JFET Match Phase

[spoontex]

Hello,

I have 8 2n5952 and I try to match them for a phase 90.

I use this matcher:




And this are the values:

1= 1.234

2= 1.421

3= 1.360

4= 1.742

5= 1.724

6= 1.618

7= 1.391

8= 1.768

They are good for a phase? or the values are a bit mismatched.

Thanks.

[Mark Hammer]

#1, 2, 3, and 7 make a decent quartet, and 4, 5, 6 and 8 make another group of 4. 

In my experience, the matching helps with "the turnaround".  That is, all JFETs continue to change drain-source resistance at the top and bottom of the sweep, rather than, for example 3 continuing to change and one of them reaching some maximum or minimum resistance before the extremes/ends of the sweep.

When MXR released the Phase 90, there were no 4 and 5-knob phasers being sold.  The P90 had a Speed control and nothing more.  So, they picked a feedback amount, and a sweep-width amount that would work well-enough for all Speed settings in all production units.

To some extent, I think poor matching of JFETs can be compensated for by reducing the sweep width and adjusting biasing.  The resistor on the output of the LFO is 3M9 on some issues and 3M3 on others.  It sets the width of the sweep.  Lower values let the LFO sweep wider, and larger values reduce the sweep.  On my builds I have gone a bit lower than 3M and as high as around 4M.  The change in sweep width is easy to hear when you turn the speed up.

Let us say that the LFO output provides a current that goes from 1 to 10 (I'm making up numbers here).  Some of the JFETs will show no drain-source resistance change below 2, and some show no change below 3, so the bias trimmer brings them all up to 3.  Some will show resistance change from 3 to 7 but stop at 7, and one of them will show change all the way up to 9.  You don't want any of them to stop changing resistance during the sweep, so if you keep the sweep between 3 and 7, all JFETs will participate and the sweep will be smoother and more "musical".  Yes, you will waste some of the possible range in order to have all JFETs sweep in unison, but this will allow JFETs that are not "perfect" to produce a pleasing P90.  So, you might be able to have a wide sweep (i.e., reduce the resistor value from 3M9 to something a bit lower), but you might need to keep the sweep a little more modest (increase it from 3M9), depending on how much usable range the JFETs share.

In other words, even if the JFETs are not perfectly matched, as long as you ask them to do something they can all do equally well, it will sound good.  That might be a bit less than what a perfect set might be able to do.

[antonis]

I just came to argue with Mark..   

4,5,6 & 8 should make a more decent quartet..
(from mean numerical deviation point of view..)

Totally agree with the rest of analysis..!!
(my respect and greetings Mark..)

[spoontex]

Hi Mark!!

Awesome explanation. It's a pleasure!!

Many thanks!!