Testing Idss of fet and not getting what I hoped for

You're complaining that Idss is too high? Normally that's considered a good thing. It appears your parts are tightly grouped together. That's good. A good way to design JFET circuits is to characterize your run of parts, and then design for that.

Yes, except I'm not a designer, I build other peoples work. I'm gathering parts for the low noise amp discussed earlier. It calls for the fet to have an Idss of 7ma, I don't know if there is room for an adjustment to work with a high Idss.

Thank, Mikek

PS, High Idss might be a good thing, but I doubt these fets are the low noise fet I wanted.

A long shot, but it might be worth verifying that your circuit is not oscillating. Depending on wire length and placement I think it is not impossible for that to happen. Maybe put a scope probe near it, or 1MOhm in series with the gate.

The data sheet says that the max Idss at Vds=10V is 14mA MAX. I don't know what part you have in hand, but I don't think it is the 2sk117 BL.

First, it's likely that those JFETs are actually more quiet than the ones you wanted. The 2sk117 was discontinued by Toshiba long ago. Although Paul and I made a nice set of measurements for them, I removed them from our JFET tables for that reason. Perhaps somebody attempted to make a replacement part, and missed on the Idss. This is the reality for JFETs, production Idss values are all over the place. That's why we teach that JFET amplifier designs should have a configuration and included circuit elements that are tolerant of actual Idss values. Your circuit is not.

However, you may be able to make it work. First, adjust the ~1.2k drain resistor as they say. For 20mA it'd be about 400 ohms. Second, although gm will be higher at 20mA than 7mA, it likely won't be proportionally higher. This means you may not be able to have the original overall gain of 100, but you could select the ~1.5k feedback resistor for a gain of 50, more like 600 to 700 ohms. Or you can increase the open-loop gain by reducing the 215-ohm resistor, maybe to 100 ohms.

A more flexible design would have made the ~1.5k fixed, and added a selectable resistor in series with the JFET's 10 ohms source resistor, bypassed by a BFC cap. Or better yet, a proper current- sink in the source to set its operating current.

The amplifier has several low-frequency rolloff capacitors inside the feedback loop, the 470uF with the 215 ohms, and 220uF with the feedback resistor. If one is not careful, this invites a low-frequency oscillation, motorboating. But I note that omitting the 220uF would only add 80mV of Vgs change to the 2sk117, which really wouldn't matter. So it's not needed. That would also allow you to more safely add a BFC, also inside the loop, if you chose that route.

Not to toot my own horn, but my own Altoid mixer's shown at:

This truly isn't to toot my own horn. The addition of eight new audio isolation transformers to the mixer's front end and a TEA2025 with a volume control pot to its back end is now in the works. All of those new components make my new fangled mixer-amp too large to fit inside an Altoids tin anymore. So a 2" x 4" x 2.5" Bud box will to be used instead, provided there's enough empty space in it to accomodate all of the new components. My goal at this initial stage of design is to see if it's feasible to eliminate capacitors C3-C8 shown in the schematic and directly couple the iso transformers to the TL082s, which will be reconfigured as "unity gain summing difference amplifiers" (so to speak) if that's even possible. Another possibility is to use something along the lines of a BA3121F-E2 instead of isos. But thus far, Bill Whitlock's papers lead me to believe that iso transformers are superior to silicon solutions.

Thank you, 73,

But it worked in LTSpice!!

Tim

For a low-noise design remember that the TL082 power supply rejection ratio (PSRR) for the VDD line may be worse than for the VCC line. The internal reference point of the chip is VDD not VCC. It could be better to swap the stabilization and filter circuit from the VCC side to the VDD side and R1 to the VCC side. And add a filter for VCC, for better noise suppression. And add a BFC over the R10 and R13 resistors.

Regards, Arie

A lot of amateur circuits worked on the bench... once.

Your constructive criticism is appreciated. My intention to eliminate as much of D1, C1, C2, Z1, R1 as possible and replace that sub-circuit with a TLE2426 Rail Splitter is missing from my original followup. Any words of wisdom on the TLE2426 approach is also appreciated. It goes without saying that replacing the TL082 with a single-supply op-amp would simplify things. But such a single-supply op-amp with the desired characteristics didn't jump out at me during my search for an alternative. Is BFC an acronym for Band Filter CutOff? Is it the same thing as a Band Stop filter?

Thank you, 73,

My circuit earlier this week had me scratching my head, Spice wouldn't run it. It took a few goes looking over it to notice I'd connected the power rails together via 2 base junctions. Oops. A base-ic error.

NT

I always thought it was Big Effing Capacitor, so if I have that wrong let me know. Mikek

Sometimes I'll copy/paste part of a circuit and everything goes to hell. LT Spice copies the net names too, so it can short all sorts of things to one another.

A lot of the time the circuit only needs to work for the one time build. In this case the author warned all that you need to to select the BL fet for an Idss of 7 ma. It seems to me he built a high input impedance low noise amp with a gain of 100, and showed how to wire it into an Altoids box for shielding. Did some actualy testing of the noise and put it out to help others. The only downside is the availability of the 2sk117 BL. The circuit first shows up in Feb 2012, so 7 years for the part to age out, not a long time. I appreciate people that put circuits online, for those of us who are more techs than electronics designers it is very helpful. Pu pu it all you want, Win tried to help, but he said his fix would drop the gain down to 50, so I decided to try and find the proper 2sk117 BL fet, or maybe 2sk209 BL an equivalent. Or Win mentioned that the fets I received with much higher Idss may be even lower noise than what I ordered. Win or anyone want to explain that?

Thanks, Mikek

Simple solution. Use netnames with their associated reference designations. For example, junction of R1 and C3 : R1C3. Q1 Base : Q1B

When you copy and paste, The reference designations change. So it's easy to spot the old netnames and update them.

Fully explained in the JFET section of AoE's low-noise chapter.

Tweak-until-they-work circuits might force some people to understand them enough to make them work. But some other people might build them, see that they don't work, and decide that electronics isn't for them.

We design for production, so we want every copy to work with worst-case part variations. Manufacturing/test can't change a part value without a formal ECO to authorize it.

A high Idss fet is equivalent to putting N smaller fets in parallel. That reduces voltage noise by the square root of N, assuming gate current is very small, which it usually is.

And that the extrinsic gate resistance doesn't dominate, which it often does.

Cheers

Phil Hobbs