Ampeg Phazzer (1974) traced - schematic, looking for advice on modifications

[aion]

To my knowledge the Ampeg Phazzer hasn't been traced before, so here's a world premiere:



The basic topology is a Phase 90 signal path almost verbatim, including the weird PNP transistor output stage. But instead of matched JFETs it uses the individual inverters of a CD4069 to control the gain stages - not dissimilar from the EHX Bad Stone. The LFO is very different to either the P90 or Bad Stone, more like the one used in the Boss DC-2.

The bias voltage is a bit different as well. The "+V-zener-resistor-0V" arrangement is reversed from the Phase 90's "+V-resistor-zener-0V" - am I right that this is so the LFO can sweep at a higher voltage? So Vref is relative to the +V supply instead of ground, and the LFO is taken from the top half of the supply instead of the bottom? What would be the purpose of this, and is it related to the usage of the CMOS chip instead of JFETs?

I haven't measured D2 yet (zener diode for bias) and can't read the markings, but I'll assume it's in the range of 3.9V to 5.1V.

With that said: I'm not crazy about one-knob pedals because they don't make for a very secure PCB mount. So let's talk possible mods.

1. Regen/feedback: The standard Phase 90 feedback resistor (22k or 33k resistor from the 4th phase shift stage back to the 2nd). Any reason this wouldn't work? CMOS chips can be noisy, and regen can increase the noise level - but would the CMOS cause any more noise than a normal JFET or MOSFET in this application, since it's not really in the signal path?

2. The standard "depth" or "intensity" mod from the Phase 90 is to replace R17 with a pot. However, the LFO is very different than the Phase 90, and the DC-2 (which uses basically the same LFO as the Phazzer) adjusts the equivalent of R18 (resistor coming off the LFO) to get the same depth effect. Which of these would be better to replace with a pot, and should it have a minimum-value resistor?

3. Univibe mods for asymmetrical phase stages. That one seems pretty straightforward.

Anything else? Or other precautions that should be heeded due to the use of the CMOS chip? I noticed there's no parallel resistor (setting the max resistance) as on the Phase 90, and no voltage divider as on the Bad Stone, so it seems like it could be less forgiving in the areas of biasing and distortion.

One other note as postscript: the CD4069 is marked "CD4069BE" rather than "UBE" (unbuffered) - but I came across something from PRR on here a few years back where he said that before a certain point, all CMOS chips were unbuffered, so old chips without the "U" suffix may be identical to newer chips with the "U". So based on the application, I assume the chip used in the Phazzer is equivalent to the CD4069UBE. (And a thank you to past-Paul, and everyone else who takes the time to share those tidbits of information that will come in handy even years later.)

[iainpunk]

the LFO is currently a triangle wave, you could change it to either a TAPLFO for mulitple wave shapes or add a asymmetry diode pair switch for ramp up or ramp down shapes and/or a clipping pair to top off the triangles and make them more ''round''.

maybe add a bass bypass path so the bass isn't lost, but this can be done by adding an easily accessible LFO bias control and moving the sweep higher in the frequency range.

add a wet only switch for some nice warble and vibrato instead of phasing only

cheers, Iain

[Fancy Lime]

+1 on using a microcontroller as LFO. Beware of FCC rules, though, if this is going to go commercial.

I've been meaning to experiment with Shepard/barber pole phasing but never got around to it. You have two notches so you can move them independently. Lots of possibilities that I have strangely never seen or heard, like counter swept notches or sweeping each notch with a completely separate LFO, or as I mentioned, the barber pole.

BTW thanks for tracing this. I've never seen CMOS inverters used as variable resistors although seeing it done it seems obvious. Neat. Have a feeling I might want to remember that trick.

Cheers,
Andy

[Rob Strand]

The LFO is very different to either the P90 or Bad Stone, more like the one used in the Boss DC-2.

That produces a triangle LFO vs exponential for the MXR's.   Boss use triangle.  Output impedance on that one is low so no need to worry about loading.   As for the sound it is what it is.

The bias voltage is a bit different as well. The "+V-zener-resistor-0V" arrangement is reversed from the Phase 90's "+V-resistor-zener-0V" - am I right that this is so the LFO can sweep at a higher voltage? So Vref is relative to the +V supply instead of ground, and the LFO is taken from the top half of the supply instead of the bottom? What would be the purpose of this, and is it related to the usage of the CMOS chip instead of JFETs?

MOSFETs require positive gate voltages relative to the source.  The source is tied to VB so the gate bias needs to be some stable voltage above VB.

1. Regen/feedback: The standard Phase 90 feedback resistor (22k or 33k resistor from the 4th phase shift stage back to the 2nd). Any reason this wouldn't work? CMOS chips can be noisy, and regen can increase the noise level - but would the CMOS cause any more noise than a normal JFET or MOSFET in this application, since it's not really in the signal path?

Best to use a switch so the feedback resistor can be removed from the circuit.  The simple feedback scheme like that on the MXR stuffs-up the notches.    If you want an adjustable regen that's a different matter.    Making the MXR scheme adjustable affects the vintage mode even when the feedback is zero.    I've got a scheme which retains vintage mode when feedback is zero.   The Boss scheme is the best but needs another opamp.

2. The standard "depth" or "intensity" mod from the Phase 90 is to replace R17 with a pot. However, the LFO is very different than the Phase 90, and the DC-2 (which uses basically the same LFO as the Phazzer) adjusts the equivalent of R18 (resistor coming off the LFO) to get the same depth effect. Which of these would be better to replace with a pot, and should it have a minimum-value resistor?

Small refinement but best to bridge the depth pot to a DC voltage so the depth = zero point has the nicest sounding notches.

You can see this bridging schem on the Boss PH-2 and the MXR Flanger,

http://superphasermods.blogspot.com/2009/08/schematics.html
https://im3.diagramasde.com/5054_mxr-flanger-pedal-schematicpdf.jpg

One other note as postscript: the CD4069 is marked "CD4069BE" rather than "UBE" (unbuffered) - but I came across something from PRR on here a few years back where he said that before a certain point, all CMOS chips were unbuffered, so old chips without the "U" suffix may be identical to newer chips with the "U". So based on the application, I assume the chip used in the Phazzer is equivalent to the CD4069UBE. (And a thank you to past-Paul, and everyone else who takes the time to share those tidbits of information that will come in handy even years later.)

Very weird set-up.   There's also the ETI 447 (ETI au Sept 1976) which used a CD4049UB.     Both ETI and the EHX Bad Stone drop the level to the phase-shifters to prevent distortion from the MOSFETs.    (The ETI might have used linearization - can't remember) Then boost it at the output.   The problem with this scheme is the signal to noise isn't good at all.   Especially on the ETI which has 100k resistors and LM741.

CMOS gates usually have protection diodes from the input to each supply rail.   The 4049 is one of the few that do not have the common pattern.    What that means is if you don't power Vdd  there's no diode to short out the gate signal on the 4049.  However the 4069 has the diode and can short out the gate.  So what the Phazzer does is leave the power rail hanging.   Whatever happens MOSFETs to the top-side P-channels is up for grabs.

https://cdn-reichelt.de/documents/datenblatt/A200/CD4049UB_ENG_TDS.pdf
https://www.ti.com/lit/ds/symlink/cd4069ub.pdf

As for the unbuffered thing.   Yet another weirdness unless they are actually unbuffered.  The chip date on the pic link you posted looks like 1975.

[Mark Hammer]

Kinda like the 6-stager that showed up in ETI, using a 4049 as the control elements.  From what I understand, it was a little on the hissy side, and because the 4049 could distort easily, you couldn't use any boost on the input to optimize the S/N ratio.

[PRR]

...So what the Phazzer does is leave the power rail hanging.   Whatever happens MOSFETs to the top-side P-channels is up for grabs.....

Right. They are just using the N-types and let the P-sides float. Usually no harm will come. IIRC(?) PAiA did a similar thing.

I never heard of "buffered" CMOS until this century. There were dedicated Buffers, but plain gates just did logic, did not try to do more. Therefore I say there CAN'T be "UBE" until B became the common type. All old CMOS are U by default.

[Rob Strand]

I never heard of "buffered" CMOS until this century. There were dedicated Buffers, but plain gates just did logic, did not try to do more. Therefore I say there CAN'T be "UBE" until B became the common type. All old CMOS are U by default.

There definitely was an era for Buffered and Non-buffered.   

Here's an official app note from Harris Semiconductors, originating 1983,
https://www.ti.com/lit/an/scha004/scha004.pdf

Figure 2 on page 5 shows the internal structure of buffered and unbuffered.  The buffered type has buffers on the input and output.  I also remember seeing pics which had two inverting "buffers" in cascade at the output buffers which give buffering by double inversion.    It's possible this later case is someones interpretation of buffered types.   

I had some notes from mid to late 70's with similar pics of internals but I can't remember if the input buffers were present or not.

Around 1976 ETI had two projects which used the 4049 as a MOSFET, a compressor and the phaser.   Those projects made me aware of the differences;  even though I had no idea what a phaser did in 1976.

From mid 70's to 1980 a lot more digital projects came about and I have a feeling there were projects which occasionally asked for buffered types and specific brands of CMOS chips.    That was the bad old days of digital where people would mix CMOS and TTL.

Funny thing is most of the parts around my desk have a B suffix; many are MSI parts.  The ones that are unbuffered are the 4007's.   There's one nat-semi CD4049CN from 1980 which I'm pretty sure is unbuffered.

[aion]

Update - D2 is a 1N5232, 5.6V zener, which measured 5.33V (within tolerance - but at the far low end).

Best to use a switch so the feedback resistor can be removed from the circuit.  The simple feedback scheme like that on the MXR stuffs-up the notches.

Yep, I've used a SPDT on-off-on switch in the past to give two fixed settings with the center position being entirely disconnected. I'll try that one out here.

Small refinement but best to bridge the depth pot to a DC voltage so the depth = zero point has the nicest sounding notches.

You can see this bridging schem on the Boss PH-2 and the MXR Flanger,

http://superphasermods.blogspot.com/2009/08/schematics.html
https://im3.diagramasde.com/5054_mxr-flanger-pedal-schematicpdf.jpg

Thanks for the heads up, I hadn't noticed the PH-2 similarity. I think I'll try out that Depth control, it should be directly transferable here.

[11-90-an]

Perhaps you can get some feedback ideas here... (thanks, Rob.. )

https://www.diystompboxes.com/smfforum/index.php?topic=125983.msg1202872#msg1202872

Since it would have to be adjusted by a trimpot i think it would make sense to just toggle it with a switch (or maybe a pot in series with the trimpot)

[PRR]

Update - D2 is a 1N5232, 5.6V zener, which measured 5.33V (within tolerance - but at the far low end)....

Test spec is 20mA. In circuit it runs 0.4mA. 80:1 difference. The test-current 11 ohms resistance would imply most of the difference.

[rankot]

What are the voltages at Vb/Vc/Vd points? What is the purpose of the 250k trim pot?

[Rob Strand]

Test spec is 20mA. In circuit it runs 0.4mA. 80:1 difference. The test-current 11 ohms resistance would imply most of the difference.

IIRC the MXR phase 90 uses 1N5231 which are 5.1V and end-up measuring 4.8V or so, they also run at 0.4mA.   So almost a whole zener step out.   So the 5.3V vs 5.6V for the 1N5232 has about the same ratio.  (Some schematics show a zener voltage and part number which is one step too low because it's based on in-circuit measurement.)

[garcho]

What are the voltages at Vb/Vc/Vd points? What is the purpose of the 250k trim pot?

Isolating the LFO, trim is for bias

[pinkjimiphoton]

i've got the schematic and service docs for this, if you guys want i can upload them.
say the word.

here's the real-deal, honest to goodness schematic for the unit.
i'll try and upload the rest of the doc when i can scan it.

[Rob Strand]

i've got the schematic and service docs for this, if you guys want i can upload them.
say the word.

Good one pink.

Be interesting to see how they set the trim in production.

[pinkjimiphoton]

will try and get to it this morning

[Rob Strand]

will try and get to it this morning

Thanks, be cool to see but no hurry.

[StephenGiles]

Kinda like the 6-stager that showed up in ETI, using a 4049 as the control elements.  From what I understand, it was a little on the hissy side, and because the 4049 could distort easily, you couldn't use any boost on the input to optimize the S/N ratio.

Mark, it was a little on the very hissy side!!

[pinkjimiphoton]

sorry guys, i totally spaced this til now. i posted it at fsb too. any question of the diode type etc should be in the BOM.

PEACE!

[Paul Marossy]

I built this circuit this weekend using some pictures of the PCB & the factory info that I obtained at the other forum. Made a quasi-reproduction PCB in AutoCAD like I usually do, and etched it myself. I must say that I think it sounds great! Kind of like a Uni-Vibe. Pretty tasty phasing going on, depending on how you set the bias control, which I'm not sure how is supposed to be set exactly. Also looked at it with the scope and the waveforms look a little asymmetrical, and there's like four different cycles that it goes thru. Sort of like EH Bad Stone in a way but different. I used Harris CD4069UBE hex inverter that I had lying around and a 2N3906 since I don't have any 2N4126s. Works fine with the 2N3906 since the specs are pretty similar to 2N4126. Seems to have a slight boost when I engage the effect. Also used low voltage ceramic caps because I like how they sound in a circuit like this, sounds more "complex" to my ears. I think the minimum rate is maybe a little faster than Phase 90, but it can be slowed down by messing with the size of the cap in the LFO. Phasing sounds the deepest when the test point voltage on Pin 6 of CD4069 and "VB" are the same voltage. IIRC that was around 4.3V when I was supplying circuit with 9.5V. Output of 1N5232 is considerably less than 5.6V (4.3V)... not sure why but but does not appear to be a problem. Used 500K reverse log pot for speed control, which seems to work pretty nice. Overall I like it a lot but wish I had a depth control to make it just a little less in your face.

So I wanted to add a depth control like the MXR Phase 90 ("R28 mod") but there does not seem to be a simple way to do it like you can with the current MXR Phase 90s (re-issue?). When I was comparing Phazzer & Phase 90 schematics I noticed a few differences. One was the input section, Phase 90 has different input stage than the Phazzer with the single opamp and unity gain arrangement. The LFO is different animal. Another one is that there is no feedback resistor as in the Phase 90, which is where you could normally add a depth control. I tried just for kicks to add a 22K resistor in same location as would occur in the Phase 90 and that did not sound good at all. I guess this circuit just doesn't need it? So I suppose I'd have to do something like the Boss PH-2 depth control arrangement in order to get a depth control? I was hoping it could be something simpler.

EDIT: Here is waveform on the scope https://youtu.be/gJBVgQhDWhA