Why are JFETs such a Pain in the Ass?

How many times do we see post in the Troubleshooting forum where the problem turns out to be a JFET?

Like Germanium transistors, JFETs are finicky beasts and buying the right part number does not guarantee success.

First, a very brief primer on how JFETs work.
JFETs are voltage-controlled devices. The drain current depends on Vgs (the voltage between gate & source). The gate current is so close to zero that we can safely assume it is zero. We'll talk about N-channel JFETs (arrow on the gate pointing in) because 99% of the time that's what are used in pedals. When Vgs is zero, the JFET is fully on; the drain current is Idss. When we make Vgs go negative, the drain current is reduced. When Vgs reaches Vp, the drain current is zero. When we using JFETs to amplify signals, Vgs will be somewhere between 0 and Vp.

JFET specs
The two JFET specs we care about are Vp and Idss. Those two number tell the pedal circuit designer all they need to know to set the bias and determine the gain of the JFET. Here's where it gets ugly. The Vp and Idss specs on JFETs are extremely loose. Some more than others. Example: 2N5457 has a Vp spec from 0.5V to 6.0V. The Idss spec is 1mA to 5mA. Parts at the extreme end of those ranges will probably not work in a given pedal. As good as JFET production methods are, there is still a significant variation from lot-to-lot. The manufacturers test and sort the JFETs, but to keep yields up and costs down, the specs are left pretty loose. It is very common for OEMs (Original Equipment Manufacturers), to either pay the JFET manufacturers to cherry-pick JFETs to fit a narrower spec window, or buy lots of extras and do it themselves. We're in the position where we have to do it ourselves. Sometimes, the pedal designer is clever enough to design a circuit that can tolerate the variations in Vp or Idss. Of all the JFET pedals sold on this site, only a few fall into that category. That's why JFET pedals have trimmers in them. But the trimmer can only do so much and a JFET that is near the extremes for Vp or Idss may not work for any trimmer setting.

Why do pedal designers use JFETs if they're so problematic?
JFETs have certain advantages. In some circuits, they can have more gain and/or less noise than bipolar transistors. Their input impedance is very large, so they don't load down pickups or other stages in the pedal. Their biggest appeal to pedal builders is that their transfer function is similar to that of a vacuum tube. They tend to generate lower-order harmonics and overload more gracefully in a properly designed circuit, compared to bipolar transistors. Many of the amp-in-a-box pedals replicate a tube amp's preamp section, with JFETs standing in for vacuum tubes.

What do we do about it?
Well, first you go to college for four years and get a degree in electrical engineering. Ok, so that's not going to work for most people. The alternatives are you find someone who can analyze the circuit and recommend a range for Vp and Idss, or you find someone who has measured a production pedal and try to match their readings. Luckily, there are forums inside and outside this website that can help you with that. Mr. PedalPCB might be persuaded to include JFET spec requirements in the Build Docs for pedals that are picky about such things. You need to own and understand how to use a transistor tester. You can get a perfectly adequate one for about $20US on eBay. The other thing you need to do is you buy quality parts from a reputable vendor*. I've said it before and it bears repeating here: "If you buy transistors on eBay, you need to have the heart of a gambler and the skill to test them." If there was ever a place that demonstrates the adage "If it's too good to be true, it probably is." it's eBay. There are some good semiconductor vendors on eBay, but they are in the minority. The rest are either ignorant or crooks because there are a TON of counterfeit parts sold there.

* A high score on eBay does not a "reputable vendor" make. They might have earned that high score selling umbrellas.

The Bottom Line
Building pedals containing JFETs is not necessarily a paint-by-numbers activity and you need to know that going in. Unless you screen your JFETs prior to assembly, sockets are mandatory. Buy from a reputable vendor and buy at least 3x as many as you will need so you end up with enough that work in your pedal.

Next Time: Biasing JFETs
 
I would be extremely interested in what you make of the King of the Britains.
I just breadboarded KotB using PF5102 FETs. Got the bias dialed-in by tweaking the source resistors. Lotta gain there. On headphones it's pretty harsh, even with GAIN, TREBLE and MASTER dialed way down. Tried running it at 18V; it got louder but the tone didn't change much. I'll recheck my wiring and plug it into an amp & speakers tomorrow, see if it sounds any better.
 
OK, today I spent some time with KotB. Played my Chibson Res Paur thru it and straight into a tube amp. Sounds pretty good, but not exceptional (to me anyway). The Marshall tone stack doesn't do all that much, no surprise there. Having two Gain controls is a plus. I have two other Marshall emulator pedals: DLS mk 3 (Covert) and ROG Thunderbird. To me the KotB is similar to the Covert, but the Thunderbird is superior to the other two IMHO. One thing that the Covert has over the KotB is its mu-amps respond well to increased power supply voltage, whereas the KotB sounds the same at higher voltage. I'll play thru it some more, but I doubt it will get beyond the breadboard stage. I made a few minor mods. Here is the as-built:

King of the Britons cb mod v1.1.png

I replaced 5 of the J201s with PF5102s and one (Q4) with a 2N5089. Most of the changes were to adjust the bias for the PF5102s. D1 was added to correct Q1's bias. I added R5 because there is no need to dial the GAIN all the way down to zero. R9 was increased to 4.7K to correct Q2's bias. R13 was increased to 2.7K to correct Q3's bias. R14 was 100K, which to me was way too high and could have been a typo, so I changed it to 10K. Q4 was originally a source follower. I tried an emitter follower instead and saved a JFET. Out of necessity (not enough A1M pots), I scaled the impedances of the tone stack down by approx 5x. Rescaling the tone stack might cause some small tonal differences when Q4 is overdriven hard. Where R18 is, there used to be two 1K resistors with R17 connected in the middle. To get Q5's bias right, I moved R17 to the bottom of the pair and replaced the two 1Ks with a 2.2K. R23 was increased to 750Ω to dial-in Q6's bias. The VOLUME pot was decreased to A100K and C17 was increased accordingly. Did I build an exact replica of the KotB? No. Is it pretty damned close? Yes.
 
Last edited:
Just for S&G, I cannibalized some pots from another breadboard and assembled the tone stack in accordance with the original schematic. Sounds the same. It's a decent pedal, but it's not for me. I definitely like the Covert & Thunderbird better. Next up... the EQD Monarch, with the bias issues corrected.
 
😁
😁
😁 Thank you Chuck D Bones I have a much better perspective now very cool way of explaining that quantum string theory Jfet mystery I had going on that you were unaware of
1 question is this or will this work for the info needed obviously I'm still not fully grasping all the electronic lingo but I snapped a pic of my tester sadly I dont understand any of the info it provides maybe someone could explain ? Does what you said about the jfets include mmbfj201 for instance?I mean should all jfets be auditioned per circuit even SMT ones that are in spec aside from the tolerant circuits? And would the PAL 800 be considered a tolerant circuit?
 

Attachments

  • 45B17CCB-8D39-4673-9017-72E32B5B15A9.jpeg
    45B17CCB-8D39-4673-9017-72E32B5B15A9.jpeg
    1.3 MB · Views: 16
  • 6720EBDF-FCD2-4D54-8C41-63DC9EEB1421.jpeg
    6720EBDF-FCD2-4D54-8C41-63DC9EEB1421.jpeg
    1.2 MB · Views: 16
OK, today I spent some time with KotB. Played my Chibson Res Paur thru it and straight into a tube amp. Sounds pretty good, but not exceptional (to me anyway). The Marshall tone stack doesn't do all that much, no surprise there. Having two Gain controls is a plus. I have two other Marshall emulator pedals: DLS mk 3 (Covert) and ROG Thunderbird. To me the KotB is similar to the Covert, but the Thunderbird is superior to the other two IMHO. One thing that the Covert has over the KotB is its mu-amps respond well to increased power supply voltage, whereas the KotB sounds the same at higher voltage. I'll play thru it some more, but I doubt it will get beyond the breadboard stage. I made a few minor mods. Here is the as-built:

View attachment 8600

I replaced 5 of the J201s with PF5102s and one (Q4) with a 2N5089. Most of the changes were to adjust the bias for the PF5102s. D1 was added to correct Q1's bias. I added R5 because there is no need to dial the GAIN all the way down to zero. R9 was increased to 4.7K to correct Q2's bias. R13 was increased to 2.7K to correct Q3's bias. R14 was 100K, which to me was way too high and could have been a typo, so I changed it to 10K. Q4 was originally a source follower. I tried an emitter follower instead and saved a JFET. Out of necessity (not enough A1M pots), I scaled the impedances of the tone stack down by approx 5x. Rescaling the tone stack might cause some small tonal differences when Q4 is overdriven hard. Where R18 is, there used to be two 1K resistors with R17 connected in the middle. To get Q5's bias right, I moved R17 to the bottom of the pair and replaced the two 1Ks with a 2.2K. R23 was increased to 750Ω to dial-in Q6's bias. The VOLUME pot was decreased to A100K and C17 was increased accordingly. Did I build an exact replica of the KotB? No. Is it pretty damned close? Yes.
Interesting I pulled mine out of the meh box last I built the I believe 1st version 3 knobs 2 toggles when it’s paired it’s nice but yeah overall back to meh pile I’ve only heard great things of ROG designs and yet to build one
 
Unfortunately, those testers do not measure Idss or Vp. In a few other threads, I described how to measure Idss and Vp using a DMM. The bottom line is very few circuits will tolerate JFETs over their entire spec range. The PAL800 is tolerant to the extent that the trimmers have a large adjustment range. Still, no guarantees every in-spec J201 will work. The pros have to test and select their JFETs if they want every pedal to work.
 
Cooder & I designed a universal JFET tester board that measures Vp & Idss.

NB: The Vp values given in the datasheet are at a specific (very low) drain current. There is no "industry standard" Id for measuring Vp. This tester uses the DMM's input impedance & Vp to set Id. It's "close enough" to the datasheet test conditions. If you find JFETs that are just barely out of spec when measuring Vp, it may well be due to fact the the Id we use to measure Vp is not the same Id the manufacturer uses. Sure, we could make a tester that uses the correct Id, but it complicates the circuit and requires the user to look up and then set the correct Id for each part number. 99% of the time, this is unnecessary.

This tester has 6 JFET footprints. Three for Vp and three for Idss. One footprint is for source in the middle (most JFETs), One for gate in the middle. One for SMD. This board should work with any part number, provided you plug the JFET into the correct socket, facing the correct way. When in doubt, consult the datasheet. Instructions are printed on the board. "what about JFETs with drain in the middle?" you ask. For low freq applications, source & drain are interchangeable. 🧐


JFET tester board 02.jpg

JFEET tester sch.png
 
Cooder & I designed a universal JFET tester board that measures Vp & Idss.

NB: The Vp values given in the datasheet are at a specific (very low) drain current. There is no "industry standard" Id for measuring Vp. This tester uses the DMM's input impedance & Vp to set Id. It's "close enough" to the datasheet test conditions. If you find JFETs that are just barely out of spec when measuring Vp, it may well be due to fact the the Id we use to measure Vp is not the same Id the manufacturer uses. Sure, we could make a tester that uses the correct Id, but it complicates the circuit and requires the user to look up and then set the correct Id for each part number. 99% of the time, this is unnecessary.

This tester has 6 JFET footprints. Three for Vp and three for Idss. One footprint is for source in the middle (most JFETs), One for gate in the middle. One for SMD. This board should work with any part number, provided you plug the JFET into the correct socket, facing the correct way. When in doubt, consult the datasheet. Instructions are printed on the board. "what about JFETs with drain in the middle?" you ask. For low freq applications, source & drain are interchangeable. 🧐


View attachment 53444

View attachment 53445
Awesome. I’ll have to make one of these too.
 
Cooder & I designed a universal JFET tester board that measures Vp & Idss.

NB: The Vp values given in the datasheet are at a specific (very low) drain current. There is no "industry standard" Id for measuring Vp. This tester uses the DMM's input impedance & Vp to set Id. It's "close enough" to the datasheet test conditions. If you find JFETs that are just barely out of spec when measuring Vp, it may well be due to fact the the Id we use to measure Vp is not the same Id the manufacturer uses. Sure, we could make a tester that uses the correct Id, but it complicates the circuit and requires the user to look up and then set the correct Id for each part number. 99% of the time, this is unnecessary.

This tester has 6 JFET footprints. Three for Vp and three for Idss. One footprint is for source in the middle (most JFETs), One for gate in the middle. One for SMD. This board should work with any part number, provided you plug the JFET into the correct socket, facing the correct way. When in doubt, consult the datasheet. Instructions are printed on the board. "what about JFETs with drain in the middle?" you ask. For low freq applications, source & drain are interchangeable. 🧐


View attachment 53444

View attachment 53445
I need one of these asap!
 
Last edited:
Isn't the Vgs(Off) an important parameter to measure as well?

Personally, I invested in a Peak Atlas DCA75 and it measures all the pertinent parameters of JFETs without issue. And if you have a Windows PC, the DCA75 comes with software for displaying the curves. The DCA75 will also Si & Ge BJTs - showing whether it is Ge or Si.
 
Last edited:
Sorry if it’s poor form to resurrect older threads - I’m new here (and to electronics in general). From the whole internet, digging through this here charnel has answered the most evasive of my questions, so big thanks to Mr Bones for being so willing to share his knowledge (and I suspect for the time spent rendering it digestible to plebs like me). From a hitherto outsider, you’re a huge asset and draw to PedalPCB.

Getting to an actual point…

find someone who can analyze the circuit and recommend a range for Vp and Idss, or you find someone who has measured a production pedal and try to match their readings

Any chance you might give a précis on the former? It’s probably backwards to want to select a JFET for a given circuit (rather than adapt a circuit to a given JFET), but I’ve been trying to bully a schematic into telling me Vp and Idss without success (for even a single optimal value, much less a tolerance range). I can guess at Vp, but my best efforts at Idss are at odds with reported measurements (and it’s become too much of an academic exercise to just use those).

I’d rather learn to fish than be given one (sorry fellow vegos: tofu analogies confuse people), but if it helps I’m looking at a self-biased common source amp where the schematic tells me the resistance values and voltage points. The JFET used is long out of production, so it’s all about determining a substitute.

I’ve mostly been looking at the transfer function
1722130397888.png
but with 2 unknowns, I wonder if there’s another equation I need to consider… I didn’t get far trying to sub values in based on definitions for Vp and Idss.

Thanks in advance for entertaining or ignoring my nonsense as you see fit!
 
You have correctly identified the problem. Two parameters define a JFET's transfer function. It's always a parabolic curve, but where the curve intersects the X and Y axes can cover a range of values. The DC bias point of a JFET is one point on the curve. We can draw a myriad of parabolic curves that all intersect that bias point, but they will all have different Vp (X-axis intercept) and Idss (Y-axis intercept). To make matters worse, different JFETs with the same part number will have different curves.

There is another parameter we probably care about: transconductance (gain). The gain capabilities of a JFET depend on the ratio Idss/Vp. If we want the maximum gain from a given JFET stage, then we want one with a large value for Idss/Vp. We don't always need high gain, but to know when we do need it requires an understanding of how the circuit works. It would also be nice to understand the designer's intent, but that information is hard to come by.

Another thing that matters in a lot of circuits is input voltage range. If we drive an N-channel JFET's Vgs above zero, then the JFET makes some very unpleasant distortion (a similar thing happens with vacuum tubes). That means that Vp has to be greater than the peak Vgs applied to the JFET. MOSFETs are much more forgiving in this regard.

What all this boils down to is the circuit imposes enough constraints on the JFET that we can't just pick any JFET, fiddle resistor values until the bias point is some magic value and expect the circuit to work correctly. The hard part is understanding how to use those constraints to arrive at a an acceptable range for Vp and Idss. Even then we will have to experiment to see which combinations provide the best results. AFAIK, there is no cookbook formula for doing that. I make informed guesses based on experience, run simulations, build and test breadboards, repeating as necessary. There could be several part numbers that will fit the bill, but they still have to be hand selected and maybe some resistor values tweaked. Oh, and they have to be available.

As much as possible, I try to design circuits with parts that are not too hard to get. I buy JFETs of various part numbers in bulk so I have plenty of options. But let's be realistic, there are a lot of vintage pedals containing parts that have been out of production for years. We have to improvise.
 
Last edited:
Back
Top