2SK97 dual N-channel JFET spec measurement & substitute

[smurfer77]

Well the funny thing about me writing this thread is that I don't even own an amp that uses the 2SK97. I got interested in the topic because I was doing some measurements of the 2SK129A (another dual N-JFET) which appears in my Sansui AU-X11. Anyway, it looks like the 2SK97 is used in some Sansui, Sony and other units, but specs are unconfirmed and subs therefore not certain.

There are some specs over on a Stereonet thread, reposted from here, which seems to be by far the most detailed account of the 2SK97 specs. In fact, it looks like it was from a spec sheet. But nobody can find the spec sheet to confirm, and also those FET spec ranges are always big and it would be good to know typical values for choosing a sub.

From those threads we have:
2SK97 dual N-Channel Junction FET
in dual in line case.
Pins:
1 : Drain 2
2 : Gate 2
3 : Source 2
4 : Source 1
5 : Gate 1
6 : Drain 1

Absolute Maximum Ratings:
Vdgo = 30 V
Vsgo = 30 V
Id = 20 mA (1 FET)
Ig = 5 mA ( 1 FET)
Ptot = 210 mW (2 FETs)
Channel Temp. 100 grdC.
Storage Temp.: -50 to 120 grdC.

Characteristics:
Idss at Vds=10V, Igs=0 : min 0.9 mA, max 14.3 mA
Pinch off Voltage at Vds=10V, Id=30uA : min 0.18V, max 1.49 V
Forward Transfer Conductance at Vgs=0V, Vds=10V : min 6.3 mS
Reverse Transfer Capacitance (same conds.): typ 2.4 pF
Input noise voltage at Rg=10kOhm, f=1kHz:
typ 13 nV / sqareroot Hz

Vgs difference between 2 FETs at Id=1mA: max 70 mV

Classification:
Rank 1 : Idss = 0.9 to 5.5 mA
Rank 2 : Idss = 4.5 to 9.9 mA
Rank 3 : Idss = 8.1 to 14.3 mA

@Ronito6 sent me a 2SK97 to measure and here. It is marked "KC22" and I guess one of those numbers might refer to "rank 2" Idss. Ronito could confirm for us which amp this device came from for reference.

Ok, so first thing I did was confirm the pin-out. It does actually follow the pin-out noted in the specs posted above. In the pic here I probably have the chip upside down, but it doesn't matter - see the white bar printed on the chip, that is the end with the drains; gates are in the middle, and sources are on the other end. The unit passed quick diode check too. So I hooked up two DC supplies and off we went with some measurements.

I ran through with Vgs = 0 V and Vds = 10 V on both FETs and got very similar results, so the data below is only for one of the FETs, not both. There are a couple of weird 'kinks' in the data, which appeared for both FETs.

You might notice one apparently weird Vds choice of 13.3V. I put that in because a quick glance at the Sansui AU-717 manual shows about 13.3V across drain and source. Also you can see from the measured data that Idss is just over 8mA.

Note that at Vds 15V and above the temperature of the device starts to shoot up as the current initially way higher than 8mA, then it heats up and the negative temp dependence brings the current back down over time. At 15V the device went up to 35 C and at 20V it went up to 50C and I didn't push it any further even though specs previously listed suggest 30V is possible. If your 2SK129 is heat sinked, you are going to get a bit different results to what I have here, as evidenced by the effect of putting my thumb on the device to bring it up/down to 37C. FYI, ambient when taking measurements was about 20C

I took a few extra data points to get a smooth curve for current vs v_gs.

Below is the data for the above graphs, and you can see transconducance calculated and Vgs_off (= - V_pinch_off). Transconductance averages around 12, and you can see in the data and graph below it swings a lot depending on where you are operating - beware there is a bunch of uncertainty in those data points and higher level of uncertainty in the slope (since we have the errors from 4 data points for each value), so just take that as a rough guide. As a guide I would say the transconductance is in the 10-20 range typically.

I took some averages of the calculated V_gs_off from the usable part of the data (small values or current have large error) and get about -0.8V (i.e. V_p = 0.8V). V_p was 0.78 as measured by another technique (simply taking the voltage over gs without a Vgs voltage source (but with Vds in place) exploiting the internal large impedance of the DVM as a load). Let's just say V_p ~ 0.8V.

Summary:
- pins confirmed. Drains are next to the white bar printed on top (there is typical marking underneath too, which I would always check in case the top was printed on backwards).
1 : Drain 2
2 : Gate 2
3 : Source 2
4 : Source 1
5 : Gate 1
6 : Drain 1

- Also specs others have quoted mention 30V is okay, I wouldn't use this thing past 15V Vds.

- Matching between FETs was good enough it wasn't worth my showing both sets of data. But there was up to 10% difference. The 1%

- Idss from previous reports was said to be: Vds=10V, Igs=0 : min 0.9 mA, max 14.3 mA. I measured just over 8mA for both FETs. It seems likely that one of the "2"s in the "KC22" on my device do in fact indicate rank 2. Whether or my Idss is typical for rank 2 or hand-picked for this amp (Ronito6, it was a Sansui AU-717 this came from?) I couldn't say. So I think the classifications (Rank 1 : Idss = 0.9 to 5.5 mA; Rank 2 : Idss = 4.5 to 9.9 mA; Rank 3 : Idss = 8.1 to 14.3 mA) posted by others could also be pretty accurate.

- Pinch off voltage was previously quoted as (at Vds=10V, Id=30uA): min 0.18V, max 1.49 V. That is quite a large range. I measure about 0.8V for both FETs. (again, this could be well matched across the pair, but might not be typical if we measure other packages).

- Transconductance (at Vgs=0V, Vds=10V) was previously quoted as "min 6.3 mS". This is probably true too. I am calling the my device 10-20 mS transconductance - see graph, the max goes just over 20 mS.

So it looks to me like those specs quoted could be legit from a spec sheet originally, and we have filled in some gaps here, especially with regard to typical values.

In terms of a substitute, if you look over at another thread of mine you will find some number on currently available dual N-JFETs. In particular:
LSK389B: Idss ~7.8mA (Vds=10V,Vgs=0V), Gm ~ 21 (Vds=10V,Vgs=0V), Vp ~ 0.62 V (Vds=10v) It looks to be a good substitute. Make sure you get the "B" version/rank if your 2SK97 is rank 2. Rank 1 will likely be closer to the 389A and rank 3 could be either B or C version depending on the actualy Idss needed in your machine.

Keep in mind that FETs parameters vary from sample to sample, much more than BJTs. Matching within the monolithic pair will be good, but from package to package, it can get wild. I recommend ordering a few units, and measuring (or if you don't have that capability, at least you could swap out for another if it just doesn't sound quite right).

 

[dlucy]

Would the Atlas Peak DCA75's ability to trace FETs on the PC screen like this help you any?

 

[dlucy]

From their website, here is what the DCA75 and its software can do:

Curve Tracing - When connected to a PC using the supplied USB cable, a range of low current curve-tracing functions can be performed. Various graph types are available, with more to follow:

• Bipolar transistor output characteristics, IC vs VCE.
• Bipolar transistor gain characteristics, hFE vs VCE.
• Bipolar transistor gain characteristics, hFE vs IC.
• MOSFET and IGBT output function, ID vs VDS.
• MOSFET and IGBT transfer function, ID vs VGS.
• JFET output function, ID vs VDS.
• JFET transfer function, ID vs VGS.
• Voltage regulator, VOUT vs VIN.
• Voltage regulator, IQ vs VIN.
• PN junction I/V curves, forward and reverse options (for Zener diodes).

Curve tracing is performed using test parameters in the range of +/-12V, +/-12mA. All curve-tracing data can be instantly pasted into Excel© for further graphing and analysis. PC Software is included with the DCA Pro on a Peak USB memory stick. Software designed for Windows XP, Vista, 7, 8 and 10.​

 

[smurfer77]

Would the Atlas Peak DCA75's ability to trace FETs on the PC screen like this help you any?

View attachment 888791

Long rant below, but basically, doing it with 2 power supplies manually is superior to curve tracers in terms of accuracy and device safety. (well, assuming it's done properly )

I haven't used that specific model, but my experience with similar devices for curve tracing and component checkers/testers is that they are good only for a quick test of if a device is blown or not, but values are not to be relied upon for this type of work IMO. Those types of devices make nice smooth curves (note, the lack of smoothness in my curves in this example, is due mainly to quirks of the FET, not the technique) but the curves/values can really be off (not just a little bit) and sometimes they are just plain wrong, and don't supply high enough voltage or current to test devices properly. Also if you read around, you will see some cases where devices were destroyed by testers/tracers. I suppose this happens during the device identification stage, where inappropriate voltage could be place across junctions (?).

I do think such tools are invaluable timesavers and a must-own, especially for quickly checking if a part is working or not, but in this case where I want to fully trust the results and not have errors due to unknown tester limitations, it is IMO necessary to do it manually with lab power supplies. If you have 2 lab supplies there is not better way but to do it yourself manually, knowing exactly what is being done... except the fact it's a bit time consuming... but the DIY/hobbyist isn't against spending time to do things properly .

EDIT: maybe I should pick up one of the newer models like you mentioned and compare the result and see how it goes.

 

[dlucy]

Long rant below, but basically, doing it with 2 power supplies manually is superior to curve tracers in terms of accuracy and device safety. (well, assuming it's done properly )

I haven't used that specific model, but my experience with similar devices for curve tracing and component checkers/testers is that they are good only for a quick test of if a device is blown or not, but values are not to be relied upon for this type of work IMO. Those types of devices make nice smooth curves (note, the lack of smoothness in my curves in this example, is due mainly to quirks of the FET, not the technique) but the curves/values can really be off (not just a little bit) and sometimes they are just plain wrong, and don't supply high enough voltage or current to test devices properly. Also if you read around, you will see some cases where devices were destroyed by testers/tracers. I suppose this happens during the device identification stage, where inappropriate voltage could be place across junctions (?).

I do think such tools are invaluable timesavers and a must-own, especially for quickly checking if a part is working or not, but in this case where I want to fully trust the results and not have errors due to unknown tester limitations, it is IMO necessary to do it manually with lab power supplies. If you have 2 lab supplies there is not better way but to do it yourself manually, knowing exactly what is being done... except the fact it's a bit time consuming... but the DIY/hobbyist isn't against spending time to do things properly .

EDIT: maybe I should pick up one of the newer models like you mentioned and compare the result and see how it goes.

I have the DCA75 and a small set of new and used JFETs. Is there something I could test for you and forward that would help?

I'm actually just now in the process of replacing 2SK30's in a Pioneer SX-626. I've done what research I can and selected the 2N5486 as a replacement. I would love to have a method of verifying I've got a good replacement, operational performance wise, before I button up the restored SX-626.

 

[Alan0354]

This is just my 2cents, I don't have much experience with JFET as it's hard to find.

I assume this is for the input pair of the power amp ( LTP as long tail pair). For LTP, the important things I look for are:

1) low input capacitance, at least replace with similar input capacitance. High input capacitance causes oscillation by lowering the pole frequency and might eat the phase margin.
2) High enough Vdg so you don't burn the transistors.
3) Comparable offset between the two transistors. That is Vgs1-Vgs2 is comparable to the original one you replace.
4) Cgd should be comparable with the original one.
5) Transconductance is not that important as long as it's within +/-200%.

Don't worry about the curve tracer and all the detail. I designed plenty of transistor circuits, never ever once I use a curve tracer. What make you think you can do better than what the datasheet provided?

Almost all the schematics of amplifiers I saw ( I saw a lot), the voltage swing on the drain ( collector of bjt) of the LTP does not swing much, maybe 1V. Curve tracer is irrelevant to see what is the slope of the drain curve. Transconductance is not super important as long as it's withing reasonable range.

 

[smurfer77]

I have the DCA75 and a small set of new and used JFETs. Is there something I could test for you and forward that would help?

I'm actually just now in the process of replacing 2SK30's in a Pioneer SX-626. I've done what research I can and selected the 2N5486 as a replacement. I would love to have a method of verifying I've got a good replacement, operational performance wise, before I button up the restored SX-626.

Your curve tracer will probably do the trick.

This is just my 2cents, I don't have much experience with JFET as it's hard to find.

I assume this is for the input pair of the power amp ( LTP as long tail pair). For LTP, the important things I look for are:

1) low input capacitance, at least replace with similar input capacitance. High input capacitance causes oscillation by lowering the pole frequency and might eat the phase margin.
2) High enough Vdg so you don't burn the transistors.
3) Comparable offset between the two transistors. That is Vgs1-Vgs2 is comparable to the original one you replace.
4) Cgd should be comparable with the original one.
5) Transconductance is not that important as long as it's within +/-200%.

Don't worry about the curve tracer and all the detail. I designed plenty of transistor circuits, never ever once I use a curve tracer. What make you think you can do better than what the datasheet provided?

I can't say much about the capacitance as I didn't measure it, but I take your point about the phase. The original and replacement I'm talking about are monolithic fab; the offset is really small (that's why they are used instead of pairs of single packages).

I think the details are curve tracing are necessary in this particular case where we didn't have confirmed specs for the transistor being discussed (2sk97 (2sk30 might be in a similar situation?)), based on which we could choose a sub, or design circuits. And when you work with JFETs you get in the habit of curve tracing or at least measuring Idss and Vp as they can be different by 200% from device to device, or even more actually.

 

[Alan0354]

I don't even think Idss or Vp is that important as long as it's reasonable.

I expect it is a monolithic match transistor, they still have different offset, Until I see the schematic, I don't know how critical is the offset. More amps have DC gain of 1 using AC coupled feedback path, then a few mV offset is not important. BUT if the feedback is DC, with gain of 20( a lot of amp I've seen), even if you have 5mV offset, output is 100mV and that's high. So depend on the schematic, something you need to look out.

That's the reason I don't design amp with FET input LTP. I thought of it a few times and went back to BJT. FETs are expensive and hard to find.

Regarding to transconductance, do you have source resistors? That is resistor at the source of the two transistors. If so, transconductance is not that important. If not, you just take the average of the curve tracer, then buy one withing that range.

 

[smurfer77]

I don't even think Idss or Vp is that important as long as it's reasonable.

I expect it is a monolithic match transistor, they still have different offset, Until I see the schematic, I don't know how critical is the offset. More amps have DC gain of 1 using AC coupled feedback path, then a few mV offset is not important. BUT if the feedback is DC, with gain of 20( a lot of amp I've seen), even if you have 5mV offset, output is 100mV and that's high. So depend on the schematic, something you need to look out.

That's the reason I don't design amp with FET input LTP. I thought of it a few times and went back to BJT. FETs are expensive and hard to find.

Regarding to transconductance, do you have source resistors? That is resistor at the source of the two transistors. If so, transconductance is not that important. If not, you just take the average of the curve tracer, then buy one withing that range.

I don't actually own an amp that uses this JFET, so I can't speak about the specific circuits, resistors and so on in the applications. About the offset, with these monolithic devices it's typically <1mA on spec sheets and in practice often more like <<1mA. The biggest difference I saw on the pair I have is about 0.5mA. when the throttle is nearly fully open, but down at reasonable levels, it is usually <0.1mA offset.

 

[Alan0354]

Well, if this is a theoretical discussion, if I were to design using FET LTP, I would design so the parameter variation is not as important. I would use cascode to reduce the Vdg requirement to get more choices. This also get rid of the effect of the Cgd and Cd. This is the main thing. I would use current mirror load to increase gain of the IPS stage, then use source resistor like 100ohm to get rid of the variation of transconductance. Then I don't worry about all these. This also help in offset problem. If you use FET, you can have higher tail current so more voltage drop across the source resistor to further reduce the effect of offset and transconductance variation. I would make sure the DC gain is 1, together with source degenerate resistor, I can even use single FET that I can have a lot more choices. I would just hand match and use a copper tape to tie them together to get good temp matching.

I just design a BJT LTP amp with single transistors, I did all these and I measure the final offset, the worst is 3.5mV, most are below 1mV.

You don't design at the mercy of the transistors.

 

[Ronito6]

Fantastic work there Smurfer77! Very appreciated. Thank you!!

BTW: That 2sk97 came off a Sansui X1 diver board.

 

[Hyperion]

Thank you Ron for providing the specimen 2SK97

 

[Ronito6]

Thank you Ron for providing the specimen 2SK97

No problemo. Good to have a mystery sorted.

 

[Hyperion]

And also let me say - what brilliant work smurfer ! - really relevant and so, so useful - well done indeed.

Conveniently, one of the UK distributors for Linear Integrated Systems is located a couple of miles from where my Daughter is at Uni - so potentially I could save on P&P with a little sweetener passed her way
(the devices themselves are expensive enough! - drat £25:00 minimum order!)

https://www.micross.com/ecommerce/shop-default.aspx?v=2 (enter 'lsk389' in the search box)

 

[Ronito6]

I'm actually just now in the process of replacing 2SK30's in a Pioneer SX-626. I've done what research I can and selected the 2N5486 as a replacement. I would love to have a method of verifying I've got a good replacement, operational performance wise, before I button up the restored SX-626.

2SK30's are another FET I would love to have a contemporary replacement for. If dlucy can definitely confirm that 2N5486 is indeed a good substitute, that would be second Christmas for me.

 

[dlucy]

Here are the curves for the 2N5486 (my choice for replacing the 2SK30) and a J109 (just for comparison, something I had lying around):

Curve Traces for 2N5486 from Atlas Peak DCA75





Curve Traces for J109 from Atlas Peak DCA75

 

[smurfer77]

Here are the curves for the 2N5486 (my choice for replacing the 2SK30) and a J109 (just for comparison, something I had lying around):

Curve Traces for 2N5486 from Atlas Peak DCA75

View attachment 891086
View attachment 891085
View attachment 891084

Curve Traces for J109 from Atlas Peak DCA75

View attachment 891089
View attachment 891087
View attachment 891088

Do you have curves for your 2SK30 too?

 

[dlucy]

Here are the curves from a 2SK30A(Y) that I pulled from a Pioneer SX-626

 

[dlucy]

Dang. The DCA75 can compare two or more devices by overlaying curves back to back... AND it shows (to my untrained eye, at least) that the 2N5486 operates very differently from the 2SK30A(Y). Back to the drawing board... ideas anyone?

 

[dlucy]

Just for the sake of completeness (and my sanity) here are the images of the device (2SK30A(Y)) and the board (Pioneer SX-626 Control Amp AWG-008) from whence it came: