Substitute for HP dual PNP transistor, part no. 1853-0075 ?

[SilverSolder]

I am working on an HP 3325B with a non-functional rear signal output, and traced the problem to a blown transistor - A22Q810 on the diagram,  a dual PNP item made by Motorola, with HP part number 1853-0075. 

Any bids on what might be a reasonable substitute for this hard-to-find component?

[mawyatt]

If you can't find an exact replacement, or a dual transistor, I would try a matched pair of general purpose PNPs like the 2N3906.

Hope you can get it back to working condition, have fond memories of the HP3325, great function generator

Best,

[SeanB]

2N 5796 is likely the transistor, but the price is a little steep for the each, as the only ones at Mouser are the JANTX versions.

FFB2907A is a SMD one, which will work as well, and is a lot cheaper, 37c US each, though you will have to do some fine soldering to attach 6 tiny wires to the package to get it on the board, or make a small PCB to do same.

[David Hess]

If you can't find an exact replacement, or a dual transistor, I would try a matched pair of general purpose PNPs like the 2N3906.

That is exactly what I would do; match the Vbe of a pair of 2N3906s to better than 500 microvolts and bind them together.  This will likely be better than a common dual transistor unless you can find a matched or monolithic part.

[coromonadalix]

I would do the same as david wrote

[SilverSolder]

Thanks for all the advice, I'll try the "DIY Dual" 3906, sounds a cheap and cheerful solution!

I'm trying to wrap my head around how this circuit works...  is my understanding correct?  -  Q810 acts as a comparator, it is biased "slightly to the left" so the output state is LOW when the input is at zero volt.  When the input sine wave exceeds some small voltage, the comparator "flips" to the High state.   Q820 is an emitter follower that basically just copies the output from Q810.

Assuming this is broadly right so far, I don't understand why the emitter follower is necessary in this circuit -  the resistors are nearly the same on both sides, so what are we actually gaining?  Not current, and not voltage...  There is no heavy load being driven...   Increased magic smoke, perhaps? 

The other thing I don't understand is the black magic around the use of U832 to drive VHIGH.  Is this some kind of turn-on delay to make sure the circuitry starts up in a known state?  Why are the inputs to those gates not connected...  does it depend on some internal pull-up or pull-down resistor?

[bob91343]

If the FFB2907A is suitable in SMD package, I would assume the old standby 2N2907A would work and be easy to install.  Also very cheap or at least it used to be.  I have a bunch of those.

https://www.ebay.com/itm/2N2907A-2N2907-TO-92-PNP-Transistor-20-Pack/233260885447?hash=item364f6ea5c7:g:dFkAAOSwvJ9dCEx9

[SilverSolder]

If the FFB2907A is suitable in SMD package, I would assume the old standby 2N2907A would work and be easy to install.  Also very cheap or at least it used to be.  I have a bunch of those.

https://www.ebay.com/itm/2N2907A-2N2907-TO-92-PNP-Transistor-20-Pack/233260885447?hash=item364f6ea5c7:g:dFkAAOSwvJ9dCEx9

Are these significantly different from the 2N3906?   -  the benefit of the latter is, I already have some! 

[bdunham7]

If the glued-TO92 doesn't work out, the real thing is $10 here:

https://www.ebay.com/itm/MOT-2N5796-CAN6-Leaded-Small-Signal-Transistor-Dual/362555638304?hash=item546a000220:g:OBIAAOSwA3dYDjpE

[floobydust]

HP original 1853-0075 was Motorola MD5000 if you need datasheets.
I would get a modern dual-PNP in SOT-23 package and an SMT adapter breakout board to fit.

[SilverSolder]

HP original 1853-0075 was Motorola MD5000 if you need datasheets.
I would get a modern dual-PNP in SOT-23 package and an SMT adapter breakout board to fit.

I was just looking at a dual 3906 in that packaging.

I had a quick look at MD5000, thanks for the tip.  This looks right, it even has the same power rating (400mW) as stated in the parts list.  The frequency response also looks right, with a 600MHz - 900MHz GBP which seems credible for something that's supposed to confidently deal with 60MHz.  The MD5000 looks like it performs significantly better than a 2N3906, though, perhaps there are better choices (of modern transistors to replace it)?

Looks like MD5000 is not unobtanium or even unreasonably expensive, so could always go with a real one...

[Stray Electron]

If the glued-TO92 doesn't work out, the real thing is $10 here:

https://www.ebay.com/itm/MOT-2N5796-CAN6-Leaded-Small-Signal-Transistor-Dual/362555638304?hash=item546a000220:g:OBIAAOSwA3dYDjpE

    The ad doesn't say where the seller is located but they're "World Electronics" so I'll give you three guesses as to where that part was made, and the last two don't count.

[SilverSolder]

I have located a genuine MD5000,  will update the thread when it is installed.

[SeanB]

U832 is used as a convenient way to get a logic 1 signal that is just above the logic 1 threshold, so the other gates will switch slightly faster, as there is less charge stored in them, plus will clamp the rising edge sooner.. As it will automatically bias with temperature it is likely used as a cheap arrangement over having 2 resistors and a diode to set it, and uses the otherwise unused gates in the packages. Gate used is close to the inputs, at least on the critical XOR so less noise there on what is a relatively slow rising input.

[SilverSolder]

U832 is used as a convenient way to get a logic 1 signal that is just above the logic 1 threshold, so the other gates will switch slightly faster, as there is less charge stored in them, plus will clamp the rising edge sooner.. As it will automatically bias with temperature it is likely used as a cheap arrangement over having 2 resistors and a diode to set it, and uses the otherwise unused gates in the packages. Gate used is close to the inputs, at least on the critical XOR so less noise there on what is a relatively slow rising input.

Are you saying, for example, the level of VHIGH on Pin 4 affects the speed of switching on Pin 5, and that is why Pin 4 is receiving the "special high"?

[David Hess]

I had a quick look at MD5000, thanks for the tip.  This looks right, it even has the same power rating (400mW) as stated in the parts list.  The frequency response also looks right, with a 600MHz - 900MHz GBP which seems credible for something that's supposed to confidently deal with 60MHz.  The MD5000 looks like it performs significantly better than a 2N3906, though, perhaps there are better choices (of modern transistors to replace it)?

I suspected it was a faster part but 2N3906s would probably still be fast enough.  Tektronix often used the 600 MHz minimum MPSH81 PNP RF transistor in similar applications.

[Kleinstein]

The circuit looks like a kind of ECL to TLL converter. So it is about a digital signal - not DC precision. So I would not worry so much about the matching or offset drift. Speed could be an issue, especially if the transistor runs into saturation.

The circuit looks like he current may go up to some 15 mA. So the 2n3906 should be large enough - the 2N2907 may be slow due to the large size. Fast PNPs are a bit tricky. The MMBTH81 seems to be the SMD version of the MPSH81 and is still available.

[SilverSolder]

I had a quick look at MD5000, thanks for the tip.  This looks right, it even has the same power rating (400mW) as stated in the parts list.  The frequency response also looks right, with a 600MHz - 900MHz GBP which seems credible for something that's supposed to confidently deal with 60MHz.  The MD5000 looks like it performs significantly better than a 2N3906, though, perhaps there are better choices (of modern transistors to replace it)?

I suspected it was a faster part but 2N3906s would probably still be fast enough.  Tektronix often used the 600 MHz minimum MPSH81 PNP RF transistor in similar applications.

2N3906 is specced to 250MHz from what I can glean, so it's not like it is out by an order of magnitude or anything.

Almost calls for a breadboard experiment (once the 3325B is back up and running)! 

[David Hess]

2N3906 is specced to 250MHz from what I can glean, so it's not like it is out by an order of magnitude or anything.

Almost calls for a breadboard experiment (once the 3325B is back up and running)! 

I have done a little work on that to find that RF transistors are *not* direct replacements for fast saturated switches.

[SilverSolder]

The circuit looks like a kind of ECL to TLL converter. So it is about a digital signal - not DC precision. So I would not worry so much about the matching or offset drift. Speed could be an issue, especially if the transistor runs into saturation.

The circuit looks like he current may go up to some 15 mA. So the 2n3906 should be large enough - the 2N2907 may be slow due to the large size. Fast PNPs are a bit tricky. The MMBTH81 seems to be the SMD version of the MPSH81 and is still available.

Looks like the MD5000 (i.e. the original part) is an annular transistor, I don't know if that makes it "special" in any way?  I noted that in addition to having a higher fT, the input and output capacitance are both much smaller on MD5000 than 2N3906  - that might also round the edges, perhaps too much?


The datasheet says: "Dual PNP silicon annular transistor designed for ultra-high frequency oscillator and amplifier applications requiring a matched pair of transistors with a high degree of parameter uniformity under varying environmental conditions."

[coromonadalix]

Could it be used like a gate or comparator ?  not sure its used like a differential amplifier

Could be special :  seems to have a very low capacitance output and really high operation frequencies ...

[Jay_Diddy_B]

Hi,

I have used LTspice to take a look at this. Here is the model:




The bases of Q1 and Q2 are biased 100mV apart. This means that V_be matching of the transistors is not required.

Switching speed of the transistor is important. I believe that the MD5000 was chosen because of its high f_T.

The circuit should work with two high f_T PNP BJTs.

The emitter current is about 20mA.

I have attached the LTspice model.

Regards,
Jay_Diddy_B
 sync circuit.asc (2.98 kB - downloaded 85 times.)

[SilverSolder]

Very nice @Jay_Diddy_B, the voltages correspond to what I am actually measuring EXCEPT that my broken MD5000 has about 2.5V on its collector permanently, leading me to think that Q2 (in your diagram) is broken in my unit.

The ultimate cheapskate would replace only the broken half of the pair, right? 

I have a new-old-stock MD5000 on its way in the post, I think that is the "least bad" way to fix this.  Even if one could suspect that all of these 40 or 50 year old transistors may be decaying by now, the rest of the HP 3325B is chock full of components of the same age -  why would you update just this one transistor, and risk having to deal with making it work exactly right, unless you absolutely have to?

I still don't understand why HP put Q3 (in your diagram) into the circuit.  Couldn't the downstream gate be adequately driven straight from the collector of Q1?

[SilverSolder]

Could it be used like a gate or comparator ?  [...]

Yes, that is exactly what it is doing.  It simply takes the sine wave output from the synthesizer board and turns it into a square wave, which is fed into the downstream gates, which in turn drives the auxiliary output. 

The 60MHz aux output on the 3325A/B is an "extra feature" that is not covered by the main 20MHz specs, so the signal is treated completely differently.  The main output (<20MHz) has its own completely separate "full blown" high bandwidth amplifier, attenuators, etc.

[Kleinstein]

The extra NPN emitter follower at the output can make the circuit faster (higher rise time) as it isolated the collector output from the load. It may work without with a single gate load, but the circuit part as a whole block may be kind of a standard building block also used for higher capacitance loads. The lenght of trace can also make a difference.

Like normal ECL gates the circuit may work without diving the transistors into saturation. Otherwise I would expect problems even with relatively fast PNP transistors.

If nothing else helps there would be the last resort and replace the 3 transistors with a ready made ECL to TTL translator chip like MC10ELT25 or similar. One would still need the biasing part for the 2nd input.