bjt current, hfe, and R.G.'s transistor test

[gaussmarkov]

because of several recent threads on R.G.'s transistor tester and my own fuzz face project, i have gone back to extending my understanding of BJTs.  to check my breadboard tester, i took a 2N3904 (silicon NPN) transistor with an hFE on my dmm of 175 and checked it for leakage and measured gain.  there was no leakage but the gain turned out to be just 124, which is a pretty big error.

so then i tried measuring the gain with different batteries, keeping the resistors in the tester unchanged and here's what i got:

battery voltage (V)	base current (uA)	Ic/Ib (gain)
8.67	3.9054	141.69
8.13	3.6622	139.28
5.81	3.2353	123.62
7.57	3.4099	135.54

if you graph these numbers, you get a pretty straight line.  if you extrapolate that line, it would take a base current of about 5.1uA to get 174.  which seems reasonable.  i guess my dmm could be using that amount of base current to read gain.  probably it's actually another value, because i am extrapolating so far outside my data and i have read that this relationship is not linear.

so what all this seems to say is that if you use R.G.'s transistor tester and you use different batteries for the PS at different times, then your gain measurements are probably not going to be comparable across times.  that is, when you sit down to build your fuzz face, you can use the tester to rank order your transistors but a gain of 100 this time may not be the same gain transistor as a 100 next time.

or here's another implication: the gain measured by small bear for those Ge trannies you just got is unlikely to be the gain that you measure.  useful to know. 

have i got this right?

(edit:  i changed the thread title to get a better description)

[Sir H C]

Hfe is also called "Beta" which is current gain.  Current gain is a byproduct of a bipolar transistor, something that is there but not necessarily wanted.  Bipolar transistors are voltage controlled current sources.  The problem with beta is that it is current and temperature sensitive.  Change the temp or the current and you change beta.  Also a problem with what you see is called the early voltage.  Transistors will have different current output depending upon the voltage on the collector.  Change the voltage, you change the current.  All these make for it to be very hard to measure beta or most other parameters unless you have a fixture that keeps all pertinent nodes at known voltages.

[gaussmarkov]

thanks for this reply.

Also a problem with what you see is called the early voltage.  Transistors will have different current output depending upon the voltage on the collector.  Change the voltage, you change the current.

o.k.  next time i am at my bench i will put a trimmer in place of the 2.22M resistor i have on the base and vary the resistance rather than the battery voltage.  it will be interesting to see how much gain variation that leaves.

if the differences in measurements were inside a 10% tolerance, i would be happy to follow one of R.G.s rules: Engineering judgement almost always uses the rule that says "10% is almost always insignificant."  but i am seeing bigger differences.

i looked at my DMM's manual.  there is no information about its hfe measurement, like what base current is applied or voltages.  is there a general standard that DMM manufacturers use?  i have not been able to find one.  if there isn't a widely accepted standard test specification, how do designers deal with the differences in the hfe measurement for different DMMs?

concerning Ge xtrs, is leakage current also sensitive to voltage?  how should we interpret someone's report (e.g., Rangemaster with AC128's - REPORT or Why my ToneBender MKII doesen't work?!!!) about the hfe of their transistors?  maybe "hfe=100" means "60<hfe<140 in 95 out of 100 posts on this forum?" 

[Sir H C]

Leakage is definitely linked to the collector voltage. You can look at the base-collector junction as a reverse biased diode.  As the voltage goes up across this, the leakage will go up.  If you look at the spec sheet for most bipolar transistors, Hfe/Beta is often given some vary wide range, say 50 to 200 with an average around 100.  Beta is just not a parameter that most foundaries are really trying to control too well.  There are many other paramters they are more concerned with, and beta just comes along for the ride.  So long as it is not too low (some PNP transistors in old silicon processes had betas down in the 10's), the manufacturer doesn't really care.

[gaussmarkov]

Leakage is definitely linked to the collector voltage. You can look at the base-collector junction as a reverse biased diode.  As the voltage goes up across this, the leakage will go up.

o.k.  good to know.

If you look at the spec sheet for most bipolar transistors, Hfe/Beta is often given some vary wide range, say 50 to 200 with an average around 100.

yes, i have seen this ... and the comments on the forum about measuring beta for different transistors with the same part number.  i thought that this range referred to the variation across individual transistors with the same part number.  i would like to understand ... so i am just checking ... we have been talking about the variation in beta for an individual transistor, right?

Beta is just not a parameter that most foundaries are really trying to control too well.  There are many other paramters they are more concerned with, and beta just comes along for the ride.  So long as it is not too low (some PNP transistors in old silicon processes had betas down in the 10's), the manufacturer doesn't really care.

i now understand this.  thanks!     and i also see a lot of focus on beta/hfe in our discussions here on the forum.  so i have the impression that even though manufacturers don't care, and (many) designers make there circuits robust to beta/hfe variation across individual transistors, the diy effects community does care.  particularly when it comes to a fuzz face circuit.  i am thinking particularly of the whole piggybacking development.

i followed up with some reading about the "Early effect," using Sir H C's reference to "Early voltage."  i could only find one reference to this on the forum.  does anyone think that this is an interesting way to manipulate bjt hfe for a fuzz face?  i am guessing that the answer is "no."   if so, then i'd love to know why.

cheers, gm

[Sir H C]

Well beta shouldn't change that dramatically for a particular transistor unless you either overstress it or run it at wild temperatures (-40c to 125c).  For the manufacturer they are content if the value is within that wide range, they don't want to have to guarantee better as they might have to toss otherwise good parts.

I keep hearing all the questions about Hfe in regards to pedal gain and all that.  Where beta comes in is first input impedance.  The lower the hfe/beta the lower the input impedance of a fuzz face.  This input resistance makes a high-pass RC with the input capacitor so you lose more bass on a circuit with the input device being low hfe.  For the second spot, I am not sure how a lower hfe or higher one affects you.  Again it could well be that you have a lower overall gain by the reduction in input impedance (it is in parallel with the first transistor's collector resistance, so if the hfe is too low you can kill all the gain here).

Pretty much (IIRC) the input resistance for a transistor = beta * (Re + Re-external)

Re = emitter resistance of the transistor = .05*Ic

Re-external is the external resistance from the emitter to ground, if there is a bypass cap this can be ignored.  So if this input impedance is too low, you get low gain in the fuzz face.

Early voltage is not much of a concern in circuits with resistive loads.  The resistive load masks it for the most part.  It is when you get a circuit with a current source load that you really see this come into play and people start doing things like cascodes to quash it out.

Since the transistor in a fuzz face or other such circuit is swinging so wildly at the collector, you can not really use the early voltage to do much.  I would say you are better off trying to modulate the currents running through the circuit with LDRs as opposed to any other way to manipulate hfe.  (hfe is related to collector current).

[R.G.]

Well beta shouldn't change that dramatically for a particular transistor unless you either overstress it or run it at wild temperatures

Beta is much more consistent in today's transistors than it was when germaniums were mainstream. But it is very much a variable.

Look at http://www.fairchildsemi.com/ds/2N%2F2N5088.pdf, at the graph, "Typical Common Emitter Characteristics".

Hfe is shown at about 0.4 of its 1ma value for Ic = 100uA. In fact, the downward trend continues from there, it's just not shown. It rises to a peak of about 150% of its 1ma value at 3ma Ic, then nosedives. At 20 ma it is only 5% of the 1ma value.

In http://www.fairchildsemi.com/ds/2N/2N3904.pdf the graph for Hfe - small signal current gain - varies from about 70 at 100uA to about 180 at 10ma.

DC gain HFE is much more consistent for today's devices, but current gain does vary, and it's good to conservatively assume so.

Re = emitter resistance of the transistor = .05*Ic

Close - it's Rc = 25mV / Ie.

[gaussmarkov]

you beat me to it.    i was looking at the 3904 data sheet also.  what i noticed was that it said dc current gain hFE had a minimum value of 40 at Ic=0.1mA, Vce=1V which increases to a minimum of 70 at Ic=1mA, Vce=1V, and then a minimum of 100 at Ic=10mA, Vce=1V.  the minimum falls back to 30 by the time you reach Ic=100mA, Vce=1V.

[gaussmarkov]

in trying to reconcile my dmm reading with my breadboard reading on my 3904, i have seen the temperature sensitivity.  i thought that this was just a Ge thing, but holding the transistor between my fingers while it was in the dmm the hfe went from 161 up to 174.  and then blowing on the transistor brought it back down.  so i need to be more careful about temperature in getting a reconciliation.

is there also a difference in what is being measured?  since no signals, large or small, are involved in R.G.'s setup, that seems to make it pretty sure that i am measuring beta (dc current gain) with that.  my dmm manual says that it is reading hFE, which i take to be "the large signal gain when the transistor's input takes it from saturation to cutoff and back."  (and what i am probably really interested in is hfe.  )  so besides differences in operating conditions, does a difference in what's being measured also figure here?

with those things taken into account, it still seems like one should take gain measurements as relative values at a particular bench session, not absolute across possibly different temperatures and battery voltages and resistor values (i breadboard my version of R.G.'s tester), for making fuzz face choices.  that way the factors that cause absolute variation will not prevent mistaking a relatively low gain part for a relatively high gain part and vice versa.  have i got this right?

thanks for any help.  i feel like i am being a hair-splitter. 

[Sir H C]

RG-

My  curse in life is working with IC processes.  Often it is minimum geometry so I am always seeing the worst case for Beta variation.  I think you are right.  You still rule.  All I can say is that I have to find this wonderful paper by Barrie Gilbert (of the Gilbert cell) which complains about the overreliance on beta.

So what do you think of my theories on the result of Hfe/beta for the effects on the FF?

[gaussmarkov]

one other item:  i've been doing some homework on the "early effect" and i am wondering how good an explanation it is of what i am seeing.  i have the impression that this effect is important for "large" values of the base current.  i was able to reproduce the graph that i found often with descriptions of the effect.  i used LTSPICE.  here's my schem:



i have a current source on the base-emitter junction and a voltage source on the collector-emitter junction.  i varied both to get this graph:



you can see the early effect in the increasing slope as the calculations move to higher values of the base current, running from 1uA to 10uA.  then i altered the thing that was graphed to the beta:  Ic(Q1)/Ib(Q1); this is the graph that i got for the same data:



all of the graphs are on top of each other so that there is not much difference in beta as the early effect grows.  there is variation with Vce but that would seem to be something other than the early effect.  at least as far as the SPICE model is concerned.

thanks in advance for any help, gm

[Sir H C]

Spice models for bipolar transistors are quite good.  MOSFETs still have some issues and that is why they are continuously making new models but the good old Gummel-Poon (spelling is probably off here) works great for bipolars.

I forget the mechanism for the early effect, but the way to view it is if you took all the flat parts of the curves for the devices and continued their lines to the left, all the lines would intersect at one point.  This is called the early voltage for the device and is related to the output impedance of the transistor.  Higher the early voltage, the higher the output impedance.

A 10:1 range for base current is not really that great, 100:1 is where you start to see variances in beta really come to the forefront.

[gaussmarkov]

thanks again, Sir H C!

Spice models for bipolar transistors are quite good.  MOSFETs still have some issues and that is why they are continuously making new models but the good old Gummel-Poon (spelling is probably off here) works great for bipolars.

great!  it's hard for the beginner to know what to trust in a Spice model.  some time ago i tried biasing a Si transistor fuzz face with Spice and failed to get confirmation of common resistor values for that circuit.  so i have become a little suspicioius.   i should probably go back and try again.

I forget the mechanism for the early effect, but the way to view it is if you took all the flat parts of the curves for the devices and continued their lines to the left, all the lines would intersect at one point.  This is called the early voltage for the device and is related to the output impedance of the transistor.  Higher the early voltage, the higher the output impedance.

yes, in the first graph, as i understand it, if you extrapolate the flat part of those lines to the left they are supposed to intersect at a common point on the horizontal (x) axis.  so that graph shows the early effect in the increasing slopes of the flat parts.   

if beta were constant, then all of those lines would be horizontal and equally spaced apart, because each successuve line corresponds to a 1uA increment in base current.  because i chose 1uA, you can read the spice model value for initial beta right off that graph: 300.  every line is .3mA above or below its neighbor.  .3mA/1uA=300.

and that's also what you see in my second graph, where beta starts at 300 in the "active region" after saturation.

A 10:1 range for base current is not really that great, 100:1 is where you start to see variances in beta really come to the forefront.

that's what i was beginning to suspect.  thanks for noting this, because it also suggests that the effects that i have been seeing are not due to the early effect.  my base current varied from 3.2uA to 3.9uA and yet i measured changes in beta from 123 to 141.  the graphs show that for values like mine, beta is not changing for a change in base current alone.  all of the change is due to the change Vce.  there is an unseen, corresponding change in Vbe, as well.  but that is caused by the change in Vce in these graphs.

hmmm.   i suppose that in my experiment, the early effect would also show up as an independent influence on Vbe as i varied the voltage in the battery.  so perhaps i should take that approximate conclusion back.   

[Sir H C]

Odd that Spice didn't give the right values.  I have used a lot of simulators over the years but not LTSpice and gotten good correlation with real circuits (I design analog ICs for a day job so I live/die by the simulator).

I think the change in beta you are seeing is due to the early effect.  You can see how having different battery voltages can give a 10% variation on the device in your simulation, and some devices are much worse than the one you simmed with.  Those can easily give you your seen variation.

[gaussmarkov]

thanks, Sir H C.  as you expected, Spice does display the same effect as i saw.  i simulated my setup for R.G.'s transistor tester with this:



and here's a plot of beta (i'm getting better at making thes graphs  ):



this shows the same order of magnitude as my measurements, though not as big an effect.  i haven't figured out a way to separate the early effect from whatever else is going on.  so i guess i will just leave it at that for now, accepting your judgement.  i appreciate your help with this.

best wishes, gm

[mac]

Maybe this helps:
http://www.diystompboxes.com/smfforum/index.php?topic=45481.0

BTW, my DDM uses 10uA of base current and VCE=2.8V.


mac

[gaussmarkov]

hey mac,

i have seen your analysis of R.G.'s test.  i have to admit that i got lost when i tried to follow it.    but having worked on this some since, maybe i should give it another go.

and, because you were able to report your dmm's hfe measurement, i went back again to the manual for mine and found another section with the info.  guess what?  same as yours:  10uA at the base and a Vce equal to 2.8V.  maybe it's a standard specification?

thanks for following up, gm

[Sir H C]

I thought it would be cool to use a current source to force the base current and read the collector current with a diode connected device.  Then you can keep the collector voltage steady and easily vary the base current using different biasing resistors.  Never gotten around to it though.

[gaussmarkov]

Sir H C:  this is a timely reply for me ... now i have learned how to make a constant current source.  i only recently understood the concept as an idealized component in Spice sims.  this yielded the sims above.  in my haste, i had not even considered the possibility of making such a thing.     it's also cool because it connects with another topic that was floating through some of my reading: compensating for transistor sensitivity to temperature variation.  and another:  the difference between "ordinary" diodes and zener diodes.

so thanks for the idea ... and the LEDs it lit.    --gm

[Sir H C]

To do one using real transistors you want a lot of emitter degeneration.  Maybe a volt.  Make the resistors match very well and you will get a very accurate mirror.