R.G. Keen was right. Again
He said don't have a charged heatsink, he says use an insulator.
I didn't. A guitar cord that was plugged into my mixer made contact with the charged heatsink and blew the chipamp to smithereens.
This was a lot of work down the toilet. Argh.
It was the amp running my subwoofer. I don't think anything else is hurt. It didn't even blow my fuse.
The smell is atrocious. The chip phisically split in half and smoke was rolling out of it.
F**k!!
RDV
You didn't use an insulated version? Or is it that they don't have an insulated version of that chip? Sorry for your loss btw. Hopefully its just the chip and you can solder in a new one on be on your merry way.
Quote from: joecool85 on April 15, 2006, 08:50:37 AM
Hopefully its just the chip and you can solder in a new one on be on your merry way.
This comment expresses the main thing why i really do not like chip amps in general. I mean, there's nothing wrong with them; they are far more advanced than the basic discrete designs and packed with very nice features - plus their thermal coupling is awesome: No need for thermally coupled bias circuits. And fixing them is easy, just put in a new one.
...But what about 15 to 20 years from now (maybe even less) when National has decided to discontinue some of their chip amp products and LMxxxx chips have become obsolete? This has happened to far too many chips already and the petty DIY market is not really a reason for big manufacturers to reconstruct a product line. When you blow a discontinued chip are you going to hunt through junk yards or pawn shops - or just build another power amplifier board with a new chip model? In later case you can never reproduce the tone of the original chip. This is really, well, inconvenient situation for people who should fix equipment for their living.
True, transistors become obsolete too but at least one can always replace them with one's that are close enough to specs.
Anyway, sorry to hear what happened RDV. Blowing up stuff unintentedly is never fun. I once did that two times in a row for a complete powerstage destroying every transistor in it - both times because of a probe slip. You can imagine i was ready to kill myself just out of sheer annoyment. Hopefully you didn't blow anything expensive, like transformers.
I guess one cool thing about chipamps is that they commit suicide when you screw up instead of holding fast and taking everything in line with them. The wretched thing actually physically split in half and belched forth much smoke.
RDV
Must have been cool to watch, even though sad and frustrating at the same time.
I don't see it being a huge deal with the obscelsance thing, just because everything becomes obsolete. When you build computers, 10 years from now, good luck finding parts to fix it to original quality. With an amp you could just build a new power section, the thing is with chipamps, 99% of the tone is from the preamp, not the poweramp.
The LM3886 amp I built last night(thanks again Joe) is working well with my sub. It may sound better than the LM4780 did probably due to the fact that this 96VA transformer is enough for a single LM3886 to function okay. I guess I'm gonna spring for another toroid.
RDV
Wow...96va, that is so small. I used a 250va for a single lm3886 lol. But I used it so I could add another channel later. I wouldn't go smaller than 125va per chip, and 160va is the closest you get to that.
Believe or not, over 2VA per watt is probably wasted, it doesn't hurt to over do it, but it doesn't help either, the chip will only pull what it needs.
R.G. said:
You have been baptized with smoke. It's time to take the vows of a power electronics initiate.
It is best to assume that you will have to insulate power devices from a heatsink 100% of the time. I personally was amazed when there were some few devices that did not need to be insulated.
So it's time to learn about power device insulators and mounting hardware.
The name of the game is to insulate electrically but conduct thermally. Those are conflicting requirements as you know.
The old standard is die-cut mica and heatsink compound (zinc-oxide filled silicon grease). The modern standard is kapton and compound or metal-loaded rubber (Cho-therm is a brand name). An even newer one is power semiconductors which are fully coated in hard epoxy that may simply be greased and mounted to the heat sink with ordinary hardware. There is an LM3886 like that, don't know about the 4780.
Good heat flow depends on intimate contact of flat surfaces to eliminate air pockets. A seemingly smooth metal surface that's not mirror-bright has mountains and valleys that prevent contact between the power device and the sink. Heatsink compound fills in the valleys with stuff that's not as thermally insulating as air, so it's better. However, getting heatsink compound thicker than just filling in the valleys to eliminate air adds more thermal insulation again, although not as bad as air.
The heatsink goop is just thermal conducting fillling. The insulator is in most cases a thin sheet of mica or kapton, which provides an electrical insulating layer and is not too bad at thermal.
Cho-therm rubber insulators are a step sideways. They are electrically insulating, thermally conducting, and squishy to fill in the valleys. They are convenient, neat (heatsink goop is very messy) and almost as good as goop and mica/kapton.
Then there's that hardware stuff. You are correct - the hardware cannot touch the semiconductor (unless it's one of the epoxy insulated semis). So the heatsink mounting kit usually includes a shoulder washer to fit in the semiconductor mounting hole(s) to space away a bolt from the semiconductor package. Nylon is OK, but generally does not have enough mechanical strength to tighten up semis to enough torque to hold them firmly to the sink.
Torque. You want to torque the semis down with enough force to hold them tightly to the sink but not so much that the pressure distorts the metal and causes the surfaces not to fit tightly together because they're slightly curved. Don't laugh - it CAN happen. I won't tell you how I know.
So read the datasheet about your power device and find out what the package number is - TO-220, TO-247, TO-3, etc. - and then go to Mouser or Digi-key and get the proper heatsink insulators (and goop if needed) as well as the right shoulder washers to fit the semi-package holes. Or buy pre-insulated devices.
The pre-insulated devices sound really good about now, right? As you'd guess, the pre-insulated ones have less satisfactory heat flow than insulator-and-goop or rubber insulators. But if you're not working all the way out at the edge of the possible heat dissipation and heatsink temperatures, they're great. As cheap as semiconductors are, it may be better to buy more semiconductors but insulated, so that each one dissipates less power on its own, and parallel them up so as to spread the heat out.
It's just like juggling, itsn't it?
When you're ready, we'll talk about heatsinks.
Wow, thats a lot of good advice. I think it would be nice to summarize some of that and throw it in the wiki.
There is a bright spot in all this. I found that I had another LM4780 and I cleaned up the umpteen zillion little holes so I can stick it in the old PCB. I've got another set coming so if this one will work I can build a 100+ watt per channel stereo amp with the right transformer and heatsinks. I'm going to try to solder it up and smoke test it tomorrow. Should be exciting! :trouble
RDV
w00t! Tell us how it goes.
Didn't work. It was highly unstable. I threw it in the trash. :grr :grr :grr :grr :grr :grr :grr :grr :grr :grr :grr :grr :grr :grr :grr :grr :grr :grr
RDV
:-( You didn't want to save any of the components? I'm such a cheapass.
That sucks you couldn't get it working.
I threw the whole board away because the components might be alright but might not be. It was passing DC and was about to rip the speaker out of the cabinet. I've got another new board and chip on order so it's not too big of a deal. The bad part was hours of cleaning solder out of the 27 little holes in the PCB.
RDV
Well, at least you have it worked out.
Quote from: joecool85 on April 15, 2006, 01:40:58 PM
Wow, thats a lot of good advice. I think it would be nice to summarize some of that and throw it in the wiki.
It is done. It's added to the LM4780 Paralleled article.
I dragged the PCB out of the trash and had a good look at it and I believe I may have shorted a couple of pins(there's 27 of them) partially so I carefully pryed up all the long side pins a bit and cleaned all the flux away and checked for continuity from the supply rails to the proper pins. I think I just jumped the gun when I tested the board by not cleaning up the board and stuff. I'm gonna test it again soon after I install some connectors so I don't have to do all that unsoldering and resoldering. I may also replace the bypass caps on the negative rails as they may have been damaged when the chip smoked.
It seems I'll never learn to be more patient. The thing may still not work, but there's a lot better chance that it will now.
RDV
Well, good luck.
One of the 27 pins(#16) on the LM4780 was not soldered well which caused my rebuild not to work. I troubleshot for about 4 hours or better before I got it. 27 pins is a bunch of pins to be soldering. It's working wonderfully as a guitar amp at the moment replacing my blown LM3886. I've got another parallel LM4780 PCB to build and make a stereo LM4780 amp with a 20v + 20v 400VA tranny. I picked the lower voltage to make the amps run cooler (at the sacrifice of a bit of wattage). They'll still be a good 80 clean watts per channel, and I won't have to use as big of a heatsink methinks. It'll still be pretty good size.
RDV