what to do with the substrate on a mono. dual JFET?

Help Support GroupDIY:

mikep

Well-known member
Joined
Feb 18, 2006
Messages
450
Location
Philadelphia
Im looking for a refrence that talks about this. never had to deal with it before. linear systems doesnt mention it on their data sheet. (i just got some LSK389 samples btw)

should you: leave it un-conected? bypass to ground with a cap? bias it with a clean DC voltage? drive it with a level-shifted signal-derived voltage? what is the ideal situation here? any pointers? thanks

Mike P
 

Samuel Groner

Well-known member
Joined
Aug 19, 2004
Messages
2,935
Location
Zürich, Switzerland
Floating it will almost always work, but bootstrapping might offer interesting possibilites in reducing substrate capacity. No experience on this though, I'm sure Brad can offer some.

I guess we're talking about differential inputs?

Samuel
 

mikep

Well-known member
Joined
Feb 18, 2006
Messages
450
Location
Philadelphia
[quote author="Samuel Groner"]
I guess we're talking about differential inputs?
[/quote]

yes but not just differential. I was also thinking of using both sides in parallel at the front of a CFB amp. I havent protoed that yet. I have a diff-input amp built up in front of me, with the substrate floating. how easy is it to fry the fets by experimenting with substrate bias?
 

bcarso

Well-known member
Joined
Feb 20, 2005
Messages
4,055
Location
San Fernando Valley, Los Angeles
I have indeed worked with this issue.

In the diff amp configuration with lots of global series feedback the substrate tends to be bootstrapped pretty effectively just by letting it float (which btw is Toshiba's explicit recommendation in the (obsolete) 2SK389 and 2SJ109 datatsheets).

When used single-ended and paralled the same thing will happen with series feedback, and even very local feedback (a resistor in the source(s) for example >> 1/gm) will help a lot to reduce input C.

In other configurations with those parts, such as the popular voltage follower where the upper source follower is supplied current by the lower device's drain, with usually each part having a small R in series with the source to limit the current in the lower to less than Idss and in the upper to provide a matched compensating output voltage offset, a significant fraction of the input capacitance comes from the substrate. It is then useful to drive the substrate actively. However, the breakdown voltage has to be determined. It appears that the Toshiba parts have back-to-back isolation diffusions/implants, as I get no diode conduction from the substrate to any of the other electrodes, and the breakdown voltage is pretty high (>100V for what I tested). So those I tested would be difficult to fry. There is no specification though, so YMMV.

I have not played yet with the LIS parts, so I don't know for sure what they are doing besides the interdigitated structure, but when I spoke to them after the release of their LSK170 and they mentioned the dual coming up, I suggested that they provide a substrate connection and cited why. Either the guy I spoke to didn't know that they were going to anyway, or they changed their minds, as he told me there would not be one. But lo and behold there it is, and good thing for it.
 

bcarso

Well-known member
Joined
Feb 20, 2005
Messages
4,055
Location
San Fernando Valley, Los Angeles
It might be sufficient, depending on how much loading of the follower output you have, to simply tie it directly to the upper Q source. If for some reason the LIS parts are simply reverse-biased diode substrates, then bias the substrate sufficiently negative with a resistor and negative voltage and cap-couple, giving due consideration to maximum voltage swings.

I had initially planned on having a low-C auxiliary FET follower to drive both it and a cascode FET in the drain of the upper follower Q. At one point I even considered a tube as the auxiliary buffer :green: owing to the smaller C modulation with voltage swing.

But the bootstrapped cascode using a higher pinchoff FET seemed to work so well at audio frequencies that I was talked out of the more extreme approaches.

When an uncompensated cascode affair of this sort was used as a quick-and-dirty solution so my friend could demo his speakers, some of the golden ears complained that it sounded a little sluggish or something. This was of course after they were told what was inside the box, which makes me very skeptical ("oh yeah, those parts are really big" [referring to the 389s]). But I did agree that the input C was higher than expected, and later realized it was Mr. Substrate, or about half of him, due to the lower device being given no advantage of driving voltage.

To keep things in perspective, I believe the net input C was about 20pF. Normally something fed from a typical hifi line-level source would not be particularly concerned about 20pF. But then it will have a voltage modulation which will introduce distortion. I suppose if it were a tube preamp feeding with a high output Z it could be a marginal effect.

As time went on the complaints went away and the dealer who has the one-and-only system says it sounds great now (perhaps after the ritualistic "breaking-in" :razz: ). Since then I have done literally thousands of hours of uncompensated design/development of systems to supersede it.
 

Samuel Groner

Well-known member
Joined
Aug 19, 2004
Messages
2,935
Location
Zürich, Switzerland
It might be sufficient, depending on how much loading of the follower output you have, to simply tie it directly to the upper Q source.
Just to make sure: to the source and not the output (i.e. after the offset resistor)? The source looks better as being lower impedance.

As time went on the complaints went away and the dealer who has the one-and-only system says it sounds great now (perhaps after the ritualistic "breaking-in" :razz: ).
Or perhaps they used a different power cord? :green:

Samuel
 

bcarso

Well-known member
Joined
Feb 20, 2005
Messages
4,055
Location
San Fernando Valley, Los Angeles
Yeah I would go directly to the source.

Yes audiophillery sigh...I just finished the simplified schematic and sim of a highpass stage. It uses five 6C45 supertriodes. Mixing sand and hollow would make it quite a bit simpler, but I am told that people feel that sand-state current regulators contaminate the tube sound.

At least it will glow prettily and warm your room.
 
Top