Measuring J201s with a new toy - Are Tayda's SMD J201s awful?

[schrectacular]

Hi all,

I got a GM328A component tester yesterday and I've been making some measurements. I had previously bought some SMD J201s from Tayda and I grabbed a few that I had handy and measured them. I got these results for the four that I had mounted and handy:

Id	Vg
.38mA	.27V
.33mA	.23V
.37mA	.26V
.41mA	.28V

Reading the datasheet, it appears:

• Id should range from .2mA to 1.0mA
• min Vg should be .3V and can go as high as 1.5V

I think I'm reading the datasheet correctly. If I am, then it seems like these measurements are very low, even outside the bounds for Vg. I've searched around and other threads show much higher numbers, nothing even close to what I'm reading. So, are these "bad" or just different? What should I infer from their usage besides, say, plugging my numbers in if I'm building Fetzer valve?

[Kevin Mitchell]

Have you tried other J201s or another tester?

The odd results are likely due to limitations of your GM328 tester. Noted in my recent GE tester thread, an ATMEGA 328 or similar MCU has a rather non-linear 10 bit ADC within a 5v reference - giving a max resolution of 255. You will not get accurate results like the ones shown in the link you've shared where they are using a PEAK tester. I see they also note that the tayda transistors are close to the datasheet specs.

[schrectacular]

No, unfortunately those are the only J201s and only tester I have.

I did test some MPF102s (not from Tayda, from some dodgy ebay seller):

Id	Vg
.67mA	.47V
.51mA	.36V

And some 2n5458s (from Tayda):

Id	Vg
2.0mA	1.4V
1.8mA	1.3V
2.1mA	1.5V
2.2mA	1.6V

Which are also low. So in general this device should be ok for testing Hfe on transistors, but not for JFETs? Actually I'll just go search for that thread and see the discussion there first. Thanks for the tip.

[Kevin Mitchell]

It was more of a brief observation rather than a discussion. Some folks had only mentioned the 328 testers that are on the web (ebay, amazon). I ended up using a 16bit ADC module to get around the internal limitations of the MCU. Still have to draw a final design.

Just figured I'd let you know about my experience with the atmega 328's internal ADCs. Certainly not meant to be a precision tool.

Link on the quote;

While I could get the ATTINY85 to function for this purpose, as I played with it I had learned how non-linear the ADC behaves which is not ideal. But for basic purposes and for the transistor tester it works OK - depending on what grade of precision you're looking for.

Though I only mentioned the ATTINY85 in the quote, the same goes for the ATMEGA328 and also the MEGA2560. Besides the lack of complete linearity, being 10-bit it can only scale 5/255 = ~19.6mv per bit. Not to mention how the basic ADC functions rely on the 5 volt supply as a reference voltage which in most cases isn't exactly 5 volts, causing further error.

[PRR]

"Vg" HAS to have a current specified.

What is that current on your FET datasheet?

What is that current on your tester?

Conceptually Vp is "off". But what is zero current? nanoAmps? picoAmps? femtoAmps? And actually something funny happens at real-real-real-real-low current. Instead we can plot the graph from medium to low current and extrapolate. But in mass-production testing even low-low currents are a hassle. Gotta use the best plastics (even ceramic/glass) in the jigs and clean them frequently. Often they just pick a "low" round number for testing Vp.

The guys(?) who program these PIC testers know so much code they don't have enough headroom to know device physics. They guess a test, it gives numbers for some device they found, done.

[PRR]

https://www.mouser.com/datasheet/2/676/jfet-j201-j202-interfet.r00-1649170.pdf

[anotherjim]

What may make an MCU ADC linearity seem bad, is when it's measuring from a high source impedance. When an ADC conversion is called it holds the sample on an internal capacitance before starting the successive approximation. If the conditions are changed for the DUT, then immediately taking a reading is risking the sample voltage being held before the cap has had a chance to follow the changed condition. It's also a big problem if you use multiple ADC inputs and don't give the sampling cap time to react to a new input source. You'd think the designers would allow for this, but maybe what suits a BJT test doesn't suit the JFET?

Pretty sure all Atmel MCU's have an external Aref pin for the ADC. Even the Tiny85...

...although you lose a pin so I don't suppose it gets used for an Aref often.
The Mega328 has a dedicated Aref pin. IIRC, if you use the internal reference, it appears on the Aref pin and you can hang a cap on it to clean it up a bit. Arduino probably defaults to internal 5v reference. Search for an ADC library that gets you full control.

[PRR]

Gentlemen(etc):

The 10nA test condition on J201 Vgs(off) is similar to 30Meg to 150Meg tester impedance.

Can these 6-buck PIC products do many-Megs? Even the one assembled right after lunch? Or on a hot and rainy day?

This is more than enough to turn 0.3V into 0.2V.

[Rob Strand]

Perhaps the numbers produced by the unit are not Vgs_off and IDSS at all.   

The GM328A came up once before,
https://www.diystompboxes.com/smfforum/index.php?topic=123788.msg1171826#msg1171826

It looks like its a single point on the JFET curve.    The ID and VGS values are not IDSS and Vgs_on (VP).

The only way to really know what the tester numbers mean is to perform manual measurements and compare them.   

If you suspect the tester is just showing a VGS, ID point on the JFET characteristic the best test is to bias the JFET to the current ID measured on the tester, then measure VGS.

Another way is to use the JFET equation,

    ID  = IDSS ( 1 - VGS / |VP|)^2     ; all positive quantities.

IDSS and VP are measured manually.

To see if the tester is simply showing a VGS, ID point on the JFET characteristic:
Plug VGS from the tester into  the equation. 
Compare calculated ID and ID measured on tester.

This method is likely to have more error than the direct bias experiment test.

[bushidov]

Though I only mentioned the ATTINY85 in the quote, the same goes for the ATMEGA328 and also the MEGA2560. Besides the lack of complete linearity, being 10-bit it can only scale 5/255 = ~19.6mv per bit. Not to mention how the basic ADC functions rely on the 5 volt supply as a reference voltage which in most cases isn't exactly 5 volts, causing further error.

I thought 8-bit is 5/255 = ~19.6mV per bit. (2 ^ 8 = 256 - 1 = 255). So, shouldn't 10-bit be 5/1023 = ~4.9mV per bit (2 ^ 10 = 1024 - 1 = 1023)?

[Rob Strand]

Here's the proof the tester  displays a VGS and IDS point corresponding to a 680 ohm source resistor.

In all cases the Vgs measurement / Id measurement produces a fixed number
and that number is equal to the source resistor in the test circuit.


JFET        Vgs meas [V]  Id meas [mA] Vgs/Id
J201        0.27          0.38         711
J201        0.23          0.33         697
J201        0.26          0.37         703
J201        0.28          0.41         683
MPF102      0.47          0.67         701
MPF102      0.36          0.51         706
2N5458      1.4           2.0           700
2N5458      1.3           1.8          722
2N5458      1.5           2.1          714
2N5458      1.6           2.2          727


So the conclusion is the displayed values on the GM328A component tester are a point on the VGS, ID curve for the JFET and are not the VP and IDSS  JFET parameters.


Here's one version of the schematic,
https://www.mikrocontroller.net/attachment/234468/GM328_V1-0_sch.png

You can see 680 ohm resistors.
Add 20 ohms for the micro gate impedance to get about 700 ohms.


Using the values produced by the GM328A and assuming the J201's are more or less the same
we can guestimate  *possible* values for the JFETs to be any of these values,

VP[V]     IDSS [mA]
0.75      0.86
1.0        0.675
1.5        0.54

However, there's almost no useful information here.
We need another test point.   The GM328A is only one point.

If you did something like this,
- take a JFET measurement
- add a 700 ohm or 1k resistor in parallel with the gate a source
- take another JFET measurement
  (Assuming the unit still recognizes a JFET and doesn't barf.
   If it barfs it might be possible to add resistor divider to the gate.)
- Use the two sets of numbers to calculate VP and IDSS.
  They probably won't be very accurate estimates.

[schrectacular]

Wow amazing info here, thanks for all the replies. It's late here, I'll run the tests above tomorrow and see what happens. Glad it was a mildly interesting topic at least!

[PRR]

Thanks.

Also the highest working voltage seems to be the 5V, which is 1/4 of the J201 test voltage. While JFETs are broadly insensitive to Vds, the 5V end is way low for higher pinch devices.

But as you say, the main point is that they test with a 680Ω resistor which gives a one dimensional answer to a two dimensional question. (Interestingly it DOES say what it would do in a 9V pedal with a 680 source resistor.)

[Rob Strand]

I wonder if a better second point test would be with a 680 to 1k (or whatever) resistor in series with the source.    In other words there is an air connection between the real source pin and the added series resistor.  The other end of the added resistor goes into the socket.

That's going to produce better numbers for getting VP and it's much less likely to cause the tester to barf-out from detecting something weird.

It might be a little easier to interpret the readings as well since adding parts may change the scaling of the measurements and we will have to "undo" that scaling to get the true current measurement.

[schrectacular]

Ok got my setup for the retests, and found another j201.

Strangely if I put the component in D,S,G the tester correctly labels the pins, but if I invert and put it G,S,D the tester labels the pins G,D,S. I guess this is because it's an _N_ JFET?

I used the breadboard to place a 677.8 Ohm resistor in series with the source as Rob suggested. Here are the results:

Id (direct)	Vg (direct)	Id (resistor)	Vg (resistor)
.41mA	.29V	.28mA	.20V
.38mA	.27V	.27mA	.19V
.37mA	.26V	.26mA	.18V
.35mA	.24V	.25mA	.17V
.33mA	.23V	.23mA	.16V

All the pedals are open now, so I can do additional tests. But if I'm understanding correctly we are expecting (and see) a linear relationship, so two points is enough.

[Rob Strand]

All the pedals are open now, so I can do additional tests. But if I'm understanding correctly we are expecting (and see) a linear relationship, so two points is enough.

It doesn't end-up linear because the source resistance voltage drop sets the gate voltage and it all gets complicated by the JFET characteristic.

So I plugged in your measurements from the added source resistor scheme and got estimates for the VP and IDSS JFET parameters.

The parameters look quite reasonable.  I suppose you could verify the results by measuring them manually.




I've put the details for all the calculations in the screenshot.

The VGS2 calculation works out the true gate-source voltage.   When we add the source resistor the gate-source voltage shown on the GM328A is the voltage across the internal 680 ohms (or 700 ohms).  So we need to adjust the VGS2 voltage based on the added resistor.


Another check would be to say add a larger series resistor and see if the numbers come out the same.    Sometimes you find that the numbers are close but not exact.

Large Rs usually gives a more accurate VP estimate but there may be a point where something breaks on the tester if the resistor is too large.

[schrectacular]

Amazing, thank you! You even did the algebra for me!

I suppose it was a lot to ask of a sub-$15 device to calculate so much, but also nice that at least I have a method to get the numbers ... if I ever end up needing them. I mainly bought it to be able to test transistor Hfe, I have a dream of finding some old junk radios and raiding them for germanium transistors to use in a fuzz pedal. Added bonus to not have to try to decipher color bands on blue resistors... is that orange or brown? Navy or black? It's hard to tell.

[anotherjim]

If the tester identifies the pinout by trial and error, it probably can't tell S from D. Most JFETs are symmetrical around G.

[Rob Strand]

Amazing, thank you! You even did the algebra for me!

I suppose it was a lot to ask of a sub-$15 device to calculate so much, but also nice that at least I have a method to get the numbers ... if I ever end up needing them. I mainly bought it to be able to test transistor Hfe, I have a dream of finding some old junk radios and raiding them for germanium transistors to use in a fuzz pedal. Added bonus to not have to try to decipher color bands on blue resistors... is that orange or brown? Navy or black? It's hard to tell.

No worries.

Well, $15 is a crazy cheap price but they could do it no problems.    There's a small difference between testing something is OK and extracting the parameters.   The test circuit they use will produce valid numbers for most JFETs  but trying to extract parameters on all JFETs might not work because the required voltage or current gets too high.   I saw another unit which tries to do the "correct" test but falls back onto another test mode if it runs out of stick.

Strangely if I put the component in D,S,G the tester correctly labels the pins, but if I invert and put it G,S,D the tester labels the pins G,D,S. I guess this is because it's an _N_ JFET?

I forgot to answer that before (it was 3:30 am).   Most JFETs are symmetrical so you can swap D and S.  There's no way the unit can tell the difference between D and S.   In reality the capacitance between the gate and one of D or S is larger than the other so for high frequency work there's a "correct" way to use the JFET pins. [jim just answered anyway.]

[Rob Strand]

I did a test to see how "tough" the estimates for VP and IDSS are using this method.

I took the readings of the last JFET

Id (direct) Vg (direct) Id (resistor) Vg (resistor)
0.33mA   0.23V   0.23mA   0.16V

then randomly bobbled the values in the last digit of each measurement by adding a random number between -0.01 and +0.01 for 50 samples.

VP        IDSS [mA]
0.7501   0.6864              Values predicted from original measurement

VP        IDSS [mA]
0.7524   0.6954              Average of estimates the 50 samples.
0.6102   0.5734              minimum of estimates over 50 samples
0.9220   0.8957              maximum of estimates over 50 samples

(Min VP and Min IDSS do not necessarily occur at the same time.  They are the independent Min values over the 50 samples.)

We can see the estimated VP and IDSS can have the potential to have a large error.
Small deviations in the measurements can produce large changes in the results.

The reason that happens is the test points are too close together, mainly ID.
That could be improved by making the added source resistor as large as possible. 
How large depends on what you can get away with on the unit.   Nonetheless every ounce helps.

As far as ball-parking the general VP and IDSS values it's probably OK with the added 680 ohm.
I means it's a lot better than (incorrectly) interpreting the raw measurements as VP and IDSS.