close to 42Vct at 80-100VA (2 to 2.5A) and toroidal. That gives you +/- 30V dc (about) and lets you run both 4 ohm and 8 ohm speakers from one chip without having it go into thermal limiting too badly on either 4 or 8. The maximum applied voltage is a practical limitation on this chip, because it will self-limit thermally. +/-28 give the maximum power on 4 ohm loads, +/-35 give max power on 8 ohms. Using a higher power voltage causes the chip to dissipate more power internally, and causes it to internally limit because it gets too hot before it puts out the max power.

The transformer current rating is solidly debatable. For continuous sine-wave power testing as opposed to listening, the power transformer needs to be able to supply the maximum DC power continuously. This is pretty severe. For, say, 50W out, you need 50W plus the power spent in the chip heating, plus any conversion losses in the rectifiers and transformer. Generally, you can make 50W rms under continuous sine wave testing with an 80-100 VA transformer without overheating the transfomer.

Music is not continuous sine waves. It has peaks and quiet places. So the actual long term average, which is what a power transformer with a 1 hour time constant will see, is probably more like 1/5th to 1/3rd of the max sine wave power, and a transformer rated for the actual power output will not be noticeable under actual listening from one with a bigger to much bigger rating. So it's probably OK to run one chip from a 50VA power transformer or even slightly less. For stereo use, the crest factor (ratio of average to peak power) is even less, so most home stereos use power transformers of as little as 30% to 50% of the average output power. Music equipment will be treated more harshly, so using 100% of the average is likely OK. 180-200% will get you to rock solid 100% duty cycle operation with constant signals, like for shaker table duty.

If you go with multiple parallel or bridged chips, multiply the transformer VA you used for one chip by the number of chips. So if you were OK with 80VA for one chip (good choice!) then for two chips in parallel to power lower-impedance speakers, use 160VA. For two chips in bridge, also use 160VA. For four chips either parallel or bridged, use 320VA. This will get you to the maximum practical output from the chips without overheating the transformer.

Rectifiers are cheap - use big ones. Buy the 200-600V diode bridges rated for 25A. They're only about $2.50. Solder four 0.01uF ceramic disk caps to the bridge, one between each pair of terminals. Use at least 4700uF/35V for each power supply cap, preferably 10,000 uF/50V, for each chip in the power amp. Use snap in capacitors or surplus computer grade (screw post type) for the higher ripple voltage rating.

[Steve Conner]

Am I the only one that remembers the disastrous reliability record of Marshall's MF350? That should be a warning against paralleling or bridging LM3886 type chips. (Though I think the Marshall amp used the TDA7293?)

[J M Fahey]

I do, for one, and add:
1) SS power amplifiers are *NOT* , repeat *NOT* meant to have their outputs paralleled. NEVER.
I DON'T F****** CARE WHAT THE DATASEET SAYS !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! NEVER !!!!!!!!!!!!!
If you need to drive a lower impedance load than what's realistically possible for your chipamp, go discrete and double (or triple) the amount of devices, all parallel driven from the SAME "brain" , receiving only ONE feedback signal. Period. No two brains, no two feedback signals.
I don't care that you get 1% resistors for free and can use them everywhere. It's not close enough.
2) Chipamps *can* be used bridged, each "brain" handles his own speaker terminal.
Problem is a psychological one.
Musicians love to add extra speakers, now or in the future.
Even if they don't, musical instrument speakers often have lower DC resistance than usual, I guess they do that to give an impression of even higher efficiency.
Personal measurements: I find most good quality 8 ohm "audio" speakers around 6.5 ohms DC; a few even almost reaching 7 ohm; but many "own brand" MI speakers, almost always made by Eminence but labelled "Fender/Ampeg/Crate/Randall/etc."., the generic black frame one found everywhere, tends to be around 5.5 ohms, and a sizable quantity plain 5.
Car audio speakers, of course, are a mess, their DC resistance is often 1/2 the rated impedance.
Bridging 2 LM3886 turns their individual 4 ohm minimum rating into 8 ohms minimum *combined*, so an 8 ohm speaker is already on the limit , and, as I said above, a MI speaker is even tougher to drive
With care, bridged chipamps *can* be used safely, but can not be trusted "in the wild", in an amp sold for use by anybody.
The otherwise excellent Crate Power Block (very good sound) had 2 75W/4r chipamps which could be bridged for 150W/8r : they dissapeared from the market with the last ones sold for peanuts.
I guess they were overloaded to death very often.
Other users complained that they would make a "tick ... tick ... "sound, turning on and off cyclically, a surefire indicator of their SMPS trying to cope with overload.
Oh well.
Yet, "one chipamp per speaker" as was suggested in an earlier answer sounds as sage advice to me.
An LM3886 driving an 8 ohm speaker is well within ratings, a future extra 8 ohm one takes it to the edge, but not across it.

[imaradiostar]

I thought I'd chime in as I've been toying with some solid state ideas. There are quite a few available PA amps on ebay- plenty of pre-built chip amps but the most interesting ones are discrete mosfet or bipolar transistors. If you have an assortment of transformers and heat sinks around from old PA amps (as I do from ones I took apart many years ago) it can be quite inexpensive to build a few power amps using the pre-assembled power amp boards.

Check this Ebay store.

Lots of interesting listings. I personally have three 60 volt center tapped transformers that are begging to be made into 150 watt/channel PA amps. A few of those would be enough to run separate mixes for monitor wedges or multiple speakers in my shop. I have a Peavey CS800 power transformer that's begging to be made into a large amp for running subs. One day...when I've got the time.

jamie

[Enzo]

Well I certainly will never forget the MF350, but we refuse to take them in for repair so it is academic. Yes, that was TDA7293 chips. They were not in parallel really. There were two pairs in bridge. And in each pair one was wired as a power follower. (Or some kind of follower) I never want to see another one.

[R.G.]

From all I read, the TDA7293 is a very touchy chip. Good concept, but fragile in real applications. I have not heard the same of the LM3886.

On paralleling SS power amps: All practical parallelling schemes use high precision gain setting resistors and low value/high current series resistors to take up that last few millivolts of offset and non-identicality of signal. It can be handled. But one amp per speaker is just so much easier, as well as more robust, as long as you can live with only 50W or so per speaker.

And to second guess myself even more, a discrete amp is really not all that touchy. It takes about $2.00 worth of semiconductors and passives and a good PCB. Once the PCB and compensation is done, it's pretty straightforward. Frankly, the hard part is having to mount two power devices instead of one, and then worry if your protection circuits work right.

[lowell]

All good info, thanks. Unfortunately the GGG kit is unavailable now. I found one on Chipamps.com but the guy has not responded to my query regarding a few questions. I may just order it and see.

Actually on second thought the chipamp.com kit is too much $$... at $90. Looks like I'll be making my own board and working this out myself. Woulda been nice to use a kit, but wouldn't learn as much from a kit.

[lowell]

I think I'll go with the GGG design but make the board and assemble all components myself. Doesn't look too bad. No more difficult than building a pedal really. The power supply will be the hardest part, but not too hard.

I'll be using a Champ power transformer for the preamp in my amp. I'm thinking I could have 2 rectifiers. One for the tube preamp and another for the +/- 18v supply. FYI the typical quiescent current for the LM3886 V+ is 50ma.

Will this work as a power supply post 2nd rectifier? (one for the + supply and one for the - supply?)

300v
220k/15k divider
=20.45v
18v regulator and filter caps

Not sure if current demands from the chip will be an issue with this sort of supply.

[J M Fahey]

Not sure to understand your last post.
¿Do you mean you intend to feed the LM3886 from the 300V "Champ" supply?
Thanks.

[lowell]

No. What I meant was: Can I use the Champ power transformer as a power transformer for my tube preamp AND for the bipolar 18v supply for the LM3886?

I'm thinking I could tap off the secondaries with a separate rectifier and divide the 300 volts down to about 20v, then into some filter caps and an 18v regulator. Like this:

300v
220k/15k divider
=20.45v
18v regulator and filter caps
=18v

I'm thinking I would need separate regulators and voltage dividers for each polarity.

[Enzo]

But I thought you wanted like 50 watts from this? Is that 300v winding going to provide 50 watts worth of current at 18 volts? I doubt that.

[lowell]

That's 2.8a of current right? Not sure what you're getting at. Is a champ pt not capable of that?

[J M Fahey]

Dear Lowell. Please search and post here what are the PSU requirements (voltage and current) to feed an LM3886 power amplifier. Only after that we can discuss where can we pull that kind of power from. Thanks.

[lowell]

The quiescent current is 50ma and the supply voltages range from 24-80v from what I remember. I don't have access to it right now but I think that's correct.