Testing unknown transistors for max frequency response
 

Hi all,

I have quite a lot of vintage unmarked Ge transistors that I would like to find out their max operating frequency, I'd like to know whether I have anything suitable that would operate at FM I.F. so 10.7Mhz?

If I used a signal generator and scope with the circuit below, would I be right in thinking that I would see the amplified trace on the scope die once I exceed the max operating frequency of the device?

Cheers
Lee

Re: Testing unknown transistors for max frequency response
 

Yes-ish.

Bypass the emitter resistor and add a couple of base bias resistors to give around 1mA collector current and add dc blocker cap to input.

Emitter R 1k, base bias 12k down and 82k up, collector 3.9k and 10n caps should do nicely.....

Re: Testing unknown transistors for max frequency response
 

Look up 'Gain Bandwidth Product'.

The gain at higher frequencies depends on the transistor parameters and the circuit design.

If you use Herald1360s circuit you should be able to measure (a) the low frequency gain, (b) the frequency at which it drops to 30% of the low frequency gain, and (c) the frequency at which the gain is 1.

Re: Testing unknown transistors for max frequency response
 

Hello,

What is a 'scope die? :laugh2: Could't resist!

I used to have some AF139's somewhere. They were solid way into UHF...

Re: Testing unknown transistors for max frequency response
 

You can find a few marked transistors and check them against spec to prove your test rig.

Re: Testing unknown transistors for max frequency response
 

Hi Chaps,

This is all very good stuff and I think I'll knock up a test rig on the breadboard and give things a try, I have plenty of NOS marked transistors too, Ge & Si so will compare my tests with manufactures data.

Thanks again.

Re: Testing unknown transistors for max frequency response
 

Watch out for stray capacitance.

The collector-base capacitance of an AF117 is only 3pF. This influences the high-frequency performance of a transistor, as it provides negative feedback (and the apparent value is increased by Miller effect). If your test rig adds any more stray capacitance, then the results will be much poorer than you expect. Watch out for solderless prototype boards, as these will have a similar sort of capacitance between adjacent rows. Space the base and collector out by several rows to reduce the effect...

Re: Testing unknown transistors for max frequency response
 

Another dragon:

Watch out for scope probe capacitance. Your average not rubbish probe has a 12pF load capacitance. At more than a few MHz it's going to look like a pretty annoying resistor in the circuit across collector to ground. This will, with the collector load resistor affect the frequency response in a low-pass fashion. It's worth adding a follower on the output with a known good high fT transistor (2n5179 is a good one if you can get it to stop oscillating) to isolate the circuit from the probe. Also keep the probe ground very short - use the little spring clip if you have one.

Re: Testing unknown transistors for max frequency response
 

Yes, having a 3900 ohm resistor in the collector will limit the bandwidth once you try and probe the circuit with anything.

Ideally you would also want to measure the Ft of the transistor. I tried to think of a quick and dirty way to do this and the circuit for this is below. It is far from ideal but I think it might give reasonable results if built with a tight HF layout and decent components. At low frequencies the AC voltage at the base will be tiny, too small to measure but as the frequency increases the AC voltage at the base will begin to increase and I think the AC voltages at the base and collector will be the same at frequency Ft.

The best tool for measuring the voltage at the base and collector will be a sensitive RF probe with low capacitance. This could be a simple diode probe as long as it is sensitive enough to measure maybe 10mV pkpk at maybe 50-100MHz. A scope with a x10 scope probe will probably be (just) too deaf to see the signals. Alternatively a decent commercial RF millivoltmeter could be used here. Don't overdrive it with the sig gen or distortion will occur.


The Vce is set by the supply voltage V1 and the Ic is set by the negative voltage supply (V2) at the 1k emitter resistor. Because the emitter is fixed at -0.7V and the emitter resistor is 1k you need a -1.7V supply at V2 for 1mA Ic, -2.7V for 2mA Ic and -10.7V for 10mA Ic etc. So with this circuit you should be able to measure how Ft changes with Vce and Ic.

Hope this is OK, it's late and I'm a bit tired. I'll build and test this circuit tomorrow or over the weekend. I've posted it up in advance in case anyone can suggest improvements. It really is just a starting point for something a bit more elegant :)

Note that I don't have any design experience with Ge transistors so I don't know how much the 'leaky' nature of these devices will affect the results of the above test. But the circuit should be reasonable with silicon BJT devices?

Re: Testing unknown transistors for max frequency response
 

Note that a more formal test circuit for Ft would test it as a common emitter amp so maybe I'll try changing the circuit to something closer to the correct configuration. I'll have a play tomorrow.

Re: Testing unknown transistors for max frequency response
 

You could make a generalised test circuit to determine the transistion frequency.

However, since you already know the circuit you want to use the transistors in, a 10.7MHz IF, one quick way to solve your problem is to set up the transistor as a single 10.7MHz IF amplifier stage with a 10.7MHz IF transformer.

Connect the transformer secondary into a detector diode (low capacitance germanium or schottky type) feeding a 1Meg resistor and a parallel 0.01uF filter cap. THe diode gets around the capacitive loading issue.

Couple into the transistor's base circuit from a generator with a low value cap like 10 to 50pF. Adjust the 10.7MHz generator for about 300mV to 500mV output from the detector when the transformer is peaked on 10.7MHz

Measure the DC developed across the detector output on a high impedance meter or scope.

A single tuned stage like this can easily have 30dB or more gain , but it is very very sensitive to transistor parameters such as the Ft and hfe and collector to base feedback capacity, which, in conjunction with the relatively low input coupling capacitance creates a divider which will result in much lower signal outputs for transistors with high range C-B feedback capacitance, compared to a known good for the application "reference transistor"

Use a reference known good transistor for this 10.7MHz application , an AF127 would be fine.

You will soon find out by substitution if any of your unmarked transistors are suited to the task like a 10.7MHz IF. If any are like OC45's you'll get no output and if others , like say AF178 , have lower C-B feedback capacity you will find the output is higher and gets even better with a small tweak of the transformer tuning.

I have tested dozens of unmarked germaniums in similar test setups and I regret to report that many are poor even when the test jig is for a 455kHz IF.

So I suspect you won't find many that are much chop at 10.7MHz , but I could be wrong depending on what parts you have. But you will see right away if you build such a test jig for IF amplifier transistors, it's very easy to tell which ones are suited.

Of course if you set up a test IF stage with a 10.7MHz resonant circuit in the collector and the base, right away you could eliminate all the transistors with high C-B capacitance as the stage would be unstable and oscillate.

Re: Testing unknown transistors for max frequency response
 

There was a circuit in one of the RSGB publications (perhaps Pat Hawker's Tech Tips compilations?) for fT measuring. It was a single-transistor oscillator, with controlled (and 'calibrated') amounts of feedback.

I still have my instantiation of it, and it works reasonably well. But I've lost all traces of the actual source (even though I'm fairly certain I must still have the book - somewhere :-[). Does this ring a bell with anyone?

dave

Re: Testing unknown transistors for max frequency response
 

Yes, I may have misinterpreted the requirement. I assumed that Lee wanted to know the frequency response (good or bad) of all of the transistors. Otherwise it's best to just make a 10.7MHz test circuit. My little test circuit should work (for silicon BJTs at least) from a few MHz up to maybe 400MHz depending on layout and component quality.

I tried building my little test circuit today and it seems to give realistic results. I've only tried a jellybean silicon BJT (BC547B) but I tried testing it at 5Vce across collector current Ic ranging from 0.7mA through to 10mA. The Ft varied from about 85MHz at 0.7mA Ic to about 350MHz when Ic was up at 10mA. This seems about right?

It's also possible to predict Fmax from Ft if the device capacitance and base resistance is known.

To feed and probe my test circuit I used a Marconi RF sig gen and I used a Marconi 2380 1GHz RF probe feeding to an old HP 54540C 500MHz digital scope. However, for lower frequency Ge parts the 10R resistors could be replaced with 47R and a regular x10 scope probe with a Tek 465 scope could probably measure the Ft up to 50MHz or so. It should be able to display the RF waveform OK on the most sensitive setting on the 465 scope?

It's not an ideal test circuit for Ft but the general idea is as per the image below. This is a simple AC representation showing electron flow up from the emitter. When close to Ft the current sharing should be equal in the base path and the collector path and so the RF voltage at the collector and the base will be the same. This RF voltage can be measured with a decent RF probe.

Re: Testing unknown transistors for max frequency response
 

To really test the setup I have here I tried testing a 2N2369A transistor. This is from an old bag of samples of the ST (SGS Thompson) version of the 2N2369A. I have had these samples in my parts stash for many years. I think I used these for crystal oscillator prototyping at work back in the early 1990s and so these are genuine 2N2369A parts and not modern clones/fakes. To get a sensible result from the scope up at 700MHz I had to set it to repetitive sampling and I tried testing at 10mA Ic and 10V Vce. The test jig showed the Ft was 715MHz. Note that my old 500MHz scope doesn't have to be accurate in absolute terms as it is just looking for the same AC voltage at the base and collector when at Ft. The Marconi 2388 1GHz RF probe has an input capacitance of about 2pF so it should be OK here when probing across a 10R resistor.

This really is stretching my lashup test setup to its limits but it does seem to agree fairly well with the 2N2369A datasheet as it states 675MHz Ft at 10V Vce and 10mA Ic. However, I think it is better (as in more traditional) to measure the 'gain' at a lower frequency and extrapolate the Ft using equations. This would put less stress on the accuracy of my test setup as long as care is taken with the measurements and the maths. I've never tried this though. I always just look at the datasheet for this stuff :)

Re: Testing unknown transistors for max frequency response
 

I had not thought of testing the transistor's high frequency capability in what amounts to a grounded base circuit (10R base resistor is effectively grounded base) because it circumvents the Miller effect of the collector-base capacity because the voltage swing at the base is so small. But I guess it is the way it might be done to determine the Ft value.

I have never been sure how the manufacturer did the test for the Ft and just accepted the published value.Obviously in a grounded emitter circuit (from the AC perspective) the source impedance at the base will have a significant effect on the high frequency performance, the lower the impedance the better.

Most likely the test of a transistor for its high frequency capability should be in some sort of circuit that resembles the one it will actually be used in.

Of unmarked germanium transistors I have tested, most of them have probably been intended for audio work and were little use above 50 to 300kHz. So I have never trusted unmarked transistors to be of any use for RF work, but they could be I guess and worth testing them just in case they turned out to be RF types.

Apart from the AF178 which is now my preferred replacement for AF11x, there was another very good transistor in the same package that was used in some UK made radios (some car radios too), this was the AF185 suited to 12V radios. I collected some a few years back but oddly even then they had become incredibly difficult to find. I even got sent some fakes that I think we're re-labelled other devices. But the AF185 is a very good replacement for AF11x, that is is you can find any genuine ones.

Re: Testing unknown transistors for max frequency response
 

They would use something like a General Radio 1607A Transfer Function and Immitance Bridge, introduced in 1958. This could be fitted with one of several transistor adaptors for grounded base and grounded emitter. Worked from 25MHz to 1.5GHz. Needs practice to use this beast.

http://www.ietlabs.com/pdf/Manuals/G...20Function.pdf

Or a Boonton 250A or HP250B with transistor test adaptors for 9MHz to 250MHz. Similar vintage to the GR

http://bee.mif.pg.gda.pl/ciasteczkow...onton/250a.pdf
http://www.jamminpower.com/PDF/HP/HP%20250B.pdf

Also described here http://hpmemoryproject.org/wa_pages/wall_a_page_14.htm

For lower fT, simpler test jigs were appropriate - some of the Ge transistors were lucky to get to 1MHz.

Craig

Re: Testing unknown transistors for max frequency response
 

Thanks Craig. It would be good to have the HP 1GHz vector voltmeter ! I guess transistor manufacturers could afford these tools.

Re: Testing unknown transistors for max frequency response
 

I've been remembering how the HP 250B works, and how it is used to measure current gain beta, and hfe in transistors. I also misquoted the specs - it goes from 500kHz to 250MHz.

Transistor measurements using this are in issues of Boonton Radio's "The Notebook" http://www.hparchive.com/Boonton.htm

The relevant issues are 6, 19, 20, 26 and 27.

Re: Testing unknown transistors for max frequency response
 

It's really just a quick and dirty test jig that can indicate when the RF base current is about the same as the RF collector current. It compromises the AC 'short' at the collector with a 10R resistor and it excites the transistor at the emitter. It really is as crude as finding the frequency where the current leaving the base pin is the same as the current leaving the collector pin. This obviously isn't the correct/accepted method to measure Ft but I think it will give reasonable results even though it is quite unconventional. However, it's performance/accuracy might suffer badly with slow Ge transistors. I really have no design experience with old school Ge BJTs. Maybe I'll buy a few online somewhere and have a play.

The idea of my simple jig was to be able to quickly and crudely predict the Ft of small signal transistors with Vce of maybe 5 to 10V and Ic ranging from maybe 0.7mA to 10mA. The aim was to achieve this using just a few resistors around the transistor under test and a dual PSU. I think it should easily be able to tell apart a 2N3053 from a BC547B or a 2N2369A by looking at the results for Ft across these ranges of Vce and Ic. I'm not sure how much value this gives but some people might be interested to see how Ft varies with Vce and Ic. On a datasheet you usually just get a prediction of Ft at one operating point.

I did come up with a common emitter version but it became quite complex and lost the simplicity of the first circuit. Also, I'm not sure if the uncertainty introduced by the extra circuitry in the CE version would be greater than any improvement in accuracy wrt Ft.

I don't think the published Ft is really the true Ft anyway because I think the published Ft will be a value that allows the designer to better predict where the 6dB/octave gain slope is for a given operating point. I think it's reasonable to assume the 6dB/octave straight line rule will begin to suffer as true Ft is approached. I think this is why they extrapolate to a theoretical Ft from lower frequency measurements. Also, for today's super fast transistors it would obviously be challenging to directly measure an Ft of (say) 40GHz.

Re: Testing unknown transistors for max frequency response
 

I've got one of these old VVMs here but I also have the modern equivalent in the form of a 4 port VNA. If anyone is interested I could measure a (Farnell sourced Fairchild) BC547B and export s2p data for various Vce and Ic operating points.

It might be fun to convert these S parameters to H parameters and try and predict where Ft is and where the 6dB/octave slope is? I could post up the s2p data files here so people can have a play at home using tools like RFSIM99 or QUCS?

There are also various (free) GNU Octave tools available online for manipulating or converting s parameter data. Maybe this is all a bit OTT but some people on here might not have access to a decent VNA but they might still want to play with the data? If I buy some popular Ge RF transistors I could do the same with these.