Shame indeed. Some Chinese eBay vendors still sell them, but with jfets there's a better solution. In RF projects I started using the BF987, a small signal depletion mosfet. Much more linear than jfet, has gfs up to 16 ma/V, Cgss = 2.7 pF, Cdg = 35 fF, Cdss = 1p. NF (200MHz) = 1 dB. The surprise was that even at low frequences (radio goes down to 9 KHz) an amp with the BF987 outperforms everything I had simmed and built before. It's one of the few devices that can be used for active resistors (with less noise than physical resistors). Still available:
BF987 Siemens | N-Channel MOSFET amplifier | Buy on-line | rf-microwave.com
I used my radios (G31 DCC, RSP1) for evaluation. The first has a display going down to -150 dBm and the second serves (temporarily) for FM. The noise of the BF987 push-pull amp is too low to show on the receivers, it's below 0.1 uV.
Sorry to rain on your parade, but On Semi owns Fairchild. And when Daddy says stop, it's only a matter of time...
I know that. However the manufacturing plants are still separate, so my point stands - the "Fairchild" manufactured ones are being continued, and the "Onsemi" ones are not.
Wow. So firm conclusions based on so few facts.
The FETs in my table above are NEW in the ON SEMI portfolio, not on their kill list.
When I bought some last winter they were tagged as "new product" at DigiKey.
And the line in red would be my personal favorite: highest gm and lowest Ciss.
That ratio is the figure of merit for low voltage noise.
Crss is not so important since it can be made meaningless with a cascode,
bootstrapped if necessary.
JFET gate current increases if you have high voltage drain to gate. If you are using a JFET to get low gate current, you don't want high Vdg. There used to be 300V JFETs, maybe only as samples, because the vast majority are nominal 30V.
DMOS, MOSFET, has a reputation for high 1/f noise, audio random rumble. This may not be fundamental, but process-related. Since few MOSFETs are marketed for low audio noise, there's no strong urge to control 1/f noise.
On-Semi moved it back to "active". Interesting, so it's fine for now, but I think the writing is on the wall that the days of high-performance easy-to-get JFETs are numbered.
The only EE students who even learn what a JFET is at my school are those who take the optional Solid State Electronics class- and I'm going to have to fight pretty hard to even get them to teach it. A little sad, really, but most of my classmates hate analog design with a passion.
I've heard the 2SK3557 is decent, though it is SMD which makes prototyping a bear.
The only EE students who even learn what a JFET is at my school are those who take the optional Solid State Electronics class- and I'm going to have to fight pretty hard to even get them to teach it. A little sad, really, but most of my classmates hate analog design with a passion.
I've heard the 2SK3557 is decent, though it is SMD which makes prototyping a bear.
That is useful - for the moment. It seems that the general trend is to eliminate TO-92 packages, unfortunately. However, nearly all TO-92 devices being obsoleted have SMD alternatives. It appears that the SMD equivalent is BF556.
For a recent high impedance amplifier I bought 50 BF256B from Farnell (still available at present). In measuring them for a matched pair I found none in the first 25 ... but in the next 25 devices, some close matches appeared. I ended up with 10 very close matched pairs (within a couple of percent) and 2 next-best matches (about 10%).
I laid the PCB out so that the flat surfaces faced each other and superglued the pair needed for the input stage together. (And the other device pairs in the circuit too). Overall the drift performance of the circuit was very good. Less than 100uV over periods of 15 mins, although using a 100M input resistor did cause a slight offset open circuit, probably due to a very slight gate leakage current. It is quite something to use a 1G resistor attenuator and get an output that is within the limits of the tolerance of the high value resistors!
For a recent high impedance amplifier I bought 50 BF256B from Farnell (still available at present). In measuring them for a matched pair I found none in the first 25 ... but in the next 25 devices, some close matches appeared. I ended up with 10 very close matched pairs (within a couple of percent) and 2 next-best matches (about 10%).
I laid the PCB out so that the flat surfaces faced each other and superglued the pair needed for the input stage together. (And the other device pairs in the circuit too). Overall the drift performance of the circuit was very good. Less than 100uV over periods of 15 mins, although using a 100M input resistor did cause a slight offset open circuit, probably due to a very slight gate leakage current. It is quite something to use a 1G resistor attenuator and get an output that is within the limits of the tolerance of the high value resistors!
This effect (impact ionization) occurs primarily in n-channel JFETs and rises exponentially with Vdg usually from about 1/4 of the gate-source breakdown voltage. Its also depends on the source-drain current level. The increase in leakage can be 5 orders of magnitude, a real trap for the unwary.
This is above and beyond the exponential increase of gate leakage with temperature due to thermally generated pairs.
It affects n-channel JFETs worse because electrons accelerate further than holes in the channel, and the effect is due to energetic charge carriers impacting the lattice and creating electron-hole pairs near the gate junction.
The moral is use p-channel JFETs if possible, and keep drain currents small, and use generously voltage rated parts, or avoid the high Vdg in the first place.
Note this applies to JFET input opamps so opamp designers have to try to defeat the problem in their input sections, cascoding being a common technique used I think.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.