Paranormal Compressor Schematic - Guitar Magazine

[Rob Strand]

I'm just fixing up my schem before I send you the file.  I noticed these part errors on your schematic (my had some errors also):

R5     10k
R6     68k
R10   1k

As far as the sim goes it looks like it does work.

The bias points of Q1 and Q6 look reasonable.

With the MIX set to DRY:  When the OUTPUT control set half way, I'm getting about -0.7dB gain which corresponds to about 5dB with the OUTPUT control set to full.  The article mentions a 6dB gain.

With the MIX set to Compress: When OUTPUT control set half way the unity gain point is at about 50mV peak input.   It compresses but it has some weird behaviour around 50mV pk input, like it is expanding not compressing.

The MOSFET output for large signals isn't clean.  It is cleaner at 200Hz than 1kHz so I'm suspicious this is related to the Q2's high input capacitance combined with the larger 1M resistors feeding the gate.

Here's my revised schematic.  I'm not going to post any more schematics as your schematic should be the master copy for the thread:

[rankot]

It's definitely interesting, I've never seen MOSFET based compressor before.

[Rob Strand]

It's definitely interesting, I've never seen MOSFET based compressor before.

The problem with the 2N7000's and similar is they have high Drain-to-Gate capacitance. Also, the resistance is quite low compared to a JFET so the gate voltage needs to operate over a narrow range.  IIRC they have more distortion because the on-window is small compare to a large VGS_off JFET.

Some time around March this year I'm sure I posted some tricks to increase RDS and to linearize the MOSFET.  Here's the idea,



BTW I worked out why the Paranormal Compressor goes a bit weird around 50mV peak input.  That is when the MOSFET is fully turned on.

[FlyingWild]

Try to measure voltages on Q1 pins.

Voltages for the two FETs, but please bare in mind something keeps making Q1 stop working!

Q1
Pin 1 (Source) 9v
Pin 2 (Gate) 9v
Pin 3 (Drain) 9v

Q2
Pin 1 (Source) 9v
Pin 2 (Gate) 7.3
Pin 3 (Drain) 8.5v

All Pots are 100K, there is no A or B designation, so is it safe to conclude they are linear?

If it would be helpful I'll try and work out the pot wiring tonight?

[anotherjim]

Q2 source should be hardwired to ground 0v, so a connection to 0v is missing somewhere. If testing on battery, the negative might be via the ring of a TRS input socket and needs a jack plug inserted to switch the power negative on.

This is a "parallel" compressor. There is a mix between compressed and clean at the output. That means you can still have natural transient peaks coming through to keep thing interesting while the lower level "body" of the audio can have a more consistent average level due to compression. It's a trick commonly employed by record mastering engineers on final tracks, not usually on individual instruments. To some extent, if the compressor action was too snappy/obvious, it can be masked somewhat by having some clean signal present.

[Rob Strand]

Voltages for the two FETs, but please bare in mind something keeps making Q1 stop working!

Q1
Pin 1 (Source) 9v
Pin 2 (Gate) 9v
Pin 3 (Drain) 9v

What the hell!

I'd be checking for shorts to 9V all around Q1.   You might even have a badly etched PCB which is shorting everything.  It's very weird to have all those voltages 9V!

Q2
Pin 1 (Source) 9v
Pin 2 (Gate) 7.3
Pin 3 (Drain) 8.5v

Those voltages also look weird.

Are you sure you have wrote Q1 and Q2 down correctly?

Maybe you have a wiring problem?  Check your jack wiring.

[Rob Strand]

FWIW,  the voltages out of the simulator are:

Q1
D  5.8V
G 2.9V
S  0.83V

Q2
D 1.2V
G 1.9V
S   0V

The Q2 voltage could vary quite a bit and the meter loading could cause them to drop to 80% or 90%.

You *have* to have Q2 source at 0V  if not there's a problem with the ground or jack wiring.

What you can do to bypass the whole jack wiring issue is to wire the battery directly to the power terminals on the PCB.  They re-measure the voltages.

[iefes]

It's definitely interesting, I've never seen MOSFET based compressor before.

I think midwayfair's (Jon) bearhug compressor uses a MOSFET in a very compact design.

[Rob Strand]

I think midwayfair's (Jon) bearhug compressor uses a MOSFET in a very compact design.

It has a MOSFET but it's the amplifier.  The voltage controlled resistance is done with a JFET.

[FlyingWild]

Apologies regarding the obvious way out there voltages I measured, I'm still trying to figure where I went wrong, and an only assume I didn't connect to a ground, a problem of trying to measure something while having breakfast at the same time no doubt. Anyway in the mean time metal fatigue from my repeated turning the board over has resulted in three of the wires going to the pots breaking free from the PCB, two of which I do not know if I've reattached them the right way round!

For what it's worth here at the voltages I've just measured, the circuit is powered from a power supply set to 9v, however Q1 voltages still looks very questionable, but maybe this is indicating why my pedal stopped working.

Q1
Pin 1 (Source) 1.5v
Pin 2 (Gate) 2v
Pin 3 (Drain) 2.9v

Q2
Pin 1 (Source) 0v
Pin 2 (Gate) 1.5v
Pin 3 (Drain) 0.6v

This pedal has worked fine for years, well since 2005 when I made it, with nothing to compare it to I can't comment on how good it was, but I was able to dial in what I wanted, and nearly always used the mix knob to keep some of the original signal there which I thought made it sound natural. I would just like to get it going again, but am also preparing to build an Engineer's Thumb.

[PRR]

> 2N7000's and similar is they have high Drain-to-Gate capacitance.

5pFd. Plus 60pFd D to S. Max.

For the 10K+68K source, this gives nearly 200KHz bandwidth.

The large-signal response is typically limited by the body diode. For that source impedance you want to be under 0.4V peak. The paranormal output stage is gain of 20, so under 8V peak output; we run into the 9V supply before the MOSFET seriously hurts.

Voltage at Gate would be near 3V with a perfect meter. The observed 1.5V would fit if meter is 1Meg, a not atypical value for generic DMMs. Voltage at Drain is a Meg away so should be even lower, tho I figure 1V not the observed 0.7V.

[Rob Strand]

5pFd. Plus 60pFd D to S. Max.

For the 10K+68K source, this gives nearly 200KHz bandwidth.

The capacitance goes up quite a bit when the voltages across the device decrease  There's a graph in the datasheet.  I was thinking more the linearization gets stuffed up because the 2x1MEG are feeding that (larger) capacitance.

The large-signal response is typically limited by the body diode. For that source impedance you want to be under 0.4V peak. The paranormal output stage is gain of 20, so under 8V peak output; we run into the 9V supply before the MOSFET seriously hurts.

It's quite possible.  The body diode issue actually crossed my mind after I posted those series MOSFET configurations.   I checked at 200Hz and 1kHz and there was no weird waveform at 200Hz.  It looked like some sort of capacitive effect however it could very well be the cap in series with the MOSFET charging-up through the body diode.   I'll have to have another look.

Another thing I've seen using those MOSFETs as VCR is the control ripple can cause an on-off effect when you get near the threshold.  You can get some really weird waveforms.  I don't think that's going on here though.

[Rob Strand]

For what it's worth here at the voltages I've just measured, the circuit is powered from a power supply set to 9v, however Q1 voltages still looks very questionable, but maybe this is indicating why my pedal stopped working.

Q1
Pin 1 (Source) 1.5v
Pin 2 (Gate) 2v
Pin 3 (Drain) 2.9v

Yes they look like something is wrong. 


Initially I was thinking extra current could come back down from Q3 but I don't think that's what is happening.

If you calculate the drain current from the drain voltage,
Id = (9-2.9) / 10k = 0.61mA
Then the expected source voltage (roughly estimate 2.7k in parallel with the 10k etc as 2.55k),
Id *2.55k = 1.56V
Which agrees with your measurement.

What that is saying is Q3 isn't doing something funny and the source & drain circuits are operating as expected.

The voltage divider measurement on the gate is off but that' probably meter loading if your meter is 1Meg input impedance.

So that leaves the MOSFET.  So it's either fried or the gate threshold voltage is way off what the circuit is designed for.

One test you can do is to short the gate of Q1 to the drain then measure the voltage between the drain and source.  It should be around 2.1V.

Another test is to replace Q1, or swap Q1 and Q2, and re-measure.

[anotherjim]

There have been MOSFET compressors before, at least as DIY projects, however, the  MOSFET's were in a CD4007 array chip, and sidesteps some of the issues of body diode and input protection.

It might be worth suspecting Q1 because there is the possibility of damage from an electrostatic discharge at the input. If it turns out it was damaged that way (hard to prove but easy to suspect), there is a simple add-on remedy...
http://www.muzique.com/schem/mosfet.htm

[rankot]

There have been MOSFET compressors before, at least as DIY projects, however, the  MOSFET's were in a CD4007 array chip, and sidesteps some of the issues of body diode and input protection.

Interesting, never seen one!

[anotherjim]

There was a Penfold project. IIRC only used 1 MOSFET from the 4007 as the control.
Not a compressor, but we did have a trem design here where the 4007 provided the pre-amp, the PSO oscillator and the control resistor.

[Rob Strand]

There's also the ETI 446 Limiter (from Electronics Today International, Aug 1976)
https://www.diystompboxes.com/smfforum/index.php?topic=101953.0

Then there's some phasers like the ETI 447 Audio Phaser (Sep 76) and Electro-harmonix Bad Stone Phaser ( EH-5800).

Because MOSFETs have body diodes the designs usually try to limit the voltage across the MOSFET.  You can see this more explicitly on the phasers as they will attenuate the signal, pass it through the circuit with the MOSFETs, then boost the signal back up.   This process obviously isn't good for noise performance.

Those funny looking two MOSFET circuits I posted earlier try to get a bit more out of it, as well as increasing the resistance which make it more useable in audio circuits.  The idea can't be used directly on CMOS devices.

[rankot]

I tried to import Pspice model into ltSpice, but it's a no go, so I drew schematic in ltSpice from the beginning. And the result is that I have "gain" of 50%. Tried with different MOSFET models, and the one I have for CD4007 performs the best, but still with not enough gain.

[Rob Strand]

Tried with different MOSFET models, and the one I have for CD4007 performs the best, but still with not enough gain.

There's many bad models for the 2N7000.  They often don't work at low currents and all the stuff we use for effects pedals.   If you search "Win Hill" 2N7000 or "Winfield Hill" 2N7000 you might find something.

The CD4007 is a *very* different device, much higher resistance.

These models are OK for small signal stuff.
It's a long while since I verified the ROB model.

* Win Hills version of this has M1 without L and W modified
* so Win's relies on default L and W.
.SUBCKT MZVN4106_ZTX 3 4 5
* Nodes         D G S
M1 3 2 5 5 MOD1 L=1u W=1u
RG 4 2 343
RL 3 5 6E6
D1 5 3 DIODE1
.MODEL MOD1 NMOS VTO=2.474 RS=1.68 RD=0.0 IS=1E-15 KP=0.296
+CGSO=23.5u CGDO=4.5u CBD=53.5P PB=1 LAMBDA=267E-6
.MODEL DIODE1 D IS=1.254E-13 N=1.0207 RS=0.222
.ENDS MZVN4106_ZTX
*


* This is my best model so far for low currents (=M2N7000_AVGSS)
*
* Average 2N7000/BS170 device for small signal
* Based on some playing around.
* VTO is in the 1.8 to 2.0V region
* KP is in the 20e-3 to 30e-3 region.
* Originally Zetex Generic (N-channel DMOS, 2n7000, 2n7002 etc)
* CGDO increased from 4.5u to 12.5u to match measurements
.SUBCKT M2N7000_ROB 3 4 5
* Nodes         D G S
M1 3 2 5 5 MOD1
RG 4 2 343
RL 3 5 6E6
D1 5 3 DIODE1
.MODEL MOD1 NMOS VTO=1.9 RS=1.68 RD=0.0 IS=1E-15 KP=26e-3
+CGSO=23.5u CGDO=12.5u CBD=53.5P PB=1 LAMBDA=267E-6
+L=1u W=1u
.MODEL DIODE1 D IS=1.254E-13 N=1.0207 RS=0.222
.ENDS  M2N7000_ROB
*

[rankot]

Thanks Rob, you're great as always! Will try those models in the morning, now time to sleep