Optimist, I presume. The 2SK170 data sheet shows for 5ma and 14.5V, an input current of 1 nano amp, at 25 degrees C Look it up. That would give you 5V across a 5G resistor. So, 5V/20mV means that your jfets have input current 250 times lower than the data sheet, OR you are using lower voltage, perhaps 9V or less. Kind of restricting, isn't it? ;-)
I would never operate these FET's in the impact ionization region for this application. No point in running it at 5mA either.
I believe that Scott Wurcer introduced those jfets to us, and they have been shown by him, and Kirkwood Rough, (who will be at Burning Amp this week in a big way) by measurement with a Quan-Tech noise analyzer, to be very good for microphone input. However, as I recall, they have very low breakdown voltage, so low voltage operation is mandatory.
It is enlightening to directly compare, as much as possible, the data sheets of the BF862 and the 2SK170. Then let us also presume that we are going to make a simple source follower with one or the other of these parts. Let us presume that we are making a 'quality' calibration level microphone, so that we can test loudspeaker distortion, for example. Fair enough?
Now, what voltage would we use, based on the data sheets of these two devices would we use across the jfet we chose for optimum dynamic range and low distortion?
First, it might be noted that the 2SK170 has a voltage rating of 40V. The BF862 of 20V. This should be a clue, but that is not all. Secondly, in general, the nonlinear input capacitance of each jfet should drop with INCREASED voltage as shown on the Ciss-Vds and Crss-Vgd graphs on the 2SK170 data sheet. It fact, Ciss drops from 50pf to 30 pf, IF you put 10V across it. Dare you use any less? Perhaps, but there will be a penalty of 'excessive' nonlinear capacitance, part of which will be minimized by the 'bootstrapping' of the source follower and its relatively high Gm. However this is but a 'band-aid' and not a cure for the non-linear input capacitance, and lower would be better, implying that higher voltage across the jfet would result in lower distortion.
However, what about leakage? Now look at the Igsx-Vgs graph for the 2SK170. It says that you can expect about 10pa at 10V, which is pretty good, and only 50 mV across a 5G resistor. Now, what about a BF862 in the SAME circuit, instead of the 2SK170. The BF862 would have about 10nA or 1000 times more leakage from the data sheet. Therefore, it would have 50 V across the 5G resistor. Perhaps I made a mistake, but this would not work.
It would seem that I have to reduce the supply to less than 7.5V just to make the BF862 work at all and that 6V or less would be even better, yet this is a 20V rated jfet!
Well, what does that do to the circuit if I reduced the voltage to 5V instead?
Then the maximum output of the circuit would be reduced by at least 6dB and the residual distortion might increase somewhat, due to increased input capacitance. Still, the BF862 might work out to be pretty good, overall, perhaps better than the 2SK170, so long as dynamic range is not of great concern.
I built just such a preamp about 38 years ago, and I still have it today. It will measure better than a standard Neumann KM84 mike, that was popular at the time.
I did not know everything then, that I have shown here, but I sort of 'lucked out'. I hope that others don't fall into the gate leakage 'trap' that many interesting jfets have.
Now, what voltage would we use, based on the data sheets of these two devices would we use across the jfet we chose for optimum dynamic range and low distortion?
First, it might be noted that the 2SK170 has a voltage rating of 40V. The BF862 of 20V. This should be a clue, but that is not all. Secondly, in general, the nonlinear input capacitance of each jfet should drop with INCREASED voltage as shown on the Ciss-Vds and Crss-Vgd graphs on the 2SK170 data sheet. It fact, Ciss drops from 50pf to 30 pf, IF you put 10V across it. Dare you use any less? Perhaps, but there will be a penalty of 'excessive' nonlinear capacitance, part of which will be minimized by the 'bootstrapping' of the source follower and its relatively high Gm. However this is but a 'band-aid' and not a cure for the non-linear input capacitance, and lower would be better, implying that higher voltage across the jfet would result in lower distortion.
However, what about leakage? Now look at the Igsx-Vgs graph for the 2SK170. It says that you can expect about 10pa at 10V, which is pretty good, and only 50 mV across a 5G resistor. Now, what about a BF862 in the SAME circuit, instead of the 2SK170. The BF862 would have about 10nA or 1000 times more leakage from the data sheet. Therefore, it would have 50 V across the 5G resistor. Perhaps I made a mistake, but this would not work.
It would seem that I have to reduce the supply to less than 7.5V just to make the BF862 work at all and that 6V or less would be even better, yet this is a 20V rated jfet!
Well, what does that do to the circuit if I reduced the voltage to 5V instead?
Then the maximum output of the circuit would be reduced by at least 6dB and the residual distortion might increase somewhat, due to increased input capacitance. Still, the BF862 might work out to be pretty good, overall, perhaps better than the 2SK170, so long as dynamic range is not of great concern.
I built just such a preamp about 38 years ago, and I still have it today. It will measure better than a standard Neumann KM84 mike, that was popular at the time.
I did not know everything then, that I have shown here, but I sort of 'lucked out'. I hope that others don't fall into the gate leakage 'trap' that many interesting jfets have.
Good illustration John, once again thanks.
I was thinking of replacing the sk170 with the BF862 in the first gain stage from a phono amp - MM. So do I see it correct that it could work better here using the BF with lower voltage supply?
What let me linger is the dynamic range you mention. Music is very dynamic, I mean the micro and the macro dynamics, and maybe a change to a BF could be a step back in view of sonic reason.
I was thinking of replacing the sk170 with the BF862 in the first gain stage from a phono amp - MM. So do I see it correct that it could work better here using the BF with lower voltage supply?
What let me linger is the dynamic range you mention. Music is very dynamic, I mean the micro and the macro dynamics, and maybe a change to a BF could be a step back in view of sonic reason.
Sonic reasons are not allowed to be discussed here. I will say that the BF would probably be OK, and operating voltage with only 50K max on the gate is certainly a different condition from 5G, or about 100,000 times different. ;-) Therefore, some care is still prudent, but not so much as in the microphone case.
Still, I would like to thank Ed Simon for the tip on the Dale resistors. My present employer found a cost effective Dale resistor that is completely non-magnetic, that works just great. In our reality, these questions are as serious as what topology to use.
Still, I would like to thank Ed Simon for the tip on the Dale resistors. My present employer found a cost effective Dale resistor that is completely non-magnetic, that works just great. In our reality, these questions are as serious as what topology to use.
Joao, it is still important to CASCODE the input part, if you have more than 10V on the BF862. The 2SK170 is a very good part for audio inputs, and I use them for just about everything that I make, including phono stages, line stages and power amp inputs. Their capacitance is not low, but their transconductance is good enough and their noise is reasonably consistent. They also have a complement. Usually, I operate the 2SK170 from 10-15V in order to get the lowest and most linear range of the non-linear input capacitance, but this would be BAD for a BF862. Anything above about 7.5V gets scary, as does above 18V for the 2SK170.
Neumann is a bad example they used selected 2N3819’s as common source amplifiers with the source resistor 100% bypassed so they need a high Vp FET to get the DC Vgs high enough or the FET is very nonlinear. Both the 2SK170 and BF862 would be abysmal in this circuit. There are several topologies that avoid many of these problems and I don’t think a cascode is a band aid. Then again replacing class A circuits with ones that use feedback will be dismissed without examination.
Translation: "If I assert things with no backup and no physical plausibility, some people criticize the assertions."
You have to learn to think like John.
I found that the 2N3819 made an excellent jfet follower. My associate, just arrived, who worked with me on that project 38 years ago.(we are working on servos, today) We had a lot of fun DIYing that project in 1972. What puzzled me was that the 2N3819 usually worked better than the 2N4416 in our circuit. (for some reason) Now, maybe, I have a clue to search out as to why, and explain what I could not, decades ago.
Sonic reasons
This forum is called diyaudio so why not discuss sonic reasons???
It's the # 1 reason I am involved in these things, making better sound, not for smelling solder fumes!
Translation: "If I assert things with no backup and no physical plausibility, some people criticize the assertions."
You have to learn to think like John.
This forum is called diyaudio so why not discuss sonic reasons???
It's the # 1 reason I am involved in these things, making better sound, not for smelling solder fumes!
- Status
- Not open for further replies.