Small Signal Depletion Mode FET

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

Samuel Groner

Well-known member
Joined
Aug 19, 2004
Messages
2,940
Location
Zürich, Switzerland
Hi

I'm looking for an easy to source small signal depletion mode MOSFET (or JFET, if it exists) to build self-biased current limiters (100 uA to 10 mA range) and to cascode JFETs (at about 3 V source-to-drain with a few mA drain current). Should be TO-92 package, come with at least 60 V breakdown voltage and preferably low capacity. What I found so far is (it seems not to be easily sourceable though): DN2540

Well, that's it--any suggestions?

Samuel
 
Infineon has a 250V depletion mode MOSFET, BSS229, although the threshold voltage seems too low for your app. Nonetheless it is a part to know about I suppose.

I'm doubtful Toshiba still makes them, but the 2SK373 is a 100V JFET with other specs on the high end of the Idss bracketing that might work. I found it appealing because based on the curves there is a zero tempco point with a useful amount of current for some typical device.

It looks like Interfet still makes the 2N6449 and 6450 JFETs, but their minimum order is US$250. This is an interesting company though, catering to the nuclear science community among others.
 
Thanks--I've seen the Toshiba part but Idss is really a bit low, I need 5 mA current limiters and want some degeneration.

It looks like Supertex is the only company offering a wide range of small-signal depletion mode parts: d_mode_mosfets_SG_device.pdf

I just checked some Apex opamp pictures and they use DN3545. This part and the DN2540 are available from Mouser, so this might be the way to go. At least until someone suggest another part!

Samuel
 
Samuel, did you ever get one of these depletion-mode DMOS parts and happen to measure the voltage noise? Also Rowan, I think you have mentioned using these as active loads?

A famous personage in a private comm asserts that an I source (or I limiter as some prefer) made from these (DN2540) parts has exemplary performance, including low current noise, presumably. I have not seen the circuit per se, but I was told it only uses two FETs (both DMOS) and a couple of resistors. So one would suppose it's a composite two-terminal cascode FET structure, with one resistor in the source to set the current based on the threshold voltage.

Since MOS has a reputation for high 1/f noise, moreover setting in at fairly high "corner" frequencies, I am somewhat skeptical that such a circuit will have particularly low noise. But technology has gotten steadily better over the years, so who knows. I ordered some samples of the recommended substitute you mentioned, the DN3545, to test. But I'm also curious about others' experiences.

The other interesting thing from the datasheet for the 2540 is the behavior shown for various device capacitances versus voltage. Frankly, after a few tens of volts the curve looks "drylab"---like the drafter said Oh this is close enough to a zero slope that I'll just make it one.

The majority of the capacitance from drain to source for these devices is the body diode's. Although the big changes in this C occur at low voltages, according to the datasheet and theory, that same theory does not say C becomes constant at high voltages. When I plug other DMOS parts into sim I see the signature of the C getting smaller and changing less at higher voltages---but the slope doesn't ever get to zero before reaching the specified breakdown voltage.
 
[quote author="mikep"][quote author="Samuel Groner"]T
I just checked some Apex opamp pictures and they use [/quote]

interesting. where do I find those pictures? I just looked at the Apex website. lots of good info there, each of the high-volt opamps seems to have a (unique) representative schematic on the data sheet. eg:
http://apex.cirrus.com/en/pubs/proDatasheet/PA83U_P.pdf

expensive stuff![/quote]

Yes, I'd like to know where such a picture would be too, as all I know of that outfit is those pricey hybrids, which AFAIK do not come with descriptions by part number of their innards.

I shall be getting some samples of the DN3545 tomorrow and should be able to do a crude but reasonably indicative measurement of their noise.

Meanwhile I wonder where Mr Groner is hiding? In the middle of exams perhaps?
 
Rowan, thanks. EDIT: I'm checking that part out---interesting---it looks like about the smallest area DMOS part I've ever seen.

I got ten pieces of the larger part mentioned by Supertex as a sub for the DN2540 (DN3545) today and preliminary results from bench testing indicate that the noise is less than I expected from being a MOS part, perhaps around 2.4-3 nV per sq rt Hz midband; significantly higher at low frequencies, but the major noise action is mostly infrasonic, so of no great consequence for audio use.

Since the threshold voltage is about 2V, there will be some significant noise from the resistor used to set the current, for a fairly simple circuit using the threshold voltage as the "reference".

I made an amp with one FET as the amplifier, with a bypassed source resistor setting the current to about 18mA, and a similar stage using a second device as current source load. Each device had about 10V across (overall a 20V supply).

The output of the amp thus made had an inverting gain of about 2700, and an output impedance of about 12k. I looked at successive results from loading the output with different values of capacitance, to get an estimate of where the low frequency noise dominates.

I notice that the DN3545 is a little higher capacitance than the DN2540. This probably means that the 2540 is a somewhat smaller area device and possibly a little higher noise.

Either way, I am impressed by how quiet these MOS parts are. I still think, based on calculations, that more complex circuits with bipolars and/or JFETs could be still relatively advantageous---but although not more expensive, they are definitely more complex.
 
Did you ever get one of these depletion-mode DMOS parts and happen to measure the voltage noise?
No, Mouser messed my order (DN2540) up. Just ordered again today.

Where do I find those pictures?
I think they removed it from the webpage once they got eaten by Cirrus. Here we go: APEXMP108MP111PRESSPHOTO.jpg

Meanwhile I wonder where Mr. Groner is hiding? In the middle of exams perhaps?
No, exams were passed two weeks ago, which meant time for some holidays!

Samuel
 
By the way, the threshold voltage for the 3545 parts measured so far is about 2V. So if used as a driven-gate stage in an input circuit, the JFET had better be low pinchoff, which of course the SK170 eminently is. Some other biasing arrangement could be devised, but then the simplicity of the FET arrangement is lost.

At least the DMOS can handle plenty of voltage on the drain.

I watched the output voltage of my test circuit from time to time on the scope yesterday. I saw some excursions in the range of 1.5uV (referred to the input) at times, with a passband defined by the scope LF cutoff at 1.6 Hz and a simple lowpass at about 26 Hz. They did not appear to be due to power supply fluctuations or thermals. So things do get pretty bouncy as you go low. Again, perhaps of no consequence for audio apps.
 
Just did some measurements of the output capacitance of a DN3545 sample in the off state. I biased the gate with a 9V battery, cap-coupled the drain to a C meter, and ran a bias voltage into the drain through a 10M resistor. One uncertainty is the stray C in the setup, which would require a "dummy" part in place of the FET to get precise but which I estimate at 2 pF. The results depart appreciably from the datasheet's for Coss. The estimated stray fixture C has not been subtracted.

Vds ..... C
0 ...... 130pF
10 .... 28.2pF
15 .... 17.8pF
20 .... 15.2pF
30 .... 12.7pF
40 .... 11.4pF
50 .... 10.4pF
100 ... 8.3pF
200 ... 7.1pF

EDIT: PS, the cap meter excitation is about 400mV p-p.

Also: the motivation here is to assess distortion mechanisms from voltage-variable output capacitance, particularly that not readily correctible with certain techniques under study.
 
I tried out a few different parts as I was developing my tube preamp module. I've used the Supertex DN2530, DN2540, maybe the DN2535, in both TO-92 and SOT-89 packages. I also tried a Clare part in an SOT-223 package but I can't find the part number right now. I got all of these from either Mouser or Digikey, so they're very much available. I had pretty good results with all of them, and I'm using the 2540 in the SOT-89 package for production thus far. I killed a few in prototyping, but in general they do seem pretty durable. I suspect this part is probably the most significant noise contributor in my preamp, but I'm not entirely sure just yet. There does seem to be a bit of subsonic junk, but it doesn't seem to matter much. Overall, the noise performance of my preamp is pretty decent so I'm not too worked up about it.

Other than that, these things pretty much act like high-voltage JFETs as far as I can tell. They're similar in that their Vgs varies a bit, and varies even more with temp, and you pretty much have to trim them. It seems kind of silly to have a bias adjustment on a preamp tube, but when you're trying to squeeze modern preamp specs out of a single bottle, you have to optimize everything.
 
Thanks for that information Justin.

One could establish the contribution of the FET to the noise pretty easily by substituting a resistor for the plate load, one with low excess (current) noise, just for an experiment, and accounting for the gain reduction. The experimental plate supply would have to be quite quiet.

It sounds like an interesting product in any case. Good on you to get it into the small package (I'm assuming this is your RMS5A7).

EDIT: to put the importance of the current source in perspective, note that for the rule-of-thumb triode noise resistance of 2.5/gm, a 7500uS gm tube (typical 6N1P for example) will have an output current noise above the 1/f region of about 17.6 pA/sq rt Hz. Since noise sources add root-sum-of-squares, a few-times-lower I source noise will be a small contribution. If the FETs discussed are about as quiet as I measured this condition should be achievable. It certainly is with some more complex I source designs, which should also contribute less distortion.
 
I'm actually not using them as a plain old current source in my circuit (yes, it's the RMS5A7) so I can't just swap it out for a resistor. I do find that varying specimens of tube (12DW7 for those keeping score at home) have a fairly minor impact on the noise floor, which leads me to believe the tube's noise is being swarmped by something else. And there ain't much else in there. But like I said, it's still pretty good so I'm not too worried about it.
 

Latest posts

Back
Top