Transistor Gain Test Circuit. I'm about to give up on all this.

[digitalzombie]

OK, so I did three tests and got three very different results, and now I'm wondering which, if any, is correct.

I'm using a 2N3904 which has a max gain rating of 300.

First thing I did was try out the plug -n- read method using the adapter. There's no markings for CBE so I just went with matching the shape to the front ( see above giant picture) and it gave me a reading of 326.

Then I tried Steve's method, took my line voltage of 9.45-0.65. I read 3100 mA. 3100/8.8=352.

Then I tried R.G.'s method from the above link using a 2.2m to base and 2.2k to collector, reading voltage across the smaller resistor. I read 0.446v. So 446mV/2200=202.

What. The heck.

To make matters worse I just went to double check the plug-in reading and... 15! I flipped it upside down and it read 965. WHAT IS GOING ON?!?!

[digitalzombie]

Good morning. In case you're keeping along with the drama that is "That week that guy obsessed over testing transistor gain", here's another wrench in the gears.

I popped an unused 2N3904 in the DMM adapter and that read 313, and the test tranny from last night still read 15. I checked them both with the diode/continuity function. Long story short the bad tranny gets continuity between the base and emitter, so somehow one of my breadboard tests blew out that junction. Can anyone take a guess as to how that happened?

[Transmogrifox]

Your number from your meter's plug-in measurement and Steve's method look reasonable.

HFE is not a constant value over all conditions.  Most datasheets I see for 3904 state this value at collector voltage = 1.0V and sometimes state or plot over a range of collector currents. 

The closest thing I can find is hFE, small signal gain at 1 kHz with a collector voltage of 10V.  For that value it states up to 400.  This gives a hint that the DC gain is higher with higher collector voltage, and, in fact, this is the general trend.

The answer is you're not comparing apples to apples, so I would say Steve's method is getting you pretty close to reality under the conditions being tested.

As for RG's method, I would expect slightly different results because the transistor is being operated at a different current, but your measurements indicate it was done incorrectly:
http://www.geofex.com/Article_Folders/ffselect.htm

If you measured 0.446V across a 2.2k resistor (wrong value resistor to begin with), the current is 0.446/2.2k = 202.7 uA, and the base current is about 8.8/2.2Meg = 4uA. 
Gain = 202 uA / 4 uA = 50.7

If you had followed those directions exactly and had a 1% 2.2Meg, and a 1% 2.49k resistor measuring 2.472k you would have come up with the following:
202uA*2.472k = 501 mV

Per RG's instruction, it is 100x the measured voltage,
gain = 501mV*100 = 50.1, which is really darn close to what I would have predicted with my interpretation.  RG scaled these resistors this way so that it works out pretty close without needing to use twiddle factors. 

This is the beauty of RG's method if you use the correct resistor values:  gain is always 100x the measured voltage so you can do it in your head at a glance. It's clever, and it "just works".

There is almost certainly something wrong with how you performed RG's method because it would otherwise have given you something close to what you got from the other 2 methods. 

Here's a guess:
RG's method is for a PNP transistor.  You are measuring an NPN transistor.  If you literally applied the circuit as shown for an NPN device, then the emitter-base junction was reverse-biased beyond its absolute maximum of 5V and avalanched the junction.  Is 200 uA enough to damage it?  My guess is no, but at the very least this would have given you a strange measurement like what you reported.

Using Steve's method, with the transistor the correct direction, my LTSpice simulation gives me something very close to what you are measuring (gain of 328).  I think this and your meter are probably within reasonable agreement.

Another opportunity to damage the component was when you put the transistor into the meter backwards.  Not knowing what kind of circuit is being used in the meter it's hard to tell, but if there is a voltage applied from emitter to base > 5V  it will probably avalanche this junction, and if it isn't power limited it can do permanent damage to the device.

Try another transistor and double check the pinout before sticking it into the socket.  I think you will find after measuring 10 or so transistors that your meter socket, Steve's method, and RG's method all agree within +/- 10% if you are doing it correctly.

A good way to see if it's your meter that damaged the device or if it's putting it into a test jig backwards, first with your meter:
Put the good one in forward, confirm.
Put it in backward.
Then put it back in forward again.  If it's damaged it will be a really weird number and you will know your meter socket damages the devices if put in wrong.

If your meter doesn't damage it then that leave's RG's method done incorrectly as the suspect.

It will be good to know this so you know you need to throw away any devices that you accidentally put into your test jig backwards.

As for accuracy and simplicity, I would tip the hat to RG's method.

[R.G.]

Welcome to the study of experimental error and the tricky issues of measuring semiconductors.

1. You have to set up an absolutely reliable testing setup, so that every time you measure something, you know it's not an artifact of some bug in the measurement setup.

The ugly truth is that there are no absolutely reliable test setups in the real world so as a newbie to this, you are getting the lesson that testing is always going to be involved with worrying about whether your test results are accurate or if some quirk of the test setup is overwhelming what you're trying to measure. If you ever read scientific papers, you'll find that a lot of the papers are given over to trying to figure out if what they measure is real or some experimental error in the setup. Even if you have a meter that reads exactly what you want, you always have to be wondering whether there is some hidden artifact going on. Worse, you have to not get too upset about it, and just get on with things, trying to understand the sources of error.

In getting my EE degree, I had to take a whole course in measurements. Sounds silly, right, just hook up the meter. Wrong-o. The universe is more complex that you can sense on any one meter. Test setups are designed to severely limit what is being measured so you can focus on it. Some measurements are like trying to nail Jello to a board.

2. Semiconductors in general don't have one number for **anything**.

Transistors have a gain that varies with almost everything, including the voltage and current conditions you measure them under and including especially the temperature. You cannot accurately say that a transistor has a gain of X without also specifying the conditions under which the gain was measured. Not only do transistors not have a single gain, it takes more than three dimensions to construct an accurate graph of how the gain varies and why.  It's instructive to set up your measurement circuit and get some reading, then place your fingertip on the device being measured for a while and watch what the reading does. Your fingertip's warmth will change the reading. In fact if your fingertip touching the device doesn't change the reading you're not reading it right. They're that temperature sensitive.

This ought to be freeing to you. There is no single number. The number varies by the local temperature, how long you held the transistor putting it into the socket, the (usually unspecified) current and voltage at which the meter reads the gain, and other things. A single-number gain measurement is an indicator, not something that is cast in stone.

There are professional engineers that have devoted entire careers to the task of figuring out how to measure semiconductor parameters. It's that complicated.

Learn to flow with the currents and still get where you need to be.

[digitalzombie]

As for RG's method, I would expect slightly different results because the transistor is being operated at a different current, but your measurements indicate it was done incorrectly:
http://www.geofex.com/Article_Folders/ffselect.htm

Here's a guess:
RG's method is for a PNP transistor.  You are measuring an NPN transistor.  If you literally applied the circuit as shown for an NPN device, then the emitter-base junction was reverse-biased beyond its absolute maximum of 5V and avalanched the junction.

Hey TM. I was going off of this thread where R.G. says to do the test with an NPN you can leave all the components where they are and reverse the voltage, so that's what I did. He also says using a 2.2K is fine, just takes some extra math which I'm fine with. I'm not going to pay $6 shipping from Mouser just to get one odd value resistor. I really do appreciate the context you're putting things into though.

Welcome to the study of experimental error and the tricky issues of measuring semiconductors.

Learn to flow with the currents and still get where you need to be.

Thank you oh wise R.G.-san *bows* I totally get that I shouldn't "obsess". This last couple weeks I've been reading and reading about the subject and I'm just waiting for that "AH-HA" moment where it all clicks which - armed with only an "A" in High School Chemistry in the 1990's, a copy of Getting Started in Electronics, and without having an EE degree myself - is proving... difficult. I want to understand it but I don't necessarily need to dominate the subject. Thus the reason for three different tests - I'm fine with a "range", as long as the results make sense.

[digitalzombie]

I'm at a point now where I'm feeling more confident about the testing setup, and out of four different 2N3904's tested the numbers where "within range" of each other between the tester jig and my adapted R.G.'s method. Here's what I got testing with socket/breadboard:
Q1: 324/334
Q2: 265/237
Q3: 322/327
Q4: 310/310

I know all sorts of stuff needs to be taken into account and on a scientific level these numbers are meaningless, but I think for the application we're all here to use them for I think this gets me into the ballpark of what I need to know going forward.

My results for Q2 are interesting to me though. Not only did test higher on the socket when the others tested higher on the breadboard, but it was also the only one under 300. While I get that's "within spec" for these, I thought silicon was supposed to be more consistent.

And yes, I tested the socket again if it was the culprit behind damaging the other tranny by flipping the polarity forward and back again on a good tranny, and it read fine. In fact, I brought a BC337 with me since the pinout is opposite that of a 2N3904, and the socket definitely only works L->R E-B-C. Would've been too convenient for them to mark that on the socket itself I guess.

I'm excited to try the full version of the test at home with an NTE germanium I picked up to experiment with. Thanks guys.

[Transmogrifox]

It looks like you have the hang of it now

Don't any longer think semiconductors/silicon is more consistent than this.  They specify a broad range because the process isn't any more deterministic than that.  As RG mentioned, it's also very sensitive to temperature and operating conditions...and the temperature of the part is affected by its power dissipation so then voltage, current and temperature are all interactive.

This is why most BJT circuits use some form of feedback to self-adjust the unknowns out of the mix.

[Fast Pistoleros]

To get down to the Bare Bones, set your meter measure a low CURRENT and then set up what's shown in these two pics:





The second pic shows a little more detail of the breadboard. The device is a 2N2222A. Maybe I should first repeat what Transmografix said, with this emphasis: To measure the flow of current directly, the current has to flow Through your instrument. The meter does not show current flow from Collector to Emitter yet, because there is no current going into the transistor Base. Now connect a 1 meg resistor from the Battery + to the transistor Base:



The meter now shows a current flow of about 1.6 ma (1600 microamps) from Collector to Emitter. You don't need a meter in the Base circuit to get a very close estimate of the current flow going into the Base, because the resistance from Base to Emitter is tiny relative to 1 Meg. So by Ohm's Law, the Base current I is 9 Volts (E) / 1,000,000 ohms (R) or .000009 amps, most easily expressed as 9 microamps. If you are not clear about Ohm's Law, find a basic reference and get straight with it.

The gain of the device is now 1600 / 9 or 177. I would expect a 2N3904 to run in the low 200s.

I hope this has made sense.

makes perfect sense. every transistor I have on hand matches right up with the Peak analyser with incredible accuracy using this bear bones method !

[digitalzombie]

Hi y'all. I'm back. I'm testing some NPN germs now and I want to make sure I'm doing this right.

So I built up my R.G. test circuit, and for the 2.4**k resistor I'm using pins 1 & 2 of a 10k trimmer dialed in as close as I can get to 2,472 ohms. It's sitting at 2.468k.

So the first thing I measured was this NTE 103A I picked up from my local Fry's Electronics. Measuring on either side of the trimmer w/out current going to the base I'm measuring 1.433V. So I divide that by the actual resistance going to the collector (2.468k) and I get 580.6uA. Now I know this is supposed to be my leakage current and I understand >500 probably isn't usable but I wasn't expecting miracles from a store bought tranny.

So next I flipped the switch that connects my 2.2m to the base, and my DMM now reads 2.31v. Divided by my collector voltage resistance and I get 936. 936-580=356. 356/4=89. So that means my gain is 89 with a leakage of >500... should I just throw this one away?


I also just received a bag of MP38A trannies from the Ukraine (woohoo, look at me!) so I threw a couple of those on the tester. I can't read the datasheet because, Russian, but here are the numbers I got:

326mV. /2.468k=132uA. w/Base= 0.82v. /2.468=332.  -132= 200. /4uA=50hFE.

Seems kinda low, or is that normal for germanium? I don't have a single silicon transistor that measures under 200hFE so is 50 even usable? Does it look like I'm doing the test correctly?

[Kipper4]

So is 50hfe even usable.
Correct me if I'm wrong guys but can't you just scale down say a fuzz faces ideal transistor values of 70hfe 110hfe to say 50hfe 75hfe and still get that sound albeit a lower overall headroom.
Right or do I need to go back to school?

[Ben Lyman]

So is 50hfe even usable.
Correct me if I'm wrong guys but can't you just scale down say a fuzz faces ideal transistor values of 70hfe 110hfe to say 50hfe 75hfe and still get that sound albeit a lower overall headroom.
Right or do I need to go back to school?

You can always punch the numbers in this and breadboard it to find out:
http://www.diystompboxes.com/analogalchemy/emh/emh.html

[duck_arse]

dz - did you take the base resistor to +supply for the npn measures?

and recent extensive consultations suggest the pinouts for the GT20/21/37/38 are the same as the TO1, like an AC128. the datasheet I'm gazing upon says "h_21e 45~100" for the 38A

[digitalzombie]

I'm following the instructions from GEO's ffselect page. So base isn't getting supply voltage unless that's what you call it when i put that 2M2 in series. And I was able to find the pinout with my DMM by one of the helpful videos posted earlier.

[duck_arse]

now that I look at the circuit again, to test npn you just reverse the battery. the push button will then apply positive bias to the base, via the 2M2.

[digitalzombie]

now that I look at the circuit again, to test npn you just reverse the battery. the push button will then apply positive bias to the base, via the 2M2.

Right. So I'm asking if someone more qualified than myself could look at the numbers I posted and let me know if it looks like I'm doing this correctly.

[Fast Pistoleros]

I did a tutorial on this , it might help some new people out

[PRR]

> is 50 even usable?

If not, then there were NO transistor circuits before maybe 1963.

An awful lot of transistor radios used "reject parts" with hFE nearer 20 than 50.

Raytheon sold heaps of rejects as "Experimenter Transistors" CK722. Datasheet Note hFE might be 12 or 22, more or less, depending on raw materials and baking; but you were "sure" not to get a lot more (because higher-gain CK72_ were marked CK721 and sold for much higher price). Hundreds of thousands of these transistors were bought, at prices above $7 in 1953.

My EE prof told us to "assume hFE>50". This meant our quiz solutions must work for ANY transistor with hFE from 50 to infinity. At that time, this was a reasonable assumption. At the time Silicon processing was pretty random. Result could be hFE from 10 to 1,000. The >500 exceptions sold for more. Random 10-500 specs were fairly cheap. Paying another 20% they would leave-out the lowest 10% parts, giving hFE=50-500.

You CAN get audio gain with hFE of 1 or less. (Ponder the "common base" amplifier where current gain is always less than unity.) The other factor (which nobody ever mentions) is what hollow-heads called "Amplification Factor", maximum theoretical voltage gain. On modern parts this is 10,000 or higher (and we can never realize this, which is why it isn't worth mentioning). But you can have current gain under 1 and voltage gain of 10 or 100. (It does become difficult to cascade without transformers.)

No, do NOT throw-out non-dead transistors with specs you don't know how to use.

Anyway, two hFE=50 parts can be Darlingtoned for hFE far over 500. (Not 2,500, because the input device runs at low-low current and hFE will be far down from 50.)

[digitalzombie]

> is 50 even usable?

If not, then there were NO transistor circuits before maybe 1963.

An awful lot of transistor radios used "reject parts" with hFE nearer 20 than 50.

Raytheon sold heaps of rejects as "Experimenter Transistors" CK722. Datasheet Note hFE might be 12 or 22, more or less, depending on raw materials and baking; but you were "sure" not to get a lot more (because higher-gain CK72_ were marked CK721 and sold for much higher price). Hundreds of thousands of these transistors were bought, at prices above $7 in 1953.

My EE prof told us to "assume hFE>50". This meant our quiz solutions must work for ANY transistor with hFE from 50 to infinity. At that time, this was a reasonable assumption. At the time Silicon processing was pretty random. Result could be hFE from 10 to 1,000. The >500 exceptions sold for more. Random 10-500 specs were fairly cheap. Paying another 20% they would leave-out the lowest 10% parts, giving hFE=50-500.

You CAN get audio gain with hFE of 1 or less. (Ponder the "common base" amplifier where current gain is always less than unity.) The other factor (which nobody ever mentions) is what hollow-heads called "Amplification Factor", maximum theoretical voltage gain. On modern parts this is 10,000 or higher (and we can never realize this, which is why it isn't worth mentioning). But you can have current gain under 1 and voltage gain of 10 or 100. (It does become difficult to cascade without transformers.)

No, do NOT throw-out non-dead transistors with specs you don't know how to use.

Anyway, two hFE=50 parts can be Darlingtoned for hFE far over 500. (Not 2,500, because the input device runs at low-low current and hFE will be far down from 50.)

BRB. Gonna go get my EE degree.

[PRR]

> Gonna go get my EE degree.

I didn't.

Don't think many post-1980 EE programs would help this question directly.

[PRR]

BTW: here is a 1957 Pop Mech article on a guitar amplifier using mostly those hFE=12-22 CK722 parts.

https://books.google.com/books?id=oeEDAAAAMBAJ&pg=PA160&hl=en#v=onepage&q&f=false

Wise-guys will note it is a simplified "Darcy".

(Lou Garner later became well-known at Popular Electronics, with a column on both new and DIY transistor ideas.)