Best JFETS for solid state switching?

[edvard]

I'm looking at solid-state switching schemes, and I'm wondering which JFET's would work best as switches?
I would guess I'd look for something with a very low "R_DS (ON)" value, but not all JFET's have such a spec listed.
J105's are listed on Mouser as down in the ~3 ohms area, which I imagine is about as low as can be gotten, but it looks like they are being phased out.

Any others?

[R.G.]

I'm looking at solid-state switching schemes, and I'm wondering which JFET's would work best as switches?
I would guess I'd look for something with a very low "R_DS (ON)" value, but not all JFET's have such a spec listed.
J105's are listed on Mouser as down in the ~3 ohms area, which I imagine is about as low as can be gotten, but it looks like they are being phased out.

First off, look up the datasheets. The datasheets will have the term "switching" prominently featured in the the stuff at the top of the datasheet.

Yes, low R_DS (ON) is an important parameter, and it's something EEs obsess over (at least the EEs that know enough analog design to know what that means), but it may not be a big deal.

What matters most is the R_DS (ON) compared to the other stuff in the switching. Imagine a JFET as a switch with a resistor equal to Rds in series with it. If you are switching a signal from, say, an opamp output with an internal source impedance of 10 ohms into a 100K resistor,  and you use a JFET with 10 ohms Rds, then the signal which gets to the input is decreased by the voltage divider of 10 ohms from the opamp, 10 ohms from the JFET, and 100K ohms.
Twenty ohms in a divider with 100K is so close to zero that you can ignore it. Ditto if the JFET was 100 ohms. In fact, if the JFET Rds was 1K, you'd only lose 1% of your signal. But if you're switching signal into a 1K input, then even 100 ohms, and possibly 10 ohms in the JFET gets to be something that you will want to worry about.

Unless I'm doing something really unusual, I ignore the Rds if it's under 50 ohms, and I don't worry all that much if it's 100.

[merlinb]

Douglas Self recommends J112 and J111

http://books.google.com/books?id=oido-UMWNZIC&lpg=PP1&dq=douglas%20self&pg=PA156#v=onepage&q=j112&f=false

I use J112. I don't think it makes a heap of difference though in a pedal. I use BJTs just as frequently.

[Mark Hammer]

DOD uses J113s, which are somewhere in the 50 ohm zone when on.

[R.G.]

I don't think it makes a heap of difference though in a pedal. I use BJTs just as frequently.

BJTs are especially good for shunt switching where the non-zero offset voltage and asymmetry of incremental resistance around zero doesn't matter so much. For series switching, the zero offset of JFETs is a real advantage, if the current being switched is low enough for the JFET's Idss-limited current capacity to not be exceeded.

[edvard]

First off, look up the datasheets. The datasheets will have the term "switching" prominently featured in the the stuff at the top of the datasheet.

Yes, I've done that and most JFET's seem to have been designed for amplification or switching, though experience shows they're just fine for either.
You answered my question about the ON-state resistance, I guess anything that comes in less than 100 ohms into a reasonable impedance should work just fine.
It's just many datasheets don't state this property directly unless that particular one is specifically for switching.

For series switching, the zero offset of JFETs is a real advantage, if the current being switched is low enough for the JFET's Idss-limited current capacity to not be exceeded.

As far as that goes, the J-series MPF102 and 2N3819 can take 20 mA at 15V in the ON-state.
Do any pedals get that hot?

Looking up on Mouser for the various J-series FET's, it looks like the better the ON-state resistance, the more expensive the part.
J113's go for 16 cents, and the data sheet says they go down to 100 ohms, while J105's cost 53 cents apiece. 

Is there a way to measure this in-circuit so I can test some of the garden-variety FET's I've got kicking around (MPF102, 2N3819, etc.)?

[Mark Hammer]

It may create more problems than it solves, but why not use two cheaper JFETs in parallel, if the intent is to use them simply as series resistors?  I suppose the strongest argument against doing so would be that when in an off state, those two high resistances are also still in parallel (hence not as high a series resistance).

[R.G.]

Yes, I've done that and most JFET's seem to have been designed for amplification or switching, though experience shows they're just fine for either.

That's a consequence of the fact that the worst JFETs from the purely switching point of view have an Rdson of a few hundred ohms. This is trivial compared to the typically 10K or higher loads they work into in effects pedals. CMOS analog switches are usually a few hundred ohms too, and they work fine. If the series resistance is negligible (that is, on average, 1% or less) of the impedance they switch into, you can ignore their resistance as irrelevant for audio signals except in special cases.

It's just many datasheets don't state this property directly unless that particular one is specifically for switching.

That's its own answer. If it's a switching speciallized JFET, it specifies Rdson. If no Rdson spec, it's not primarily designed for switching. Notice that you cannot rely on the measured Rdson that's not specified in a datasheet. They may all work fine, but the manufacturer probably does not even look at the Rdson for devices which do not have that specified on the datasheet. It may be fine, no guarantees.

As far as that goes, the J-series MPF102 and 2N3819 can take 20 mA at 15V in the ON-state. Do any pedals get that hot?

JFETs cannot, under normal circumstances, carry currents larger than their specified Idss. That is on every JFET datasheet. A JFET with its gate and source connected together makes a fine constant current diode which conducts its Idss. And in pedals, unless you're switching current into an LED (I'm the only one I've ever seen do that in a pedal, with the Millenium Bypass   ) then the current through the JFET is quite small, because you're going to make sure that the impedance you switch the signal into is quite large. So the load on the JFET is so big that currents in the tens of milliamperes are impossible. Generally this current will be microamps.

Is there a way to measure this in-circuit so I can test some of the garden-variety FET's I've got kicking around (MPF102, 2N3819, etc.)?

You have to measure two things to get the numbers you need for deciding about a JFET. You need to know the Rdson, just to check that it's low enough. However, this is usually trivial, as JFETs in general get to a few hundred ohms, even the worst ones for switching, and the impedances you switch into are over 10K, generally over 100K. The bigger issue is Vgsoff. This is the reverse voltage on the gate needed to force the JFET to turn off so it's open.

When Vgs is zero, gate shorted to source, or even through a resistor, the JFET switch is fully on. It will conduct any current up to Idss (it's more complicated than that, but that's where to start).  To make it turn really, really off, you need to force the gate more negative than the source (for N-channel, positive for P-channel) by enough voltage so the JFET is much higher resistance than the load it switches into. This is typically high, which is good for the on-resistance, but bad for the off-resistance. You have to make the JFET much, much higher resistance than the load, which may be up to 1M or more in pedals. The JFET has to get to 100 megohms or more to do a good switching job. So you have to get to Vgsoff.

That means, back to the datasheet and read the MAXIMUM Vgsoff. You need to ensure the reverse voltage on the gate-source is bigger than that to ensure it turns off. This is one reason the 2SK30, 2N5485, 2N5292 and similar JFETs are popular in pedal use. Vgsoff for these is in the 3V to 6V range, so there is enough voltage to put their drain and source at Vbias of about 4.5V, and still be able to turn them off by grounding their gate. Fairchild's datasheet for the MPF102 and 2N3819 both list Vgsoff as a maximum of -8V, and gives no minimum. The MPF102 or 2N3819 you get may turn off with Vgs of -0.1V, or it may take -8V to turn it off. You can't tell from the datasheet, so you could count on some of them not turning off in a typical pedal circuit at all.

You can build a test rig to test Vgsoff and Rdson. But if what you are trying to do is use the JFETs you have, put in a socket and try them out, being careful to get the pins in the correct socket holes. And if what you are trying to do is ensure that you buy JFETs that work, look at the datasheets for Vgsoff for the devices which are cheap enough and available, and buy the ones which fit your needs and pocketbook. These are two very different objectives.

[edvard]

Awesome answer, thanks.
I just learned a ton...

It never occurred to me to think of VgsOFF and what that meant for switching.
I'm assuming that's why in most FET switching arrangements, you have drain and source at 4.5V, so the gate control has swing either way for maximum off-signal?
I did a few Spice models and the thing just didn't work without being arranged like that.

You can build a test rig to test Vgsoff and Rdson. But if what you are trying to do is use the JFETs you have, put in a socket and try them out, being careful to get the pins in the correct socket holes. And if what you are trying to do is ensure that you buy JFETs that work, look at the datasheets for Vgsoff for the devices which are cheap enough and available, and buy the ones which fit your needs and pocketbook. These are two very different objectives.

Gotcha.
I think I'm just going to swing for some J112's or 113's come parts-buying time, and save my MPF102's and 2N3819's for buffers.
The J201 has a VgsOFF of max -1.5V, so it might do double duty...

Thanks again.

[R.G.]

I'm assuming that's why in most FET switching arrangements, you have drain and source at 4.5V, so the gate control has swing either way for maximum off-signal?
I did a few Spice models and the thing just didn't work without being arranged like that.

Actually, the gate has to swing between the same voltage as the source and drain, and 'way negative (for N-channel). It has to not swing positive above the source. "ON" for a JFET switch means the gate can be at the same voltage as the channel between source and drain, and one way to do that is to short it to the source; another is a resistor connecting them, and yet another is to simply open the gate. With the gate open, the channel acts like the resistively-doped silicon path it is. When the gate is pulled way negative, the depletion region caused by the reverse biased gate-channel diode literally pinches the area of the channel that can conduct electricity until it's pinched off entirely, like squeezing down a garden hose until it's completely shut off. So "ON" is gate open/tied to source, and "OFF" is "more than Vgsoff negative". Biasing the drain and source up at Vbias is convenient, because that means when you ground the gate, it's Vbias negative with respect to the voltage on the channel.

[merlinb]

BJTs are especially good for shunt switching where the non-zero offset voltage and asymmetry of incremental resistance around zero doesn't matter so much. For series switching, the zero offset of JFETs is a real advantage, if the current being switched is low enough for the JFET's Idss-limited current capacity to not be exceeded.

Interestingly I go the other way. I generally avoid BJTs for shunt switching as they introduce too much 2nd harmonic at high signal levels (volts), and the offness is poor unless you balance the source and base resistance just right (and even then its not stellar).
As series switches they are excellent, however, provided there is a DC path from the source, and provided the source resistance is low. I do most of my switching with opamps and series BJTs now.

[R.G.]

BJTs are especially good for shunt switching where the non-zero offset voltage and asymmetry of incremental resistance around zero doesn't matter so much. For series switching, the zero offset of JFETs is a real advantage, if the current being switched is low enough for the JFET's Idss-limited current capacity to not be exceeded.

Interestingly I go the other way. I generally avoid BJTs for shunt switching as they introduce too much 2nd harmonic at high signal levels (volts), and the offness is poor unless you balance the source and base resistance just right (and even then its not stellar).
As series switches they are excellent, however, provided there is a DC path from the source, and provided the source resistance is low. I do most of my switching with opamps and series BJTs now.

Probably depends on the app and whether you allow DC offsets - or, I guess, whether the rest of your circuit can reject any changes in DC offsets. Pesky base current.

[edvard]

OK, after looking into the 555 flip-flop and doing some sims in LTSpice and reading this thread again, I have one question.

Actually, the gate has to swing between the same voltage as the source and drain, and 'way negative (for N-channel). It has to not swing positive above the source.

My plan is to use N-channel and P-channel FETs with the 555 circuit, as it only outputs the one control voltage.
The drawback is that it swings all the way to 9V and back, which would put the gate at 4.5V ABOVE the source, which is divider-biased at 4.5V (This was why in my original post about the 555, I had a R divider to provide a 4.5V control, so it would never go more positive).
Now however I see that the diodes at the gates of these switching schemes prevents the N gates from seeing positive voltage and vice-versa for the P gates, so a 9V control signal should actually be ok, amirite?

One last thing; when I sim'd this in LTSpice, I noticed that the switching action takes about 1/2 a second, is this normal?