6 channel LME49830 power amp

[cuibono]

I'm back, and ready to roll.

I'll be putting the first module through more rigorous testing this weekend, so expect some results!

[cuibono]

Well, things are looking good.

I ran the amp into a load, with signal, no real problems. Very nice.

I popped the 20pF comp cap back into place (I was using a 30pF), and had no problems! Nice. I could try and push things farther (bandwidth vs. stability), but I think for my purposes, things are fine.

I did notice one problem though - the temperature/bias voltage slowly slips upward. Past a certain temperature, the Qvbe transistor starts putting out more bias voltage, and so things get hotter and hotter. Luckly it happens slowly, but I assume it would eventually self destruct. So I need to adjust the resistors around Qvbe, particularly Rb3, which sets the slope of the temperature/bias curve. I guess I have to drop it some. Looks like it will take some playing with...

But otherwise, things look good. I feel like I'm ready to build up the rest of the modules, and close things up.

[cuibono]

moving Rb3 down a little helped the temp. compensation - just slightly negative slope, which I think is good. But I messed up the resistor ratios, and was getting oscillations, and then blew my only fuse....

So hopefully, I get the fuse soon, and get the ratios back to where they should be, all works well, and this story is over!

[cuibono]

Damb, those resistors are finicky.

I after about 4 hours of trying to get just the right set, I blew the power supply somehow - the soft start relay wont switch on, and the thing just wants to smoke

Oh well, maybe later..

[cuibono]

Well, I guess I've got a short in the amp somewhere - the psu seems to be okay. It looks like I've got a short on the negative rail somewhere - either a mosfet or maybe the protection diode? Hmmm..

I guess I'll have to take it apart to check the fets.

[cuibono]

Well, two fets were blown. I think I 'accidentally' drew about 3A per device of bias current through them when they died. BUT! I replaced them all (twice), and the thing very much isn't working - the rails won't pull up (in fact, the positive rail looks like its trying to swing negative), and at least one fet is pulling about 500mA through the source resistor (.1ohm). So I replaced all the fets twice, checked for shorts, and still the same. The output stage is pretty simple, what else is there to break? The LME chip?

Anyone know what else I should be looking for?

[cuibono]

Back in business.

Turns out I had killed the LME chip somehow...

Notes to self: don't work from memory of schematic... Use the light bulb current limiter... DONT keep putting in bigger fuses when they keep blowing! When will I learn?


I've gotten the resistors/bias sorted, at least for now. Here are the values that work for me:

Rb3 = 2.61k
Rb2 = 2k
Rb1 = 330
Rp = 200

These values are very finicky. When adjusting them, work in changes of 10 - 20%, and only do one at a time. I highly suggest getting a multiturn trimmer, rather than the single turn I've been using. Also, use a 300 to 500 ohm trimmer, 200 ohms is too little. With Rb3 = 2.61, the bias current is slightly overcompensated for - the effect of which should be checked in any final design, to make sure it is acceptable - changing any other resistor, including the trimmer to some degree, changes its behavior.

This strange thing happens, maybe because I've got a single turn pot - at a certain point, the bias jumps really fast. In the setup above, it jumps from 540mA/side to 1400mA/side. I won't push it farther yet. I think a multiturn trimmer would fix this... Just be warned.

The light bulb current limiter proved fairly helpful, once I started using it - with it in series with the amp, you can try different resistor combinations, and see what range you have with your pot (measuring gate to gate voltage) because the FETs cant get any current through them. I found the limiter to have a small but not important change on the gate voltage.

The important thing is knowing what gate to gate voltages give what bias current - this has to be done with the amp plugged into the wall, so be careful. For my FETs (I suspect any irfp240 or 9240s), 7.36V gave 31mA/side, 7.54V gave 108mA/side, 7.97v gave 253mA/side, 8.13V gave 620mA/side. Above 8V, things quickly get out of hand (at least with a single turn trimmer). Anyway, I suspect the above resistor values and gate voltages should be very useful for anyone using IR FETs.

Another note - R1 + Rp must be greater than 250ohms or so - when you drop below that (with any other resistors for R2 and 3), things act funny.

I haven't heard the amp, but it seems to be running nicely with a 10 ohm load on a scope!

[cuibono]

Okay, not super pretty, but working...

[cuibono]

just the board. Notice the thermal track transistor mounted between the FETs. Also, the black piece of tape next to the FETs is where I tape my temperaure probe. Haveing a thermocouple has been very handy.

[cuibono]

this is a mock up of the heatsink with case (upside down) - I don't particularly like large expensive metal cases, and keep trying different things for enclosures. For these, I went to the local 'cheap stuff' store and got some 'foot stools' for about ten bucks. Everything is already wired up inside. If you look at pictures above, you can see I lined them with aluminum roof flashing, to provide some shielding. Noise seems low even with the amp open and case ungrounded (3mVpp, with no mains hum)