Strange results measuring beta

[Analog Kid]

[my first post here]

I'm hoping someone here can show me what I'm doing wrong, because I'm getting some screwy results trying to do something really simple.

This is an experiment, for my benefit, measuring the beta of a transistor. Should be simple, right? Drive the DUT (device under test) with a known base current, measure the collector current, divide I_C by I_B and voila. Except no voila here, just a bunch of values that make no sense.

The rig I'm using is this:


Pretty simple: I'm using a current mirror to supply the base current, adjusting it to a small known amount (using 20 μA). I'm using 2 meters, a DMM (cheap one but accurate) for the base current, an analog one for collector current. It's breadboarded on a board that seems to have good connections. Using a fresh 9-volt battery for the tests.

I'll post a table of result values when I do some more systematic tests, but to give you an idea, I tested a 2N2222, applied 20 μA base current, got an I_C of 25 mA. That would be a beta of 1,250, which seems ... unlikely.

Another thing: some of the transistors being tested get really, really hot, even though the I_C would indicate that they're well within their current and power limits. That 100Ω resistor is an attempt to limit current enough to prevent heating; I don't like it being there, and it's probably interfering with accurate measurement.

I tried using a 5-volt supply (well-regulated, using a 7805); the collector current barely registered at all (30 μA vs. 20 μA base current, beta basically zilch).

My understanding of things transistor tells me that in order to make a beta test that's anywhere near accurate, the transistor needs to be within its active region. I'm not sure that's the case here.

My intent here is to teach myself transistor basics through a hands-on experiment; I'd also like to come out of this with a simple, accurate-enough beta test rig to go through transistors and weed out bad ones. (After heating up some of them I'm pretty sure a lot of my 2222s and 3904s are fried!)

So what am I doing wrong here?

[mawyatt]

Why bother with the 2N3906 current mirror, just use 100K pot across Vcc and 470K to DUT base from pot wiper.

Check for DUT oscillations. A small cap (100pF) from DUT C to B should help, as does ferrite beads on the C and B DUT leads.

Best

[RoGeorge]

Another thing: some of the transistors being tested get really, really hot, even though the I_C would indicate that they're well within their current and power limits.

If the DUT heats at apparently low Ic, usually it means there are undetected RF oscillations.

The measuring instruments might not see the oscillations.  Sometimes the RF is so high that not even the oscilloscope can sense them.  Usually the counter measure against oscillations is a ferrite bead inserted on the base terminal, but that usually happens for common collector, not for common emitter.  Though breadboard might show enough parasitic L and/or parasitic C to make an unintended oscillator.

I would add decoupling capacitors on the power supply, too, 10uF in parallel with 0.1uF placed near the circuit (the exact value is not critical, any values in that range should do it, decoupling the power supply might avoid undesired oscillations).

Don't know if RF oscillations are the reason for heating your transistors.

If you go near the max power specified in the datasheet you may see considerable heating.  Often the max power in the datasheet is specified for an infinite radiator.  To calculate the expected temperature for a given power, look in the datasheet for the thermal resistance between junction and capsule + thermal resistance between capsule and air.

[Analog Kid]

Thanks for both those replies.

I guess I was overthinking things with the current mirror. Using the pot as a voltage divider actually works better, give more range. And since the current source is adjustable, there's really no need for a constant current source.

At this point I honestly don't think I have to worry about RF oscillations; it's a simple DC circuit with no feedback path. (There may have been one with the current mirror.) Now I'm getting believable measurements, in the 100-350 range.

There was also a very weird thing--varying the base current had zero effect on collector current--until I learned (better late than never!) that a PN2222 isn't the same thing as a 2N2222.

[Analog Kid]

My next question: parameters. I kind of randomly ended up using 20 μA as a base current, a nice low round number to give a fairly low collector current, but I'm not sure it's ideal. My guess™ is that a guy would want the highest possible value that wouldn't risk overdriving the DUT with excessive current. Can someone help me pick such a value? For now, this is only for small jellybean xistors, no power ones, no 3055s, TIP32s, etc., so I guess I_C can be limited to 50-100 mA max? (Hmm, maybe not that much: 100 mA = ~900 mW, way too much for some Qs.)

How do the pro transistor testers do this?

[RoGeorge]

a guy would want the highest possible value that wouldn't risk overdriving the DUT with excessive current. Can someone help me pick such a value?

What your guy really needs is the beta at the specific Ic used in a given schematic. 

Beta varies with Ic (and varies with frequency, too, that's another story).  So, a generic beta would be better than nothing, but what I would really want to know is the beta for the particular Ic I am using in my schematic.

Often there are charts in the transistor's datasheet, chart(s) that represent typical beta vs Ic.  In your case, you'll have to adjust the potentiometer until the Ic becomes the one you intend to use in your schematic, then read the Ib and make the ratio Ic/Ib.

I don't know what is the typical Ib used by generic DMMs capable to measure beta, but your example with 20uA for Ib seems fine.  Alternatively, you can search for a few DMMs schematics, and see what typical Ib they are using.

[T3sl4co1l]

Easier to let the circuit set base current for you.

Set up a base voltage divider, and use an emitter resistor.  Measure Ib and Ic.  This sets Vce (approximately) and Ic as well.  Ib is then simply whatever's necessary to achieve that condition, usually a small fraction, as expected.

You didn't mention what V, I you tested at, nor of what all devices; setting Ib on a Darlington for example could be disastrous, though how hot a part gets in relation to its size, and the load resistor, isn't clear.

Beware also a MOSFET, if you accidentally test one, won't draw "Ib", and (unless it's an RF type, that is very old for one, but also likely ESD-zapped in the process of handling, unless you took considerable precautions?) draws more or less as much "Ic" as it can (the highest Rds(on) parts commonly available are 5 or 10 ohms still, quite low).

Tim

[jwet]

Beta is a messy and kind of non-sense parameter.  It useful to give you some feel for what's happening but the actual operation of a bipolar is much different.  Biploar transistor are actually "transconductive" meaning that the collector current depends logarithmically on the base emitter voltage.  Google "Ebers-Moll" model, this is a reasonable low frequency model that will illuminate a lot of the operation of transistors.  A proper model for a bipolar should be a transconductance (gm) controlled by Vbe.  You can find resources online that can explain this way better than I can.

I applaud your quest for first principles knowledge.  I would recommend a couple of books- Art of Electronics is one of the very finest. (oider editions are available to download online for free- the bipolar transistor hasn't changed since 1980 so you won't miss much).  Another good intro is Electronics Circuits- Discrete and Integrated by Shilling and Belove, a relatively low theoretical content (minor calculus) but solid into to transistor and fets.  Ironically, it covers a lot of topics but doesn't cover current mirrors.  There are others that people may recommend here.  The realization that transistor beta is mostly a fantasy driven by incomplete understanding of the real physics is a real eye opener for anyone in the field.  Few people actually appreciate the real operation of these devices.

[Smokey]

LTSpice if you aren't already playing with that.  It's significantly easier and quicker to "experiment" with theory in the simulator than it is on the bench.

[magic]

Yeah, you will quickly learn theory and then have no idea why it doesn't apply in real circuits.

In this case, my bet on the transistor being reversed (collector and emitter swapped). Or maybe some other pinout oopsie. This would easily explain impossibly high β at 9V and very poor at 5V. And, by the way, SPICE doesn't simulate the 9V case correctly.

Note that 2N2222 is a metal can device and there are various TO92 equivalents with subtly different part numbers and two common pinouts (EBC/CBE). And when it comes to "made in China", all bets are off.

[PCB.Wiz]

I'll post a table of result values when I do some more systematic tests, but to give you an idea, I tested a 2N2222, applied 20 μA base current, got an I_C of 25 mA. That would be a beta of 1,250, which seems ... unlikely.

It is unlikely..
Did you confirm your meters are properly in spec ?
If you have a multimeter with a HFE testing socket, you can sanity check your results.

Self heating can affect the numbers as HFE also varies with temperature, so I'd suggest simple fixed high value resistors for base feed, with momentary push buttons.
eg Base current tests of 1uA 10uA 100uA would be useful for most small signal transistors.

An alternative approach to HFE is to feedback regulate for a given emitter or collector current, at a defined VCE and then measure the base current that requires.
A series base resistor can give a simple means to check base current, as usually multimeters can measure voltages over more ranges, easier than current.

[Analog Kid]

Thanks for all the replies. I'm impressed just by the amount of traffic here!
Few notes:

o I have The Art of Electronics. Great book, although a little heavy on the math for my taste (I took calculus but it makes my head hurt). The guy who gave me his old copy is actually a good friend of Paul Horowitz.

o i know from that book and others that beta (or H_FE) is a parameter to be avoided when designing circuits, as it varies from part to part and in circuit operation. I'm mainly interested here in learning principles firsthand, and also in building a tester to separate decent working parts from junk.

o No MOSFETs here, at least not yet. I'll burn that bridge when I come to it.

[mawyatt]

Thanks for all the replies. I'm impressed just by the amount of traffic here!
Few notes:

o I have The Art of Electronics. Great book, although a little heavy on the math for my taste (I took calculus but it makes my head hurt). The guy who gave me his old copy is actually a good friend of Paul Horowitz.

o i know from that book and others that beta (or H_FE) is a parameter to be avoided when designing circuits, as it varies from part to part and in circuit operation. I'm mainly interested here in learning principles firsthand, and also in building a tester to separate decent working parts from junk.

o No MOSFETs here, at least not yet. I'll burn that bridge when I come to it.

Please note post by jwet, learning the bipolar transistor is a formidable task that very few have mastered. 

Quality analog circuit designs (and designers) don't rely on Beta but on the semiconductor physics relationship between junction voltage and current which is fundamental in analog IC design involving bipolar transistors. The genius of Bob Widlar, George Erdi, and Dave Fullagar is shown in their brilliant analog IC designs and understanding of the bipolar transistor.

Studying the details of transistor behavior from the inside out rather than outside reveals the remarkable characteristics that gets "covered up" with parameters like Beta and HFE (hfe) which BTW are not the same parameters.

The bipolar transistor hasn't changed much over the years being displaced with the incredible shrinking CMOS, with only the introduction of Ge in the Si process for the SiGe device, and the use of InP rather than Si, both of which were to increase transistor speed.

We've been fortunate enough to design in Si, SiGe and InP over our career and altho they are different when viewed from the outside their internal bipolar behavior isn't that much different when observed from a semiconductor physics standpoint inside.

You are venturing into a fascinating trip wrt the bipolar transistor, please don't stop at the outside details, dive deep into the physics of the wonderful devices, they are absolutely amazing

Best 

[PGPG]

I'm mainly interested here in learning principles firsthand, and also in building a tester to separate decent working parts from junk.

In my opinion...

1. You don't need to measure Ib.
Vbe is around 0.6 to 0.7V so if you power base by resistor from 9V you can simply count Ib=(9-0.7)/R and it is more precise than beta measurement precision needed.

2. I think measuring beta for collector current about 1..5mA for signal transistor is good enough.

3. Learn to understand Θja and Θjc parameters. If you dissipate 300mW in 2N2222 and its Θja (junction-ambient) is 350°C/W then junction is 0.3*350=105°C higher than ambient temperature. If air temperature near transistor is 25°C than junction temperature is 130°C. If Θjc is 97°C/W then case is 0.3*97=27°C lower than junction so case is 130-27=103°C. Certainly too high to be touched even it is only 300mW.

Beta is a messy and kind of non-sense parameter.
....
Biploar transistor are actually "transconductive" meaning that the collector current depends logarithmically on the base emitter voltage.  Google "Ebers-Moll" model,

Small signal transistor analyse is next step in transistor understanding than its DC polarization. DC polarization should also be made such a way that beta 'don't cares' but for DC working point calculations beta is fully sense parameter to be used to count what about is base current when collector current is as you decided you want to be in your circuit.
Many transistor circuits can be designed and enough correctly counted not using small signal analyses. In oscilloscope I made when was young vertical deflection plates required 30V/cm. Designing whole its Y amplifier can be done without thinking about small signal transistor model at all.

[TimFox]

Here's an elementary circuit, which I originally used for measuring grid current for vacuum tubes.
You probably don't need all the meters, depending on what two power supplies you used, but you need to measure base current and emitter current (or collector current).
I originally used a constant current supply for the emitter (or cathode), but that is not necessary.

[Analog Kid]

Yes. Here's the current test rig after removing the current mirror:


Measuring emitter current would make it be off by the amount of base current, yes? Although I_B would be pretty much swamped by I_C ...

[TimFox]

The reason for my circuit is to measure I_base as a function of I_emitter, which is safer.  It’s also useful sometimes to look at different V_ce values.
For critical applications, you can correct I_c for the base current.  The negative power supply allows you to set the emitter current.

[Analog Kid]

The reason for my circuit is to measure I_base as a function of I_emitter, which is safer.

Safer? How?
Not challenging you, just want to know.

[TimFox]

The base current is a much slower function of the emitter or collector current than vice-versa, so it is safer to control the emitter current and measure the base current.
If you apply enough base current to saturate the transistor, the collector current from a “stiff” voltage source can be too high for the device

[Analog Kid]

So we can assume, for the sake of close-enough testing, that I_C ≈ I_E, right?

[Someone]

The reason for my circuit is to measure I_base as a function of I_emitter, which is safer.

Safer? How?
Not challenging you, just want to know.

This is safer by directly adjusting the parameter that is most likely to result in failure, limiting the emitter current directly. If you only adjust the base current and leave the emitter current unbounded/unlimited then it could end up (unexpectedly) far above "safe" conditions.

[PCB.Wiz]

Yes. Here's the current test rig after removing the current mirror:

I would swap the 470k and meter.
Multimeters plus leads can pick up a lot of hum and general noise, which can affect sensitive measurements like this, and the base-emitter junction is a diode.

Avoiding the effect of long leads is another reason for the suggestion of fixed switched base resistors.

[Someone]

So we can assume, for the sake of close-enough testing, that I_C ≈ I_E, right?

That is one approximation that could be valid. It depends on the accuracy of the measurements being taken. That approximation error might be less than the error of the meters being used, or in reverse; some accurate meters could measure 2 of the parameters and still have an accurate result of the 3rd parameter.

[TimFox]

Algebra:  if you measure I_b and I_e, then the ratio
I_c/I_b = 1 + I_e/I_b
is accurate.

[Analog Kid]

Ah, yes, because ... alpha.