simple JFET Buffer, what is the minimum Source resistor/current required ?

[MoruyaGuitarist]

Hi All,
What effect does the source resistor have in the below circuit, is it mostly only to do with the load you are driving ?

I built a straight forward JFET buffer (JFET common drain amplifier) with the super common schematic below. Most use something like 1k to 10k for the Source resistor. I tried it with a 100k for R3 and the performance seems similar (testing by ear). Possibly some slight EQ difference, guitar amp and pedals all 500k to 1M input impedance.
I only have MPF102 and 2N5485 and was no real difference. Will be using it for an inboard guitar buffer. With the 10k R3 it's still quite low current draw for a 9volt battery. Just curious what the limits are. I was also testing for noise with high gain stages and the 100k R3 wasn't any different. I'm sure other JFET's will make a big difference, such as J201 which is unavailable here. Thru hole MPF102's and 2N5485 are still easy to get here, around 3 bucks though.

[merlinb]

A large source resistor forces you to operate the JFET at low current, which means it will run out of headroom (hit cutoff) sooner when driving lower-impedance loads. So the choice of source resistor depends on whether the stage is driving a fixed, known load, or whether it driving something unknown like the input of whatever amplifier someone plugs into it today.

[antonis]

I'm sure other JFET's will make a big difference,

As big as the difference among particular devices transconductance..

Source follower voltage gain is (g_m * Rs) / (1 + g_m * Rs) so the higher the g_m (for a given working current) the more close to unity the voltage gain..

To add on what Merlin very well said:
R3 and Load form a voltage divider during  signal negative waveform, resulting into "saturation" (in the mean of BJTs term..) earlier than Source going all down to GND..
(e.g. A follower with 100k Source resistor, biased at Vcc/2 and driving a 100k load, can go all the way up to Vcc but only down to Vcc/4..)
Source followers of small resistors exhibit larger symmetrical swinging margins than big ones..

P.S.
Of course, you can overcome the above mentioned issue by "off-setting" Source voltage (e.g. by making R2 bigger than R1..)
(it can be easily calculated for a known load value..)

[Rob Strand]

Source follower voltage gain is (g_m * Rs) / (1 + g_m * Rs) so the higher the g_m (for a given working current) the more close to unity the voltage gain.

More current gives a higher gm but to get higher current you need a lower RS - For 4.5V on the drain ID = 4.5V / RS.  It turns out the gain is lower (worse than unity) with small RS.   With a 2N5457 and RS=1k it should be a noticeable drop in level.

With low source resistance RS you can also get output swing problems in the positive direction.  At full positive swing the JFET needs to pull the source up close 9V.  To do that the drain current must reach ID = 9V/RS.   However the JFET current limits at IDSS.   For a 2N5457 IDSS is about 3mA  so RS needs to be greater than 3k ohm.   As it happens with RS = 3k the gain loss in the buffer is about 1dB, about as much as you would want to lose.  So both the +ve swing and the gain loss occurs at about RS=3k or 3k3.

For low currents you probably want to target a source resistance RS about 1/10th to 1/5th the expected load in order not to lose negative swing.   For 47k load that's 4k7 to 10k.

You can see that the low and high limits on RS are fairly constrained, perhaps why you see mostly 10k.

[MoruyaGuitarist]

For low currents you probably want to target a source resistance RS about 1/10th to 1/5th the expected load in order not to lose negative swing.   For 47k load that's 4k7 to 10k.

You can see that the low and high limits on RS are fairly constrained, perhaps why you see mostly 10k.

Thanks All !
So the fairly simple by ear test I did going into a 1M load it makes sense that I couldn't head the difference with a 100k Source resistor. I have no vintage amps or wacky germanium pedals. Think I may have one pedal that's 500K but most are 1M. So even up to a 50k Source resistor should be pretty safe and have a smaller battery drain then. I have seen in one of the "Guitar cable JFET Buffer" circuits it used a 30K source resistor. Different JFET I think, need to check.

I can also slightly increase the R2/R1 ratio to get the voltage swing a little higher like what  anotonis suggested. I still think JFET's are voodo !

Wonder if it isn't just easier just to use a JFET buffer followed by a simple BJT stage with a 20 cent transistor to get a clean gain....

[antonis]

More current gives a higher gm but to get higher current you need a lower RS - For 4.5V on the drain ID = 4.5V / RS.  It turns out the gain is lower (worse than unity) with small RS.

Sure but what counts is the product of g_m and R_s, isn't it..??

[antonis]

Wonder if it isn't just easier just to use a JFET buffer followed by a simple BJT stage with a 20 cent transistor to get a clean gain....

You can't go "clean" with voodo involved.. 

>off topic ON<
I'd propose a BJTs complementary feedback pair with boostraped input, like below:
(but I'm well known for my disgust concerning JFETs..)  



R41 & C21 may be omitted..
(though they are highly recommendend for oscillation and RF rectification prevention..)
>off topic OFF<

[Clint Eastwood]

With the jfets you have, I think you can do without R1 and C1.

[Rob Strand]

Sure but what counts is the product of g_m and R_s, isn't it..??

When you double the current Rs halves but gm only goes up by sqrt(2) = 1.41, so at high currents gm*Rs is 1.41/2 = 1/1.41 = 0.707 the reference value.  So you lose out on gm*Rs at high currents and you get more attenuation.   By the same argument at low currents gm*Rs increases and will have less attenuation.

To avoid the gain loss you can use a current source instead of a drain resistor.  The gain is pretty much unity but there is some attenuation when you place an AC load on the output.   In this case operating the JFET at higher currents can reduce the loss.

So it comes down to specifics.

[antonis]

With the jfets you have, I think you can do without R1 and C1.

Gate grounded -> Increased sensitivity of the DC operating point to device parameter variations..
No C1 -> V_GS unpredictable..
(Gate should stand at any input DC level..)

[antonis]

When you double the current Rs halves but gm only goes up by sqrt(2) = 1.41, so at high currents gm*Rs is 1.41/2 = 1/1.41 = 0.707 the reference value.

Well said, Rob..!!

(I had in mind g_m=1/r_e - not enough coffee for transconductane and current gain discrimination..  )

[Rob Strand]

(I had in mind g_m=1/r_e - not enough coffee for transconductane and current gain discrimination.. 

I've argued this with myself so many times now I know the answer before I go through the reasons  .

[Clint Eastwood]

Gate grounded -> Increased sensitivity of the DC operating point to device parameter variations..
No C1 -> V_GS unpredictable..
(Gate should stand at any input DC level..)

It will be used as an inboard guitar buffer, and guitar pickups don't generate DC. The gate will be 0 volts, very predictable.

You just have to select a Jfet with a Vgs(off) of about 1.5 volt or more. That only will take a minute to measure, no voodoo. Most mpf102 or 2n5485 will be ok.

Another option is to build the buffer into the jack that plugs into the guitar, so you don't have to alter anything to the instrument:

[antonis]

It will be used as an inboard guitar buffer

If you say so..
(didn't read anything relevant in OP's post..)

There's nothing bad in JFET self-bias configuration but there's less than nothing in voltage divider one..

P.S.
Guitar picups don't gererate DC but also dislike DC coming from bias resistor off-set due to Gate leakage current at elevated temperatures..

[Clint Eastwood]

also dislike DC coming from bias resistor off-set due to Gate leakage current at elevated temperatures..

Can you put a number on this, Antonis? How much offset do you fear? I think the OP would have to set his/her guitar on fire to have a noticeable effect. (may be a good idea for a rock concert!!)

[antonis]

Can you put a number on this, Antonis? How much offset do you fear?

I think you will fear if you calculate the voltage drop across a 2M2 resistor of 10nA (at 25^oC) current in a 105^oC junction temperature..
(just double its value for every 10^oC..)

I think the OP would have to set his/her guitar on fire to have a noticeable effect.

Who told you that JFET will sit on (or inside) the guitar..??  

[MoruyaGuitarist]

Thanks all !
I understand it a bit better now with the constraints of the source resistor. Yes it will be used as an in guitar buffer so that's why I questioned about the negative impact of a high Source resistor / low current. I had to downsize my guitar collection so only have one main one so jfet buffer cable is not needed. It's all been soldered as I was testing for noise but I used a socket for the source resistor to let me change it. I understand that self bias works great for gain stages but will just go with what I have now. I actually used a "noiseless bias" with a 1 uF capacitor and extra resistor, probably not needed but was just to try keeping low noise. Was just too lazy to draw it :-P

Without being too radical will go with a 20k source resistor to give an under 200uA current draw so heaps of battery life, easily 2000+ hours. (I can't measure it exact as when putting my meter in series it also impacts the current draw.) To be fair if I plugged in my guitar into an amp with a less then 100k input impedance right now the tone will definitely be very different, so being able to drive low impedance hasn't really changed.

[MoruyaGuitarist]

So final circuit as it is now. Not deviating from standard JFET buffer schematic and just going with 20K source resistor to give slightly better battery life. I couldn't hear any difference going into a 1M load with a 100k source resistor so is still very conservative. "noiseless" bias is used, probably unnecessary but have already soldered it. I was trying for as low noise as possible. Thanks for all the suggestions and will use something better later but it works well as it is and am soldering it into the guitar for now. Will get some JST connectors to allow easy swaps in future for an inboard buffer/boost later.

[Eb7+9]

I still think JFET's are voodo !

you can't drive blind with jFET's like you can with BJT's ...
Vgs(off) puts a cap on input swing and Idss puts a (rough) cap on drain current

assuming that loading is not an issue (tube amp input or buffer)
have a look at the jFET triodifier curves posted here:

http://www.lynx.net/~jc/transferCurvature-TubeSimulation.html

as you increase Rs the transfer curve gets stretched out to the right indefinitely
relative to a give signal strength you get a more linear transfer (ie., less added harmonics per pound of signal)
if you are playing with jFET's that have high Vgs(off) values relative to your signal swing you'll hardly hear a difference

it's when you go LOW in Rs value that added colour is most perceivable

QUIZ TIME: assuming a perfect quadratic transfer what's the minimum Rs value that still provides max headroom (ie. full 0<Id<Idss range AND symmetric input)

[MoruyaGuitarist]

Have put the JFET buffer in the guitar and it works well. Just wanted a clean buffer with a low current draw. I measured around 120uAmps but again don't know how accurate the that is with the meter in series. It's also very small, could be way smaller but was still playing around with parts and layout.

It was also to resolve a problem I had. Without getting into detail I had some weird high pitched squealing sounds while using a combination of digital and analogue pedals. Mostly with high gain and to do with send & return loops of the digital unit. ( low cost amp modeller / IR). Was not power supply related at all. Anyway problem is now fixed !
The order of true bypass pedals and buffered ones is now also less important.


I wasn't trying to use it for getting a "FET sound" which really only occurs close to clipping. This requires a totally different design and using more current would no doubt help. And the obvious 2 or more stages....