"Sonic Death Ray" Phaser - a phase insurrection!

[frequencycentral]

I'm pretty obsessed with phasers (among other things!) I've looked at a few vintage FET phaser schematics recently that have made me want to try a few experiments and see what happens, namely:

Maestro: http://analogguru.an.ohost.de/193/schematics/Maestro_MPS-2.gif

Univox: http://analogguru.an.ohost.de/193/schematics/Univox_MicroFazer.gif

Bad Stone: http://www.freeinfosociety.com/electronics/schemview.php?id=380

What I find interesting about these phasers is the drain/source resistor value (or lack of any in the Maestro). As I understand it, this resistor's job is to reduce distortion, the higher it's value the deeper the deeper the phasing but the more distortion. The MXR P90 uses 22k, while the P45 is slighly more elaborate again, and slightly higher quality than the P90.

Also take a look at this: http://www5b.biglobe.ne.jp/~houshu/synth/PhaseFet0205.GIF

So how does the Maestro get away with no drain/source resistor? I'm guessing by using FETs with a higher vgs than one would find suitable for use in a P45/P90 (I've built varients of both). I have loads of FETs, and my starting point is a matched sets with vgs of -2.5v, which were too high for my previous P90 build.

Here's my schematic, now phasing on the breadboard:



As you can see, it's pretty stripped down, but owes something to the P45 as well as the Maestro. I'm using 2n5485 FET's, the opamps are two TL084's. It sounds really nice, initially I got a touch of distortion as the LFO reaches the bottom of it's cycle, but I was able to limit the sweep by choosing a cap value of 0.01uF between the gates and sources. As there are six stages, there are a few choices for regen - I tried them all and found that the best ones were 4 and 5 stages apart, so there's a mode switch for this. 3 stages apart also sounded good, so I may end up using a three way switch. The different regen modes each give a very different feel, much more so than switching the phaser itself between 4 and 6 stages. The 2n5485 I'm using are matched at -2.5 Vgs, I also tried another set matched at -2.3 Vgs which also sounded good. I even tried half a dozen unmatched 2n5485 with equally good results. The cap values in the phase stages are markedly highly than one would expect, I initially tried smaller values but the notches were too high. I also tried Univibe values, but the 0.47uF just sounded best.

So there you have it, a stripped phaser which breaks a few rules - I should have called it Phase Insurrection. I don't know how it will work with different FET types, or FETs with lower Vgs than I used, but using the values specified it works just fine. Soundclips when I finally get my mic back!

[StephenGiles]

Interesting, you could try a static phase shift stage or 2 in the regeneration path to see what other rules you might break!!

[Mark Hammer]

The P45 has the RC network between drain and gate which reduces distortion.  Mike Irwin told me that it does not prevent distortion, though.  It merely moves the threshold for distortion upwards.  His take on it was that although the threshold was moved upwards, when distortion DID occur, it was of an ugly variety, and ultimately he preferred the more gradually onset of distortion than the all-or-none arrangement, so opted for omitting the RC network.  At the same time, I note that Mike's predilection is for synths, which have a higher amplitude signal level than guitars (usually line-level or higher).  So, his decision model may be a bit different than most people here.

As for the drain-source resistance, keep in mind that the needed resistance is going to be a function of the cap value, where you want the notches to be, and what the drain-source resistance is when the JFET is biased "in the zone".  The function of the drain-source resistor is just to set a minimum notch frequency.  So, for the MPS-2, if those FETs had  a d-s resistance of 150k when biased just right, the lowest possible notch would be at 106hz. Note that although the P90 uses 22k d-s resistors in parallel with the JFET, they also use .05uf caps.  If the JFET d-s resistance is high enough that the combined parallel resistance is 22k, then the lowest notch location is at 145hz.  Since the d-s resistance is probably going to be in megohms, rather than infinity, we can safely assume that the lowest notch frequency is likely somewhere around 180-200hz.

[earthtonesaudio]

When the LFO output voltage is most positive, you're almost certainly forward-biasing the JFETs.  I would either shift the audio portion's bias more positive, or the LFO's bias more negative (or both!), to try and reduce distortion while increasing the range of the sweep.

[MikeH]

I like the low parts count; seems like you could do a lot more stages and still keep the size reasonable.  I like it.  Too bad I still haven't gotten around to building your other phaser yet.

[stringsthings]

awesome ! .... phasers rule ! .... ( i'm still working on the Causality 4 builds .... )

[frequencycentral]

When the LFO output voltage is most positive, you're almost certainly forward-biasing the JFETs.  I would either shift the audio portion's bias more positive, or the LFO's bias more negative (or both!), to try and reduce distortion while increasing the range of the sweep.

I'll give that another try, playing with the audio bias does increase the range but also adds distortion at the bottom of the LFO sweep. I've no distortion now, I've had to reduce the range of the sweep a little to get that.

doc drop tells me he's going to breadboard it this weekend, maybe panterafanatic will too? I'll be interested in a second opinion with a different set of FETs, on the other side of the world.

After another play with this thing, it's clear to me that there are four possible taps for the regen, each one markedly different and useable, so I'm going to include a 4 way switch, with 4 LED's to indicate which mode is on. The only four way switches I have are quite large rotary, so it's going to have to be in a bigger enclosure than a 1590B - maybe an excuse for a bigger circuit with a couple more swept stages and two static stages in the regen path?

[MikeH]

...maybe an excuse for a bigger circuit with a couple more swept stages and two static stages in the regen path?

Now you're thinking buddy!

[KazooMan]

Small Bear sells a really compact 1p6T rotary switch.  It has a screwy layout for the connections to the pins.  I have used it for switching caps and it works great.  I made a small pcb to mount it and make it easier to add the caps.  It can easily stand upright in a BB box.

You could cut down the board and connect your leads where I have the numbers for the caps.  Note that positions 5 and 6 are flipped due to the weird arrangement of contacts on the switch.

[Eb7+9]

What I find interesting about these phasers is the drain/source resistor value (or lack of any in the Maestro). As I understand it, this resistor's job is to reduce distortion, the higher it's value the deeper the deeper the phasing but the more distortion. The MXR P90 uses 22k, while the P45 is slighly more elaborate again, and slightly higher quality than the P90.

ok, so you understand that there is a trade-off of "some" sort ...
the trade-off is transfer linearity to resistance range (slope of transfer around zero-crossings)

So how does the Maestro get away with no drain/source resistor? I'm guessing by using FETs with a higher vgs than one would find suitable for use in a ...

that's where you loose it ... you need to read the Vischay "jFET-VCR" paper I've mentioned numerous times in this silly forum
read it 100 times or more, so it starts to sink in ...

I know you guys are allergic to math, but that's where it's ... at
I hate to say it but your post is once again saying that something fundamental is being overlooked, n'est-ce pas ?

---

so then, there's the delusional concept of "matching" that has been introduced by a chap who obvisouly can't swim his way out of a glass of circuit analysis (the flying mulcher guy) - but I've already said anough about this already ... my educated guess is this is greatly part of the reason why you guys are still to this day whacking your head on the wall of this cage ...

read the paper yourself instead of seeking advice from others who obviously haven't read it,
or have but can't "really" figure it out ...

remember, it's impossible work for those who lack the proper conceptual tools

app notes are written for design engineers - ie., people who are potentially trained to read between the lines ...
people who, more importantly, know why the math does what it does - as opposed to someone memorizing and regurgitating "trick" equations ... I have shamed several senior design engineers in my work who do just that to appear knowledgeable

unfortunately, it's very common in analogue design circles ...
I see it here all the time

[KazooMan]

Only my 123rd post.

Let me guess.  You have a lot of friends on this forum.

Read this 100 times or more so it starts to sink in.......

"I will not be a pompous a**hole".

[soggybag]

Please shame me with a short sweet summation of what we are all missing. I'm, to this day, still under the impression that FETs would need to be matched for this purpose do to their loose tolerances. Which all comes from what I've read here.

[frequencycentral]

Please shame me with a short sweet summation of what we are all missing. I'm, to this day, still under the impression that FETs would need to be matched for this purpose do to their loose tolerances. Which all comes from what I've read here.

Let's not go there. Life is far too short to take so seriously what should just be fun. Please let's not spoil the fun of this thread with a flame war.

--------------------------------------------------------------------------------------------------------------------------------------------------------------------

Anyway, back to our scheduled programme.    


I'm sure I recall Mark Hammer posting a link to a phaser with fixed stages in the regen path, but I can't for the life of me remember which phaser it was. Mark?

Tomorrow I'm going to add a third TL084 to the circuit -  2 extra swept stages, and 2 fixed stages in the regen path. As I'm getting 4 choices of regen flavour currently, two extra stages should add another two regen options to play with. I can't quite imagine what two fixed stages in the regen path will sound like, but I'm gonna find out for sure - maybe I could use a DPDT toggle to select or bypass these two stages for further sonic possibilities. I just have to get some more 10k resistors...........fkg phasers, so many 10k resistors............!

[soggybag]

To be honest I'm trying to ruffle any feathers. I'd really like to hear a short explanation about getting around the idea of having to use matching FETs for things like Phase 45. Which is what I thought Eb+7 was getting at.

[woolley]

...maybe an excuse for a bigger circuit with a couple more swept stages and two static stages in the regen path?

Now you're thinking buddy!

bi-phase we're all thinking it. you know you want to!

[StephenGiles]

To be honest I'm trying to ruffle any feathers. I'd really like to hear a short explanation about getting around the idea of having to use matching FETs for things like Phase 45. Which is what I thought Eb+7 was getting at.

Just don't bother, it really doesn't make that much difference

[stringsthings]

.... The 2n5485's I'm using are matched at -2.5 Vgs, I also tried another set matched at -2.3 Vgs which also sounded good. I even tried half a dozen unmatched 2n5485 with equally good results .....

how would i go about matching the 2n5485's ?  ... can i do this without a scope ?

[Mark Hammer]

I'm sure I recall Mark Hammer posting a link to a phaser with fixed stages in the regen path, but I can't for the life of me remember which phaser it was. Mark?

Well, your memory took you 95% of the way there.   Here's the other 5%.  It's this one (many thanks to Rob Strand).  Note that there is a fixed stage in the feedback path to mimic what would, more or less, be true if it was a 6-stage phaser and the feedback was going from the 6th to 2nd stage.  Since each phase-shift stage (fixed or variable) is inverting, the feedback needs to pass through an odd number of stages to present an inverted signal to mix back in.  Because the fixed stage accomplishes that inversion, that permits the feedback to be directed to the first stage, instead of the 2nd as most of us are accustomed to seeing.

The AM97C11 is long gone (At least as far as anyone knows.  If you spy some, please report back.), and only has 4 FETs on board, which more or less explains why Ross adopted the approach they did. 

I said "more or less be true" above, because there is more phase shift introduced in the feedback path than there is in the direct path.  I've never tried this out. so I have no opinion on what it sounds like, but it may be interesting, and a little different at the lower end of the sweep.  You will note that R23/C10 form a lowpass filter going to the non-inverting pin, instead of the highpass filter we see throughout the rest of the swept stages here (C2/C4/C6/C8 each form highpass filters with their respective drain-source resistances).  As such, it adds more phase shift the lower the frequency.  In this instance that added phase shift commences a bit above 300hz.  This might be an interesting add-on to insert into a P90, though my instinct would be to use a dual op-amp on a daughter board, with one cap to ground of maybe .015uf, and the other as .05uf, just to spread out that extra phase shift a little farther across the spectrum.  In one of my Ropez phasers, I used the little patch-point pads that Francisco graciously provides to insert a pair of fixed stages, making 6 in total, and quite like the result.  My added stages are of the lead (CR) rather than lag (RC) type, so the additional phase shift is added as you sweep upward.

[frequencycentral]

Hmmm, I'm wondering how worthwhile it would actually be to.....

.....add a third TL084 to the circuit -  2 extra swept stages, and 2 fixed stages in the regen path.

.....given that this sounds pretty rich already. And that as Mark states, the Ross used fixed stages in the regen path to accomodate rich 'six stage style' regen in what is otherwise a four stage phaser. Performing those mods would make the circuit half as big again - for maybe just a little more whoosh, law of diminishing returns. I guess I have to try it to find out, but I imagine I'll end up sticking with what I have already (with the addition of more switch regen options) unless the mods are really groundbreaking.

I'm finding the different regen taps really interesting. Conventional wisdom is that 3 stages apart is the minimum, and that it must be an odd number of stages. puretube turned me on to the fact that it doesn't have to be an odd number in the Causality 4 thread.............

But it has to be an odd number of stages right?

Who says: "ODD" anymore, when such a circuit is shifting its phase-relations from at least zero° to almost or even over 360°
for almost any frequency  

....and he's correct of course. What surprises me is that even 2 stages apart sounds really good. The four regen modes really do make it sound like four different phasers, some low and bass heavy, others stratospheric.

.... The 2n5485's I'm using are matched at -2.5 Vgs, I also tried another set matched at -2.3 Vgs which also sounded good. I even tried half a dozen unmatched 2n5485 with equally good results .....

how would i go about matching the 2n5485's ?  ... can i do this without a scope ?

The GEOFEX FET matching fixture, just a very small circuit and a DMM, build one onto your breadboard, or even make a hardwired perf/vero/PCB fixture:

http://www.geofex.com/Article_Folders/fetmatch/fetmatch.gif

[stringsthings]

The GEOFEX FET matching fixture, just a very small circuit and a DMM, build one onto your breadboard, or even make a hardwired perf/vero/PCB fixture:
http://www.geofex.com/Article_Folders/fetmatch/fetmatch.gif

thank you ! ... that's a good read on GEO ...