Jfet matching for phase 45 build

[Locrian99]

Hey,

So my next project is going to be a phase 45.   Started doing the Jfet matching using this:




I can't help but feel like I'm doing something wrong here.  I decided not to use the switch and just remove the 100 ohm resistor, and jumper from the gate to ground.   It looks to me like that is all the switch is doing.   Which then makes sense that my two reading are very similar.
Within a few mV.






Lastly if this is correct how close am I aiming for here.   I'm assuming two at 1551mv and 1554mv would be considered matched.   But would 1538 and 1547?

Thanks again, you all have been super helpful in me getting started in this.

[Locrian99]

Oh the switch makes it so it by passes the 1M resistor I think. 

[Locrian99]

Ok that makes more sense now.   My two best out of 8 are 1558mv with an idss of 19.2ma 1556 mv with an idss 18.8ma.   I'm assuming those are close enough. 

[Phend]

I just completed a 45 a month ago.
I purchased a matched set of 5485's from General Guitar Gadgets.
Here is a quote from their site.
"2N5485 Matched set of 2 JFET transistors that are matched to within one-tenth of a volt. They are tested, matched and sorted by hand. A very good set of Fairchild transistors for use in a Phase 45 or other 2-stage phase shifter."
1/10 seems like alot. But if that is what is needed then I am not the one to comment on that issue.

[Locrian99]

I have seen articles on how to match them not much on what is "matched".   At .1v I'd have 7 of 8 being able to be called matching, haven't done the idss on all of those yet as I just took took my two best. 

[Rob Strand]

I have seen articles on how to match them not much on what is "matched".   At .1v I'd have 7 of 8 being able to be called matching, haven't done the idss on all of those yet as I just took took my two best. 

Less than 50mV is excellent. 100mV is very workable, in fact that's what Ibanez/Maxon use as worst case.

It's a good idea to connect a 470nF to 1uF poly cap across the source and gate to check if noise isn't affecting the measurements.    The measurements with the cap is likely to be more correct.

[Locrian99]

I have seen articles on how to match them not much on what is "matched".   At .1v I'd have 7 of 8 being able to be called matching, haven't done the idss on all of those yet as I just took took my two best. 

Less than 50mV is excellent. 100mV is very workable, in fact that's what Ibanez/Maxon use as worst case.

It's a good idea to connect a 470nF to 1uF poly cap across the source and gate to check if noise isn't affecting the measurements.    The measurements with the cap is likely to be more correct.

Good to know, I'll re measure with that.   The impression I was getting was I'd be lucky to get a pair from 8.   Only 1 is more than 100 mv off I've got four within 10 of each other. 

[Rob Strand]

The impression I was getting was I'd be lucky to get a pair from 8.   Only 1 is more than 100 mv off I've got four within 10 of each other. 

Sometimes that's not far from the truth!

The people that sell the matched sets have the luxury of buying a whole stack from the *same batch*.   The large sample size and the same batch greatly improves the chances of matching.

When you buy stuff from your local electronics store you could get JFETs from three different batches.  If you buy 10 the chances getting a match can be quite low!   Even if you get some from the same batch the chances of a match is reduced due to the low sample.  That's where it can be more economical to just buy matched ones.   Parts from larger supplies are likely to come from the same batch.

[Locrian99]

Ah.  I got these from small bear or pedal parts.   I'd have to look, I bought jfets from both last time I ordered.  I rarely by at my local store because it's ridiculously expensive.  And they don't have a lot for pedal purposes. 

[Locrian99]

I have seen articles on how to match them not much on what is "matched".   At .1v I'd have 7 of 8 being able to be called matching, haven't done the idss on all of those yet as I just took took my two best. 

Less than 50mV is excellent. 100mV is very workable, in fact that's what Ibanez/Maxon use as worst case.

It's a good idea to connect a 470nF to 1uF poly cap across the source and gate to check if noise isn't affecting the measurements.    The measurements with the cap is likely to be more correct.

Is it a good idea as well to have a power filtering cap on the positive and ground rails?   Had it initially and took it off as I was disconnecting power when I changed jfets, and discharging every time made it annoying. 

[Rob Strand]

Is it a good idea as well to have a power filtering cap on the positive and ground rails?   Had it initially and took it off as I was disconnecting power when I changed jfets, and discharging every time made it annoying. 

It doesn't hurt that's for sure but normally you can get away without it because the power rails have a low impedance.   

I've done experiments which check the measurement consistency on different setups.   I've also checked touching the various leads in the test set-up to make sure there was minimal effect from noise.

- I powered from battery and a number DC adaptors.  It is definitely true that power supplies which connect to the mains are more susceptible to noise than battery power.  Some supplies are worse than others.   With the 470nF/1uF cap usually the measurements are OK.

- Adding caps across the power wasn't very effective at fixing noise issues.  (kind of expected)

- Adding caps across the drain source and gate were very effective.   You need at least 470nF, 1uF preferred.
  The cap needs to be a film type to ensure the leakage is low.
  (Junk can get in from the relatively long meter leads.)

- Some JFETs oscillate in the test jig, especially the low Rds_on/high Yfs types.  A 1k resistor in series with the gate was very effective at preventing such oscillations.   Power supply caps less effective.  (With the 1k present I think I had the 1uF cap across the DMM terminals and not the gate as such).   

- Measuring VP (Vgs off) at higher currents is less susceptible to noise than at low currents.   That means a lower Rgs resistor.   However, testing at higher currents means the Vgs measurement is somewhat less than the VP (Vgs_off) JFET parameter.   The Vgs measurement is OK for matching but not for parameter measurement.   I don't mean crazy high test currents, maybe 100uA max.   I still test at low currents.

Other stuff:
- the 1k gate resistor give added protection to the test jig if the JFET is inserted incorrectly.
- it's a good idea to keep 100 ohms in series with the power rail to help prevent frying the JFET if the JFET is inserted incorrectly.

I think I drew a diagram of this set-up a year or two back but I can't find the posts.   The schem has a kharki/olive background.

FWIW, I don't think I ended up drawing a complete schematic.   I drew a schematic explaining some technical issues and mentioned the cap later on.

[Locrian99]

Thanks again,

Added the items you mentioned definitely saw more stability in the readings than I did before.   
Now I'm debating if I want to make the 90 instead.   My readings for the 8 were (I didn't do the IDSS) yet.
In mV
1501
1471
1247
1535
1548
1544
1544
1571

4-7 there should be pretty solid I'd think. 

[Rob Strand]

Now I'm debating if I want to make the 90 instead.

Why not!

4-7 there should be pretty solid I'd think. 

Those look quite good.

My readings for the 8 were (I didn't do the IDSS) yet.

For a phaser IDSS is unimportant compared to VP (Vgs_off).
If you use the same part numbers matching VP will pretty much guarantee the other parameters
are close enough.

[Rob Strand]

Adding caps across the drain source and gate were very effective.

FWIW, I actually added the cap across the source and gate.

[Locrian99]

Adding caps across the drain source and gate were very effective.

FWIW, I actually added the cap across the source and gate.

Yea I put it between source and gate, I don't even know if I really took in the typo.   I just put it there as that is where I figured it would go I guess?   Anyways realized I didn't have any 15u caps or a 5.1 zener they will be here Friday so I decided to move stuff around in my pedal board to make some room for a few of these I've made in the last couple weeks and broke the batter snap off on my Black Russian big muff.   Guess I found what I'll be doing tonight lol. 

[Rob Strand]

Anyways realized I didn't have any 15u caps or a 5.1 zener

The 5.1V zener should be a 500mW typ if you use a 1W type the voltage
will be a little low.   I think you end-up with around 4.8V across the zener
when you use the correct 5.1V 500mW and an ill-defined 4.0V if you use
a 5.1V 1W.  If you can only get 1W zeners you might be better off with
a 5.6V 1W.   There's quite a few posts on this forum which elaborate
on the precise voltages.

[Locrian99]

Darn it!   Didn't realize the wattage on them mattered.  Little confused by the product description in stompboxparts





Product description notes the 5239, but then says 5.1v.   Didn't even look at that when I ordered.    Guess I'll see what I get

[Rob Strand]

Product description notes the 5239, but then says 5.1v.   Didn't even look at that when I ordered.    Guess I'll see what I get

I think you might end-up with 1N4733A's  5.1V 1W.

It will still work but you lose a bit of head-room and the LFO doesn't quite work the same.
No an enormous problem.

If you really want you can reduce the 10k between +9V and the zener.   Probably needs to be about 2.2k or so.
That will bump the zener voltage up a bit.    Another way is to leave the 10k and put a 1N4148/1N914
silicon diode in series with the zener.   That will bump the voltage up by 0.6V.   The zener "points up" and
the 1N4148/1N914 "points down".

[Locrian99]

quote author=Mark Hammer link=topic=82132.msg681143#msg681143 date=1265331970]
He's right, of course.  The zener is there simply to stabilize the voltage supplying the bias circuit in the face of an everchanging battery voltage.  If you use an external PS, then all those pesky battery-related issues simply vanish.
[/quote]

Reading this from a previous makes me wonder if the diode is necessary?   I use power supplies that always give the 9.3v.  You bias it off the 250k trim pot, correct.   Wouldn't just a 78l05 be a better way of doing this? 

[Rob Strand]

He's right, of course.  The zener is there simply to stabilize the voltage supplying the bias circuit in the face of an everchanging battery voltage.  If you use an external PS, then all those pesky battery-related issues simply vanish.

Reading this from a previous makes me wonder if the diode is necessary?   I use power supplies that always give the 9.3v.  You bias it off the 250k trim pot, correct.   Wouldn't just a 78l05 be a better way of doing this?

It's largely correct but not 100% correct.   If you have an unregulated external supply the voltage doesn't stay the same.  If you have a nominal 9V regulated supply it could be anything from 8.5V to 10V.  Each individual supply consistently stays at it's own voltage but different supplies may have slightly different voltages.    If you use a zener it keeps the *local* "5V" rail relatively constant.  When you set-up the phaser it is adjusted relative to this local "5V" rail, so it should stay in adjustment for the life of the product.

In principle the 78L05 would work.  You might need to add a small dummy load because in-circuit the LFO can feed current *back into* the regulator.    The main reason the zener is used is because the MXR circuit originate back in the days before 3-terminal regulators were common and they weren't that cheap.    The 78L05's actually draw more current than the zener circuit.  So for battery operation it's a consideration but not so much when using external supplies.   The regulation of the 78L05 is much better than the zener.

If you think about, if you match your JFETs to 50mV you might expect the JFET adjustments need to be within say 100mV.   And from that point of view it would be good if the 5V rail could be kept within 50mV to 100mV over the range of DC supply voltages.

The thing that puts a spanner in the works is the LFO is powered from the 9V rail, so if your DC supply voltages vary from say 8.5V to 10V you are going to see variations at the output of the LFO anyway.   In other words there's a point where the LFO variations swamp any good intentions from better regulation on the 5V rail.