High power amp project

[megajocke]

First, a warning: This will project will probably not be finished very soon. It might not even become an audio amplifier... But probably.

It will most probably be a class H (like in stepping the rails). It is supposed to be used to drive PA subs. Preferably it should be able to drive 2 ohms with music or music-like signals. I'll accept 4 ohms if I can't get big enough heatsinks or the transformer proves to be too small.

I've just got the transformer - a 30 lbs UI core rated at about 2000VA, don't know exactly as this is estimated from the DC current it says it is supposed to deliver. But right now it just has one 80V secondary and some lower powered ones, not very useful.

The transformer is the large one in this picture

Now I have to rewind it. I'll have to choose 2(.5) tier with rails that can drop to 0V or 3 tier. I already have 10 63V 10000uF caps and am thinking of using these to start with and getting 2 more giving 30000uF per tier. Rail voltage will be +-115V or so when idle at nominal line voltage with a tap halfway.

With this rail voltage I guess about 600-700 something watts at 8 ohms and 1kW/channel or a bit more with 4 ohms load does this sound reasonable? Maybe the transformer size could support a bit more power than this for music but I think I'll take that out by lowering impedance rather than increasing voltage as I'm not going to afford loudspeaker drivers with insane power handlings.

The 2 tier configuration has the advantage that I can get good transformer utilization even when running only from the low rails. One rectifier per polarity, center tap going to the low tier. One channel will probably be used inverted to make the PSU utilization even better. With rails that can drop to 0V reactive load output transistor stress is very low.

If I go for 3 tiers I will need to buy more caps but then 50V types would be OK. I don't know if the little better efficiency and lower peak transistor dissipation is worth it. The power supply will be more complex. More cables everywhere...

Posted by wg_ski in http://www.diyaudio.com/forums/showt...52#post1600652

I just use hexfets for class H. IRF1407, 2907, and more recently the Fairchild 75344. They switch plenty fast enough for any woofer amplifier, and I've never tried to hard switch more than 50V at a time. With quad 63V rails, It was a class-G retrofitted S-Leach type amp so I alreday had upper transistor banks in place. Now with anything that big it's 3-tier. I've got some of the Fairchild 20N60's around, but I was saving those for experimenting with some bigger SMPS (assuming I can get some decent size cores).

I'm probably going to use these for rail switches: Fairchild FDP75N08A-ND. They don't have as good on resistance as the ones you suggested but are a little cheaper and if needed they could easily be parallelled.

Output devices: Probably fairchild FJL4215/FJL4315 if I decide to use plastic devices or OnSemi MJ21195/6 if metal cans. The plastics are interesting as the excellent high voltage SOA of the OnSemis aren't needed with rail switching, they are pretty fast, they are cheap and I already have a tube of each polarity. More of them will be needed though than metal cans, possibly making it impractical to fit them all to the heatsink. They will have a lot of area though and spread the heat very well.

OPS protection will most probably be analog multipliers connected to RC-circuits simulating the junction temperature of the output transistors. The rail switching should make second breakdown a non-problem. The protection will probably control a limiter circuit decreasing input amplitude. If the limiter isn't enough to stop the rising temperature (say, hard short on output or a strange load) another comparator with higher temperature set point should trip the channel off for a short delay before trying again.

More information to follow...

[wg_ski]

Originally posted by megajocke
First, a warning: This will project will probably not be finished very soon. It might not even become an audio amplifier... But probably.

Yes. Be prepared for something like this to take several months. Something can be thrown together quickly, but to make it reproducible and/or road ready takes time. I'm still designing the protection circuits for mine and figuring out where the PCB and trafo for that are going to go in the chassis. I'm rapidly running out of room, and the result needs to be buildable/serviceable.

Your ideas seem quite workable, but that much at 4 ohms is a little optimistic. The rails will drop and you'll be more like 700-900W at 4R. The +/-115 will make something about as big as an RMX2450 is *supposed* to be. 1200+ honest watts at 2R should be possible, and with decent reservior caps you'll even get that at 20 Hz. The RMX2450 only has a 900-1kVA trafo at the same unloaded rail voltage. 4R and 8R power are limited by the voltage drop under load, but 2R power is limited by the circuit topology. It does about 950 honest watts at 2R at low frequency, but could do more with decent reservoir caps and one miserable extra current gain stage.

[megajocke]

Yes, I figured out after writing the post that higher voltage will be needed for that power output.

If I stick with my original idea I will have the same rail voltage as the RMX2450 but double the VA rating of the transformer and three times the effective capacitance per tier. Especially the low impedance driving capability will be better.

Going for three tiers isn't that much more work though. In that case I will probably use 50V caps and buy them all new. The question is if the extra 1.5 or so dB are worth it.

[megajocke]

I found some heatsinks I've bought now:
http://cgi.ebay.de/ws/eBayISAPI.dll?...m=260281934433

Standing on end on a table 100W dissipation makes the sink about 95 degrees C without fan cooling. I haven't tried with a fan yet because the transistors that were mounted on the heatsink blew up at only 50W dissipation per piece at 100 degrees heatsink temperature and without insulators...

The "MJ15025" transistors have the lettering the wrong way and has a circle exactly like the fakes in this link:
http://sound.westhost.com/fake/counterfeit-p1.htm

The heatsinks are excellent quality though. The transistor mounting surface is milled and there are countersunk and tapped holes for transistor mounting. I've heard that milled surfaces have better flatness than just extruded, is this true?

So 2 per channel will be used, NPN on one and PNP on the other. The transistors will be 5 pairs of MJ2119x most likely mounted without insulators. Drivers will be Fairchild FJL4x15 or possibly FJA4x13. I'll just have to make sure the sinks are big enough before going further.

I'm having trouble deciding though: Common collector or common emitter output?

Common collector has the drawback of having the heatsinks connected to the switched power supply voltages and I'm afraid capacitive coupling from them will increase distortion or cause EMI problems. Maybe not a problem if the rise/fall times are kept reasonable though. The heatsinks could probably be mounted on the other side of a metal plate, having heatsinks in a tunnel that's fan cooled and circuit boards outside the metal walls of the tunnel.

Pluses are that it's easy for the rail switch circuit to sense voltage at transistor emitters or bases making it take into account the voltage drop over emitter resistors. Maybe easier to stabilize too.

The other alternative is common emitter which will make the heatsinks (both of them) sit at output potential. Capacitive coupling from this will not cause distortion, just changed frequency response and stability. If I do it like QSC in the PLX amps it is easy to drive too but I'm worried it might be hard to stabilize and maybe glitch on rail switching.

I saw you used heatsinks connected to switched voltage wg_ski, did you have any problems with this?

Here is a link to wg_ski's amp:
http://www.diyaudio.com/forums/showt...postid=1509573

[AndrewT]

Hi,
that high current 80Vac winding is ideal for the first stage voltage supply.
Split it exactly in half to get 40-0,40-0, giving about +-58Vdc and about 300W into 4r0.
Now add on a smaller diameter wire for the next voltage stage, again a pair of 40Vac windings. Bringing up the second stage to +-116Vdc.
You should get at least 1000W into 4r0 on low duty sinusoid signals and 600 to 680W into 8r0.
I would use your existing 63V 10mF caps but put at least +-40mF on the rails, requiring another 6off 63V caps.

Get that lot working and consider if you need another pair of 30Vac windings for a third stage of voltage supply (using 50V 40mFcaps), if there is room in the toroid for yet more copper.

[megajocke]

You are right, that would probably have worked fine if it was a toroid but this is a UI-core transformer and there isn't any room left now for any more wire. When I have unwound the lower voltage secondaries there will be a little but I doubt it is enough.

Right now the high current winding is split with 40V on each leg and then put in series. Primary winding is half on each and then put in series. I don't believe using the centerpoint as a centertap will be a very good idea, leakage inductance (and thus the radiated magnetic field) will be high if the halves are loaded assymetrically like when there is different current draw from positive and negative supply.

I'm most likely going to do it like this: Two ~80VAC center tapped (exact value so that at max line tolerance and unloaded output the voltage will be below cap max rating) per leg. Then they are parallelled, one from each leg are put in parallell. Now there are two 80V AC center tapped winding, one for positive and one for negative.

I'm also planning letting one channel run with opposite polarity than the other, effectively doubling the useful power supply capacitance on signals that are mostly in phase. (One channel draws from negative supply while the other from positive)

[wg_ski]

Originally posted by megajocke

I saw you used heatsinks connected to switched voltage wg_ski, did you have any problems with this?

No problems *yet*, but the whole amp is far from completed. One PCB stuffed and it runs. No full power test yet - just lightly loaded with the light bulb in the mains line. I've run into a few problems fitting everything in the chassis. I need to find a spot to install a 50VA toroid to run the fans and soft-start, then I can start the final wire-up of the power supply, I/O etc. I also need to re-spin the daughterboard PCB - Ive decided to put the temp sensors on it instead of the soft-start board, becasue I've linked the temp sensors directly to the muting circuit. When I get the whole thing together, with both channels, I'll get a clearer picture of what the switched bus heat sinks do to me. Probably not much - the front end is well away from the sinks, and this isn't the first big amp to do this.