Stuff about CMOS based phasers

[bioroids]

Hi!

I've been breadboarding a 4-stage phaser, using the CD4069UB for the variable resistors (Craig Anderton style). The Bad Stone uses this trick too. I used a TL074 for the stages.

I found it works very nice, has a lot of depth (apparently more than the Small Stone I'm comparing it with), but it has one drawback: you need to set the bias for mosfets in a very precise point, to get a good sounding range.

This bias point has to be so precise than a change in the supply voltage (dying battery,  switching to other power adapter) throws it all away. I mean a small change in supply voltage renders the phaser from great to non-working.

One solution: derive all the reference voltages from a regulator. I used a 78L05 with great results. I had to derive the Vref to the opamps with a resistor divider from the 5v output, and the bias voltage for the mosfets with a 15k pot between the 5v and Vref. Is very important to also power the LFO with this regulated voltage, as a change in the LFO amplitude will certainly drive the phaser out of wack (I used a LM358 for the classic two opamp LFO).

I tried regulating with a simple Zener but found it isn't constant enough for this .

Another solution: it is possible to use one of the mosfets as a reference to dynamically addapt the bias, with some circuity. I mean something like this http://www5b.biglobe.ne.jp/~houshu/synth/mike_irwin/cmos-vcr.gif which I don't totally understand and couldn't try (I can't get hold of those dual transistors). Maybe simplifying this a little is enough for a phaser where you don't need V/oct or V/hz control. On the Bad Stone there is a little bit of circuit (opamp F1 and surrounding components on http://topopiccione.atspace.com/PJ05EHBadStone.html ) that seems to serve this purpose. I have to try to understand it yet 

The ETI 447 Phaser doesn't seem to have any control regarding this.

I noticed the MXR Phase 90 also derives the bias and Vref from a Zener regulated voltage. I suppose the FET's allow more variation than the MOSFET?

On a side note, I tried also using the P-Channel Mosfets on the 4069 instead of the traditional N-channel. It worked but with an awful noise (similar to a ring modulator) I couldn't get rid no matter how I biased it. Maybe the P-channel works as variable resistor but for the opposite carriers?

I will hopefully post more info once I get my new soundcard, as I plan to do some scope and spectrum analysis on this.

Anyone has any experience with this circuits? I'd like to hear it, as this trick is proving very usefull (goodbye LM13600 ??). Anyone tried the Bad Stone or the ETI Phaser with different supplies?

Luck!

Miguel

PS: that spell checker makes me laugh. This time it suggested Undertone for Anderton ...

[gez]

I only ever breadboarded the circuit in this thread, but here's one way to regulate:

http://www.diystompboxes.com/smfforum/index.php?topic=26740.0

[christian]

Badstone uses 4009 with little different setup.
The Pearl phaser is also using 4069, schematic,
might wanna check that out.

ch.

[bioroids]

Gez: that's more or less what I have done, but with a standard regulator. Did you like the sound when you breadboard it?

Christian: I had not seen that Pearl schem, thanks!

Seems like there are a lot of CMOS packages that can be used this way: 4009, 4049, 4069 and 4007.

Luck

Miguel

[gez]

Gez: that's more or less what I have done, but with a standard regulator. Did you like the sound when you breadboard it?

I can't remember much about it Miguel.  It was a test to see what difference a sine wave makes.  It takes you one step closer to getting more of a Leslie vibe (not that the circuit sounded anything like one), so it was a useful exercise. 

The op-amps are biased from the regulator too so the circuit is consistent regardless of what the supply is doing.  Only drawback is positive output swings are more prone to clipping as a battery flattens...rail-to-rail output op-amps help though.

[Mark Hammer]

Mike Irwin built the ETI circuit a while back, and told me that it sounded nice but that it was quite noisy.  That is perhaps one of the reasons why the oriinal circuit did not include any regeneration capabilities, because adding regen just increased the noise.  He partially remedied this by using the RC network between drain and gate to increase immunity to clipping for the FETs in the 4049.  This permitted increasing input level a bit so as to improve S/N ratio (i.e., noise is constant, but ratio improves if signal level is increased).

More recently, though, he remarked to me that his attitude towards use of the RC network between drain and gate had done a 180.  While it reduced distortion, he noted, when distortion finally did kick in, it was a substantial rather than gradual increase, and of a much uglier sort than occurred without.  His current view is that he'd rather live with the hiss and very modest, more tolerable disortion, than deal with the sudden change in distortion when using the network  That's an entirely fair view, but I think it is worth noting that Mike works primarily with synthesizers, and given their generally higher signal levels, relative to guitars, that may be much more of an issue with synths than it is with guitars.  Of course, once you plug your axe into a high-gain booster before sending that to a phaser, synth/guitar differences may well vanish, and Mike's objections become just as valid for guitar as for synth.  Personally speaking, I haven't yet heard the sort of distortion Mike found so unpleasant, and I also don't know what the input level is at which it occurs.  Perhaps if I had, I might hold a similar view.

You can see a real-world example of this sort of compensation here: http://www.freeinfosociety.com/electronics/schematics/audio/ibanezpt909.pdf
As always, here is Osamu Hoshuyama's layout comparing different (MOS)FET-use schemes: http://www5b.biglobe.ne.jp/~houshu/synth/PhaseFet0205.GIF

[bioroids]

Interesting. I don't know how noisy can this be, as on the breadboard it's hard to tell.

I'm currently dividing the signal down by 10  to avoid distortion as on the ETI schem, but I think that's too much for guitar uses. Probably dividing by 4 or 5 is enough.

The Pearl Phaser seems to use a kind of different aproach, modulating the reference at pin 7 (GND) of the IC, supposedly a 4069

I have yet to try a sine LFO, if you Gez say it's closer to a Leslie then it's worth the try!

I think regulating everything is the best aproach for consistency. I'm running the Vref at aprox 3.5v (1k
and 2k2 resistor divider from the 5v regulator). That improves headroom and there's still 1,5v left to bias the Mosfets (which apparently is enough)

Luck

Miguel

[christian]

I redrew the part of badstone that connects to the inverters. It basically uses the same servo-approach as Mike Irwin.
clickie


Pins 1 and 16 on the 4009 are +Vcc and "ground" connections, pin 8 is the -Vcc.
Note that 4009 itself is totally different from 4069 or 4049. 4007 can be connected as 4069/4049, but I dont know about the input diode biasing on that chip, which has some effect on the voltage-control too.

ch.

[George Giblet]

IIRC,the badstone was a good example of a reliable way to create the MOSFET bias.  One advantage is it used MOSFETs on the same chip which have a good match to the controlled MOSFETS.

I believe the 4009 has a similar input network to the 4069.   

The 4049 has a different input network to the 4069 which makes it the most suitable for controlled resistances.

The 4007 is a different beast, the documentation shows no protection networks but I seemed to remember there was something there (I could be getting confusing with one of the other chips that didn't match the documentation).  I wrote all this stuff down but I archieved it ages ago and can't remember where it is.

[bioroids]

Thanks for posting that schem christian, it is a lot clearer than the "original". So this servo approach has the benefit of not needing trimpots, at the cost of an extra opamp.

Looking at the 4009 datasheet, seems that this is more different that the other inverters, using 2 P mosfets and 3 N mosfets on each inverter stage. How can this affect the behavior as variable resistor? On the 4069/4049 the access to the N-mosfet seems more direct:



The 4007 provides access to the individual mosfets without tricks, but sadly you only got three of each, so you are one short for a 4 stage phaser.

Luck!

Miguel

[George Giblet]

> Looking at the 4009 datasheet,

For the 4009 I think you will find there is a picture in the data sheet that shows a protection network which applies to each inverter - it's drawn separately.

[Zero the hero]

Although having a different circuit,4009 and 4049 share the same pinouts and I swapped them in my Bad Stone clone with no appreciable difference.
Hope that it could help.

[gez]

The 4007 is a different beast, the documentation shows no protection networks but I seemed to remember there was something there (I could be getting confusing with one of the other chips that didn't match the documentation). 

I think it depends on which data sheet you look at.  Can't be 100% sure but I think ST's 4007s have diode protection...don't quote me on that though! 

[christian]

At least in this data-sheet they have drawn protection circuit.
I thought it was necessary to have this on CMOS inputs?
Each of these inverters probably have their own characteristics, but in this case they shouldn't matter that much, as long as you provide this servo-circuitry that the badstone has.
What I'm interested in knowing is that, do you need to add 1 servo for each additional inverter chip that you add, i.e, each 5-6 stages?
Wouldn't be so bad to build a lotta stages with this one.

ch.

[bioroids]

For the 4009 I think you will find there is a picture in the data sheet that shows a protection network which applies to each inverter - it's drawn separately.

Yes I found it, there's a diode network protecting each pin. Besides that the inverter stage seems a lot different than the others, but according to Zero commentaries that doesn't affect the usage as variable resistors, at least in the Bad Stone circuit.

I think it depends on which data sheet you look at.  Can't be 100% sure but I think ST's 4007s have diode protection...don't quote me on that though! 

On National and Motorola ds it says clearly that each input is protected by clamp diodes, though not drawn on the schematic.

Each of these inverters probably have their own characteristics, but in this case they shouldn't matter that much, as long as you provide this servo-circuitry that the badstone has.
What I'm interested in knowing is that, do you need to add 1 servo for each additional inverter chip that you add, i.e, each 5-6 stages?

Yes I think you would need a servo for each chip, because it is there partly to accommodate variations between ICs. If you go with the bias trimpot approach you would need a separate trimpot for each chip.

The mosfets inside a single chip are supposed to be matched, but I don't know where this assumption came from. It seems to be true though, at least on my limited experience.

Miguel

[christian]

One more thing that isn't clear on the badstone schematic is what is the lfo's output voltages?
The manual control there goes from -Vcc to (around)(virtual) ground, so is the lfo made to sweep in that range too?
Its also very strange looking lfo, so it just might do that too.

ch.

[bioroids]

I guess the best is to breadboard it and see.

It is a strange LFO to me, what kind of waveform would it be?

Regards

Miguel

[Zero the hero]

It's a hypertriangular wave.
I took a pic of this LFO on my scope: http://topopiccione.atspace.com/pjimages/EHBadStone.osc1.jpg
The up-going side of the wave is a sine wave, and the down-going side is triangular.
This LFO give the phaser a nice accent, it emphasizes the up-going side... the result is a uuuaaaaaaaaahhhhhhhuuuaaaaaaaaaahhhhhhhhhh phasing, instead of the usual boring uuuaaahhhaaahhhuuuhhhaaahhhuuu...

[Brian Marshall]

Thanks for posting that schem christian, it is a lot clearer than the "original". So this servo approach has the benefit of not needing trimpots, at the cost of an extra opamp.

Looking at the 4009 datasheet, seems that this is more different that the other inverters, using 2 P mosfets and 3 N mosfets on each inverter stage. How can this affect the behavior as variable resistor? On the 4069/4049 the access to the N-mosfet seems more direct:



The 4007 provides access to the individual mosfets without tricks, but sadly you only got three of each, so you are one short for a 4 stage phaser.

Luck!

Miguel

sounds perfect for a 6 stage phaser

[Brian Marshall]

Mike Irwin built the ETI circuit a while back, and told me that it sounded nice but that it was quite noisy.  That is perhaps one of the reasons why the oriinal circuit did not include any regeneration capabilities, because adding regen just increased the noise.  He partially remedied this by using the RC network between drain and gate to increase immunity to clipping for the FETs in the 4049.  This permitted increasing input level a bit so as to improve S/N ratio (i.e., noise is constant, but ratio improves if signal level is increased).

More recently, though, he remarked to me that his attitude towards use of the RC network between drain and gate had done a 180.  While it reduced distortion, he noted, when distortion finally did kick in, it was a substantial rather than gradual increase, and of a much uglier sort than occurred without.  His current view is that he'd rather live with the hiss and very modest, more tolerable disortion, than deal with the sudden change in distortion when using the network  That's an entirely fair view, but I think it is worth noting that Mike works primarily with synthesizers, and given their generally higher signal levels, relative to guitars, that may be much more of an issue with synths than it is with guitars.  Of course, once you plug your axe into a high-gain booster before sending that to a phaser, synth/guitar differences may well vanish, and Mike's objections become just as valid for guitar as for synth.  Personally speaking, I haven't yet heard the sort of distortion Mike found so unpleasant, and I also don't know what the input level is at which it occurs.  Perhaps if I had, I might hold a similar view.

You can see a real-world example of this sort of compensation here: http://www.freeinfosociety.com/electronics/schematics/audio/ibanezpt909.pdf
As always, here is Osamu Hoshuyama's layout comparing different (MOS)FET-use schemes: http://www5b.biglobe.ne.jp/~houshu/synth/PhaseFet0205.GIF

Just a guess, but perhaps some of the distortion is slew rate related.  I have found that once you put feedback in to a 4+ stage pahser circuits there is a lot of noise beyond our hearing range.  You can see it on a scope.  you cant really hear it, but you can hear the loss of headroom.  I was always pretty sure that this was due to the slew rate of the opamps, and noticed that a cap in the feedback loop of the first stage usually would cure it.