# jfet

• posted

This looks similar to the BF862. Noise is characterized as nf relative to 1K ohms, around 1 dB. What is that in nv/rthz?

The 1/f corner looks pretty low.

--
John Larkin         Highland Technology, Inc
picosecond timing   precision measurement
• posted

Awesome, thanks. I'll have to get some and see.

Cheers

Phil Hobbs

• posted

ON will generously send me 5 40-cent JFETs free, as long as I pay \$6 shipping and \$5 handling.

I'll wait till they're in distributor stock.

Morons.

Cheers

Phil Hobbs

• posted

The BF862 has gm = 26mS at 5mA, and 12mS at 1mA, according to our measurements, see Table 3.7, page 217 and Table 8.2, page 516 and noise curve E in Figure 8.52 on page 518. (Note, the BF862 appears in 24 places in our book.)

ON Semi's NSVJ2394SA3, with gm = 24 at 5mA, should have a similar noise. Both sot-23 packages use the same pinout. The feedback /output capacitance looks about 1pF higher.

--
Thanks,
- Win
• posted

According to one reference, the 1 dB / 1K ohm NF corresponds to a tad under 2 nV/rthz, much worse than the BF862. But I'm not sure about that.

--
John Larkin         Highland Technology, Inc

lunatic fringe electronics
• posted

** You don't know how to do the conversion or just too lazy ?

Start with the voltage noise of a 1kohm resistor = 4.04 nV/rtHz.

Increasing that number by 1dB gives 4.53 nV/rtHz.

Squaring and subtracting gives 2.05 nV/rtHz.

( Assuming the FET has negligible current noise. )

.... Phil

• posted

Friday night. Had a beer and burger at Zeitgeist for lunch, salmon cake sandwich and another beer for dinner. Not in the mood for algebra.

My nomograph came in about 1.7ish, so BF862 is still the winner.

At least some people are still inventing jfets.

--
John Larkin         Highland Technology, Inc

lunatic fringe electronics
• posted

Uh, work it out? The power in a resistor at room temperature is -174dBm/ Hz; surely you can go from there. Look at their test circuit -- the best impedance for a noise match is often not the best impedance for a power match. Dunno if JFETs can be counted on to have all voltage noise, although it's hard to imagine where current noise would go at low frequencies.

--
Tim Wescott
Control systems, embedded software and circuit design
• posted

The spec is at 1 mA, so it ought to get quieter at higher currents.

I have a couple of reels of BF862s for insurance.

Cheers

Phil Hobbs

• posted

The noise values in spec sheets are nearly useless.

By now Paul and I have measured hundreds of parts, and find the noise levels pretty well follow the standard g_m formulas (spelled out in Chapter 8), and are consistent part-to-part, and year-to-year as well, as long as you're well above the 1/f freq. We detail these measurements in Tables 8.1 and 8.2.

You can get good noise estimates for your specific parts, simply by measuring g_m.

The rising 1/f low-freq noise scene for JFETs is another matter entirely. We pick relevant parts and have graphs of what you might encounter.

ON Semi's new family of JFETs may in fact be nearly as good as NXP's BF962, that's nice, but unless you need the smaller package version, or the new dual part, why not stick with NXP?

--
Thanks,
- Win
• posted

Sure, they supposedly have noise equivalent to a resistance of

2/(3*g_M). I'd expect there to be a contribution analogous to R_bb' at some level, though, especially for the higher transconductance devices.

I think it's pretty cool to have more choices, especially larger devices that avoid the necessity of parallelling BF862s for higher capacitance transducers. Also given the very low test frequency of the noise voltage, there's hope that the 1/f corner is lower than the BF862's moderately obnoxious 1 kHz. (That doesn't matter so much in AC-coupled bootstraps, of course.)

The dual part is potentially interesting as well, particularly if the FETs have any sort of vague matching. DC-coupled TIAs and bootstraps running at very high impedance make it difficult to control drift. The usual trick of snooping the summing junction with a FET op amp contributes a whole lot of low frequency noise. Vaguely matched JFETs running at the same bias should have vaguely the same drift, so it can be controlled open-loop.

And the existence of multiple sources for what was previously a dangerously unique part is a great comfort.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
• posted

Hey, Win - Is AoE 3rd edition the latest? Thanks, John

• posted

None appear to be for sale yet, except for the NSVJ3557SA3, for which Octopart shows 108,000 available at Verical, in 3k reels, MOQ. At 20 cents each, that'd be a \$600 purchase, for an unknown performance.

--
Thanks,
- Win
• posted

OK, it may be as quiet as the BFT at a more realistic current.

I'd guess that jfets is jfets, that noise follows DC specs pretty much. What causes the 1/f corner? It's below 100 Hz for the part that I posted.

I regularly peruse the "new parts" bits of all the major semi sources, to keep up with what's happening. Seeing a new jfet was unusual.

Are you going to measure the noise of those samples?

--
John Larkin         Highland Technology, Inc

lunatic fringe electronics
• posted

Take a walk to the end of a local pier; surely you can go from there.

Try.

--
John Larkin         Highland Technology, Inc

lunatic fringe electronics
• posted

I may cough up the \$11 if they'll send me samples of all of them. Certainly I'll try them out.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
• posted

Whoa! You're wondering if there's a 4th edition? Paul and I spent 20 man-years writing the 3rd ed.

--
Thanks,
- Win
• posted

I'm sorry, I didn't mean to be insulting. I just wanted to make sure I purchased the latest and greatest. My memory is not very good these days. I just looked and my copy is 2nd edition, so I should have looked first before asking.

Again, I apologize.

• posted

Okay, samples of 5 different devices ordered, same \$11 fee. Now I just have to build a simple noise meter.

I'll probably use a 1 kHz pilot tone oscillator and a really low-Z voltage divider on the gate, another low-Z divider on the source, CMOS mux lock-in to detect the pilot tone, a 3-stage RC lowpass filter plus AM detector for the noise, and an analog divider. (I have a bunch of well-aged AD534s gathering dust in the drawer.)

That way I can calibrate it in nV/sqrt(Hz), with banana jacks for connecting to a nice big analog meter, like a Simpson 230 or an Avometer Model 8 Mk IV.

If I feel fancy, I might put in a second lock-in so I can measure the spot noise, but that's a bit of a calibration headache.

I've been looking for an HP4470A transistor noise tester for a long time (Win and Paul suggested it in their earlier editions) but none has appeared on eBay in the last decade or so. So I guess they're all dead now, and I'll have to build my own.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
• posted

Is the pilot tone to measure gain? Why not just crank the level up?

A spectrum analyzer or a scope FFT or a sound card could get you the noise spectrum, with a pilot tone used to calibrate gain. Pity that most spectrum analyzers don't work below 9 KHz.

Your problem is that you're not lazy enough. It takes a lot of hard work and experience to be efficiently lazy.

--
John Larkin         Highland Technology, Inc

lunatic fringe electronics

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.