DIYHiFi.org

Recent diy projects have emerged that reintroduce this topology back into the diy community so it would probably be useful to produce a table of active device complements here as reference. If we make it complete enough and add the data sheets, it would make a good stickie. I'm submitting a few below for a start but I'm sure the members here could flesh list this out completely. The idea here is to identify the complementary pair with a description alongside.

IRFP250 and IRF9250 (double die packages very useful for decreasing the number of devices)
IRFP240 and IRF9240 (fairly linear but less so working below +/-25V)
IRFP150 and IRF9150 (fairly linear even below +/-20V with exeptional transconductance to Ciss ratio)
IRFP140 and IRF9140 (very linear even at lower voltages, excellent transconductance to Ciss ratio)
IRFP610 and IRF9610 (fairly linear with exceptional Ciss but with comparitively weak transconductance)

Please feel free to input your experiences with other push pull complementary pairs or offer suggestions on mixing and matching the above.

What are the desirable properties of a FET in a complementary push-pull output? It can't be transconductance or capacitance, because FETs are worse in both regards, compared to power BJTs. Is it ease of driving? Durability?

No experience with complementary FET outputs, so I'm genuinely asking.

jwb
I don't know enough about BJT or FET devices to be an authority but here is what I know. The capacitance of a FET is just high enough to limit bandwidth out of the audio application range and they can avalanche much more current than any BJT I know of so that makes the common FET a very workable part. If you know of a BJT pair that suits a circuit we can build, by all means include them here and state their merits and limitations. I started this thread to identify and describe working candidates for push pull designs that recently became public. Moderators, should I have posted this in the technical forum since these devices are not passive and the aim is to put together reference material for those looking to make their own push pull variation.

Two great things about FET outputs:

1) They are stable with a capacitive load. A BJT is not and requires special measures not to blow up when it oscillates at 10 MHz.

2) The drive requirements are *much* lower than a BJT. Most people use a BJT double emitter follower. This is a huge mistake as it loads down the front end and causes distortion. (Oh, no problem! Just add some feedback -- that'll cure the problem!) In contrast, most front ends can drive a FET output stage directly.

The reason *not* to use a FET output stage is that most people use the US made vertical MOSFETs. These have several problems:

1) They have to be temperature compensated, just like a bipolar output stage.

2) The most common brand is IR. They have a defective process and all of their P-channel are fucked up. The transconductance changes in the first millisecond of operation, which puts the problem right in the middle of the audio band. This is easily seen on a curve tracer. The problem seems to disappear at high temperatures, so I guess you could run them hot and hope that you've side-stepped the problem.

3) The input capacitance is not only relatively high, but also extremely non-linear. So it applies a non-linear load to the front end, causing distortion at high frequencies. (Sure, just apply some feedback -- that'll fix the problem!)

If you want to use FETs and money isn't a problem, the best thing to do is buy *lateral* MOSFETs. Hitachi (Renesas) makes some, but their P-channel parts don't leave the "triode region" until they get about 20 volts across them. The best ones are the Semelab/Magnatec clones made in the UK. The problem with lateral parts is their transconductance is low. This means their output impedance is high. (Just apply some feedback -- that'll fix the problem!) The best solution is to parallel a boat-load of them. For a given transconductance, the lateral parts will have less than half the capacitance, *and* it will be about 10x more linear compared to vertical parts.

The lateral parts also have a zero tempco point that fortuitously happens to be right where you want to operate them for a class AB amp. And if you want to go class A, the tempco is negative, so there is no chance of runaway.

Have fun!

Charles Hansen
Thanks for the definitive reply on the subject of the BJT part. I knew the bandwidth would probably make them unwieldy but I could not have identified the real world consequences so plainly. It is both an honor and pleasure to read your concise posts. Is there a reference online for a circuit to test and identify the defective P channel parts before putting them into service or better yet, a list of the defective process masks? I found the following:
http://www.seattlerobotics.org/encoder/ ... tracer.htm
but it seems to require a storage oscilloscope.
I have read that some of the new Fairchild parts have reintroduced these older defective masks into production since they were purchased from IR. This page:
http://www.fairchildsemi.com/products/d ... b6mos.html
shows some of the dies used by Fairchild. Do you know if the IRF9150 equivalent described on the page is one of the sucky ones?

Feedback fixes everything! Just ask.......uh, better not get started..........

Jocko

_________________
"Because, I hate you guys. I hate you guys so very, very much.

Yours,
General Cartman Lee"

Sure:

NJW3281G (NPN) and NJW1302G (PNP), $3.10 for the pair.

Advantages: nice, flat gain in the vicinity of 1A collector current. Capacitance much lower than older MJ15003 you may have used before. Generous safe operating area. More convenient than old metal can package.

Disadvantages: still a BJT. Subject to thermal runaway outside of SOA.

The latest thinking apparently is that feedback is a great thing, as long as the stage(s) you are applying feedback around is(are) already almost perfectly linear. I read that just today.

Two things:

1) When I first put the parts on a curve tracer, I couldn't believe what I was seeing. So I tested them a different way. I made a simple circuit where the DUT was set up as a cascode (or a folded cascode) and looked at the waveforms when driven with a square wave. What you see at various spots in the circuit will show you the problem very quickly. Much more quickly than building a curve tracer! All you need is a cheap square wave generator and a cheap 'scope.

2) Don't bother. Vertical MOSFETs for audio suck. Even when they work right. They are designed for one thing, and one thing only -- switching power supplies. They would be last on my list of active devices. Except maybe an op-amp....

Exactly. Just like a boob job looks great on an 18 year old girl that doesn't really need a boob job to look great.

Most people put feedback around circuits that wouldn't work well without it. Like using a BJT output stage with only two sets of transistors. Folks, they need *three* sets to avoid loading down the front end. Or using vertical MOSFETs with their super non-linear capacitances.

It's like giving a boob job to a 75 year old woman. Ain't gonna fix the problem.

Charles Hansen
If you set them up in a cascode, folded or not, how do you identify which is the faulty DUT? Your advice to avoid vertical MOSFETS is hard to take. The cost is a key concern for me and there are many excellent commercial examples of their use. I suspect you intend to push us toward new circuits and new levels of achievment with your advice but I am not ready to break new ground at this time. Many great audio designers chose the vertical part, warts and all. Do they get around the non-linear capacitances by forcing the amperes?

Thanks for the nice visual. I'll never think about amplifiers the same again.

Thank you Charles, for the "boob job". I didn't know what it is, but LEO is my friend. And now, I've extended my knowledge a bit...
Tino

For those of you wondering about existing circuits with the lateral FET, Rod Elliot used them in his P101 and Dave Hafler used them in his DH line of amps. The wide bandwith probably makes these parts extremely sensitive to layout but since they are so linear and rugged, you can probably cap the pins to keep them under control. Not my idea of a simple build but maybe the rewards are there for those more courageous.

I've only heard one amp that uses verticals I like. Only one. This amp uses a slighly different output arrangement, not a common source.

Please don't ask.

If not for this one amp I would be agreeing with Charles whole heartily. I'm a dinosaur, I still like bipolar outputs.