ampage Tube Amps / Music Electronics	For current discussions, please visit Music Electronics Forum.

ampage archive

Vintage threads from the first ten years

Millenium Bypass Redux

:

4/22/1999 6:46 PM

Millenium Bypass Redux

When I designed the original Millenium, I considered using Nchannel enhancement MOS devices before using the JFETs. However, I reasoned that the delicate MOS gates would be connected directly to the wires for the switch, and as such vulnerable sooner or later to static electricity death. The gate is clamped only to the +9V supply, not on the ground side, so negative transients would kill the MOS device.

And so I went with JFETs even though they have the problem of not quite turning off completely sometimes (you can get around this by putting a resistor in parallel with the LED...) depending on the JFET and the LED.

However, a reader from Sweden prompted me to reconsider this by asking why I didn't use the N-channel MOSFET devices.

My mind put this together with another fact that I picked up since I did the original Millenium. The central point of the Millenium is that reverse biased diodes have an almost constant and very tiny leakage, so the pullup current can be supplied properly. I resisted using two diodes on the original Millenium, one from the gate to +9, one from ground to the gate, because the leakages might or might not be proper, in series as they would be, and the gate operation would be unpredictable, even if you could now not kill the device.

Since then, I have found that signal diodes like the 1N914 are gold-doped, and have real leakages much smaller than the few nanoamperes they are specified at. With a "normal" diode like a 1N4001 to +9 and a 1N914 from ground, the gate can be both protected and the leakages will probably make it function as you note, and now you can protect the gate with the two diodes and a series resistor.

I'll be testing this and if it goes as I believe, we'll have a Millenium II, improved by having no residual LED leakage.

More to come later.

4/23/1999 6:46 AM

Didn't get the operation I thought I would, so I went back and checked my sources. 1N914's and 1N4148 have actual leakages of about the spec value, 10-25 nA. It was normal diodes that have much less leakage than spec. I had to reverse the two diodes, putting the 1N914/1N4148 from +9 to gate, and the normal diode from gate to ground. This seems to work much better.

There's a bit more tinkering to get it perfect, but it's close. The difference between two diode leakages will probably work as well as just a diode's leakage, and will make for more "off" LED's.

4/23/1999 9:12 AM
Jack Orman	If more leakage current is needed, you might consider a small value tantalum capacitor which should have more than a diode. -Jack

4/23/1999 9:28 AM

There are several ways to get more leakage, including germanium diodes.

I did test a 1N34A. It leaked about the same current as a 10M resistor, 'way too much. A capacitor would of course slow down the transition too much.

The difference between gold-doped signal diodes like the 1N914/1N4148 and normal diodes or even low leakage jinctions like the collector-base of a transistor is about right. The low leaker is one to two orders of magnitude less leakage than the gold doped ones, so it has no real effect on how fast the FET switches, and the two diodes protect the gate, so you can use a MOSFET instead of a JFET and get more off offs.

Really, they're two variations of the theme, and probably both have their applications.

5/1/1999 6:38 PM

One idea to consider is: that since it is desirable to have low leakage on the bottom diode (to ground), an FET gate-source junction could be used in place of that diode as protection. While any FET is going to have quite low leakage (probably lower than the 1N914), the 2N4117A is especially suited to this use due to its small junction - leakage will run around about 0.1 pA or so.

This doesn't add to the parts count and only puts a few cents on the cost. It also allows a wider selection for the other diode, will ensure reliable switching, and still provides protection for the mosfet.

best regards, Jack

5/1/1999 9:03 PM

I appreciate your consideration.

The real poop is this:
1N914, 1N4148 and their ilk have gold doped junctions to enhance recombination in the depletion region of the junction. This does two things - first it cuts the switching time dramatically, and second, as a side effect, it increases the nominal leakage to the relatively huge (for silicon) range of 10-20nA.

Normal non-gold doped silicon junctions are usually much lower, and the smaller the junction area and the more lightly doped they are, the lower the leakage is. Very cheap NPN bipolar silicon transistor collector-base junctions are lower leakage than most power diodes because the junctions are much smaller. Junctions that are very lightly doped are even smaller. JFET gate-channel diodes are in general better than bipolar base-collector junctions for this reason. JFETs especially made for low gate leakage are even better.

The bipolar collector-base junction is, however quite good enough when combined with the gold doped (and cheap) 1N914/1N4148 style diodes. In my judgement, a very cheap and highly available part that did the job was a better choice than ordering a special low leakage part to get that last 0.001% of perfection. While 2N4117A's are not all that hard to find, 2N3904's are even easier and cheaper, and may in fact be free, found in the sweepings on the workbench. But I do encourage you to try different devices in any Millenium clones you try out.

Speaking of which, I notice that my post about the need for a higher leakage diode at the supply and the low leakage diode at the ground side of your CMOS chip variant of the Millenium-2 style switch with a MOSFET input seems to have vanished from AMZ's discussion board.

Did you have some technical difficulty that lost some posts?

5/2/1999 12:47 AM
aron	R.G., I believe that Jack archived the posts - I think... Aron

Page 1 of 2