JFETs being phased out?

As has been pointed out, it's more the TO-92 packages that are suffering poor sales and being discontinued. May I mention our Table 8.2 in AoE III, page 516. Here you'll find about 70 parts, with the SMT part numbers corresponding to the TO-92 parts. You'll also find measured values of noise data, with plots.

They may still be good for audio, give them a try. For example, the large-die parts with low Ron are often very well suited for low-noise amplifiers.

It depends a lot on the Vds you're applying. For many of these JFETs, running them at Idss at more than a few volts Vds will increase the power dissipation in the part beyond the safe continuous limit. So, you either have to run them at low voltage (at or near zero gate bias), or bias them to run them at a lower current.

For starters, divide the part's maximum continuous power dissipation, by the largest voltage you expect to have across the part. Then, derate as appropriate.

Look at the thermal characteristics; DC and pulsed.

Plus measured data in Table 3.7, pg 217: actual Idss, Vgs at Id (typically 1 or 5mA), g_m, and g_os, with our special highly-useful Gmax term. To get this data, I purchased a hundred types of JFETs, Paul spent months taking measurements, I made huge analysis spreadsheets and created the tables. It's well worth taking a look at. About the 2N5458 you were looking at, this has no sellers listed on Octopart, but the MMBF5458, that we call out in the table, made by Fairchild, has many distributors, lots of inventory, and is 15 cents at Newark. Ditto for dozens of others.

We found the sot-23 SMT JFETs measurements to be the same as the TO-92 parts (same die), and we also found parameter measurements to be constant over several decades of date codes, for a given manufacturer. Nice!

For the '5458 we measured Idss 4.1mA, Vgs -0.97V at 1mA, g_m 2.2mS at 1mA and Gmax = gm/gos = 190. Gmax is independent of Id for low to moderate Id. That's because the output resistance scales with current, tracking the transconductance.

As we all know, the variation in Vgs vs Id is a big problem, but thankfully the other parameters track Id quite well (see our graphs in Chapter 3), so it's very useful to use a circuit that biases your part at your desired Id.

Gmax = gm/gos is what the tube guys called "amplification factor."

These days, unless one lucks into a supply of "new old stock", the least-awful solution to getting some through-hole parts for prototyping may well be to buy the MMBF (or similar) SOT-23 parts, and a handful of "SOT-24 to SIP" adapter cards, and just hand- or toaster-oven-solder up a bunch of TO-92-workalikes.

I got 50 little adapter PCBs of this sort from an eBay seller, delivered from China, for under $6. That's roughly the same price as the JFET itself (MMBF5458 is about $0.12 from Mouser, quantity 100) but the total is still quite reasonable compared to the cost of the few remaining TO-92 new-old-stock parts available (if you can get 'em).

Another option is to buy from Linear Integrated Systems - they have through-hole second-source versions of many of the popular parts - but as they're a specialty provider their prices aren't going to be wonderful. However, if you want really _good_ JFETs (tightly- matched monolithic duals, very low noise single or dual, etc.) they're kinda the only game in town these days, I think.