Preventing a BJT from exploding using only base current

Hi all!

TL;DR: when a BJT sits naked (no DC resistances on either the emitter or collector, but well controlled base current) between Vcc and GND, is it safe to use ONLY the base current to prevent it from releasing magic smoke? Details below.

I'm building a transmitter for a wired interface (Ethernet). I chose to drive the transmission line itself with a transformer which is in turn driven by a common-collector BJT (a voltage follower/buffer), but with no DC current limiting resistors.

The inductance of the transformer has been selected to provide a good impedance at the low end of the bandwidth, but that left me with such an impedance at the high end of the range that I can not really afford to put a series limiting resistor to control the current (and thus power) through the BJT when the transmitter is idling.

Stated differently, when the transmitter is not sending any frames, the output of the voltage follower is pure DC. It passes only through a transformer primary whose impedance at DC is near-zero.

I can not put a resistor in series with the transformer because the amplitude of the output oscillations would be too low. I can also not put a resistor in series with the collector for the same reason.

The option of using a DC resistor with a capacitor AC bypass is, again, DOA because during the operation the capacitor will get charged up to maybe Vrms of the output signal and will clip the output signal when the transistors input signal falls below that value.

I want to avoid the common-emitter conformation if anyhow possible because of problems with the signal distortion. I'm also not yet willing to experiment with switching the Vcc wire on and off to provide power-on- demand, so to speak. Same goes with the baseline base voltage.

So I'm basically stuck with a transistor which, when idling, sits naked between Vcc and GND. To enable proper voltage following, its base needs to be biased to a pretty high voltage - I'm keeping it at 1-1.2V - but this opens up a possibility of a short circuit.

The only way I figured out to prevent it from exploding is to use a well- selected series resistor with the base to moderate the baseline current and basically abuse the BJTs core operating principle (moderation of resistance via current) for fun and profit.

How reliable do you judge this to be, given random operating conditions the users are sure to put the interface in? Does inter-transistor difference (especially hFE differences between individual transistors) become important? Will I need to tightly match the base series resistance for each individual transistor?

Reply to
Aleksandar Kuktin
Loading thread data ...

On a sunny day (Sat, 25 Apr 2015 17:01:39 +0000 (UTC)) it happened Aleksandar Kuktin wrote in :

I have done it and it still works. Beta is reliable within reason. Of course above some temperature the thing will melt anyways.

Reply to
Jan Panteltje

No load would be pointless. In fact that's not what you have.

Well, what are a tr's pertinent limits? Ic Ib Pdiss Vce Theta SOA Does controlling Ib control all the above? No... unless you either a) monitor those things & use that to control Ib. b) don't monitor, just provide a heavily overrated tr

How about R,C in parallel? Or less good inductance at low end? Or higher transformer resistance?

You can't add a supply current limit?

monitor the copper drop across the transformer?

diode drop or R.... as you suggest. But you either need to monitor the tr's limits or be very conservative.

Hfe varies according to Ic, Vce, and varies widely between samples. Vbe varies according to theta, which generally varies Ib too. I don't know what the hfe range is, but perhaps 5:1 or 10:1 for the 1st 2,

5:1 for the 3rd, giving a /very/ wide spread of hfe.

So its hopeless unless you select each resistor for each transistor, checki ng it works for all conditions, and giving the tr lots of room. Given enoug h Pdiss & Ic ratings I daresay it could work. If you think it couldn't, wha t failure mode are you proposing? Other than gross financial & energy ineff iciency. I'd be looking for another solution.

NT

Reply to
tabbypurr

I don't entirely understand what you are doing. Can you post a schematic?

Since a transformer doesn't pass DC, why not capacitively couple into the transformer, and avoid frying the transistor?

--
John Larkin         Highland Technology, Inc 
picosecond timing   laser drivers and controllers 
 Click to see the full signature
Reply to
John Larkin

That was my thought.

--
Tim Wescott 
Wescott Design Services 
 Click to see the full signature
Reply to
Tim Wescott

Drive the transformer differentially.

Reply to
bloggs.fredbloggs.fred

I was going to use a BFR93A, but I've since changed my mind and now I'm going with 2N3904. The BFR93A is too fragile to be used without continuous oversight (version a above), and 2N3904 should be able to withstand even high temperatures and high betas without overheating (version b above)

Haven't tried it. I'm worried it will produce ringing with some signals (link test pulses, in particular).

Lower inductance doesn't help with idling. I can't control transformer resistance.

I might, but it would increase complexity.

Too complicated. Easier and cheaper to just overrate the transistor.

Hmm... I've read about 3x differences, but 5x or 10x seems a little much..

Well, the only failure mode I'm really worried about is thermal runaway. There is no negative feedback loop anywhere in Ib control (unless I specifically add it) so if for any reason power to be dissipated - during idling - overcomes the ability of the transistor and circuit to dissipate it, the transistor will enter a vicious circle (higher temperature increases Ic leakage, which increases temperature). Normally, there would be a resistor somewhere and it would limit the current and prevent the transistor from running off, but because of transformer problems, during idling, the transistor sits alone between Vcc and GND and has no current limiting scheme other than the thermally vulnerable Ib-Ic relationship.

For the record, by using a 2N3904 and low Ib, I was eventually able to construct a circuit which is - on paper - not supposed to smoke even on silicon temperatures exceeding 100 degC (which works out to 50-60 degC ambient).

Now my only problem is that the transmitter output is not high enough.

Reply to
Aleksandar Kuktin

Thinking about it, and trying it a bit. Unfortunately, a true differential circuit greatly increases both the complexity and part count and - so far - hasn't given me a noticeable benefit.

Reply to
Aleksandar Kuktin

Thanks!

Reply to
Aleksandar Kuktin

Umm.. I don't really have a ready place on the 'net where I could post it. Is a SPICE netlist fine?

Vcc 1 0 3.3 Cin 100 110 680e-12 Rbias_up 110 1 15000 Rbias_dw 110 0 18000 Qdriver 1 110 200 Q2N3904 Xbalun 200 0 1024 1025 BALUN l=20e-6 c=0.998

Balun consists of two 20uH coils with a coupling of 99.8%. Turn ratio 1:1.

Not entirely sure what you're suggesting... You mean have a capacitor in series with the transformer? I can't do that because the capacitor charges up during transmission. Then, when Vb falls below the charge voltage on the capacitor, the transistor stops conducting and the output waveform is has one of its amplitudes cut. Even if I break the standard and let an asymmetric waveform onto the line, the cut is below (actually above, because the downstroke is cut) the critical 0.6V level. The receiver will reject such a signal.

Reply to
Aleksandar Kuktin

if R is right...

Silicon doesnt suffer from thermal leakage current the way germanium does, the issue is Vbe lowering. So current drive is fairly stable, voltage drive is suicidal. Drive it right & give it plenty of thermal headroom and it sh ould be good. Design it to be happy even if heated to 150C.

NT

Reply to
tabbypurr

, the issue is Vbe lowering. So current drive is fairly stable, voltage dri ve is suicidal. Drive it right & give it plenty of thermal headroom and it should be good. Design it to be happy even if heated to 150C.

Or design it to take advantage of the hfe plummet as Ic increases

NT

Reply to
tabbypurr

LTspice .asc files are easiest for most of us. You just cut and paste into a text editor, save the file, and launch LTspice. Also has the advantage that we can play with the circuit easily.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
 Click to see the full signature
Reply to
Phil Hobbs

Yeah, that does give me some problems...

Reply to
Aleksandar Kuktin

I'll conjure one up later today. I'm using ngspice so I have to translate it LTspice.

Reply to
Aleksandar Kuktin

I'll conjure one up later today. I'm using ngspice so I have to translate it LTspice.

Reply to
Aleksandar Kuktin

Doesn't work. The BJT can handle the load, that part of the solution is fine, but no signal passes through to the line.

I'm gonna switch over and use a common-emitter line driver, like all the normal people. No idea how hard it will be to fix distortion issues though...

Reply to
Aleksandar Kuktin

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.