Mixing fets with similar Vgs in a Phase 90

[mac]

I had a lot of troubles with the Phase 90, distortion mostly, still do not know why. Maybe the breadboard?

Anyway that force me to do a thousand mods, including using lots of fets with different Vgs, bf245a, 2sk246, 2sk30, 2sk117. All with Vgs below one volt.
I have 3 almost identical mpf102 and one bf245c, all with the same Vgs=3.17v at 10k. I put them in the breadboard, set the zener to 6v, and to my surprise they phase nicely, wide and deep. To be honest I was not expecting any good result.
I used RG matcher and tested them with different values of Rset, 1k, 3.3k, 6.8k, 10k, 22k. All 4 measured almost the same with the 5 resistors. I wonder if it is a happy accident or as long as Vgs is the same across the whole range it does not matter what type the fet is. I mean, Vgs more important than Yfs, similar in this case, and other properties.
Thoughts?

mac

[R.G.]

Mac,
It is somewhere between a happy accident and similar conditions forcing similar results.

What you want to match in a phaser is rds, the AC resistance between drain and source, for all Vgs that the LFO puts out.

This can be done, but it is complicated and time consuming. It also requires a much more complicated JFET matching setup to do the matching. I deliberately designed the JFET matcher to be as simple as I could make it and let most people get good results.

The off resistance of all JFETs is the same - essentially an open circuit. This is reached at Vgsoff, which varies from a fraction of a volt to maybe -12V or more. The on resistance of JFETs is not the same, but is usually between 1K and 10 ohms. This is reached at Vgs=0. My circuit selects JFETs that have the same DC current flow at the same Vgs. This is related through the semiconductor physics to the rds on. It is not perfect, but it's close, and it's about as easy as a good test can be.

If you get similar JFETs, even if they are not the same, and select for ones with the same Vgs at a certain drain current, you are selecting for devices which have a resistance versus Vgs curve which passes through the same point. the slope of their rds versus Vgs may be different, but all the ones you selected are the same at that one point. If you further measure them with a different set of Rset resistors, you are selecting for devices which have similar slopes to the rds versus Vgs curve. So it's not terribly suprising that they should come out very similar, even if the device numbers are not the same.

So you have a result that's part happy accident (in that you had devices which happened to do that) and part that you selected for similar conditions. It certainly does not follow that same-Vgs devices of all types will have the same rds in the circuit. However, the more points you force to match in a DC test, the more likely the devices are to be the same in an AC test.

Vgs is important to matching JFETs for phasers, more important that Yfs. But what you REALLY want to measure is AC resistance at several places on Vgs. The more points you match rds and Vgs, the more identical the JFETs are, no matter what number is stamped on them.

[mac]

Thanks RG.
I'm going to measure one mpf102 and the bf245c from 100R to say 47k-100k at regular intervals and post the results for educational purposes.
The good news is that with a little patient and some luck, different brand fets we thought useless could match for at least a Phase 45.

mac

[Gus]

Reading this and looking at the tester at Geofex
RG why not make one of the voltage divider resistors switchable to do a two point or more match

[R.G.]

Yeah, I thought about that.

Primarily because the results are almost always good enough if you have the same type number JFETs and get a one point match, and it's much simpler for people who are pushing the envelope to remember which end of the soldering iron to hold.

There are a lot of elaborations that could be done; a good one would be to actually measure AC rds at a given Vgs with a representative AC voltage, maybe 500hz/100mV. A really good tester would measure Vgs for two or more rds values. Even better, you could match the curves of rds per Vgs, which is REALLY what you want. But I doubt that more than three people would ever use something like that.

[mac]

But I doubt that more than three people would ever use something like that.

To be honest, I'm not one of them.

Before finding RG matcher I used to compare fets connecting them as a buffer, drain to Vcc, gate to 1M to gnd, source to Rs to gnd, and measured voltage across Rs for different values of Rs.

One con, pinout is a problem for those who buy traced boards.

mac

[RedHouse]

Reading this and looking at the tester at Geofex
RG why not make one of the voltage divider resistors switchable to do a two point or more match

Our brethren over at prodigy_pro have been there already:


http://www.prodigy-pro.com/forum/viewtopic.php?t=4531&highlight=matching

one post mentions RG's rig, and of course good 'ol PRR up's the ante' as is his way. 

Sure we have our local hero's over here but it's good for us to get out once in a while, run on over there and mingle with these guy's now-n-then they've got alot of good brainpower at work over there.

[Gus]

I am aware of "the lab"

I was posting about a simple change for R.G's circuit that people here might be more aware of because of the link to geofex (top left). Yes there are other test circuits one can build.

[mac]

I measured the transistors with RG matcher. I just simply changed the source resistors, but I guess one could replace Rset with a pot, measure the resistance, power on and then read vgs. Or as I did, breadboarded the matcher and replaced resistors on the fly.

BTW, I measured as low as 10ohm... yes, I overheated the bf245C... 2.45v/10 = 0.245mA across the source for about 30secs or more until I smelt something. Pw=0.6w,  2x the max power dissipation of the fet... ¿goodbye P90?
Does overheating have some consequences? I could continue testing so I guess nothing bad happened.

Rset             Vgs BF245C Vgs MPF102 10 2.45 Not again!!! 22 1.32 1.22 (Kamikaze) 33 0.52 0.44 50 -0.38 -0.45 100 -0.85 -0.85 220 -0.76 -0.74 330 0.08 0.25 390 0.72 0.78 470 1.12 1.20 680 2.00 2.00 1k 2.60 2.58 1.2k 2.87 2.63 1.8k 3.10 3.09 2.2k 3.13 3.12 10k 3.17 3.17 inf 3.17 3.17 [/tt]

They are... or were... almost identical. They are both from the same brand, maybe this helps.

Now if you excuse me I have to see if the fet is ok...