Suitable replacements for Quad405 output transistors

[bikerhifinut]

Hi,
Given that original transistors for the 405 output stage are unobtanium, and it seems quad themselves changed the types over the production, what are the recommendations for replacing a blown pair of output transistors. The drivers seem to have survived whatever abuse it was that blew the outputs and the boards will be thoroughly checked before replacing the transistors.
The Chinese clones us ON semiconductors MJ15024, and I wondered if there other types that are a drop in replacement.

Regards.

Andy.

[greg_simons]

Hi
Herald of this forum seems to have a good supply of genuine mj15024's, I've bought a few off him and all are ok, worth contacting him if you want the best.
Greg.

[Craig Sawyers]

MJ15024G, ONSEMI are a current part from Farnell, RS and Mouser. All in stock.

Craig

[Diabolical Artificer]

"Given that original transistors for the 405 output stage are unobtanium" These are 2N3773's right? Langrex have them for about £3 each Andy.

Tother Andy.

[Craig Sawyers]

2N3773G is a current ONSEMI part, from all the usual suppliers. Supply, like the MJ15024G is not a problem. The G just refers to environmental friendly stuff, although it might be worth checking the fT to make sure it is compatible with the older devices.

Craig

Craig

[Craig Sawyers]

OK - had a quick shufty of datasheets for the alternatives Quad listed.

BDY74. 140V, 10A, fT 0.8MHz min
BDY77/2N3773. 150V, 16A, fT 0.8MHz min (old 2N3773 hometaxial spec)
2SD424. 180V, 15A, fT 5MHz typical
2SD676. 160V, 12A, fT 4MHz

So there is a wide range of fT values successfully used.

Modern parts:

2N3773G, 140V, 16A, fT not well publicised but allegedly 4MHz
MJ15024G, 200V, 16A, fT 4MHz min.

All in all the MJ15024G looks comparable to the 2SD parts in most important regards. The 2N3773, G or not, is more of a servo amp and has a really soggy hFE versus Ic. The MJ15024G is much better in this and hence will not put so much stress on the drivers as current goes up.

Good luck resurrecting the amp. Check the crowbar too (Quad's most inglorious moment)

Craig

[woodchips]

Warning, this is way outside my area of competence!

Having just re-started poking around with valves I have been doing a lot of searching on the web, and all sorts of hi-fi web sites come up. Some are interesting to read, but can never find them again.

One was talking about replacing transistors in old amps, in this case the 2N3055. They were saying that whilst 2N3055's are easy to obtain, they are not anything like the ones made in the 60's used in that amplifier. As far as I can remember it was the fT that had gone from hundreds of kHz to tens of MHz. I also remember a comment about how they were made, in the fabrication process used, not a bar of silicon any more but buried bits here and there. Sorry can't remember exactly what was said, just that they were very different.

The point being that the sound of the amplifier will be different. The suggestion was to find transistors from that era, not a particular type.

Over to the experts!

[paulsherwin]

Yes, 2N3055 replacement is problematic for the reasons you state (and because they're often faked), but that's not really the issue here.

[bikerhifinut]

Thanks for the heads up guys.
The originals were 2SD424, and as Craig points out they are a close equivalent to MJ15024, which seems to be the device seen on the various clones that abound.
Please be assured that they will be repaired to a very high standard, by a forum member friend. and one of the areas definitely covered will be the on board clamp circuit, these may well be removed/bypassed and an aftermarket output protection fitted, probably Vellemans kit which seems to tick the boxes and is really a bargain in kit form. I have seen that board fitted in a 405 so its feasible.
So from Craigs observations it looks like the MJ15024 is the best bet and my friend can easily get them from his regular supplier for a reasonable price.
He'll no doubt read this thread being a forum member and again thanks for the practical help in getting this amp breathing again.

As an aside I will be converting it and its stablemate back to stereo as I really don't need 180W (Who does in a normal domestic setup?) and I am also thinking it was an excess of power up the ESL989 that started the chain reaction that damaged the output stage as its clamp circuit was at the time faulty due to a dry joint that I discovered so I think a loud transient engaged the clamp and then it cascaded back. The boards on this amp were later ones with a built in clamp, and there's another on the output board where the outputs are paralleled and i wonder if it was a breakdown waiting to happen. It's stablemate has different boards without the clamp circuit and that amp survived the ordeal.

Again my sincere thanks to one and all.

Andy

[Craig Sawyers]

The other things to replace as a matter of course are:

Board mounted electrolytics. One in particular is above a power resistor and cooks. Mount new ones on the reverse side of the board.

And of course replace R7 and R8 (3k3) that feed the zener supplies to IC1. Quad used underrated power handling resistors, soldered in hard contact with the board. They burn out, put the output transistor to one rail or the other, and the crowbar operates and blows up the output transistors and also possibly drivers too.

Early amps used a 12V zener and the resistors dissipate 0.44W. Later ones use a 15V zener and the resistors dissipate 0.37W. I'm not certain what the dissipation of the carbon film resistors that Quad used (almost certainly 1/4W or 1/3W), but they certainly cannot cope with those dissipations in practice. Use a modern 1W, or even 2W ROX spaced a couple of mm above the board; resistor dissipation specs are all in free air.

Craig

[Mooly]

You might find post #33 here and the thread in general to be useful with regard to the 405 and use of modern and seemingly suitable replacement semiconductors:

https://www.diyaudio.com/forums/soli...st5331502.html

[Craig Sawyers]

That is interesting. According to the ONSEMI datasheets for the MJ15003G (and MJ15024G) there is no banding for hFE. Maybe Dada select them for low hFE.

The original 2SD parts *are* gain banded, but the Quad service manual does not mention this in their list of alternative devices.

Looking at the schematic evolution, every time they changed power transistors type, they messed with the details of the circuit around them. R//L circuits in the collector of the dumpers appeared then disappeared. Catch diodes appeared, and then moved position. A resistor appeared across the bridge inductor, then changed value.

So Quad clearly knew there was a problem with detailed spec of the dumper transistors, and kept changing the circuit details throughout the life of the product, presumably to ensure stability.

Craig

[Radio Wrangler]

Originally Posted by woodchips

One was talking about replacing transistors in old amps, in this case the 2N3055. They were saying that whilst 2N3055's are easy to obtain, they are not anything like the ones made in the 60's used in that amplifier. As far as I can remember it was the fT that had gone from hundreds of kHz to tens of MHz. I also remember a comment about how they were made, in the fabrication process used, not a bar of silicon any more but buried bits here and there. Sorry can't remember exactly what was said, just that they were very different.

The point being that the sound of the amplifier will be different. The suggestion was to find transistors from that era, not a particular type.

Several things happened to the 2N3055 over the years to get to where we are now.

The 2N3055 was the first decently rugged silicon power transistor to make it in the big time. It was the answer to a lot of designer's prayers. There had been several development types from RCA which came out with RCA proprietary type numbers each one a little better and slightly eclipsing its peers, then one of them looked like a point worth settling on and a JEDEC type number was applied for. It got 2N3055. RCA kept on playing around fine-tuning their process and a few related parts appeared with RCA type numbers, some of these were simply selected from the 2N3055 flow. Quad opted for one of the RCA numbered parts for their new amplifier, the 303.

These early transistors were made by starting with a P-doped wafer, uniformly P type throughout. A pattern of photo-resist was exposed and developed on one side to expose windows to allow more dopant in (N-type this time) when the wafer went through a furnace of hot gas. The other side of the wafer was fully exposed to the gas.

The N-type dopant gas diffuses into the crystalline silicon, replacing a proportion of silicon atoms in the crystal matrix. This counteracts the P dopant already in the wafer and strong diffusion over-writes it, leaving it N-type. This happens from both sides of the wafer at once.

The trick is to get the speed of dopant diffusion (temperature, gas density) and the amount of time just right, so that the two sides of the N diffusion don't quite meet in the middle. The P-type bit in the middle is the base. The smaller N diffusion is the emitter. It's smaller so that some base is left exposed on the top so it can be connected to. The big N diffusion on the other side is the collector and this side gets brazed to the TO-3 housing flange once the wafer has been broken apart into individual transistors. Add bond wires to connect emitter and base to their respective pins, stick on a cap and bob's your aunty.

The problem comes in controlling how much base thickness gets left in the middle. You want it thin for good gain, and you also need it uniform to prevent hot-spotting problems. You also want a good abrupt change. Without this control, you get poor frequency response and wide variations, especially in gain.

These transistors were better than anything which had gone before, but there was plenty of room for improvement. Their Ft was somewhat uncertain, but it was low. RCA spec'd their Ft as no less than 800kHz for the 2N3055. There was no upper limit.

Along came a new semiconductor process... It would take more operations (bad for cost) but those operations would all be from one side (good for cost)
The raw wafers were made with N-type dopant already in them. A photoresist process made windows accessible to P type dopant gas and the wafer went for a hot and toxic ride through the furnace. This diffused a P type zone into the surface of the N type wafer. Think of a bathtub shape. Strip off the photo resist and then apply a new one with smaller windows. This goes for a ride in a furnace with N typs gas. This makes a smaller bathtub situated entirely within the first one as the N dopant overwrites the P-dopant. So you have a bathyub within a bathtub. The smaller bathtub is the emitter, the larger bathtub is the base and all the silicon around them is the collector. Transistoring happens with electrons flowing downwards, the active part of the base being the thin bit left underneath the bottom of the smaller emitter bathtub.

Around this time, there were improvements in the purity of materials, time and temperature control.

The devices were much more accurately made and the gain increased, but more remarkably the frequency capability increased.

A massive error was made. 2N3055 was THE power transistor. The transit van of the power transistor world! It dominated TO-3 power transistor shipments. It killed off the Germanium ones... OC28 and NKT403 and their relatives. Some were made for replacement support, but the world wanted silicon and things switched over very rapidly. Along came the new process and there was a leap in the capabilities of silicon power transistors. But 2N3055 was the number that sold. 2N3055 has open ended specs, the Ft or gain could go infinite and they'd still meet spec.

They should have got it a new type number. But sticking 2N3055 avoided having to spend megabucks on marketing it.

All of a sudden, people ran into problems. Power supplies with a little control board and harnessed-up wires going miles to a power transistor on a heatsink started oscillating. With the new gain and Ft, there was enough coupling from wire to wire. The same thing happened with a number of 2N3055 based power amplifiers. The Quad 303 with its 'Quad Triplets' - little 3-transistor feedback loops was badly hit. Amps with simple emitter follower output stages were OK if wiring was short. The 2-transistor quasi-complementary feedback arrangement was OK. But the Quad 303 was in trouble. Manufacturing of the old type 2N3055 was run down, stocks would dry up. Quad needed to change to something more tolerant of modern transistors... They came up with the 405.

Meanwhile there was more 'progress' on the 2N3055 front. By later standards the 3N3055 was something the crudest semi conductor fabs could make. Too easy, a cheap commodity. Not sexy enough for their profit targets.

But the newer sexy devices for switch more power supplies and RF transmitters didn't have perfect yields. However, the reject wafers still met and exceeded the 2N3055 specs, so to cut waste costs, they could just dice them up, stuff them in TO-3 cans and flog them as 2N3055.

The 2N3055 had become a dumping ground. Get one of these and you may see a very surprising Ft and have a lot of trouble taming them. Still meets spec, though!

If that wasn't enough, enter he fakers. They will put ANY die in a TO-3 can and write 2N3055 on it. A lot of users don't use them to the full 2N3055 capabilities. They might never notice a die a quarter the area of a real one.... yeah.

So I avoid anything marked 2N2055 like the plague. Their era has ended and the fakers have killed any trust.

Also, in designing a circuit, I assume Ft is likely to get increased a lot and I design to accommodate it.

Hfe changes with Ic. Don't just compare spot values in choosing a replacement device, look at the graphs! they tell you a lot more about how Hfe drops off at higher currents. 2N3772 does not look good. I'd rathervgo for the Motorola/ON MJ15024 for a Quad 405

David

[eddie_ce]

David, thanks for a very interesting post.

[Leon Crampin]

My priority in the Automotive world was SOA (safe operating area). As outlined by David, the 2N3055 morphed into a kind of electronic dustbin which could contain almost anything resembling npn silicon. Of the devices I examined, the biscuit (wafer?) has to go to Fairchild who, at the time produced the smallest 2N3055 chip I have ever seen. RCA Liege was my preferred supplier of their excellent "hometaxial" devices.

So, the need for in-house numbers became obvious, and the 2N3055 saga should serve as an object lesson to those loose-specifiers of components generally (alas, many of them British designers).

A Fairchild 2N3055 was reduced to vapour in one of my lab tests with a load dump from a 28V 250A 'bus alternator running at full chat. For the benefit of my colleagues, I repeated the experiment using the same device type with the lid sawn off. Just a sooty smudge remained of the chip...

Leon.

[Radio Wrangler]

I've had to design stuff to survive field dumps from 28v aircraft alternators, as well as load dumps. Quite an interesting challenge! I know exactly what Leon's been through.
Oh, and then there are lightning induced surges. Lightning strikes the aircraft and then jumps off somewhere else to complete its path. Ludicrous currents in the airframe and rather wild transients induced in all the wiring.

Anything taking the direct hit cannot survive, but the people in the plane sort of want the engines, radios and nave equipment to keep working.

Electronics can get quite lively.

David

[bikerhifinut]

Just to steer this back round to my original question.

The boards are getting a thorough examination, they'd been serviced in the past with new electrolytics (Nichicon) of good quality and looked like some resistors had been swapped too.
It's looking like they'll get MJ15024 in, that does seem to be the device of choice for repairers.
Luckily the driver transistors survived the incident but all the silicon will be tested I am assured.
Andy

[Radio Wrangler]

There are those two resistors in the feed to the opamp and zeners, that are over-run unless beefier parts are fitted.

David

[Craig Sawyers]

Originally Posted by Radio Wrangler

There are those two resistors in the feed to the opamp and zeners, that are over-run unless beefier parts are fitted.

David

Already mentioned in post 10 with power calcs. Yes they are a guaranteed failure, and have been responsible for many a blown power transistor when the totally awful crowbar fires.

Craig

[Al (astral highway)]

Originally Posted by Leon Crampin

A Fairchild 2N3055 was reduced to vapour in one of my lab tests with a load dump from a 28V 250A 'bus alternator running at full chat...

Hi Leon , I’m not quite getting the configuration here or its purpose... is this not a predictable result if I’m reading correctly that the transistor handled this entire load??

Please can you draw a sketch or explain the configuration and why you did this. I have no experience in production, design or testing, but am extremely interested in these end-of-life scenarios for components.