U-Boat build and questions

[thomasha]

Hi all,
I finished a build of Merlin's U-Boat. What a nice circuit. Before I start with the questions, I have to thank Merlin for this circuit, it is way easier to build.

I had to deal with some problems due to my small-ish layout but it is kind of working.
It tracks nice with humbuckers, even on the 6th string, but there is some weirdness going on.
First things first: I used the original schematic: http://www.valvewizard.co.uk/uboat.html
but used TL072 instead of the LM833.

One of my first mistakes was to build it with the NE5532. For some reason the pre gain stage was drawing too much current and dropping the 4.5V to 3.8V.
I changed all IC's to TL072 and it worked! I added the 10M resistor, as suggested.

Another thing that I changed was the FET. Instead of the J112 I used the J201 in SMD shape. The switching is working.

For me it was hard to identify strong changes between the smooth and the synth configuration. I was expecting a more "in your face" kind of change.
In synth mode there is a more pronounced octave effect, though. I decided to simulate it in spice and compare it to my build. I tested it with a 800 Hz sine wave and 100 mV RMS.
Here you see how the simulation looks like after the FET, in smooth mode:

And the measured signal after the FET (cheap scope)


As you can see, the first valley looks way shorter than it should. Has someone an idea why this happens? Is it the FET (J201) or am I missing something here?

Is there a way of correcting this, so that the valley has the same size as the second one? There is some difference also visible in the simulation, but not as pronounced. In the measurement it looks like the first valley is actually clipping. Is there a way to improve the symmetry? Will it improve something in the sound?

In case you are curious of how the synth mode looks like:

So it is different indeed, it is just me who can't hear it (in my defence, I'm using a small workbench amplifier, where things might sound less bass heavy as in reality due to the smaller speaker)

Cheers, Thomas

[ElectricDruid]

It looks to me like the non-inverted and inverted versions of the signal aren't very well balanced. They'll need to be the same volume for the best effect. The mixing seems to happen with that 4K7 and J112, so perhaps replacing the 4K7 with a 10K trim would let you dial it in for the specific resistance of the FET you've got?

[thomasha]

Hi, that's a good point.

Checking the simulation, it seems that the two signals that are combined at the FET seem a bit off regarding their symmetry line, which should be at 4.5V.
One is at 4.4491V and the other at 4.532V, or, 83 mV apart. In the first figure it is possible to see that this creates a slight offset.

So not only signals have a different amplitude, but they also are centred at different levels.

[anotherjim]

What it looks like and what it sounds like might not be the same. The inverted signal appears to slew toward a different average DC level. Maybe not surprising as the source impedance changes according to whether the JFET is on or off. The 100k pot references to 0v and the 10uF cap after it has 4.5v on the other side, so the DC level will slowly change. The slewing rate should be sub-audible. Only your eyes may care about it.
You might try the 100k pot terminating to 4.5v instead of 0v?

[thomasha]

You might try the 100k pot terminating to 4.5v instead of 0v?

Yes, that helps a little bit, in the simulation at least. Instead of 4.5V I added a 100uF capacitor between the potentiometer and ground.

I tried some other changes, in the simulation at least.
The best result I can get so far is with two JFETs, controlled by the two Q's and a slight change in the bias of the inverting OpAmp. Instead of 4.5V I used 4.57V. This makes the signal look symmetric. Now if this applies to the real circuit is another point.


Switching both sides at least reduces any influence the non-inverted signal has on the inverted one after the FET. Otherwise they would add up and the DC error would decrease the swing of the lower half.

V(n002) is the signal at the potentiometer, V(n001) is the non-inverted signal, being switched by the jfet on the left and V(n004) is the inverted signal, being switched by the jfet on the right.

Without the additional switching it looks like this, where a larger decrease in V(n002) is visible when compared to V(n004):

[anotherjim]

The preamp/filter stage has some gain so could produce a significant offset from 4.5v (including any from the input buffer). Followed by the inverter will effectively increase the DC offset between the two phases. You might design the worst of that out by providing the preamp feedback -input with an adjustable reference instead of the common 4.5v so that its output corrects to match the 4.5v to the inverter +input.

[thomasha]

Yes, I tried that, and adjusting the voltage kind of corrects it in the simulation.

I wish there was an easier solution.

[anotherjim]

What I've noticed, is with anything that cut & shuts waveform cycles around, is immense difficulty in getting a textbook-perfect result at the joins. I'm surprised simulation also has trouble as that isn't something I use. The diode inside opamp precision full-wave rectifiers also come out wonky on the scope and gets worse as signal frequency increases simply, I think, because the impedances and stray capacitances are not equal before being mixed together. The good news is that it isn't anything that you can hear - and in that particular case, a perfect full wave sounds pretty nasty anyway.

[thomasha]

So, today I tried the changes in the real circuit:
I tested it with a 800Hz sine wave with 0.3 V peak-to-peak.


As you can see the two modifications:
1- 4.5V instead of ground at the trimmer
2- JFet switching the non-inverted signal
worked perfectly!
Now both signals have the same amplitude and are kind of similar, better than it was before.

I added the changes to the original schematic, I hope it's not a problem.


In the simulation adding a separate power supply for the 4.5V solves the balance problem. It seems the voltage divider using two resistors and a capacitor is way worse than just a 4.5V supply...

[thomasha]

I finally boxed this one. Wife took some professional shots from it