sot jfet

Try ON, Zetex and Philips for starters. Then let us know if you find more.

--

Tim Wescott
Wescott Design Services
http://www.wescottdesign.com

Thanks, guys, but false alarm.

I've got this box full of programmable time delays. Each one is a PNP transistor current source driving a capacitor, reset by a gaasfet, making about a 60 ns linear ramp into a comparator with a 16-bit DAC on the other side, so I get 0-60 ns delay in about 1 ps steps. The current source is a BCX71 PNP inside an opamp feedback loop (fb from source resistor off +15 rail), conventional stuff. Like an idiot, I left out a series base resistor so the PNP likes to oscillate at around 100 MHz, causing a bunch of jitter in the delays. These are very dense, rather pretty boards and I'd like to find a fix that's not an ugly kluge. A jfet would work, but the SOT-23 jfet pinouts don't match up with the PNP.

So the choices are either a small p-ch mosfet, if I can find one with low enough capacitance, or one of those "digital transistors" with a built-in base resistor. Or a really slow PNP, if anybody still makes one.

So listen up, kids: ALL EMITTER FOLLOWERS NEED BASE RESISTORS unless I give you permission otherwise.

John

Thanks, but no. The graphs show Ft as around 300 MHz in the 5-10 mA range, almost the same as the BCX71. It will probably oscillate too. This is for a $6 billion laser, so I can't take chances.

I had thought that the output impedance of the opamp would be enough base degen to prevent oscillation, but I guess the wire bonds, esd diodes, and the output transistors can form a high-enough-Q circuit to sustain oscillation. Obviously so, in retrospect.

If I ask the production folks to lift the base lead and stick an 0402 resistor in there, they'll be after me with torches and pitchforks.

John

John Larkin wrote: (snip)

Like this?

--
John Popelish

It occurs to me that the transistor may be feeding the oscillations because it is too slow to maintain the current steady, but feeds back collector voltage to the base. Have you tried a much faster transistor lower gain, lower collector to base capacitance, but higher fT)?

Something like:

--
John Popelish

IIRC Central Semi does too.

--
--
kensmith@rahul.net   forging knowledge

The feedback loop itself is stable?

--
Reply to nico@nctdevpuntnl (punt=.)
Bedrijven en winkels vindt U op www.adresboekje.nl

I used the BCX71 because in the past the RF-types (NEC, or the Moto parts when they still made them) were *really* bad about oscillating.

My first experience with this problem was a long time ago, when I needed a TTL powerup reset signal. I did this...

+---------+--------- +5 | | r | | c +--------b 2N2219 | e | | c | | +----------ttl gate------> reset sig gnd | r | gnd

which never made a high to the ttl gate, it was oscillating so hard at

100 MHz or so. A small base resistor, 33 or 47 maybe, generally fixes it.

So most small-signal bipolars are hazardous when operating as emitter followers when the base sees a low-Z source. As the transistor gets faster, things seem to get worse, with the traces and wirebonds to the base becoming microwave resonators.

Considering how often this seems to happen, at least to me, it's surprising how seldom it's mentioned. The AoE example (p 300) is similar, but blames the oscillation here on impedance in the collector circuit.

John

Yes. The opamp Ft is only about 3 MHz, and the oscillation is around

100. It's definitely local to the transistor.

John

The emitter is a 500 ohm resistor, so it's low-Q already. And there's already a ferrite in the collector! Base resistance is the solid fix; I just got sloppy and left it out.

So it's a mosfet or a digital transistor, and a pilgrimage to a Holy Shrine for penance; Fry's maybe.

John

I would love ot have seen the layout that this circuit was made with that oscillated so vigorously. I have been designing and laying out circuits that contained various uses for emitter followers, and have had only a few bad experiences with them. And in every case, I was able to cure the oscillation with a better layout. Maybe I have just been very lucky.

Would it be possible for you to post a photo or plot of the layout of the part of the circuit that is giving you fits, now?

--
John Popelish

In article , John Larkin wrote: [...]

The output impedance of an op-amp is alway such that it makes oscillation more likely. They tend to look like parallel LC circuits. The L part is because as the frequency increases, the servo loop gain is dropping causing a rising impedance. The C part is the various stray capacitances.

Can you deQ the emitter side of things or slip a bead in the collector's path?

--
--
kensmith@rahul.net   forging knowledge

I read in sci.electronics.design that John Popelish wrote (in ) about 'sot jfet', on Sun, 16 Jan

2005:

Inductance in the emitter lead creates a negative impedance at the base. I expect that's why the thing oscillates. Inductance in the collector lead could possibly also cause trouble, but the oscillation is often less persistent.

--
Regards, John Woodgate, OOO - Own Opinions Only. 
The good news is that nothing is compulsory.
The bad news is that everything is prohibited.
http://www.jmwa.demon.co.uk Also see http://www.isce.org.uk

John Larkin wrote: (snip)

collector

To be fair, that example had a d'Arsonval meter movement in the collector lead.

-- John Popelish

I cant prove it without doing the maths, but I suspect an analysis similar to that given in the Siliconix MOSPOWER app handbook will show emitter followers to be inherently unstable, just like FETs, when all the parasitics are taken into account. If you havent read the paper, its a good one - rather than evil number crunching a Routh-Hurwitz stability analysis is performed. very nice.

Cheers Terry

In article , Terry Given wrote: [...]

Perhaps more like "really want to be unstable". You have to connect any of a large number of combinations of stray Ls and Cs to make them oscillate.

Consider:

Vcc ! +-- ! ! !/ --- ---[L]---+---! --- C2 ! !\e ! --- ! ! --- +--+----- to load C1 ! ! -----+ !

This is a Colpitts common collector oscillator. If the Cbe and Cce make workable values, the transistor will oscillate with just the inductor. With fast transistors not much trace inductance is needed.

or:

Vcc ! [L] +-- ! ! !/ --- ---------+---! --- C2 ! !\e ! --- ! ! --- +--+----- to load C1 ! ! -----+ !

This is a Colpitts common base oscillator. If the Cbe and Cce make workable values, the transistor will oscillate with just the inductor. With fast transistors not much trace inductance is needed.

--
--
kensmith@rahul.net   forging knowledge

No. Looks like the best kluge is to lift the base lead and solder an

0402 resistor from the lead to its pad. I do have some p-ch mosfets and some pnp "digital transistors" coming in tomorrow, and one of those may be a nicer fix.

I will *never* make this mistake again.

John

$10000 / 1500 = $6.66 per board, pretty cheap. He got off easy.

Why didn't somebody stand up and call him a stupid fathead? Better yet, why didn't everybody call him a stupid fathead?

John

Is there a via feeding the base of the culprit transistor? If so and if you have space and if the board is 0.063" (standard) thickness than drill out that via and drop in 0603 size base resistor.

Regards,

Boris Mohar

Got Knock? - see: Viatrack Printed Circuit Designs (among other things)