Two Questions About Transistors

[jdutky]

I'm going through some online tutorials about transistors and one thing makes me wonder: we are shown diagrams of your typical BJT where the collector as a sandwich of NPN or PNP doped silicon, where the outer layers (the collector and the emitter) are of equal thickness. But we clearly treat the collector and emitter differently in actual circuits, so is the instructional diagram wrong? What is it that makes the collector different from the emitter?

A less technical, but no less important question is: given the specs on a part that I need to replace (assuming that I can't obtain the original part) how do I go about finding a transistor (or diode) that has the characteristics I'm looking for? I've tried using the component selectors on Mouser, Digi-Key, Newark, etc., but with very limited success. Is there a better way to find a transistor that meets a set of specifications (e.g. power dissipation, breakdown voltages, hFE, and fT)?

-- Jeff Dutky

[bdunham7]

The basic descriptions of transistors you are seeing are simplistic compared to most real ones, although it is possible to build symmetric transistors where the E and C are equivalent and interchangeable -- but their performance is not as good for most purposes.  The physical difference between E and C in most transistors is the doping level and physical construction.  If you look at a transistor in normal operation, the EB junction is forward biased and has about a 0.6V drop across it.  The EB junction on most transistors can only withstand a small reverse voltage, often only 5 volts or so.  The BC junction, on the other hand, is reverse-biased in normal operation and has to withstand the full rated C-E voltage.  The two junctions are made differently to best perform their designated function.

As far as specifying replacements, that's an art.  Just ask for help and value experience over theory.  You start by understanding the circuit that you are repairing to see what specs would be most important, then looking at what is available.  Be aware that often exceeding specs can be as problematic as not meeting them, especially when it comes to gain and bandwidth.  Many an audio amplifier can be turned into an RF oscillator with transistors that exceed every spec of the ones being replaced--they look good on paper.

[jdutky]

As far as specifying replacements, that's an art.  Just ask for help and value experience over theory.  You start by understanding the circuit that you are repairing to see what specs would be most important, then looking at what is available.

But how do I "look at what's available?" As I said, I've tried using the parametric search tools at the big parts houses, but have had very limited success. Is there some other place that I should be looking for available parts?

Also, a related question: what's up with the series naming schemes? Is there some logic to things like 2N3904 and 2N3906 (and similar for diodes)? Some of the help I've gotten on other forums has hinted at an underlying logic, but it's only been the vaguest of hints.

-- Jeff Dutky

[bdunham7]

There are many resources and there are problems with all of them.  Give a specific example and perhaps someone can go through the process with you.

The naming conventions, whatever they are, aren't uniform enough to be of any use to me.  The numbers just get familiar with time. 

[jdutky]

There are many resources and there are problems with all of them.  Give a specific example and perhaps someone can go through the process with you.

Well, I'm still in the very early stages of educating myself about transistors (I have an educational background in EE, but bailed in sophomore year to study CS, so there are lots of things I have only the vaguest notion of), so I can't offer much in the way of concrete examples. I am currently repairing an old oscilloscope (a Tek 475A from the early 80s) and I keep running into components that either were once commodity parts but are no longer produced, or were Tek custom parts (and are no longer produced). I've gotten a good deal of help from folks on the TekScopes group, but I'd like build up my own skills (and be less of a burden).

One of the examples that I'm currently dealing with is a part number A5T3571 which is an NPN in a TO-92 case with the following characteristics (taken from the Tek Common Design Parts Catalog for Tek part #151-0367-00):

V_cb 40V
V_ce 30V
V_eb 3V
I_c 50 mA
I_cbo 10 nA (?)
C_cb 1.5 pF
V_ce(sat.) 300 mV/20 mA (?)
h_FE(min) 100 @ 6 V/5 mA (?)
f_T(min) 1.2 GHz @ 6 V/5 mA (?)

I'm not sure about the characteristics that I put question marks after, either because the value seems improbably small (NANO-amps, really?) or because I don't fully understand what the notation is supposed to be telling me (I've only seen the V/A notation in Tek documents, not in other data sheets. I assume that this is telling us something about a family of curves by giving us a value at a specific point, but I'm lost beyond that guess).

This is a part in the Z-Axis/beam intensity amplifier. Some of the parts are subject to voltages as high as +110 V, and I assume that high bandwidth is of some concern (because this circuit controls the blanking signal, which has to be able to blank the beam in the chop mode, which operates in the MHz range), but beyond that rudimentary analysis I am well out of my depth.

Pretty much everything I've tried to search on in the parametric tools has resulted in either thousands of candidates or zero candidates, neither of which are helpful.

The naming conventions, whatever they are, aren't uniform enough to be of any use to me.  The numbers just get familiar with time. 

So, basically like the 74-series logic part numbers: you just get used to them?

-- Jeff Dutky

[tkamiya]

As to naming standard, made in Japan stuff has a convention/industry standard

2SAxxx  PNP high frequency
2SBxxx  PNP low frequency/DC
2SCxxx  NPN high-frequency
2SDxxx  NPN low-frequency/DC
I don't recall the rest, but any kind of transistors, FET, triac, etc, has industrial standard.

American made ones does not appear to have any standard.  2N<something> is an attempt at using same nomenclature.  But it didn't go very far.  To this day, any combination of alphabet can be a transistor.

I can tell you this much though....  in old days, everybody used commonly available ones like 2SC372 or 2SC945 for everything.  Only used something else when it was necessary.  These are very similar to 2N2222.  It works well for a lot of things.  I don't know why but tendency is to use NPN more than PNP. 

As to knowing how to substitute....  First thing to do is determine if choice of transistor is at all critical.  Many circuit can use anything commonly available.  PNP/NPN can even be switched if you know how.  Then I'd compare the spec of original piece to what I know to be widely available for that frequency, current, voltage, and hfe.  You'll accumulate this knowledge as you go forward with your hobby. 

As someone has said, knowing and understanding the circuit is very important.  If you are making an LED flasher, or a small power supply, anything will do.  If you are making a high performance RF amplifier at 10GHz, choices will be very limited.

[gcewing]

f_T(min) 1.2 GHz @ 6 V/5 mA (?)

I assume that high bandwidth is of some concern

Almost certainly. They wouldn't have used a part with such a high f_T unless there was a good reason.

You may have trouble finding a modern through-hole replacement. All the high frequency stuff seems to be going surface-mount these days.

[bdunham7]

One of the examples that I'm currently dealing with is a part number A5T3571 which is an NPN in a TO-92 case with the following characteristics (taken from the Tek Common Design Parts Catalog for Tek part #151-0367-00):

Well, for Tek stuff where they don't specify a common part number, I would try very hard to find an OEM part.  You certainly haven't started in an easy, forgiving arena.  And the part is available from a reliable vendor:

https://www.ebay.com/itm/NEW-5-PCs-LOT-Tektronix-151-0367-00-Custom-Transistor-/233674679586

However, since your question was a general one about substitution, not fixing a Tek scope, let's look at the specs:

V_cb 40V
V_ce 30V
V_eb 3V
I_c 50 mA
I_cbo 10 nA (?)
C_cb 1.5 pF
V_ce(sat.) 300 mV/20 mA (?)
h_FE(min) 100 @ 6 V/5 mA (?)
f_T(min) 1.2 GHz @ 6 V/5 mA (?)

So it will work in a circuit up to 30 volts (less some margin), has a tender EB junction that can only take 3V reverse bias, and is not high power (I_c only 50mA).  The 10nA looks reasonable (it is reverse leakage of the strongest junction)  and the saturation voltage looks normal-ish, so nothing odd there.  The gain is specified at least 100 at the specified conditions, well out of saturation and well below the power/current limit.  Gain varies a lot with conditions so just the gain number isn't a good way to compare--you need specific conditions.  Then you have f_T which is the transition frequency, or the highest frequency at which the transistor will work normally--your circuit bandwidth should be much below this.  You also have the capacitance spec, which also tells you that this is a high frequency device. 

So how would you sub it out?  The first thing that I would do is a cross reference, then I might try  a limited parameter search using only the transition frequency and V_ce specs and read through the results.  If I didn't get results, I would take a close look at the circuit and see what the actual operational requirements are.  Perhaps it operates at 15 volts and 20MHz of bandwidth is plenty.  So lets try a few places:

The NTE cross reference says the NTE278 (in a TO-39 case) is the replacement.  Read it and see what you think.

https://www.nteinc.com/specs/200to299/pdf/nte278.pdf

This datasheet seems to indicate that this is a 15V part and not 30V and that 2N3571 or 2N3572 would be replacements.  Neither of those seems current, but it is another part number to check for NOS on.

https://www.datasheets360.com/pdf/2979968532151235572

And I'm out of cross-references for now.  So let's look at Mouser just with the two specs.  I added 'through hole' to eliminate SMT parts.  At first I got no valid results with those specs.  Then I eliminated the voltage and used 'greater than 900MHz' for the frequency.

https://www.mouser.com/Semiconductors/Discrete-Semiconductors/Transistors/Bipolar-Transistors-BJT/_/N-ax1sh?P=1z0z63xZ1yyx4b4Z1yyx39eZ1yyzhiuZ1yyyt0wZ1yp7i0eZ1y95l6h&Ns=Gain%20Bandwidth%20Product%20fT|0

So now, read the spec sheets of the first few hits, see what you think.

So, basically like the 74-series logic part numbers: you just get used to them?

The 74-series numbers all have meanings, but either I recognize them directly without deciphering or I look them up.  I need pinouts anyway.

[dietert1]

If it's Tektronix, it's probably an American transistor and if you look up the 2N3571 it roughly fits. What is special in your transistor is the higher than average fT. Probably at Tektronix they selected them for a certain hfe in order to use them in DC to 200 MHz broad band amplifiers - and labeled them A5T3571.
https://www.ti.com/general/docs/suppproductinfo.tsp?distId=10&gotoUrl=http%3A%2F%2Fwww.ti.com%2Flit%2Fgpn%2Flm723

If you use the context like this, it will still be guesswork. There is a special web group for old Tektronix gear where you may get more precise information.

There exist "transistor equivalent" tables. E.g. you will find a recommendation to replace a 2N3571 by a 2N3012, but that one isn't going to work very well. Only 12 V and only 400 MHz may not be good enough.
https://transistordata.com/bjt/2n3571

Regards, Dieter

[jdutky]

Well, for Tek stuff where they don't specify a common part number, I would try very hard to find an OEM part.  You certainly haven't started in an easy, forgiving arena.  And the part is available from a reliable vendor:

https://www.ebay.com/itm/NEW-5-PCs-LOT-Tektronix-151-0367-00-Custom-Transistor-/233674679586

Yes, I've been doing that, though also with limited success (no show stoppers, however).

The NTE cross reference says the NTE278 (in a TO-39 case) is the replacement.  Read it and see what you think.

yes, that looks pretty good. I've bookmarked the NTE cross reference link as I'm sure it will come in handy.

And I'm out of cross-references for now.  So let's look at Mouser just with the two specs.  I added 'through hole' to eliminate SMT parts.  At first I got no valid results with those specs.  Then I eliminated the voltage and used 'greater than 900MHz' for the frequency.

https://www.mouser.com/Semiconductors/Discrete-Semiconductors/Transistors/Bipolar-Transistors-BJT/_/N-ax1sh?P=1z0z63xZ1yyx4b4Z1yyx39eZ1yyzhiuZ1yyyt0wZ1yp7i0eZ1y95l6h&Ns=Gain%20Bandwidth%20Product%20fT|0

I was able to turn up 9 good candidates from your starting point. I guess I'll just need to keep plugging away at the parametric search until I get a gut feel for it.

-- Jeff Dutky

[T3sl4co1l]

Besides doping, thickness is also very important.  Of course you can't just wire two diodes together and get a transistor -- there's a bunch of metal inbetween, and metal destroys the semiconductive effects -- but it's still not good enough to take a lump of silicon with two spots on it.  Particularly, the base needs to be very thin: under 10µm or so.  This is the distance over which free charge carriers (electrons with enough energy to ride in the conduction band, and the holes left behind in the valence band) can diffuse through bulk silicon, occasionally combining along on their way at doping sites (which give off a small excess concentration of electrons or holes, and by reciprocity, also consume them).

So, if you can take two N-type blocks and one very thin P-type sliver, clean their surfaces perfectly, and also polish them atomically flat (I know, so picky, right?) -- you can physically assemble a transistor.

Needless to say, the planar method is far easier, either diffusing dopants in from the top (with depth controlled by heating schedule and concentration), or by growing doped material on top directly (epitaxy, usually using heat or plasma to decompose silicon- and dopant-bearing gasses).

So that's how you make a transistor.



As for the substitute, PN3563 comes to mind, have seen it in plenty of circuits.  Though it's lower voltage (and also quite obsolete), which I'm sure Tek also needed for the application.  Agreed, the -3571 could be the 2N (JEDEC*) designation, then they modified or selected it from there.

*An industry standards group -- numbers are simply sequential as far as I know, and almost all are deprecated/obsolete.  For that matter, I would guess almost all were obsolete within years of their allocation.  Probably a lot of numbers were scatter-gunned based on the wide parameter spread of early transistors (particularly in hFE and Vceo or Vcbo), so there was a lot of overlap, too.  It's worth noting how this type of standardization works: anyone can produce a "2N3904" that meets or exceeds that spec.  A particular manufacturer may give information about their particular part (ON Semi (née Motorola) often does) but that doesn't mean those data apply to all manufacturers.  Whereas these days, with many single-source parts, you simply get whatever a e.g. STP6N60M2 is, and if you need a substitute from someone else, you look up whatever crosses to it, or search for something with comparable parameters.

Tim