Transistor failure and protection.

[itsmoked]

The 2N2905A is blowing at the slightest provocation. I've been thru 8 on four out of 6 channels. Generally the transistor fails with the Base shorted to the Collector and the Emitter open (5). And, in two cases, all three pins open.

I've never seen the transistor even warm. I'm looking for any suggestions as to what might be giving that transistor such a glass jaw. I'm wondering if these are V[sub](BR)EBO[/sub] failures?
In the spirit of hotrodding do you have any suggestions for protecting that transistor.

Keith Cress
kcress -

http://www.flaminsystems.com

 

[Skogsgurra]

I am always careful when I see 1N4148 in power circuits. So, I would put a 1N4004 (or -07, 'cause that's what I have available) instead.

Trying to understand the diagram: The -15 pulse turns on the 2905 via a couple of helper transistors. 2905 supplies transformer primary (center) with +28 V that are switched to GND at cirka 20 kHz to produce gating pulses to thyristor.

If diodes in transformer circuit are shorted or open, that transformer will not behave and transistor will *perhaps* die. But 80 ohms *should* save it.

If, on the other hand, the diode connected from GND to 2905 collector is shorted, there is an immediate path from +28 V via 2905 to GND and 2905 will blow immediately. With no noticable heat produced.

Hoffe damit gedient zu haben

Gunnar Englund

http://www.gke.org

--------------------------------------
Half full - Half empty? I don't mind. It's what in it that counts.

 

[itsmoked]

In der Tat. Ja.

Your understanding agrees fully with mine.

You think a 1N400x is up for the speed, especially since no one even states it's recovery time? It's a rectimfryer as opposed to a 'signal diode'. I'd sure prefer the 1A instead of 300mA rating.

Can you conjure up any way to exceed -5V across V[sub]BE[/sub]? You'd have to have more than 5mA run thru that 1k res. Or somehow get the base to 23V.




Keith Cress
kcress -

http://www.flaminsystems.com

 

[Skogsgurra]

Danke!

No, I see no possibility for that on the 2905. But the NPN can have 15 V Vbe reverse and does probably zener when supposed to be off. I am not so sure if that hurts or not - current is limited.

20 kHz is almost DC and one or two us recovery time shouldn't be a big problem. 1N4004s are used in chopped thyristor gate drives with no problems. Siemens did that for decades.

Gunnar Englund

http://www.gke.org

--------------------------------------
Half full - Half empty? I don't mind. It's what in it that counts.

 

[VEBill]

Thinking outside the box:

I notice that the 125A fuse on the AC side is connected *directly* to the transformer secondary center tap. If you're switching such large loads, is it possible that some large switching transients might be capacitively coupled across the transformer windings (secondary -||- primary) and result in voltage spikes damaging the transistor's sensitive BE junction? The 28v PS would be cconsidered to be 'ground' for such transients. Also, if L1 is one phase of the local grid, then the spikes might be external (lightning, etc.).

Even the voltage rating of the transformer should be double-checked if it is to be connected directly to the grid through such a large value fuse. The schematic hints that it is a little small signal transformer, probably with small gauge enameled wire. Obviously I'm guessing...

 

[VEBill]

Typo: L3, not L1.

 

[2dye4]

What limits the voltage on the collector of 2n2905a ??

I see its emitter tied to 28 volts but i wonder how switching transients could put much more on that collector for a few microseconds..

 

[itsmoked]

No, I see no possibility for that on the 2905. But the NPN can have 15 V Vbe reverse and does probably zener when supposed to be off. I am not so sure if that hurts or not - current is limited.

Interesting point! I shall investigate further.

20 kHz is almost DC and one or two us recovery time shouldn't be a big problem. 1N4004s are used in chopped thyristor gate drives with no problems. Siemens did that for decades.

All right! Thanks.

I notice that the 125A fuse on the AC side is connected *directly* to the transformer secondary center tap. If you're switching such large loads, is it possible that some large switching transients might be capacitively coupled across the transformer windings (secondary -||- primary) and result in voltage spikes damaging the transistor's sensitive BE junction?

Possibly, but I've seen them blow with NO load hooked up.

The 28v PS would be cconsidered to be 'ground' for such transients. Also, if L1 is one phase of the local grid, then the spikes might be external (lightning, etc.).

Not lightning. We only get it about once every two years for perhaps 3 bolts total. This is served underground too.

Even the voltage rating of the transformer should be double-checked if it is to be connected directly to the grid through such a large value fuse. The schematic hints that it is a little small signal transformer, probably with small gauge enameled wire. Obviously I'm guessing...

Excellent point. These transformers are about the size of an inch tall stack of quarters. I pulled one on a bad channel and tried to detect an issue with it but my megger died. I used a DVM that goes to 40M Ohms and it saw nothing. But..

What limits the voltage on the collector of 2n2905a ??

I see its emitter tied to 28 volts but i wonder how switching transients could put much more on that collector for a few microseconds..What limits the voltage on the collector of 2n2905a ??

Yeah.. The 2n2905A is rated V[sub]CEO[/sub] -60V. I could find a transistor with a higher rating. Oddly the first one all the way to the left has a 160V rating yet there is no way for it to see much over 30V.

I could bump up the voltage to a 100V unit. Thoughts?

If the feeling here is that it's a current blast frying the transistor I could up the rating on it, one that could see more current like 4A instead of 0.6A max.

Should I consider adding a TVS type diode somewhere to clip spikes? There are a lot of spikes evident on the shiftkeyed output to the transformers. They're random, frequent, and about 50% more than the modulation amplitude.




Keith Cress
kcress -

http://www.flaminsystems.com

 

[Skogsgurra]

This looks like a rather antique unit. Did it ever work? Is it misbehavin' lately? Moved to new habitat, perhaps?

Gunnar Englund

http://www.gke.org

--------------------------------------
Half full - Half empty? I don't mind. It's what in it that counts.

 

[itsmoked]

It's a 2000 unit. Ran for seven years. Now this.

Keith Cress
kcress -

http://www.flaminsystems.com

 

[Skogsgurra]

It has proven itself then. A late millennium bug?

Gunnar Englund

http://www.gke.org

--------------------------------------
Half full - Half empty? I don't mind. It's what in it that counts.

 

[BrianG]

You may be exceeding the peak collector current for the device. A quick calculation shows that at 20KHz the 0.47uF has a sufficiently low reactance such that the series/parallel combination of it with the 20ohm and the 80ohm equals about 25 ohms. Ignoring the inductance of the transformer this could give rise to a pulse current of more than 1 amp, whereas the 2N2905A is only rated at 800mA absolute max.

Note also that some data sheets specify a maximum base current of 200mA for the device: your base current is not specifically limited but shares from the collector load, thus the transistor may not saturate fully. I could not find any SOA curves for this device, but this condition could give rise to secondary breakdown if SOA (safe operating area) is exceeded.

 

[Skogsgurra]

The main problem is that the design seems to be OK - it has been doing its thing for around seven years - and now it fails.

Gunnar Englund

http://www.gke.org

--------------------------------------
Half full - Half empty? I don't mind. It's what in it that counts.

 

[VEBill]

The seven (or eleven) year history may be an important clue, or it may be a perfect red-herring. It's reasonably likely that the ultimate root problem might contradict this apparently-successful history.

It should be kept in mind, but don't let it guide the troubleshooting too much.

 

[BrianG]

I should check my maths before rushing into print! The impedance of the C-R collector load at 20kHz is 19.6 ohms not 25 ohms, so ignoring the transformer, this gives a potential pulse current of 1.4A - an even worse overload.

 

[VEBill]

Generally the transistor fails with the Base shorted to the Collector and the Emitter open (5). And, in two cases, all three pins open.

If the 2N2905A is in a metal can, then it might be worth the few minutes it takes to open the can and peek inside with a microscope. If a bonding wire (terminal to die) is burnt open then that would tend to confirm overcurrent.

Short circuits traditionally indicate overvoltage events.

Maybe we have multiple problems...

And, as always, keep in mind that the problem might be something that doesn't show up on the schematic (shorted cables, dirt, etc.).

 

[MacGyverS2000]

So what exactly is the schematic symbol for dirt? And no, "ground" doesn't count... ;-)

Dan - Owner

http://www.Hi-TecDesigns.com

 

[itsmoked]

Thanks for the input folks.

I've seen many automotive problems that were completely the result of design problems. You think, "but they've made millions of these wrong!" So, I never assume it can't be the design or that a design can't be improved.

Brian; I can scope across the that cap-resistor:resistor network to see the voltage across it.

VE1BLL; I have a handful of the deceased and they certainly are metal cans. I can cut off a top. I'll do that and post a micrograph of it.

As for "dirt" while this board wasn't really very dirty any dirt on it was likely silicon carbide. After my initial work on it I threw the dice and washed the board. It's now squeaky clean. But, subsequently I lost two more channels on the test bench so I don't think that was the problem.

There is one thought I don't want to believe but I must consider. I continually probe the circuit with my scope at large raised test points that are located between the first transistor on the left and the 10k resistor. My scope has a dumb ground ring around it just above the tip. Since the test points are all about 3/10" tall and 3/10" apart I may have shorted a test point to ground in a few cases just by coming in at an angle to a neighbor with the probe. I don't ever recall doing it - but? I'm not sure that would blow the channel even.

Presently all the channels are working correctly. I don't have a single spare transistor left.



Keith Cress
kcress -

http://www.flaminsystems.com

 

[itsmoked]

BrianG;
I went back and measured the modulation frequency instead of estimating it from a scope glance.
It's 3.4kHz. That changes the collector load to 48 ohms.
28/48 = 580mA.

Across this network during the ON period 17.6V are dropped.
~17.6/48 = 367mA. And this only occurs during the fractional line cycles that a particular SCR needs to be on during.




VE1BLL;
Here's the innards.

Note the missing collector wire(missing with extreme prejudice).

Here's a close up of the die. (all I need is a wirebonder and I can go into transistor repair!)

And for everyone's interest here's what the board looks like.
1- Aforementioned test points. Note the scope probe tip bottom right of pic.

2- The sequencer chip.

3- That first transistor in the schematic (top left)

4- A channels drive stack. Pretty much everything in the schematic before the RC-R group.

5- A transformer, the two diodes, and the entire output network for a single channel.

6- The RC-R group.

It's certainly a nice board to work with.

Keith Cress
kcress -

http://www.flaminsystems.com

 

[VEBill]

While you have it open, take a few minutes to carefully check the solder joints (for microscopic cracks) on the TO-220 devices on the heatsinks (just behind the electrolytics). Those sort of AL heatsinks typically expand with heat (as would be expected) and in extreme cases can expand enough, perhaps often enough, to stress the solder joints causing nearly-invisible cracks.