Tektronix P6201 FET Probe repair

[ronbryan]

I have a Tektronix P6201 FET probe that needs repair. These probes can be quite useful for looking at IF amplifiers as the low input capacitance gives much less detuning than a normal x10 probe.

This one was found to work properly on the 1kHz scope calibrator, but produced no output with a 10MHz sine-wave fed in. The manual showed that DC and LF amplification was done separately from the AC path and the two sections were combined in the output amplifier, so the AC amplifier wasthe likely suspect.

Measurements in the probe body that housed the AC amplifier showed that the +7V and -10V power supplies were present and that the first emitter follower Q120 was cut off, with its base nearly at +7V, rather than around 0V, so it seems probable that the FET Q100 had failed short circuit. The FET is inacessible until the pcb is removed from the probe body, so I decided to leave that until a new FET could be located.

The active devices in the probe have very small packages, about 2mm diameter, with tape leads, of a type unknown to me or Google. A later version of the probe has a SOT-23 packaged surface mount MMBF4416 FET fitted in the Q100 position according to the parts list and these seem to be available on eBay at a reasonable price.

Has anyone managed to successfully fit the MMBF4416 in place of the original 2mm round device, or can this only be done with a revised pcb?

I imagine that the FET is fairly vulnerable to over-voltage, so I am hoping that someone on the forum has encountered this problem and fixed it.

Ron

[Radio Wrangler]

Even the '4416 FET has quite a wide range of Idss, so it may be that these probes used selected parts.

Those small pill packages were quite common in the day but needed a hole in the board and careful fitting. Is there a variant of a possible device in the smaller SC70 package... like a little SOT 23.

This looks fixable with a bit of hunting and care.

Oh, and that capacitive divider trick for waggling the drain voltage to null some of the gate capacitance is pure evil. Wish I'd thought of it!

David

[Craig Sawyers]

Interesting. Farnell says that MMBF4416 is "no longer manufactured"

But OnSemi list it as an active part https://www.onsemi.com/products/disc...jfets/mmbf4416

Craig

[ronbryan]

Here is a picture of the underside of the probe pcb, which I have removed from the probe body. The FET is the small white round device at the sharp end. The circuit is slightly different to the published one, as the FET bias pot R152 has been replaced with a fixed 2k resistor, probably selected for the correct source voltage (0.025V to 0.050V).

The FET is shorted as expected, but also some of the bias resistors in the BJT part of the circuit have changed value significantly, particularly the 160 ohm R170 (270 ohms) and the 510 ohm R160 (460 ohms). The 10M gate resistor R140 has also gone up to 12M. I had fun trying to test the FET out of circuit as it kept flipping about - eventually I pushed it into a bit of Blu-Tak to anchor it!

The resistors are tiny Allan-Bradley BB 1/8W 5% hot moulded carbon types about 3.5mm long. I've ordered some old stock replacements, but 510 ohms is still outstanding, so if anyone has a 510 ohm BB resistor I would be pleased to make an offer for it.

It remains to be seen if the BB hot moulded carbon replacements I have ordered are still in tolerance. Alternative 1/8W metal film resistors have a spiral cut to trim the resistance value. I would appreciate opinions as to whether this extra inductance would make a significant difference to the probe performance.

I have made up series-parallel trees of Electrosil C3 1/8W resistors using values that were to hand, to replace the faulty ones so I can test the circuit, with a temporary 2N3819 FET in the front end. The probe pcb seems to work with a 10MHz input, so not much more can be done until the parts arrive. It looks as if a MMBF4416 SOT-23 package will fit on the pcb, provided that source and drain are exchanged and the drain leadout is extended a little, so the challenge will be keeping it in the right place while I solder it in.

Ron

[Radio Wrangler]

Sounds like the fet may be better upside down to get the legs in a better order.

Metal film resistors are probably OK, but if parts are difficult you can fit flat SMT ones they're not spiral cut.

David

[ronbryan]

David

Thanks for the comment about film resistors. It looks as if I shall need to find at least one film type. I'll look into the possibility of surface mount resistors, but they may need leads added to stretch the connections to fit in a pcb designed for wire-ended resistors.

I've just had a closer look at the device pinouts shown in Section 4.4 of the manual. From note b) at the bottom of the page, referring to the Q100b FET sketch, it says that "Leads are used as shown (source and drain are reversed from the vendor's designation)". So it seems that Tek were happy with reversing the channel connections when they chose to use the substitute FET the right way up. That is quite a result for me as it takes away the uncertainty about the lead swap.

Ron

[orbanp1]

Hi Ron,

Regarding swapping of terminals of a JFET.
It is not how Tek uses a JFET, but it is the structure of the FET that determines if the S and D terminals can be swapped.
The FETs where the terminals can be interchanged do have symmetrical structure.
The 4416 FET datasheets do not mention that this is the case.
The J310 JFET datasheet e.g. specifically mentions that the Source and Drain are interchangeable.
I am not suggesting that the J310 is a suitable replacement, I did not check, just bringing it up as an example of a symmetrical JFET.

Regards, Peter

[G0HZU_JMR]

If it helps I dug out a reel tape of MMBF4416A (purchased at Farnell) and stuck one in a microwave test fixture that gets used with a lab VNA to measure 2 port s-parameters of SMD parts (using bias tees).

I have two types of reversible fixture here and because the results were so close I measured it with both fixtures. The VNA power has to be set very low to prevent non-linear effects in the device. This has to be a genuine small signal measurement.

The idea is to configure the device in common gate and then swap the precision fixture around and take s2p data to 2GHz when biased at 5.000Vds and 1.000mA Id. This effectively swaps the role of drain and source.

The first surprise was that the dc operating point was exactly the same when the device was swapped around when in common (grounded) gate mode. The bias current didn't change even by 1uA. The operating point stayed within 1mV or so after the drain and source swap.

See below for a plot of the s2p data after it was used to design a common gate RF amp at 125MHz. This is a highly critical test to see how symmetric the part was. I have to confess there is a slight fudge factor cheat in the 'corrected' plot as to get perfect overlay I had to add about 0.15pF at the drain on the 'regular' pinout version. This happened exactly the same with the other fixture so I suspect that the device is quite symmetric but there is probably some stray capacitance in the die somewhere that just spoils the symmetry ever so slightly. See the second (uncorrected plot to see what I mean) It only takes just over 0.1pF to shift the amplifier response sideways like this.

I do think this difference is inside the device but it is a tiny difference in my opinion. I have a lot of experience of taking device models like this so I think the results are realistic.

[G0HZU_JMR]

In case the 'exactly equal' bias current was some freaky miracle (when drain and source are swapped) I tried the drain-source swap around test at 1mA, 2mA and 3mA bias currents and in all cases the operating point stays within about 1uA when swapped around. I even tried a second device from the same reel tape and it was exactly the same.

Are there any experts in semiconductor physics that can explain this one to me as I was not expecting to see dc bias symmetry this good!

The VNA tests do indicate a subtle lack of symmetry in the die, there seems to be less shunt capacitance at the drain part of the die compared to the source. I think the drain and source sit over the top of the gate so maybe there is a tiny bit less drain area compared to the source. However, this difference is tiny. I tried the VNA test on another MMBF4416A sample and it was the same. There seems to be a tiny bit less capacitance at the drain but only about 0.12pF - 0.15pF.

[Julesomega]

That's very impressive engineering, Jeremy - I was unaware of this property. Now Ron just needs such a FET in a SO37 package!

[G0HZU_JMR]

Thanks! The VNA I used was the 4 port Agilent E5071 and I used an Agilent N4431-60006B Ecal calibration module along with the test fixture so the s-parameter models I produce should be representative of the real device.

I think Ron is going to try the SOT-23 MMBF4416 with drain and source swapped. Sadly I don't have any SOT-23 MMBF4416 but have a lot of SOT-23 MMBF4416A parts here. I'd expect the 4416 to be very similar to the 4416A that I tested on the VNA as they are both process 50 devices.

It would be interesting to check out the MMBFJ310 on the VNA for symmetry as it is described as symmetric. However, this device is from process 92 so I'm not sure it could be used in this probe. I've got some on SMD tape so I could try this one day.

[Radio Wrangler]

There were some transceiver designs where symmetrical J FETs were exploited to make reversible grounded-gate amplifiers to make a reversible IF` strip using diode ring mixers and crystal filters for SSB.

David

[radioman]

Originally Posted by G0HZU_JMR

The VNA tests do indicate a subtle lack of symmetry in the die, there seems to be less shunt capacitance at the drain part of the die compared to the source. I think the drain and source sit over the top of the gate so maybe there is a tiny bit less drain area compared to the source.

This page from my Siliconix databook seems to confirm your hypothesis !

Andy

[G0HZU_JMR]

Interesting to see that die image, thanks Andy! I guess one way to try to prove it would be to make a very simple Cgate RF amp at 100MHz with a MMBF4416(A) and then swap the SOT-23 part around for source and drain. If there is a tiny bit less drain capacitance in the die then the tuned frequency will be 2-3% higher with the conventional pinouts but the gain and bandwidth should be the same. There would need to be time allowed for cooling after soldering though. It might be quicker and easier to press fit the part as it would only be a very brief investigation.

[G0HZU_JMR]

Of course what I should have done was just use Genesys to process the s-parameter data from the VNA to display the input shunt capacitance at the source and also at the drain and then look for a difference when the drain and source are swapped! I've now done this and the plot is below.

As you can see in the plot the 'normal' pinout has the lowest drain capacitance by just over 0.1pF. However, as expected, it is the other way around at the source. Here, the shunt capacitance is slightly higher at the source with the normal pinouts. It also shows that the difference is just over 0.1pF.

The two differences aren't quite the same (they don't cancel) but this may be because of differing dc voltage at the two pins. There may be a slight scaling effect because the drain runs 5V higher than the source in these tests and this may change the internal physics a bit. I'm just guessing though. These results look really good to me and they seem to indicate the subtle difference in internal capacitance could be caused by the differences in the die layout in Andy's datasheet diagram a few posts back.

Just a reminder, these plots are of data taken with the 2N4416A in common/grounded gate at 5Vds and 1mA Id. The plots compare the effect of swapping the drain and source pins around to see if the device is fully symmetric.

[Radio Wrangler]

I suspect that the low frequency symmetry found is evidence of good, uniform channel width and depth along its length (JFETs are long-channel devices) along with a uniform gate area symmetrically disposed between the ends of the channel.

With this structure, there is no inherent drain or source. It is only when the device is biassed-up that one end gets used as a drain and the other as a source. This is the basis of the successful use of these transistors as variable resistors handling only AC signals.

I suspect that the mounting of the die in the 4416 is asymmetric. One end of the channel is connected to a leadframe via a bondwire, as is the gate, while the other end goes via the bulk silicon to the supporting part of the leadframe. This latter could see better heat sinking, and will see different stray-C and also connection-L

There was a databook by Natsemi giving die characteristics of their various JFET 'process' parts, which ought to show parameters without packaging effects. Many Siliconix parts were the same dice in different packages.

David

[Craig Sawyers]

I recall using unprotected gate Siliconix MOSFETs (yes I know - not the subject of this thread) in the early 80's. These were TO18 package, and came with a little spring clip shorting the leads together. You only took that off once the device was safely soldered in.

JFET's are not immune to static damage either, so it is always a good idea to handle them until soldered in on an antistatic mat. Same with things like microwave diodes too.

Craig

[ronbryan]

Thanks to all for their interest in this thread, to Jeremy for taking time to experiment with the impact of D-S reversal and to David for useful advice.

When searching for a datasheeet for another device, in fact the 2N4417, an N-channel FET, I came across a Solitron datasheet actually showing the silicon die for that part, which may be of interest. It shows that the gate is the substrate and the source and drain are the bonded out electrodes.

It is true that the attached datasheet may be incorrectly referenced as 2N4417, as the typical devices listed for that die include the 2N3823 among a list of others. It also says nothing about Process 50, so it may not be relevant to the MMBF4416 construction.

I received five MMBF4416 FETs yesterday afternoon and tried to start testing. Without a test fixture, it is tricky attaching fine wires to the SOT23 package as everything is very small. I have checked Idss on a couple of the devices and it was found to be 10mA, which is in the middle of the 5-15mA Idss range specified. It is however twice the current taken by the original package FET, meaning that the source resistor will have to be reduced to around 1k from the existing 2k to achieve a max 50mV source voltage w.r.t gnd. The dissipation of the source resistor then goes up to 100mW (1/8W component) and the FET to 50mW. I expect that the remaining three FETs will have the same value for Idss.

David did wonder in post #2 if the FETs would have been selected for Idss. I don't have enough FETs to be choosey about Idss values, but it would be useful to know if Jeremy's higher voltage 4416A FETs also have similar Idss to mine and if the operating performance at the higher 10mA drain current would be inferior to the original 5mA for any reason.

Ron

[G0HZU_JMR]

I can do some more tests this evening but the Idss for the 2N4416A I have here was measured at 5Vds and it was 8.36mA. The Vp was -3.2V for <1uA Id.

The bias tee resistance used in the source was 2177 ohm.

I try and include as much info in my S2p files as possible in the filename as below and this can be quite useful if a more detailed analysis is required.

2N4416A(CG(5VDS(1MA(VP3200MV(VGS2170MV(IDSS8360UA. S2P

A quick play with an old excel design spreadsheet predicts that with 8.36mA Idss and Vp -3.2V the source resistor for 1mA bias should be 2093R so this 2N4416A JFET is working quite close to theory as I used 2177R for 1mA. See the screenshot below.

That datasheet for the die has a 2N5485 listed and this is a classic 'middle of the road for Idss' process 50 JFET.

[Radio Wrangler]

Bulk=gate is the ideal construction. That can be entirely symmetric though in modes other than grounded gate, ot leaves the can on the high-Z node.

I'd been thinking more in terms of JFET input opamps which have to have a separate substrate. I'd been looking into noise mechanisms and there's much there that single devices are blissfully unaffected by.

David