Testing unknown transistors for max frequency response

[G0HZU_JMR]

Update: I took some s2p data of a BC547B in common emitter at various bias current settings and various Vce and compared a fairly formal attempt at measuring Ft (based on s2p data) against the little test jig with the very same transistor in the jig.

The results were pretty good but once the Ft max approached 200MHz I saw creeping error and this was because the PSU decoupling caps making up C2 in my jig had a subtle resonance towards 300MHz. It's really important to try and make the impedance at the base (10R) appear to be the same as at the collector. However, the collector needs a decoupling cap in the supply and this introduces extra inductance and you only need a few nH extra to cause an imbalance at UHF. Putting several caps in parallel can cause resonances. I ended up using a decent 1nF chip cap from ATC for the UHF tests. You also need to use decent chip resistors for the 10R resistors if you want to measure Ft up towards 300MHz. Otherwise the AC voltage will appear artificially high and this will cause the Ft to appear higher than it really is.

Once I got rid of the resonance, the results were really good and were very close to the data posted below. So it really does require a tight and resonance free layout with decent quality components with low parasitics. Even a slight bit of extra inductance here can cause false voltage gain.

I used the VNA to take s2p data of a real BC547B at 0.7mA, 1mA, 3mA, 5mA and 10mA at 3V, 5V and 10V. In every case the little jig did a good job of predicting Ft. I'm not sure how well it will work at higher currents or with Ge transistors but for small signal Si BJTs at lowish collector currents it seems to work well enough.

See below for a word doc showing my initial attempts at processing the s2p data. The green line in the graph is current gain on a log scale. This should drop at 6dB/octave ( = 20dB/decade) once past the LF knee point. It does look like the rolloff rate on the graph is 20dB/decade and so this makes me feel the data is good. Ft is where the green line crosses the zero line on the gain axis. There seems to be good agreement between the S2p model and the Genesys model for the BC547B and also when the Genesys model is configured as my little test jig with an effective 10R load at the collector and base.

[Argus25]

To help answer the question for germanium RF transistors I set up a test circuit and made some measurements(images attached) I don't have a VNA but I have a well calibrated SG503 levelled sine wave generator and a 2465B scope.

So I set up a simple circuit with the transistor in common emitter mode with close to 1mA collector current and tested a few transistors from my stock.

I found, that for this circuit/test setup at least, there appears to be a magic number that allows a transistor's fT value to be predicted (if a transistor of known fT is tested and used as a reference). This configuration is probably good to 250 MHz, over that resonances might be an issue.

For example , using a fixed drive voltage and testing an AF185 I found that when the drive frequency was equal to the manufacturer's fT of 80MHz, the output voltage was 8.52 dB down (4.5mV pp compared to 12mV pp in the lower frequency range). So then I tested a range of other transistors and measured the frequencies at which they also had the output fall by -8.52 dB and lo and behold, these frequency values at that level of signal amplitude drop pretty closely matched the published transistor's specified fT.

Of note, as can be seen in the attached table, a 2N2084 is very similar to the AF117 in the high frequency department. The AF185 is a tad better and the AF178 better again at the HF end. I'm not 100% sure that I have the correct value for the manufacturer specified fT for the 2N2084.

The OC45 is poor and only suited to MW broadcast band radios, though the are good performers in 455kHz IF amplifiers if neutralised well. Still the fT was predicted fairly well by the test circuit. There is an error in the table for the OC45, the estimated fT was 3.4 MHz and the manufacturer fT is between 3 and 12MHz (its a bit late here)

For a radio like the EC-10 that goes to 30 MHz it is easy to see why the AF178 makes a good choice.

My Hacker Sovereign is populated with 2N2084, making near identical replacements for whisker affected AF11x

I also documented the frequency value which the output for this test circuit drops by -3dB which is interesting to see.

It would be very easy with this test setup to determine if an unmarked germanium RF transistor was any chop for either a domestic MW radio or a wide coverage coms radio like the EC10.

[G0HZU_JMR]

I can't compare results because I don't have access to any Ge BJTs to try and I also can't find much in the way of a decent datasheet for Ge parts which is a bit frustrating.

Normally, the Ft is quoted at a particular operating point and it will vary depending (mainly) on Ic but Vce affects it too. See below for the Ft curve for the BC546/7 at 5Vce when Ic is varied across 1mA to 100mA. I'm not sure if this is for the BC546 or the BC547 (or both) but you can see how Ft varies with Ic.

This curve agrees very closely with my measured data at 5Vce and if I use my test jig I can explore Ft at higher collector currents and I get the same steep collapse in Ft once Ic approaches 80mA. I think this is caused by high level injection effects in the transistor and it effectively slows the transistor down. It's also going to get very hot!

[Argus25]

Jeremy,

I'll test a BC547 tonight (or maybe a BC557 to suit the jig) and see what happens compared to the tested Germaniums will be interesting.

Hugo.

[G0HZU_JMR]

Here's the BC546 (BC547?) datasheet curve for Ft along with my results for the jig and the VNA for the same BC547B transistor at 5Vce. I only took s2p data across 0.7mA to 10mA with the VNA (blue dots) but I tried testing from 100uA to 80mA with the test jig (red dots). I think in theory there should be a straight line for increasing Ic for very low Ic but it develops the drooping arc as Ic increases and high level effects take over and it eventually curves enough to start falling with increasing Ic.

I found that the Ft falls like a stone after 80mA and seems to fall off a bit sooner than the datasheet. But maybe this is because this graph is labelled for the BC546 and there could be a difference in Ft response between them? The BC546 and BC547 both have the same Ft spec of 300MHz typical at 5Vce and 10mA. But this is based on a test at 100MHz that predicts Ft to be at 300MHz. But it doesn't seem to agree with the 5Vce 10mA Ft graph shown later in the datasheet. This shows Ft only gets to about 230MHz here and it never gets to 300MHz. This is for the BC546/7 and not the B versions though.

[Argus25]

Originally Posted by G0HZU_JMR

Here's the BC546 (BC547?) datasheet curve for Ft along with my results for the jig and the VNA for the same BC547B transistor at 5Vce.

That looks great.

For Germanium RF transistors, in radio RF and IF stages, most of which are grounded emitter stages, it seems that they run fairly low collector currents in the range of 0.5 to 1.5mA. So I think the value for fT at about 1mA is the important value to know for the particular specimen.

My jig of course doesn't plot the fT at all, but it does give a rough way of estimating it and suggesting if say some unmarked Ge transistor might be any use for RF work and the frequency range over which they might be useful.

The 3dB roll off point might even be a more useful value than the fT, because obviously, anywhere below that frequency (or close enough) the transistor would obviously be just fine in the application, assuming the collector current was close enough in use in a radio compared to the jig. It is 90MHz for the AF178 and 25MHz for the AF117 in my jig. The AF185 is 33MHz, and like a 2N2084 is very close to the AF117 , but as I mentioned, for some reason the AF185 became very hard to get and I'm not sure why. But there are still some out there.

[Argus25]

I tried a BC556 in my jig. Because it is a silicon with a higher Vbe, the collector current was only about 0.6 of a mA, so I had to add a 68 k parallel resistor with the 27k bias resistor to get the current back up to about 1mA (This is the problem with dropping silicons in place of germaniums in vintage radios, the bias resistors ideally need changing).

I have added the data to the table attached. One issue is that the fT for typical silicons is stated at 10mA collector current, but for germaniums it is normally around 1mA.

Using the method described the estimated fT for the BC556 is only close to 100MHz (1mA collector current), when the manufacturer stated fT is 200MHz at 10mA. So it appears to agree with Jeremy's graphs for the fT at different collector currents on post #25.

The -3dB down frequency for the BC556 is close to that of the AF117, so that probably explains the success some have had in replacing the AF117 with pnp signal silicon transistors, even without changing the bias resistors.

Finally, I considered the world's first transistor radio, the Regency TR-1. These used 2N94 transistors or similar, which are an NPN germanium. So out of interest I reversed the polarity of the jig and tested a NOS 2N94. As can be seen from the table, these are not quite as good as an OC45 (that came later in history) but almost.

[G0HZU_JMR]

Taking it to extremes, I downloaded the s parameter data for a very fast BJT, the BFP420F

https://www.infineon.com/dgdl/Infine...142763e30d0755

This is shown on the datasheet as having a typical Ft of 25GHz. The datasheet graph for Ft is also given below. It shows Ft should be just over 25GHz at 3V and 30mA.

I downloaded the manufacturer's own s2p data for the BFP420F at 3V 30mA and simulated it in Genesys.

The datasheet indicates that they measure beta at 2GHz and extrapolate to the Ft from here. See below for my Genesys simulation of beta based on their s2p data file at 3V and 30mA.

The green trace on a log scale shows a fairly good 6dB/octave rolloff rate at 2GHz. The blue graph shows a plot of beta and this measures 13.19 at 2001MHz on the marker on my graph.

If we assume 6dB/octave rolloff rate then the extrapolation to Ft is as simple as multiplying 2001 by 13.19 which gives 26400MHz or 26.4GHz.

The marker at 1000MHz predicts Ft at 1GHz*25.862 = 25.86GHz.

This is probably the correct method to measure the device and predict Ft but that is just a guess.

[G0HZU_JMR]

I wonder how well it would work if my jig was built in a tight SMD layout with 5R loads and AD8307 log amps at each test port. This would allow beta to be measured across quite a range allowing extrapolation to Ft at several GHz from beta measurements up to maybe 500MHz. So something like a BFS17 or maybe even a (5GHz Ft) BFR91 or BFR93 could be measured and checked. This may seem a worthless task at first (because we have the datasheet!) but it might be a good way to check parts to see if they are fakes or dodgy clones of the real deal.

Also, I downloaded the s2p data for the BFP420F at 3V and 5mA. The datasheet graph given in the post above shows that Ft should be 14GHz here. See below for the simulation based on their S2p data. The Ft works out to be 14.35GHz. Pretty close!

[Argus25]

Originally Posted by G0HZU_JMR

This may seem a worthless task at first (because we have the datasheet!) but it might be a good way to check parts to see if they are fakes or dodgy clones of the real deal.

I think is a really good idea for new parts, at least for a sample of some from the supplier or test maybe a random sample of them.If I was in manufacturing I would do that.

There appear to be so many fakes out there now. Luckily many vintage transistors, especially Germaniums, have not been faked yet, but some suppliers will still re-label another part and pass it off that way after you have paid for it.

For my projects I now always try to buy obviously NOS semiconductors/IC's with vintage date codes that are genuine old parts to avoid the fakes. For example currently there are Signetics MC1496's available from some 1970's stock, so I get those instead of the new ones, just in case.

[Argus25]

Jeremy,

I'm very impressed with your skill and abilities to work at GHz frequencies and envious of the equipment you have to simulate and test it.

The best I have is the 400MHz 2465B scope and a Tek SG504 leveled generator that goes to over 900Mhz. Surprisingly, the 2465B can just visualize and barely sync lock a 900MHz sine wave and at that point of course the displayed amplitude is pretty meaningless. There are no digital 400MHz bandwidth scopes that can do this that I know of.

So for the most part the highest frequencies I can work with easily are below 400 to 500MHz. When the signals are in the GHz range my test gear is about as useful as taking a piano accordion to a Deer hunt.

[Diabolical Artificer]

Watch out for your scope's band width too which can give you misleading results. I was testing a sig gen a while back on my Tek SC502. It showed attenuation at the extremes IE less than 300hz and over 10mhz. I spent ages trying to find a fault that didn't exist. There was no attenuation on another scope.

Andy.

[Argus25]

Yes, it always pays to check it. I calibrated my 2465B's myself as I'm very fussy. These are handy in that the scope also has an inbuilt 50R termination and I use original Tek 50R precision cables too. And as a final check, I feed them with the SG504 leveled sine wave generator (it has a leveling head right on its output) to make sure the bandwidth is correct after the calibration procedure.

[MotorBikeLes]

Lucky you, I can only "boast" of a Tek 191. A set of those SG50x units would be a positive step towards the 21st century, though not there yet.
Les.

[G0HZU_JMR]

Originally Posted by Argus25

I'm very impressed with your skill and abilities to work at GHz frequencies and envious of the equipment you have to simulate and test it.

Thanks
Most of my experience is up in the GHz region so I tend to have lots of microwave test gear here.

One classic technique that gets used with s parameters is to plot GMAX using an RF Simulator. This is the maximum gain that the device can give (when optimally matched) at any frequency whilst maintaining a stability (K) factor >1. Obviously, it is desirable to try and keep K >1.

A while back I did an amplifier design demo for someone up at 6GHz and this used the BPF650F BJT. This has an Ft up at 42GHz (!) and the design aim was to get as much stable gain as possible up at about 6GHz and still deliver a P1dB of about +15dBm. This device only runs from about 3.5V so it needs to run at a highish current in this case.

The first plot below is for the GMAX for the device. This is the purple trace. The red trace below it is the unmatched S21 response. This includes a typical via hole arrangement with several vias in a very thin PCB material. This shows GMAX is only 10.9dB at 6GHz with the vias. The vias will reduce the GMAX from maybe 11.5dB down to 10.9dB in this case. So straight away I know that no matter what I do I won't get more than this much gain if I want the K factor to be >1.

I also had to design an efficient L match at 6GHz at the output using a printed inductor and a decent SMD cap. This will have some loss and there will be losses in the input match and also the PCB and the connectors. But I managed to get about 9.5dB gain on the VNA and still managed to get a P1dB of +15dBm.The PCB material was Rogers 4003 and it's only 0.02" thick.
I milled a PCB and built and tested the 6GHz amp and see below for some old pictures of it. This also shows a Sonnet simulation of the PCB layout. This shows slightly less gain but I think my via inductance is less on the real PCB because of the solder around the vias and I also think the caps had less loss than my models.

Whilst these techniques to maximise gain up at 6GHz seem really exotic (and may appear off topic?) they can be applied down at a few MHz as long as the s parameters are known for the device at the chosen operating point. The manufacturers of the BPF650F produce s parameter data at 0.5V increments and each increment also has s2p data for several collector current settings. So it's easy to find a suitable s2p file.

If I manage to get hold of some RF Ge transistors I could measure s parameter data for them up to a few hundred MHz at various Vce and Ic settings. This info doesn't seem to exist anywhere on the web so it may be useful?

[Argus25]

Originally Posted by MotorBikeLes

Lucky you, I can only "boast" of a Tek 191. A set of those SG50x units would be a positive step towards the 21st century, though not there yet.

Les,

Over the years I collected a lot of different Tek Plug Ins. The main ones required for scope calibration are the PG506 amplitude calibrator & tunnel diode pulser and the TG501 time marker generator. The SG503 & 504 are really just luxuries good for checking the bandwidth later. But they are also great lab generators. I turned one of my spare SG503's into a pantry TX (photo attached) in a reversible manner. In these Tek controlled the emitter current of the oscillator transistor to level the amplitude, so I simply broke the feedback loop and injected audio.

When I started to get into Tek plug ins, the first thing I manufactured was a plug in that electrically bought the edge connector to the front panel and I have extension cables with edge plugs & sockets to power up plug ins for servicing. I also put voltage monitor circuits in that unit and LED's to show if the various power supply levels from the mainframe are correct. (see picture in post 22 on this thread, that's what all the LED's are for).

At one stage these plug in were fairly cheap as they were being abandoned in favor of modern test gear, but it appears now many have recognized the value of them and the prices seem to be going up. They are actually an absolute requirement to calibrate a 2465B scope, because the firmware protocols in the 2465B expect the TG501 and PG506 generators are being used and its very awkward with other generators unless they are set up to emulate them exactly.

[Argus25]

Originally Posted by G0HZU_JMR

If I manage to get hold of some RF Ge transistors I could measure s parameter data for them up to a few hundred MHz at various Vce and Ic settings. This info doesn't seem to exist anywhere on the web so it may be useful?

Jeremy,

Wow ! 6GHz, its like you are working in another world at those frequencies , that little amplifier board looks beautiful too.

It would be interesting to test germanium RF transistors with your gear. There are some types that might even impress you.

The AF239 for example is claimed to have an fT of 650 MHz. An AF186 an fT of 820 MHz.

It is said in the data sheet that the AF186 could be used as pre-amps and mixer osc up to 900MHz. A quoted power gain of 8.5dB at 860MHz and 2mA collector current. I think these mainly got used in vintage UHF TV tuners. Clearly these represent the peak of germanium RF transistor development before silicons took over. I think the AF186 is the or one of the highest fT germanium RF transistors. Someone might know of a better one.

[G0HZU_JMR]

I'll try and find some of those Ge devices and have a play. I can show that it's possible to use a transistor as an oscillator way beyond its Ft frequency. Ft is when beta falls to 1 but the device can still deliver power gain way beyond Ft.

I took one of my S2p files for the BC547B and the best one to use here is the 10V 10mA operating point. This has an Ft of about 300MHz on the datasheet and I got a bit less than 300MHz when tested with the VNA. By loading this into Genesys and grounding the base and looking into the collector the device produces loads of negative resistance up into the VHF region and this is tpically in series with a few pF. By adding a 1pF cap across the CE legs it is possible to optimise things so the negative resistance extends up beyond 500MHz and it appears in series with about 4.8pF. So it needs about 21nH to be resonant at 500MHz.

It's then just a case of making a 'real' circuit that biases the BC547B at 10V and 10mA and adding the required 21nH inductance in the collector (for resonance) at 500MHz. See the plot below. I also built it in dead bug style on an old scruffy bit of PCB. It's scruffy and ugly but it oscillates where expected and by pressing on the collector inductor with an insulated tool it's possible to adjust the inductance and see the frequency climb to about 600MHz and then die. A quick look at the negative resistance plot in Genesys shows that the resistance at the collector ceases to be negative at 600MHz. So no surprise it dies at 600MHz. Note that this circuit isn't a great design for various reasons and it is really just a quick and dirty (low parts count) circuit that demonstrates oscillation up towards 600MHz (2 x Ft) with the BC547B

This might not work with all manufacturers' versions of the BC547B and I used the Farnell part 2453790 made by ON Semi/Fairchild.

[Argus25]

Originally Posted by G0HZU_JMR

I'll try and find some of those Ge devices and have a play. I can show that it's possible to use a transistor as an oscillator way beyond its Ft frequency. Ft is when beta falls to 1 but the device can still deliver power gain way beyond Ft.

Jeremy,

That is very cool. I would never have guessed the humble BC547 would make a good oscillator at 500MHz. Also I like your 21nH inductance and its easy to see why with some circuit track designs one could accidentally create an oscillator.

On that basis the AF186 with a nearly 900MHz fT would probably be good for an oscillator at over 1GHz. So I guess germanium transistors really did come of age, before they were made obsolete by silicon devices.

One of the main things that led to germanium devices going obsolete, along with their additional leakage and temperature dependence issues and lower max operating temperatures, was the fact that germanium devices don't lend themselves well to integrated circuit fabrication.

I have always been intrigued by just what could be done with germanium transistors. As a thought experiment I set out to find out if it would have been possible for a company like Lucas to have used them in the early 1960's for dynamo controls, this is what I came up with:

http://nebula.wsimg.com/16f83c1c70ef...&alloworigin=1

Of course by the time alternators took over in the 1970's, most manufacturers for most electronics had abandoned germanium devices by then and gone to silicon devices. But it was sort of sad to see them go.

[Radio Wrangler]

Oscillators beyond device Ft are fairly well known if you're in the RF business.

Germanium is making a comeback, but in alliance with silicon as SiGe compound semiconductors.

Lucas was very tightly constrained by costs, more so than by technologies. The mechanical bits of alternators are cheaper to make than dynamos, so the changeover happened when semiconductors became robust enough. Beefy diodes came on stream at the same time as transistors/zeners became capable of handling the transients. Dynamos could have been controlled by germanium devices, but by then the industry was looking ahead to the alternator. A transistorised dynamo controller would have had normal development costs (both money and effort that could have gone to other projects) but the production lifetime would have been short. You only had to look at Lucas products to see that all decisions were made by accountants. Meanwhile in the world outside Lucas, we were fitting several recon RB340s per week.

David