[sdiy] FET for VCO cap discharge

[Ian Fritz]

Hi JH --

Thanks for the report on discharge FETs.  I agree that if an inexpensive 
common part works, then that's the thing to use.  But I'm having trouble 
understanding your result, at least based on the data sheets.

The BF256C is denoted as a VHF/UHF amplifier, whereas the 2N4391 is a 
switching device -- therefore they are spec'ed differently.  I have only 
ever tried switching devices, although other people have reported using 
small signal ones.

According to the data sheet I have, the 256 has a zero-bias drain current 
of under 20 mA.  They don't spec the maximum drain current, but let's 
suppose it's 20 mA.  They don't spec the drain-source resistance, either, 
so let's just call it zero.  Then the time to discharge a 10 nF cap is
t = CV/I = 2.5 usec.  Your result suggests something closer to 1 
usec.  Doesn't that mean you're driving more that 50 mA through 
it?  Perhaps under pulsed conditions this is possible!

For the 4391, let's just assume that the decay is determined by the RC time 
constant.  For a 30 Ohm series resistance, five time constants is t = 5RC = 
1.5 usec.  So three in parallel should work OK.  This calculation assumes, 
of course, that the FET can actually carry the peak current, which it may 
not.  But as a rough estimate, 5V / 30 Ohm = 160 mA is not much over the 
capability of a good device.

So I'm obviously missing something important here.  Any FET experts who can 
straighten me out on this?

   Ian


At 07:55 AM 10/22/05, JH. wrote:

>For this I need a precise saw waveform with less than 1us reset time,
>with a frequency range from < 0.1 Hz to > 16kHz.
>
>I started with this circuit (Hi Ian!):
>http://home.earthlink.net/~ijfritz/sy_cir2.htm
>But I wanted a 10nF capacitor instead of the 1nF, to stay away from leakage
>(which might distort the waveform) at very low frequencies.
>
>This is quite a demand on the discharge transistor, considering I wanted
>a JFET.
>
>I found that the 2N4391 is not good enough for this. (It's great for
>1nF, but for larger caps, I'd have to increase the discharge time,
>which I didn't want.)
>
>So I started paralleling devices. This works. But I also found a
>quite common JFET that is better than the 2N4391 in this
>application: the BF256C.
>Two BF256's in parallel were enough to discharge 10nF in
>Ian's circuit in ca. 500ns. I put in 3 in parallel just to be sure.
>
>So I thought I'd pass this on - the good old unexpensive BF256C
>may be a very good choice in that kind of oscillators.
>
>One caveat, though: JFETs have huge tolerances, so I
>_might_  just have run into some relatively poor 2N4391's,
>and into some relatively good BF256C's. I just had
>3 of each to test.
>
>The circuit from Ian's site is obviously good enough to drive
>3 gates in parallel. I've used CA3140 for the opamp and
>LM311 for the comparator.
>Ultra-precise tracking is not so important in my application,
>and of cource I had to leave out the compensation
>resistor R19.