Help with the simulation of Phase 90 on LTSpice!!! JFET's giving 0.3 Volt sweep

[Rob Strand]

A couple of things up front:
- When you hit a place which goes bad, in your case the third stage, you are best to top there an try to fix it.
  everything beyond that point can't be expect to work as the "road is blocked" by that stage.
- You should check what to expect from each point in the circuit using spice.
  The points at the + input of the opamp are expected to be high-pass.
   This doesn't appear on the output as circuit as a whole is wired as an all-pass filter.
- Check the DC voltages on the opamp outputs and inputs

So to start there.   The things that are going to stop the third-stage behaving like an all-pass are: (EDA schem)
- R10 not connected. }  In both theses cases the opamp would act like a buffer
- R11 shorted.             }  and the high-pass behaviour on the + input would appear on the output.
- Opamp dead.   Feedback from R10 and R11 clearly will stop the design all-pass behaviour.

Beyond DC checks and checks for shorts, unless you can prove to yourself otherwise, it would be easier to simply replace the opamp for the third stage.   For example if you lift the leg of the cap to R12 you would expect that stage to have a gain of -1.  You can check both the gain an polarity with an oscilloscope.

[savethewhales]

- When you hit a place which goes bad, in your case the third stage, you are best to top there an try to fix it.
  everything beyond that point can't be expect to work as the "road is blocked" by that stage.

First thank you very much for responding, Rob!
Well you are right. Actually I discovered that it's not on the 3rd stage that it starts messing up..
The 2nd op amp (1st of the phase shifting) has a bad freq response on the inverting AND non inverting inputs but the output is good, like (?) . What you're saying about the + input having this response bugs me because the - input of the 2nd also has bad freq resp (though I know you're right)...

Then the 3rd op amp has the inputs bad but the output is good, which seems even worse.

The 4th op amp also has a bad output.

You should check what to expect from each point in the circuit using spice.
  The points at the + input of the opamp are expected to be high-pass.
   This doesn't appear on the output as circuit as a whole is wired as an all-pass filter.

You're absolutely right, I've become so obsessed with debugging the circuit that I forgot what it should look like in spice..

- Check the DC voltages on the opamp outputs and inputs

For this matter, I've checked it dozens of times, and it is all good!

So to start there.   The things that are going to stop the third-stage behaving like an all-pass are: (EDA schem)
- R10 not connected. }  In both theses cases the opamp would act like a buffer
- R11 shorted.             }  and the high-pass behaviour on the + input would appear on the output.
- Opamp dead.   Feedback from R10 and R11 clearly will stop the design all-pass behaviour.

Okk! So, following these advices, I went to check continuity (doule checking) and R10 (two sides) and R11 (two sides) are good actually.
But this could help if the problem is really at the 2nd op amp as I think it might. 

Beyond DC checks and checks for shorts, unless you can prove to yourself otherwise, it would be easier to simply replace the opamp for the third stage.   For example if you lift the leg of the cap to R12 you would expect that stage to have a gain of -1.  You can check both the gain an polarity with an oscilloscope.

Maybe I'll do it on the 2nd op amp, idk. But right now I'm in fear of taking off components of the board...

I did continuity test on the whole PCB and power supply section, every single thing was alright except:
470k, R15, R8, R28: These ones might be badly connected and I'll re-solder them.

[savethewhales]

I just broke/destroyed one of the pads of my PCB trying to resolder the resistance R8.. What can I do now? Do I have to interrupt the other pad of the resistance and solder it directly on the point that I want? What should I do?

Note that I have a PCB with accidental smd pads (Which I didn't mean to order but messed up).

Thanks in advance

[Rob Strand]

I just broke/destroyed one of the pads of my PCB trying to resolder the resistance R8.. What can I do now? Do I have to interrupt the other pad of the resistance and solder it directly on the point that I want? What should I do?

These things happen.

Use thin wire to wire the parts up directly.    (The circuit won't care if the signals pass through the PCB or the wires!)

[savethewhales]

These things happen.

Use thin wire to wire the parts up directly.    (The circuit won't care if the signals pass through the PCB or the wires!)

Yeah they sure do... Unfortunately it was the resistance R8 which I think was the problem. Anyway, I'll try to do as you say, trying not to touch other components with wire.

Btw, do you think I should do anything else ike rip off what's left of the "pad"? Also, it's a node where there's 3 components connected. If I connect one leg of R8 to for example the JFET (first component), will it also be connected to the other (capacitor)?

[Rob Strand]

Btw, do you think I should do anything else ike rip off what's left of the "pad"? Also, it's a node where there's 3 components connected. If I connect one leg of R8 to for example the JFET (first component), will it also be connected to the other (capacitor)?

It depends on what damage has been done and how the PCB is routed.    Just make your own judgements with your eyes.  Anything that looks like it might not connect just wire it across to make sure it's connected.   Even using the multimeter isn't trust worthy when intermittent connections are possible.

[savethewhales]

It depends on what damage has been done and how the PCB is routed.    Just make your own judgements with your eyes.  Anything that looks like it might not connect just wire it across to make sure it's connected.   Even using the multimeter isn't trust worthy when intermittent connections are possible.

Right, that' accurate. I've checked the actual vias and paths of the PCB on EasyEda and was able to see that the Pad I lifted doesn't have any vias, it is directly connected to other component, tk god..

When you say wire it, you say wire it from the component to the desired point, even if there's a weak connection there?

Yes. Using the multimeter got me perfect connections along the whole board. However I got to see with good light and discovered some crappy ones and one that was disconnected (R15).

Now I also went to look at the freq responses on LTSpice and discovered that the bad freq resp should exist on the - and + inputs, so everything was correct. However on the 4 th op amp (3rd of phase shifting) there was a bad output, like you mentioned, and I discovered a problem between R14 and R15.. So I conclude it was always a matter of this R15 being disconected.

I already soldered it and I'll se tomorrow how the circuit behaves (of course if I can resolder the other missing connections and fix the lifted pad point)

[Rob Strand]

I already soldered it and I'll se tomorrow how the circuit behaves (of course if I can resolder the other missing connections and fix the lifted pad point)

Good luck.  Given it worked before it's probably only some small silly thing.  I guess the problem is finding it but you should be able to narrow things down with an oscilloscope.

[savethewhales]

Good luck.  Given it worked before it's probably only some small silly thing.  I guess the problem is finding it but you should be able to narrow things down with an oscilloscope.

Thanks Rob! Yes it might be.. Given that I am soldering on smd pads, there might be more than one silly thing hahahah.
I actually was doing it with the oscilloscope of the audio precision measurement system and also by getting the frequency responses.
Anyway I'll check again around R10 and R11 as you said, there might be something there too.

[savethewhales]

Look what I had to do with R8! Hahah, soldered it directly to R27 with a little wire on the tip..

Everything is now resoldered and theoretically well connected (and visually). So tomorrow I'll see if this guy behaves good or not, and I'll make sure to let you know!

[Rob Strand]

ook what I had to do with R8! Hahah, soldered it directly to R27 with a little wire on the tip..

it does the job for sure.

[savethewhales]

Today's tests with Vbias on max and Vbias on min, respectively (depth and speed on minimum):





No possible words for the satisfaction. Thanks Rob!

[Rob Strand]

No possible words for the satisfaction. Thanks Rob!

No worries man.    You did all the work  .   Anyway good it's up and running again.

[savethewhales]

No possible words for the satisfaction. Thanks Rob!

No worries man.    You did all the work  .   Anyway good it's up and running again.

I'm really very glad! You did some work too, if I had to count with only my teacher, this would probably gone bad (or not gone anywhere).

Here's some frequency responses (depth min) with some comments:



Why should the left notch be less attenuated than the right? I don't know, maybe my high pass filter..




This flatness could be explained by the LFO varying a little bit even with the depth on min? Or should I be putting more test points to make it more accurate? Or smth else?



The middle of the spectrum seems to have more uniformity on the attenuation of the notches..



This is with the feedback on the maximum value. Is it normal that the attenuation on the notches is less than without feedback? Or can I remediate it?

These are not very urgent questions btw, the circuit is working, and for sure my teacher accepts this like it is.

[Rob Strand]

Why should the left notch be less attenuated than the right? I don't know, maybe my high pass filter..

This flatness could be explained by the LFO varying a little bit even with the depth on min? Or should I be putting more test points to make it more accurate? Or smth else?

If you look closely at the plot the curves aren't smooth.   The curve is made up of lines joining together.   If the lines just happen to join at the middle of a notch you will see a deep notch.  If the line joins each side of the notch you will see a chopped-off notch.  You can see this in your second plot.

The problem should be fixable by just changing the plotting or FFT settings.     It's normal for FFTs to have a certain resolution which is set by how long you look at the waveform.   If the sample rate is high and the number of points is fixed then you don't look at the waveform very long.  So what you have to do is reduce the sample rate.   FFT resolution affect the low frequencies since a step of 10Hz at 50Hz is more jagged than a 10Hz step at 1kHz.

*However* in your case it might be a different problem because I can see there is jaggedness at 400Hz with wider frequency steps than at 40Hz.   So the jaggness has the same to the eye on the screen.  To me I'd be looking a the plot settings more than the FFT.   See if you can find the plot resolution or number of plot points.   (It still could be the FFT settings if the equipment is using a log-spaced FFT, although this isn't very common.)

This is with the feedback on the maximum value. Is it normal that the attenuation on the notches is less than without feedback? Or can I remediate it?

Feedback *does* mess with the notches.    As it turns out it's not the feedback itself that affects the circuit.  It's the way feedback is added to the circuit.  The 24k (or 22k) feedback resistor affects the way the all-pass filter works.   If for example you added an opamp just to do the feedback, so all of the all-pass filters are identical, you will see different a behaviour.   The way to check that would be in spice.

[savethewhales]

Just smth before we continue the previous subject. I am using this circuit to power up my pedal (with battery or DC jack switching when one is connected and other is not, you know):



Now, I am trying to make this work for ages, and for some reason it always refused to turn on the power with my battery (I've tried messing with the 6.3 input connector being connected or not; using only the battery; checking the connections; checking thousand times what are the pinouts of my 6.3 jack connector to make sure I don't make mistakes, etc).
For some godly reason, I happened to touch the battery - on the ground when the + was already connected to the switch, and IT WORKED!

So I'm thinking two things:
- The circuit I'm copying is wrong;
- I might be confusing the pinouts on the 6.3 connector

If the circuit is wrong, maybe what should be in place of the - battery on the ring is that - connected to the groud point of the DC switch and then on the ring, as some circuits on the internet suggest, am I right? I sincerely hope..

If you look closely at the plot the curves aren't smooth...   
...You can see this in your second plot.

True! I must try to change the number of points, and see if it helps..

*However* in your case it might be a different problem because I can see there is jaggedness at 400Hz with wider frequency steps than at 40Hz...   
... (It still could be the FFT settings if the equipment is using a log-spaced FFT, although this isn't very common.)

What do you mean with "So the jaggness has the same to the eye on the screen." ?
Anyway I'll try messing with the plot ponts and if it doesn't help, I'll go and try to mess with the FFT!

Feedback *does* mess with the notches...
...The way to check that would be in spice.

Pretty interesting! I might give it a try in Spice because I'll have to explain the working of the feedback as it is, so it would surely help.
I understand that it does affect the all pass working, because it is putting output signal on the - input of the 3rd all pass stage, so it's kinda amplifying what was in between notches.

[Rob Strand]

So I'm thinking two things:
- The circuit I'm copying is wrong;
- I might be confusing the pinouts on the 6.3 connector

It looks OK to me.

Confusing the pins won't help.

Try using a multimeter and check if the battery connect through or not.   No need to power up anything but you can plug the plugs in and out to see what connect and what disconnects.

What do you mean with "So the jaggness has the same to the eye on the screen." ?
Anyway I'll try messing with the plot ponts and if it doesn't help, I'll go and try to mess with the FFT!

The plotted curve is made up of lines.   Normally if it's the FFT  the lines a longer (causing it to be jaggered) at low frequencies and closer together at high frequencies (smoother).   However on your plot the lines seem to be roughly the same length across all frequencies.  That looks like a plotting problem.   If I look close it even looks like the curved part has short lines - that's something automatic plotting can do - the problem is all the lines need to be shorter to smooth out the curve.

Pretty interesting! I might give it a try in Spice because I'll have to explain the working of the feedback as it is, so it would surely help.
I understand that it does affect the all pass working, because it is putting output signal on the - input of the 3rd all pass stage, so it's kinda amplifying what was in between notches.

It's worse than that.   Even with a separate opamp you get signal from the output.  That's kind of the point of adding feedback.  However, if you grounded the side of the 24k/22k resistor that goes to the output (maybe through a cap so the DC doesn't get stuffed up) there will be no feedback but the all-pass filter with the 24k/22k doesn't work the same as if you removed the 24k/22k altogether.

[savethewhales]

Try using a multimeter and check if the battery connect through or not.   No need to power up anything but you can plug the plugs in and out to see what connect and what disconnects.

Ohh you're right, I've not done that with the battery connected there yet...

Anyway here's a picture of the exact 6.3 jack I'm using:



I'll try to explain what I think it works:

There are three lugs "together," and one lug connected on the other side. This one separate lug I tried to understand where it connects and didn't arrive at a good conclusion, so I chopped it off.

Of the 3 remaining lugs:
-Lug on the left: ground
-lug in middle: ring
-lug on right: tip

Now for the" 3 big legs":
-big one on the left: tip
-small one touching the big: ring
-the big on the other side: ground

Hope I'm right because it's already soldered hahahah (works with 9V transformer)

The plotted curve is made up of lines... 
... the problem is all the lines need to be shorter to smooth out the curve.

Yes, makes total sense. Later today I'll be able to tell you some kind of conclusion around this as I'll be around the APX.

[savethewhales]

Oh and one more thing: my pedal is working well with the guitar connected. BUT there's always a BUT and this time the BUT is that on the output I can hear clicking that have a constant cycle (that changes when I change the speed pot). Also when it is discomnected (true bypassed) the clicks continue (but more attenuated).

I know they are from The LFO, because it theoretically makes triangle waves all the time (non stop) but I don't know how to fix this.. Maybe low pass filtering with a sharper curve, or should there be another way to fix this?

The LFO has frequencies of 0.3 Hz to 8 Hz aproximattely.

[bluebunny]

There are three lugs "together," and one lug connected on the other side. This one separate lug I tried to understand where it connects and didn't arrive at a good conclusion, so I chopped it off.

The "separate" lug appears to be the screen connection, so removing it wasn't a good idea.  It's easy enough to work out which lug is connected to which connection using a DMM.  Having an extra switch connection isn't helping.  Try to find stereo jacks in future with only three connections.