older mosfet Trace Elliott unreliability

[Alex R]

I'm having a bit of an issue with a couple of older Trace bass heads, with power amps that use the BUZ series mosfets. These amps have burned out their mosfets twice in one case and three times in the other. Both amps work hard and the replacement mosfet jobs I have done have lasted a few months of regular use before smoking their mosfets again. I do check the other components too, and make sure the thing is idling cool before returning it to the customer.

What might be the reasons for this? Is it the mosfet matching thing? Here in the UK I'm not aware of a source for matched mosfets - I hope I don't have to spend hundreds on a batch and match them myself.

Any ideas where I might be going wrong?

[Steve Conner]

Hey Alex,

I believe Trace themselves admitted there was a design issue with these MOSFET PAs. You used to be able to get a replacement PA board with BJTs, called the "Bipolar Bear".

What kind of MOSFETs are they? I'm going to guess BUZ900/905 in metal TO3 cans. These are lateral MOSFETs for audio, and don't need matching, they share current reasonably well without it.

Make sure you're using the right ones though. If I remember right, the BUZ900P has a single die with 125W dissipation, the BUZ900DP has two dies and 250W dissipation. Obviously if it was designed to use the DP devices and you replaced them with the P ones, it wouldn't last long.

Even if you put in the same devices that you found in it, bear in mind that someone else may have fitted the wrong ones in the past. It would be tempting for a repair guy, as the single die parts are about half the price.

Exicon also make equivalents to these. I've built a hi-fi amp with the Exicon devices and it's been going strong for nearly 10 years.

If they're plastic TO-3P, the same issues apply, as single and dual die versions of these were available too. Class-D have recently brought out some TO-3P lateral MOSFETs, and I believe they're basically the same as the Exicon/Magnatec parts.

http://www.classd.ltd.uk/lateral-mosfets/

[Alex R]

Thanks for the informative reply Steve. I wondered whether there was an inherent problem. The mosfets in question are indeed BUZ 900p/905p, but in plastic TO-247 cases, spring-clipped rather than screwed to the heatsink. (Maybe I'll try drilling the heatsink and screwing them in.) Deffo the single-die types.

Those Class-D parts look well priced. I think I'll give them a try.

[Steve Conner]

Nope, the spring clips actually work better than screws in terms of even clamping force.

How many FETs does it use, and what's the rated output power? Maybe Trace were being optimistic when they designed it, and the double die parts would be a good retrofit. These FETs are not really indestructible or short-circuit proof, as the makers claimed.

I did some more research into this, and apparently Exicon, Magnatec and Class-D all use the same silicon dice from the Semefab foundry in Scotland. The only other people still making lateral FETs are Renesas (Hitachi) with their 2SK1058/2SJ162, which look roughly the same as the BUZ900/905 single die parts.

[Alex R]

Nope, the spring clips actually work better than screws in terms of even clamping force.

How many FETs does it use, and what's the rated output power? Maybe Trace were being optimistic when they designed it, and the double die parts would be a good retrofit. These FETs are not really indestructible or short-circuit proof, as the makers claimed.

I did some more research into this, and apparently Exicon, Magnatec and Class-D all use the same silicon dice from the Semefab foundry in Scotland. The only other people still making lateral FETs are Renesas (Hitachi) with their 2SK1058/2SJ162, which look roughly the same as the BUZ900/905 single die parts.

Thanks Steve, looks like we might have an answer in the Class-D double-die parts - so a set of those would be drop-in replacements? And would be tougher?

The amp I'm looking at has two BUZ900P and two BUZ905P, but some have three of each. Trace claim 300 watts into 4 ohms for the two-per-side version, or at least they call it a 300 watt amp and the minimum impedance is 4 ohm.

[Steve Conner]

Hi Alex,

Yeah the double-die parts should drop in. The only possible problem is that they have twice the input capacitance of the single die ones (can you guess why? ) and this might make the amp less stable and more vulnerable to parasitics.

300W into 4 ohms from two pairs of the single-die parts is optimistic, all right. When I did my hi-fi amp, I worked out the SOA calculations for the Exicon FETs, and 200W into 4 ohms was comfortable from two pairs of the double-die devices, at a heatsink temperature of 75'C, allowing a bit of headroom for reactive loads and so on. The rails on this amp were +/-50V and it ended up doing 180W into 4 ohms.

So Trace were trying to get 300W out of half the amount of silicon I'd have used. In light of this I think it's well worthwhile trying the double die parts.

The reason why some amps have three pairs is probably that Trace realised they overcooked the original design, and added an extra pair.

I never quite believed the old Maplin Mosfet amp's claimed 160W into 4 ohms either...

[Alex R]

Hey Steve - well that all adds up. The amp on my shelf, which has eaten two sets of (4) mosfets already, belongs to a rehearsal studio, so it takes a beating. I'll put the double-die parts in there and we'll road-test it. I'll make sure their speaker leads aren't falling apart .

[Alex R]

...except that Class-D are out of stock of both N and P DDs. Bugger. Do you know of any other sources?

[Alex R]

... to answer my own question - these seem to be the same:

Pro Audio - Alfet Lateral MOSFETs

£62 for 10. Ah well, in for a penny...