Automated Ge leakage and Hfe with data logging (or just insanity)

[boogietone]

The following is my insanity for the evening. Before getting started, it is agreed that this just may a bit of overkill. But, it amuses me...

I have a bunch of Ge PNPs to measure (see previous post) and would like to "automate" the data acquisition and provide data logging direct to Excel. My previous data collection effort on 100+ JFETs was all manual with breadboard, a 12 pole rotary switch, my DDM, and sheaf of paper and pen. Tedious describes that exercise quite well. This may or may not end up being an improvement.

The circuit is based on RGs Ge transistor tester. The idea is to socket the transistors and then push a button to trigger a microcontroller to make measurements using its analog inputs. The microcontroller's programming would cycle through the DUTs measuring first the leakage (base floating with no base current). Several sequential measurements a couple of seconds apart of each transistor are made and compared until the measured values are consistent for each DUT. This is to allow time for the transistor temperatures to settle. The relays are then closed by raising a digital output HIGH. Measurements are again made and the values dumped from the microcontroller's to the computer.

Some of the issues that took a bit to think through are, which microcontroller to used, how to close the relays with the available signal and power, available range of the microcontroller analog input, grounding, how to power each sub-circuit, separation of analog and digital signals, polarity of the circuit and sub-circuits (these are PNP transistors with positive ground), and others I have forgotten or did not think of.

The circuit brakes down into 4 sub-circuits: power, DUTs, DUT select, and DUT base current engage.

Still missing is a idle state (ready to make measurements) LED, a measurement complete LED, and a push button to trigger measurements, though both of those could be handled from a GUI from the computer. I also intend to measure temperature on the assembly at the same time as the leakage and Hfe, probably with a Ge diode calibrated to deg C.

Getting my feet wet with microcontrollers, I am currently working with an Arduino UNO. This is my first controller project and this one seemed to have a easy buy in even if it is not the cat's pajamas of microcontrollers. The USB/COM port communication and Serial library of the Arduino was relatively easy to set up. I did have the always present comma bug due to either a bad String class documentation or bad documentation reading comprehension. In any case, initial tests on the breadboard have shown that this will work using a simple test program without the MUX. The eventual programming aspect of this, including the data logging into Excel should not be a problem as one of my alter egos is C++ and Excel programming.

Any comments and questions on the above, or more importantly on the schematic below are welcome.

Thanks.



Click here for a larger, more readable image http://i.imgur.com/u2WpI.jpg.

[brett]

Hi
fantastic idea.
fyi - Japanese and Russian Ge transistors don't leak, which IMO makes them a better choice than everything else.
cheers

[PRR]

> how to close the relays with the available signal and power

Tip/suggestion: the relay driver will probably be N-type (NPN or N-MOS) so the input switches a ~~Volt above common (CPU ground and relay supply). Run this common-emitter(source). The relay coil does not have to go to ground. Run relay from V+ to collector(drain).

That way the CPU does not have to swing the full relay-coil voltage plus the transistor turn-on voltage. And the relay coil gets nearly all the relay supply.

I do not understand the 10uFd across the relay coils. It makes the N-MOS work a little harder. A BS170 will probably pull-up the 10uFd _very_ rapidly, HIGH peak current. For what purpose?

The 12K doesn't do much at all. The parallel relay coils will be a hundred ohms or so, the node will fall.

[PRR]

You are measuring hFE by applying a fixed base current and observing collector current. The collector current will vary a lot. Assuming hFE from 20 to 200, Ic can be 45uA or 450uA. "Usually" we "know" what collector current we want; then the question is what Base current is needed to get that Collector current. This could be extended to reads at 1mA, 0.1mA (the extremes of conventional Ge-pedal designs), and at zero mA (your Ico or leakage test). The mods to do that for one socket are not so bad; for eight, maybe more than you want to get into.

[Pyr0]

Hi
fantastic idea.
fyi - Japanese and Russian Ge transistors don't leak, which IMO makes them a better choice than everything else.
cheers

Are you sure ? I've come across plenty of leaky Russian Ge's.

[brett]

Hi
I really shouldn't have said Japanese and Russian devices don't leak. A more realistic comment would have been that Japanese and Russian manufacturers made high quality, low leakage Ge devices after other countries had gone for Si. By the mid 1970s most devices (Ge and Si) were manufactured quite well, and Ge leakage went very low (a few uA) and Si gain went high around this time. High power devices got cheap, too (e.g. 2N3055).
Of the huge amount of Russian devices produced from the mid 70s or in the 80s I have barely seen any that are dead or leak (as is normal for 1960s Brit, Euro, US devices). Though my experience is limited to a few manufacturers, only about 20 types (no power types) and a few thousand devices.
GT313, GT308, .... >99% good
cheers

[Pyr0]

Thanks for the detailed explanation Brett 

[boogietone]

The majority of the ones I am looking at are Russian NOS. Some are leaky some less so. In the circuits I have tried them in, FuzzFace, Rangemaster, and ToneBender, they all have sounded pretty good. Since I have a bunch, I want to have a better understanding of what makes a "good" one.

[boogietone]

Thanks for the feedback. Very helpful.

> how to close the relays with the available signal and power

Tip/suggestion: the relay driver will probably be N-type (NPN or N-MOS) so the input switches a ~~Volt above common (CPU ground and relay supply). Run this common-emitter(source). The relay coil does not have to go to ground. Run relay from V+ to collector(drain).

That way the CPU does not have to swing the full relay-coil voltage plus the transistor turn-on voltage. And the relay coil gets nearly all the relay supply. 

Took me a bit to understand this, but I think I got it and it solves a potential problem. As drawn, the voltage across the relay coils (i.e., the voltage at the BS170 source) measured on the breadboard somewhere in the range of 2-3 V. That it actually worked surprised me because the relays are 5V coils that are documented to require 3.8 V to operate. Figured I got lucky, or that it is because they are new, or the phase of the moon was right. In any case, it would be better to not depend on it. I will try them in the drain of the BS170. The LED indicator can then be the only component in the source leg and may not need the 220 current limiting resistor.

I do not understand the 10uFd across the relay coils. It makes the N-MOS work a little harder. A BS170 will probably pull-up the 10uFd _very_ rapidly, HIGH peak current. For what purpose?

The 12K doesn't do much at all. The parallel relay coils will be a hundred ohms or so, the node will fall.

I think that my idea with the capacitor was that it would soften the charging/discharging of the coils. Don't remember where I got that idea. The resistor does not belong. I had it on the schematic before I breadboarded the circuit. Neither the cap nor the resistor was used on the breadboard. I will remove both of these.

[digi2t]

I may have the hfe/datalogging part figured out;



Wavetek 235. It's what I use on my bench.

Tests transistors, and has an optically isolated RS232 output for datalogging to a PC. Just doesn't account for leakage though.

2 outta 3 ain't bad, right? 

[boogietone]

You are measuring hFE by applying a fixed base current and observing collector current. The collector current will vary a lot. Assuming hFE from 20 to 200, Ic can be 45uA or 450uA. "Usually" we "know" what collector current we want; then the question is what Base current is needed to get that Collector current. This could be extended to reads at 1mA, 0.1mA (the extremes of conventional Ge-pedal designs), and at zero mA (your Ico or leakage test). The mods to do that for one socket are not so bad; for eight, maybe more than you want to get into.

I agree that it would be nice to extend to make several measurements at different base currents. I am a bit confused by the units and values you show above. Do you mean 1mA(1000uA), 0.1mA (100uA), and 0mA (open base) of base current? RG and Smallbear's testers provide approximately 4uA (0.04mA) and 9uA (0.09mA) of base current, respectively.

One option is to use a two multiplexers back to back to select the base one of the 8 transistors and then a base resistor to ground. My concern with this, and the reason I went with relays in the first place, is any leakage through the multiplexer to ground or between the parallel in/outs could cause some small amount of base current. I have no idea if this is a real issue or not. Thus, in the expediency of eliminating the issue altogether, I popped in relays. If this was not a concern the relays could be eliminated. If it is, they could still be used to open the bases for the "leakage" measurement.

Also, are there "better" resistor values to use for either the collector or base? I based these off of what RG describes on geofex. He chose those values to make the calculations easy when done manually with a DDM. This approach requires no such setup.  Smallbear's approach is to have no collector resistor just an ammeter (which has some resistance itself) at the collector and uses a 1 meg resistor to ground the base.

All thoughts appreciated.

Thanks.

[PRR]

> of base current?

No.

Lookk at transistor specs. They give specific collector current, then the hFE at that current. Look at useful curcuits. Say collector should be 4V, there's 5K to 9V, we know collector current must be 1mA.

So first get the collector current, then find the base current needed, then do the hFE calculation (Ic/Ib).

The brute-force way with your robot-controlled rig is to vary Ib until you get desired Ic. But that's awkward. You also observe time and temperature drifts, so you want to idle the transistor at the desired Ic even if the CPU is busy elsewhere.

Here's something (drawn for NPN Si):



This plan will nail the Vce at 5V (or whatever) and Ie (nearly Ic unless hFE is super-low) at 1mA within a few percent. The opamp output reads Ib at the rate of 1V/10uA (10V=100uA and hFE=10; anything less is a dubious part) (0.1V=1uA and hFE is 1,000, we don't expect any hFE this high).

If the plan is to idle eight transistors, this does need eight opamps (2 quads) which is a lot of complication. (It also needs bipolar supply or some re-thinking).

Ah... maybe.... ground the base with 10K. Without opamp it will still idle very close to 5V 1mA. Use the MUX to switch one opamp to desired transistor. The opamp will take-over nearly all the base current (~~0.2uA error). One opamp for any number of parts.

For other currents: if "9.4K" (should be 9.7K for Ge; use 10K) is changed to 97K (100K good-enuff) the emitter current is 0.1mA (100uA). This does need 8 contacts (or a cleverer trick). Change the opamp's NFB to 1Meg, 1V per uA, to handle the lower Ib expected at lower Ie.

Opamp must be FET or "precision DC" (Ib under 1uA, Vos a couple mV).