I can never make sense of thermal numbers, but the Bergquist site has a pdf file on pad performance with TO-220 devices. They get about 3.0C/W with the material you specified. Looking at a National Semi datasheet for TO-220 regulators, plus some other web sites, gives a value for mica and grease of about 1.5C/W. No way in the world are you getting 0.3C/W, regardless of the device, so think about how many watts you're dissipating in each device, multiply by 3.0 and see how that compares with actual practice. My guess is you need a better thermal interface.
This is going nowhere fast, resistive load --- ESSENTIAL.
You also mention a 19 Hz resonance value for the speaker, is this the mechanical resonance frequency of the cone ? If these are huge drivers the inductance must be very significant, guessing at what 1mH or there abouts.
If you have a fire element measure along it's length to see what gives 4 ohms and shove a couple of wires across it and see what that does for output power. There are always ways and means
You also mention a 19 Hz resonance value for the speaker, is this the mechanical resonance frequency of the cone ? If these are huge drivers the inductance must be very significant, guessing at what 1mH or there abouts.
If you have a fire element measure along it's length to see what gives 4 ohms and shove a couple of wires across it and see what that does for output power. There are always ways and means
Mooly said:
Yes I mentioned this right in the first post, two peerless XLS 10" drivers in an open baffle, H frame.
http://www.tymphany.com/830452
for the data sheet. Yes the VC inductance is huge.
Yeah, the thermal number for the pad is correct on the Farnell site. It is, however, for one square inch of material, and the devices in question are, I think, less then half a square inch (If they're like the Hitachi plastic package). You are getting nowhere near 0.3C/W in the actual application. Rod Elliot has an excellent page on thermal issues that would be a good read.
If you have a fire element measure along it's length to see what gives 4 ohms and shove a couple of wires across it and see what that does for output power. There are always ways and means
If I use 4 ohms the thing gets way too hot, even air cooled. I cannot dump this into water. If I use 8 ohms instead of 4 and air cool it I can develop 42 volts across it, at this level the element is glowing a little. I guess the massive increase in resistance is going to kibosh any usefulness of this test?
With 42 volts across the heating element I'm getting around 4.8amps.
Which would show a resistance of 8.75 ohms.
This would = 201.6 watts
Anyway at this output level I can keep my fingers on the FETs for quite a while before having to take them off.
The input current during this time is 2.3amps.
Hmm I've left it running for a little bit longer at the 42 Volts, the amp could run all day like this, it's really not getting that hot at all.
It is possible that the parallel XLS are just a rotten load and are making life incredibly difficult for the 2 pair of output FETs?
Hi,
add a series resistor to your speaker, try 100r to start with.
Now wire your hi impedance speaker in parallel to your test load.
You should be able to hear when gross distortion is being added due to severe clipping.
If you're very sensitive, you might even hear more minor clipping.
But this is a slow method and you are risking blowing up your amp due to overheated output stage lowering your SOAR.
Buy a cheap second hand scope and a couple more DMMs.
add a series resistor to your speaker, try 100r to start with.
Now wire your hi impedance speaker in parallel to your test load.
You should be able to hear when gross distortion is being added due to severe clipping.
If you're very sensitive, you might even hear more minor clipping.
But this is a slow method and you are risking blowing up your amp due to overheated output stage lowering your SOAR.
Buy a cheap second hand scope and a couple more DMMs.
Does sound a bit like the speakers are a tough load - The reactive thing I guess.
I dont think the heating of the element will be a problem testing wise. Never measured one, but, a 3 bar fire draws about 13 amp. When cold I doubt it will be any more. Another test called for. Measure your fire with it cold on resistance. Plug it in and when red hot, unplug and read immediatly. See if any difference.
Why won't it go in a bucket
I dont think the heating of the element will be a problem testing wise. Never measured one, but, a 3 bar fire draws about 13 amp. When cold I doubt it will be any more. Another test called for. Measure your fire with it cold on resistance. Plug it in and when red hot, unplug and read immediatly. See if any difference.
Why won't it go in a bucket
Mooly said:Does sound a bit like the speakers are a tough load - The reactive thing I guess.
I dont think the heating of the element will be a problem testing wise. Never measured one, but, a 3 bar fire draws about 13 amp. When cold I doubt it will be any more. Another test called for. Measure your fire with it cold on resistance. Plug it in and when red hot, unplug and read immediatly. See if any difference.
Why won't it go in a bucket
It's not a bar heater its an electric fan heater the entire thing is inside a plastic case, so I can't really do that measurement, I had to open the thing up to use the element. It is now back together 'cause I need the heat.
However, when measuring the element, when finding the correct place to connect the amplifier, I got just over 8 ohms. Then when measuring the output current and voltage I arrived at a value of 8.75 ohms, it doesn't look as if the resistance is drifting too much with temperature, and at 8.75ohms the element was glowing slightly.
Ultimately though, if I can get 140 watts into an 8.75 ohm load on two output pairs, I'd be fairly happy with that. It's quite clear pushing 250 watts into 4 ohms on only two output pairs is pushing it. Not only this, but at just over 100hz the drivers in parallel will be reaching lower then 2.5 ohms. Granted there is a 4th order active LR at 100hz, isolating the amps somewhat from this, but still, perhaps the amps are just running out of steam.
Certainly when pushing them as hard as possible with music, the sinks get hot, I wouldn't want them much hotter, so even if the thermal performance could be slightly improved, the maximum the sinks could comfortably support wouldn't be much more anyway.
So far we've managed...
1) To eliminate the DC offset issue by altering the input filter.
2) Show the amplifier can produce 140 watts into just over an 8 ohm load before clipping.
3) Probably rule out oscillation, from the amp passing the AM radio test.
What still irks me.
1) The bias is very dependant on the value of P1 (R13), this shouldn't be the case.
2) Why was the output protection circuitry tripping prematurely.
All in all, from my point of view, things don't seem too bad and with the equipment I've got at my disposal, I am rather happy to leave this where it is.
I would just like to thank everyone who has offered advice, allowing me to sort out some of the problems I had with these, and perhaps the problems the amplifiers had weren't quite so severe, more the load I'm trying to drive is rather difficult.
Not to beat my dead horse, but it still sounds like the junction temperature of the outputs is too high, causing the problems you've seen. I'd do the full calc for junction temperature, using the data directly from the Berquist web site for the pad material you're using. Or, just replace the darn things with mica and grease. They're not as good as you think. Berquist even calls it a medium performance material in their product line up.
Conrad Hoffman said:
http://uk.farnell.com/520238/semico...g-antistatic/product.us0?sku=multicomp-mk3305
Something like this with grease?
And I agree, when driving the 8 ohm resistive load, just before clipping the temperature didn't seem to be any issue at all, but when driving the loudspeakers, the FET case temp periodically gets very hot.
And I guess you apply grease to both sides of the pad?
http://www.arcticsilver.com/ceramique.htm would be the best I guess
A quick browse at Farnell seems to show they don't stock thermal grease.- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.