BNOS 30V 30A PSU questions

[vintage port]

Hi all

I have a BNOS PSU (30V, 30A) with 8 paralleled 2N3055H dating from 1987 in it. The schematic is broadly similar to a doubled up version of that linked below, but with rather different voltages of course.

https://www.vintage-radio.net/forum/showthread.php?t=218476

It is very robustly made however the heatsinks for the two batches of four 2N3055H are quite small, so hefty derating at low voltages must have been originally specified. I fixed it up, it worked nicely, then soak tested it and one power transistor went to silicon heaven, taking the driver with it, I hope not the LM376N as well (but waiting on the TIP120 before I test that). The rest of the power transistors are OK. On the curve tracer they look like epitaxial types (and had snubber caps between base and emitter) but are branded BNOS so no idea of the manufacturer. I guess my options are one of the same or similar, or replace the lot with either 8 x new hometaxials, thusly RCA which are available albeit at a cost. They would be more rugged than the existing I imagine.

A more rugged (and expensive) alternative might be MJ15003 - thoughts or experiences anyone? Is this at all sensible?

The 2N3055H have emitter resistors of 0.1 ohm. To my inexperienced eye the emitter resistors seem rather low value to adequately compensate the currents but I would welcome opinions from those with experience?

Many thanks in advance,
James

[kalee20]

If there's 8 transistors supplying 30A then that's 3.75A each, which is reasonable for the 2N3055.

Then the emitter resistor of 0.1 ohms will drop 0.375V.

That's quite a bit of ballasting! (I presume all the bases are connected together?) So in comparison with a typical base-emitter drop of 0.8V, you have an additional 0.375V to take care of imbalances between the transistors due to device mismatches (very small) and thermal mismatches (noting that with 2mV/°C temperature coefficient of Vbe, this equates to a huge temperature mismatch of 190°C.

No, you've got nothing to worry about here. The BNOS designer has got the right value of emitter resistor.

[ortek_service]

Yes, with 2N3055's although they were rated to 15A max it was usually recommended to run them at no more than 5A each.
Some variable 2A PSU's like the Thurlby / TTI PL320 actually ran 4 but that was probably mainly to distribute the power dissipation at low output voltages when the fixed main DC rail produced a large voltage drop (so power-dissipation) on these/

IIRC the 2N3055H was meant to be more-resistant to secondary-breakdown - but wasn't really rated any higher than the standard cheaper part.

And it would be normal practice for the equipment itself to have emitter-resistors on each of the paralleled bipolar power transistors , to provide better current-sharing balancing.

[vintage port]

Thanks guys, appreciated. I'm a tube guy at heart, and no pro at electronics, so this is advice well received!
Shall I replace the one with something likely dissimilar, or fork out for all eight and beef it up a bit?

[vintage port]

...yes, all the bases are tied together. Two separate trafos, bridges, and banks of power transistors, one board controlling both.

[vintage port]

Hi all,
For future reference in case of interest, attached are some snaps of curve traces of 2N3055. The first is a hometaxial RCA, the second the BNOS which looks like an epitaxial, both scaled at 500mA/div vertical, 0.5V/div horizontal. The third is the BNOS but at 1A/div, 5V/div showing that the curves of this type do not straighten up much with increasing voltage. The hometaxial curves are essentially straight lines with increasing voltage.

I have been kindly offered a couple of RCA 2N3055H so presumably the same shape as the hometaxial traced here. Might the different curves here lead to any problem if I were to mix the two types or does the variation in gain outweigh any considerations of that nature? I am simply not experienced enough to hazard a guess and don't want to abuse a fellow enthusiast's kindness...Thoughts anyone?

[kalee20]

Originally Posted by vintage port

Shall I replace the one with something likely dissimilar, or fork out for all eight and beef it up a bit?

I'd be inclined to replace all 8, with the same type, manufacturer, etc.

The current-sharing emitter resistors ought to ensure nearly-equal sharing - but you need to set things up as best as possible for them. Having different manufacturers, different technologies of the transistors isn't exactly helping!

The curves are interesting - I take it you have Ic against Vce. But what is constant (and changes between curves on a single screenshot)? Is it Ib that's constant, or Vbe?

[Craig Sawyers]

If you really want to beef it up, the 2N3773 is a possibility.

Those curves were taken on the good old Tektronix 575 - a transistor (and diode) curve tracer that uses valves. I of course have two - one with the high voltage modification (mod 122C). I also have the rare 175 high current add-on (weighs more than the 575!)

The 175 is wired inside with welding cable. Collector current to 200A and base current to 12A . Connects to the back of a 575. The 575 supplies some power and acts as the display device for the 175.

First time I tested out the 175 I wired in a TO220 device. It blew so spectacularly the burning device shot across the room and burnt into the carpet. "What's this brown mark on the carpet?" "Well that is wierd - no idea".

Craig

[ortek_service]

Unfortunately, I've only ever used 'standard' 2N3055's rather than the usually more expensive 2N3055H version, that was meant to be more resistant to secondary breakdown.

Note: From a little reading-up, it seems that with the 'H' suffix device, it did not actually signify it was a Hometaxial device.
And the RCA 2N3055 actually started-off as a Hometaxial device, but was later moved to an Epitaxial device, to cut costs (as well as give higher, more consistent, higher frequency response?), which may have meant it was more-prone to secondary-breakdown.
So this may have lead to a more secondary-breakdown resistant 'H' version (But not necessarily a Hometaxial device?)

This previous forum thread may help: https://www.vintage-radio.net/forum/showthread.php?t=138407
As well as: https://en.wikipedia.org/wiki/2N3055 - Where it seems 'modern' 2N3055's can now be rated at 15A continuous?

Looking on Mouser, they only list one 2N3055H - Now obsolete, with 'Restricted Availability' https://www.mouser.co.uk/ProductDetail/onsemi/2N3055H?qs=vLkC5FC1VN%252BTgSHO4UI%2Fiw%3D%3D
With the datasheet for it actually being the standard 2N3055! And the same if you go directly to On-Semi website: https://www.onsemi.com/products/discrete-power-modules/gener...-and-low-vcesat-transistors/2n3055h
(With many of their other obsolete etc versions linking to current datasheet!)


And an actual 2N3055H datasheet from Farnell-Multicomp being rather rubbish, with critical-errors in secondary-breakdown current spec. (showing a frequency!) https://www.farnell.com/datasheets/89481.pdf

So maybe any 2N3055H's now bought are NOS, or possibly remarked standard 2N3055's. And exact spec. could vary between manufacturers?
I've not seen any reference to internal emitter-resistors, so maybe this was only in some 2N3055H's / was just something that some appeared to have, from their curves?
And really the BNOS circuit should have these already - as unlikely BNOS would have gain-matched all of the 2N2055H's transistors they used (which could still vary over temperature and exact operating current) / these probably still varied quite a bit within the same batch.

It seems with On-Semi that their 2N3055H was replaced by the 2N3055HG (Green 'Lead-Free') version in 2006, before they stopped adding the G suffix as they were now all RoHS compliant.
But On-semi do have a 2N3055A version, with a secondary-breakdown spec:
https://www.onsemi.com/pdf/datasheet/2n3055a-d.pdf
That is still being sold (at a price!): https://www.digikey.co.uk/en/products/detail/onsemi/2N3055AG/1475092

It could be that the long-lived (but rather budget performance) 2N3055 is generally on the way out, as not being used by manufacturers in current designs, and TO3 packages are getting more-costly to produce / buy.

From some Googling it seems that the 2N3055A has a higher Safe Operating Area, so is an improved 2N3055.
But it may also have a higher-frequency response, which can be a problem if used for its original purpose in a PSU:
https://www.diyaudio.com/community/threads/2n3055-inside-com...mentary-power-output.157430/page-29

I doubt you're going to find any genuine 2N3055H devices being sold at a sensible-price.
And the 2N3055 is now relatively-expensive device for what it is (I used to buy these new for < £1 back in the 80's).
So if wishing to change them all (at least on one board), to the same make / batch etc. to maximise chance of them behaving the same, then Farnell are about as cheap as any major distributor. Especially for their own 'Multicomp' version, that now does have a better datasheet with Secondary-breakdown spec: https://www.farnell.com/datasheets/4396245.pdf
And they also sell the 'improved' 2N3055A on-semi one
https://uk.farnell.com/w/c/semiconductors-discretes/transist...rs-bjt?st=2n3055&range=inc-in-stock

[ortek_service]

Farnell's 'CPC' company also sells the 2N3055 at similar prices
- Although they don't sell the 2N3055A, only the standard On-Semi 2N3055 + the cheaper Multicomp version.

However, Cricklewood Electronics are even-cheaper, for their basic-spec (No details of manufacturer) 2N3055's.
And they also do actually list some 2N3055H's (Described / pictured as " Genuine Motorola Heavy duty" at a currently-reasonable price of £5+VAT. See: https://cricklewoodelectronics.com/search.php?mode=search&page=1

[vintage port]

Originally Posted by kalee20

I'd be inclined to replace all 8, with the same type, manufacturer, etc.

The current-sharing emitter resistors ought to ensure nearly-equal sharing - but you need to set things up as best as possible for them. Having different manufacturers, different technologies of the transistors isn't exactly helping!

The curves are interesting - I take it you have Ic against Vce. But what is constant (and changes between curves on a single screenshot)? Is it Ib that's constant, or Vbe?

Thanks for this advice. In answer to your question, in the first two photos the curves represent step changes to Ib. These were set at 20mA/step from memory, and the lowest step looks weird because of the angle the photo was taken at. I think the last photo was 10mA per step because that device has a gain of 83

Referring to the gain of the 7 devices, it varies from 28 to 83 so fair to say, these transistors weren't selected stock!

[vintage port]

Thanks Craig and Ortek-service, appreciated. Yup, a good old 575, thank goodness I have no 175 or I'd most likely set fire to myself, a mod 122C was always on the wish list so I could mod it for valve curve tracing but these things aren't exactly common in the UK and, as for owning two, my wife has pertinent opinions about the amount of 'junk' I have accumulated already ..

@Ortek-service check the RCA Power Devices Databook, 1978, attached re 'H' suffix 2N3055.

Hmm, I might bite the bullet and order eight of the RCA, well if the TIP120/LM376N fix works anyway...gulp!

[Aub]

I see there is no mention of Ft in that chart. That might have been of interest to restorers of Quad 303 amplifiers.

Cheers

Aub

[kalee20]

No. That chart is only the absolute maximum ratings - it doesn't include things like current gain, etc. - the sort of things a designer works with - they'll be in a different table, probably entitled 'Characteristics.'

The 2N3055 (from memory) only ever had a minimum Ft, which wasn't particularly high, and production devices met it but not by a large margin. When the production process changed, the new devices vastly exceeded the Ft spec, so they were still 'to spec' but designs that relied on Ft being not very high, went unstable.

Really, it's indicative of bad design, to rely on a loosely-specified characteristic and hope that it isn't too extreme. But that's moving off-topic!

[ortek_service]

Originally Posted by vintage port

>>
>>
@Ortek-service check the RCA Power Devices Databook, 1978, attached re 'H' suffix 2N3055.
>>

That RCA datasheet seems to be a bit confusing / contradictory, as it says all of the transistors on it are Hometaxial, inc. the 2N3055 (Hometaxial) * Where * = 'Formerly 2N3055H' (so therefore later-on, no H-suffix to have Hometaxial version?)
But it then also says an Epitaxial version is also available (not specifying any suffix to order that) and to order the Hometaxial version, order 2N3055 (Hometaxial)

So it would seem that the dropped using the 2N3055H part number, and you had to spell it out in full that you wanted the Hometaxial version - Nut not clear how they then marked the device to distinguish this!
And it seems other manufacturers went their own way, with using 2N3055H for a more-robust version that wasn't necessarily Hometaxial.

[kalee20]

Originally Posted by vintage port

Referring to the gain of the 7 devices, it varies from 28 to 83 so fair to say, these transistors weren't selected stock!

No - that's a fair variation - but it won't actually matter very much!

The bases are all connected together (thanks for confirming) and the bottom ends of the 0.1 ohm emitter resistors are all connected together. So the voltage across all the 0.1 ohm resistors will all be very nearly the same (assuming that the Vbe's match well, which in practice is a very reliable characteristic of a silicon PN junction), and so the emitter currents will all be very nearly the same.

Then the device with a hfe of 28 will have a collector current 28/29 of the emitter current (96.6%) and the device with a gain of 83 will have Ic of 83/84 of the emitter current (98.8%).

The difference between the two is 2.2%... less than the likely 5% tolerance of the emitter resistors!

It might be significant in a Class B amplifier, where one transistor operates over half the cycle, and the other over the other half-cycle, because the load on the driver would change throughout the cycle leading to distortion. But here, nope, it's a non-issue.

[Trigon.]

Langrex are selling the RCA 2N3055H at £3.50 ea. +£5 p&p

https://www.langrex.co.uk/products/2n3055h-rca-to-3-npn-audio-power-transistor-x1pc-vr286/

- and at the same price through their eBay outlet, but with a reduced postage cost of £1.80...

https://www.ebay.co.uk/itm/267424612182

Cheers

[Craig Sawyers]

Originally Posted by vintage port

Thanks Craig and Ortek-service, appreciated. Yup, a good old 575, thank goodness I have no 175 or I'd most likely set fire to myself, a mod 122C was always on the wish list so I could mod it for valve curve tracing but these things aren't exactly common in the UK and, as for owning two, my wife has pertinent opinions about the amount of 'junk' I have accumulated already ..

Yeah - we're both there with the junk thing.

I acquired the 575 ages ago, and the 575mod122C rather later. I've also got a fair collection of device jigs for them.

The 175 I bought from a guy in the West Country, knowing we were about to have a holiday down there. So I just collected it. All I had to do was make an umbilical between it and the 575, and buy the 100A high current "Supercon" connectors and hefty enough cable for the collector current and high current banana plugs for the base current.

It is a truly mad and wonderful device. I really need to use it in anger with a bank of parallel transistors, and not simply as a red hot shrapnel generator.

Craig

[Diabolical Artificer]

I thought using the older, slower 2N3055's was only important in certain amp designs. As your PSU was made in 1987 they'd have probably used the later 2N3055's anyway. I have several vintage PSU's they often have a mixture of all different 2N3055's & operate without any issue.

Lastly AFAIK in PSU's 2N3055's are configured as emitter followers which have less than unity gain therefore what your devices hfe etc is won't really matter as much.

Andy.

[kalee20]

Originally Posted by Diabolical Artificer

Lastly AFAIK in PSU's 2N3055's are configured as emitter followers which have less than unity gain therefore what your devices hfe etc is won't really matter as much.

That's almost certainly true - however, in amplifiers the output devices are almost always configured as emitter followers also - and there it can matter, because it determines the load on the driver stage, when chasing sub-0.5% distortion.

In the power supply, the driver will (hopefully!) have been sized for worst-case hfe, so anything better will give it an easier time. And as my earlier post, in the power supply there's no crossover to worry about.