Lamba Octave?

[JKowalski]

Here's an idea I had. How about an octave effect using a lamda diode in a voltage divider? The negative resistance region will trigger at the peak of the signal and invert only the peak, doubling the frequency of the signal. The strength of the doubling would be porportional to the strength of the signal - the hard you hit the note, the more octave is thrown in.

Import this to Falstad's electronic app to see what I mean.

Just copy the code, go here,, wait for the javascript applet to open, go to file-import and paste it in.

Code Select Expand



$ 1 2.8E-7 34.14951009862697 65 2.0 15
j 416 176 448 176 0 -4.0
j 480 224 448 224 1 -4.0
w 480 224 480 144 0
w 480 144 448 144 0
w 448 160 448 144 0
w 448 208 448 192 0
w 448 240 448 256 0
w 448 256 416 256 0
w 416 256 416 176 0
r 448 256 448 304 0 238.0
g 448 304 448 336 0
a 384 272 288 272 1 15.0 -15.0 1000000.0
w 384 288 384 320 0
w 384 320 288 320 0
w 288 320 288 272 0
c 288 272 240 272 0 1.0E-6 0.39846802091401196
r 240 272 240 352 0 10000.0
g 240 352 240 368 0
O 240 272 192 272 0
w 384 256 416 256 0
R 448 144 448 96 0 1 800.0 3.0 3.0 0.0 0.5
o 20 64 0 42 10.0 0.003125 0 -1
o 18 64 0 42 1.1692013098647223 9.765625E-55 0 -1




I can't b-board this one, because I don't have any p-channel jfets in my stock at the moment.

Just though I would throw it out there.

[PRR]

> lamda diode

That is one sick puppy.

I'm having a hard time wrapping my head around it. But I think it "works" -only- because you know the peak signal amplitude; that you have put a signal swing and bias into the simulator which DO work. But for random size waves such as music, I bet it generally won't work this way.

Let me know if you can make it work on guitar in real life. I may be wrong.

[JKowalski]

> lamda diode

That is one sick puppy.

I'm having a hard time wrapping my head around it. But I think it "works" -only- because you know the peak signal amplitude; that you have put a signal swing and bias into the simulator which DO work. But for random size waves such as music, I bet it generally won't work this way.

Let me know if you can make it work on guitar in real life. I may be wrong.

I skipped a letter in lambda twice, different ones each time. SIGH. L-a-m-b-d-a  

Here is a better biased one. So you buffer & amplify the signal, bias it around 3.5V (sure to be different in RL with different components, if it even works well in real life at all), put it through this divider, buffer the output, get a nice octave out. It looks very promising to me! And it gives comparatively clean octaved sine waves, much cleaner then the typical rectifier configuration.

As far as varying amplitude, try it yourself - adjust the amplitude and you get an octave effect across the range.

If the signal is biased right at the point where the negative resistance region starts, then whenever the AC goes to the positive cycle, it will lower the current instead of raising it. Whenever it goes negative, it lowers the current like it normally would. Putting it through the decoupling capacitor resets it back to the AC, twice the frequency. So the signal strength doesn't matter. Only the bias voltage.

To get rid of the bias voltage on the input and output scope signals, speed up the simulation to maximum for a few seconds to let the decoupling capacitors settle and then bring it back down to normal speed. Use the slider on the right.

To see the input signal, right click on the topmost "out" and hit view in scope.

Code Select Expand



$ 1 2.8E-7 56.303023683595114 65 2.0 15
j 336 112 368 112 0 -4.0
j 400 160 368 160 1 -4.0
w 400 160 400 80 0
w 400 80 368 80 0
w 368 96 368 80 0
w 368 144 368 128 0
w 368 176 368 192 0
w 368 192 336 192 0
w 336 192 336 112 0
r 368 192 368 240 0 300.0
g 368 240 368 272 0
a 304 208 208 208 1 15.0 -15.0 1000000.0
w 304 224 304 256 0
w 304 256 208 256 0
w 208 256 208 208 0
c 208 208 160 208 0 1.0E-6 0.7309776652213429
r 160 208 160 288 0 10000.0
g 160 288 160 304 0
O 160 208 112 208 0
w 304 192 336 192 0
R 368 80 368 32 0 1 800.0 2.0 3.3 0.0 0.5
c 368 80 272 80 0 1.0E-6 3.2714379283188366
r 272 80 272 128 0 10000.0
g 272 128 272 144 0
O 272 80 240 80 0
o 18 64 0 42 0.45877444269384915 6.103515625E-205 0 -1
o 24 64 0 42 5.0 9.765625E-5 0 -1





[JKowalski]

Thought I would add a pic for a better visual - the bias point is on the top of the V-I curve. When the signal goes negative, below the bias point, the current decreases. When the signal goes positive, the current decreases again. The resistor converts that current back into a voltage. So the top of the input signal that is above the bias point gets flipped over to the other side, no matter the strength of the signal. The slope of the V/I curve is not totally symmetrical, so the octaving is not symmetrical  - every other cycle will be slightly smaller in amplitude but not by very much. And the breakover point is pretty smooth, no sharp distortion.

[jasperoosthoek]

I stumbled upon this rather old thread. Have you built any real circuit with a lambda diode yet? I've ordered some tunnel diodes that I'm going to play around with.

[JKowalski]

No, but I will try something tomorrow. 

[jasperoosthoek]

I'm looking forward to your results!

[phector2004]

Cool! I opened this up in the sim

how would the frequency "see-saw" effect affect the octave? Think it'll be chorusey?

[Hides-His-Eyes]

SOMEBODY must have the bits for trying this out!

[jasperoosthoek]

I was wondering. If all you're after is a quadratic part in an IV curve then you might as well try this with a voltage multiplier. You know one of those chips with two inputs that gives you Vout = V1 * V2.

Then connect both inputs to the same source and you've got circuit that perfectly doubles a sine wave. Just add some clipping diodes before the multiplier to compress the quadratic signal a bit. Maybe a different amounts to each input.

[JKowalski]

I was wondering. If all you're after is a quadratic part in an IV curve then you might as well try this with a voltage multiplier. You know one of those chips with two inputs that gives you Vout = V1 * V2.

Then connect both inputs to the same source and you've got circuit that perfectly doubles a sine wave. Just add some clipping diodes before the multiplier to compress the quadratic signal a bit. Maybe a different amounts to each input.

I have a few MC1496's lying around that I was going to try for that purpose. I know that suggestion has been tossed around quite a bit for octaves.

This isn't really going for the perfect octave obviously, it's gonna lend quite a bit of distortion as well (not as much as rectification, but still). I'm mainly after the dynamics/surprising sounds I might be able to get out of this. It will also probably end up fairly simple if it does work.

SOMEBODY must have the bits for trying this out!

I do, I do! I'm working on it today. Right now I am setting up a. IV curve tracer for my scope so I can test different JFET combinations or maybe even the JFET/PNP combination (yes you can use a PNP bipolar for the lambda diode instead of a P channel JFET)

[JKowalski]

Alright, the basic concept is...

A success!

The combination of a N-Channel 2N5859 and a P-Channel 2N3820 make a great curve. When set up as in the simulation, a sine wave adjusted to the right point is almost perfectly doubled. It's surprisingly clean, much cleaner then the JFET sim. Even doubling occurs for a wide range of input signal strengths.




I'll start breadboarding an quick n' dirty pedal for the breadboard to try it out with audio.

[Taylor]

Awesome.

[trad3mark]

omg omg omg!!! Just read this. Does this mean that potentially we could have a good clean octave up effect that's tiny and easy to build!?

[Taylor]

It'll still probably be a little dirty, and certainly not polyphonic. No amount of slick design will make this work like a POG. But I think it will be cool for what it is. And I like to see new stuff arising on this forum.

[JKowalski]

You simply cannot get a perfectly clean octave with non-DSP effects in a polyphonic way (even in a monophonic way really) because of the sheer complexity of a guitar signal. Even though this may give a great approximation of a doubled sine wave a guitar note is not a sine wave. Don't expect too much out of this, in the end it may be just another funky circuit trick if you ever need to double a sine wave (or turn a traingle wave into a 2x hypertriangle  )

The POG probably converts the input signal to digital, and stores one period of the guitar signal in a special place. Then, it outputs the stored period twice (repeated once) at 2x the speed of the ADC input. This would give an almost perfect octave effect.

Here's what I mean. Top left is a made up guitar signal (single notes usually look close to that). Bottom left is the ideal octave of that signal. Bottom right is what you get with this method and any other analog version (equivalent to flipping half the signal over, except the lambda diode does that with a slightly rounded response.

[trad3mark]

ah ok. Now i get it.

I've a whammy, but the tracking on it is so inconsistant. I'm going to definitely consider investing in a POG while i'm in NY, but they're so bloody expensive.

[JKowalski]

ah ok. Now i get it.

I've a whammy, but the tracking on it is so inconsistant. I'm going to definitely consider investing in a POG while i'm in NY, but they're so bloody expensive.

Sorry to kill your enthusiasm!   

[Taylor]

If it helps you get motivated, I actually have programmed and built my own digital pitch shifters and I'm still interested in this design.  I just like the idea of stretching the boundaries of DIY a little bit.

[phector2004]

I just see analog octaves as grimy little frequency doublers to play with until I can afford blowing money on something cool like a HOG

I'm hoping this'll work, of course 

Hopefully it'll sound nicer than a green ringer as well, though I can squeeze out some cool video-gamey octaves out of that by messing with my tone knobs