Hey smart people: what can we do with tunnel diodes?

[Taylor]

Occasionally I start reading about weird exotic electronics stuff. It strikes me as strange than there are components that just about nobody uses in effects.

One intriguing one is the tunnel diode. I barely understand the common components on a truly fundamental level, so theorizing about a use for tunnel diodes is way over my head. But surely they can be used for something. The obvious is to plop them in a tube screamer as feedback loop clippers, or shunt clippers. I'm sure that's all this thing is doing.

But, you smart guys are way more clever than that. I'm positive that you can come up with a more intriguing use for these things. Falstad's simulator includes the tunnel diode. I tried to play around with it to get something interesting with no luck.

Any ideas?

[jasperoosthoek]

You can connect two back to back in series to use their IV curve. This you can use as clipping diodes that generate much higher overtones. My old electronics professor told me nothing is easier than building a tunnel diode oscilator: just connect them to a power supply to get RF...

They are very hard to use in a stable circuit but it is possible to use them in a octaver/harmonizer that sounds much smoother and just different. PCB/perfboard layout is critical to avoid oscillation. They die quickly so you will have to be careful.

I've played with this idea for years as I worked with a TD test setup.. If you want I can help you design something.

[CynicalMan]

My old electronics professor told me nothing is easier than building a tunnel diode oscilator: just connect them to a power supply to get RF...

I can make a 60Hz oscillator by touching the end of my guitar cable.... 

For those of use who prefer spice programs, here's a macromodel for a tunnel diode:
www.aeroflex.com/AMS/Metelics/pdfiles/TUN_SP.pdf
Datasheet is here:
www.aeroflex.com/AMS/Metelics/pdfiles/MBD_Series_Planar_Back_Tunnel_Diodes.pdf

[Taylor]

You can connect two back to back in series to use their IV curve. This you can use as clipping diodes that generate much higher overtones. My old electronics professor told me nothing is easier than building a tunnel diode oscilator: just connect them to a power supply to get RF...

They are very hard to use in a stable circuit but it is possible to use them in a octaver/harmonizer that sounds much smoother and just different. PCB/perfboard layout is critical to avoid oscillation. They die quickly so you will have to be careful.

I've played with this idea for years as I worked with a TD test setup.. If you want I can help you design something.

Cool! I think it would be really fun for us to put together something new and interesting for the forum like this.

When you say back-to-back, do you mean the way most people use that term, i.e. the way the diodes are in a tube screamer (which I always thought should be call back-to-front) or actually with the 2 cathodes ("backs"?) connected, and then connecting the anodes to the rest of the circuit?

And when you say octave/harmonizer are you talking about rectifier things like Octavia-type circuits?

[jasperoosthoek]

Yes it would probably sound like an octavia but with 'cleaner' higher frequencies (hopefully).

Tunnel diodes have the very weird property that they conduct most current in the "backwards" region. So where a normal diode blocks current they conduct heavily. Therefore they have to connected in series instead of parallel. In this case back to back means: connected in series with their N side or P side together.

In the forward region they first conduct, then when you increase the voltage the current drops (negative differential resistance) and when the normal turn on voltage is reached they start to conduct again:

Taken from this site: http://www.mpoweruk.com/semiconductors.htm
If you connect them back to back you will get this:


Although this looks like an easy circuit it is very difficult to get to work in practice. The problem is to prevent oscillation as the voltage of the tunnel diode is in the negative differential resistance region. I'll have to look at the information from the working circuit again. I haven't looked at it for 3 years as it was a project for undergraduate students to get some insight into electronics. I was the supervisor for three years.

I think the best thing is to add some resistors in parallel so the total IV curve does not have a region that goes down. Basically it's like adding a I=V/R (linear equation) to the curve. Then this resistor has to be connected as closely as possible to avoid RF oscillation. RF oscillation causes a small jump in the IV curve which simply results in a harsher sound. Another issue is current protection. During the 3 years of supervising the tunnel diode setup I lost about 3 units. Mostly due to errors of the students but also possibly due to the fact that they just break down easily. Therefore I connected two normal diodes back to back to avoid too much voltage on the tunnel diodes.

A better design would be something like this:

The normal diodes protect against straining the tunnel diodes too much. The other resistors have to be similar to the ones my test setup. Afterwards it might be needed to remove some of the clean signal which is caused by the extra resistors.

I might just order a few and play a bit with them.

[arawn]

Yeah I agree with you about the things people don't use. But my brain is always working, makes me a little bit of an uberspazz but hey! I Found some diodes at work that i've never heard of anybody using but I think They sound pretty awesome. They are motorola MUR140 normally uised in RF applications, But they sound like a cross between silicone and Germanium.

[jasperoosthoek]

That may be nice for a switchable diode stack in a Tube Screamer or something similar. I know 'Spud' is working on a UBE Screamer (together with me) from Run Off Groove. (Posted on the AX84 forum). He likes to implement various diodes with a switch. This might be a nice addition.

[jasperoosthoek]

I've been trying to get a quote/price from a company that produce new tunnel diodes for 4 days now but they just don't want to give anything without information about my company  (I am a hobbyist without a fax, not a company!). I mailed them again begging to tell them me the price. ($0.5 or $50, just let me know).

So instead I just ordered a batch (10) of Russian GaAs tunnel diodes on ebay! 18 euros including shipping (~US$22) which is very cheap for even a single tunnel diode on some websites.  These AИ101Д give some inspiration on the character set that I'm going to use for the name of my new effect. Hmm, something Cyrillic maybe . Maybe this thing is going to be called the TЦИИEL BLДSTEЯ or something along those lines. I'll make a new thread when I get some initial results.

[Taylor]

Very cool, excited to see what you come up with.

[JKowalski]

Do you know about Lambda diodes? Two JFETs connected in a way that gives a IV curve with a negative resistance region. Cheaper to make then tunnel diodes.

I totally forgot about this since at the time I didn't have suitable JFETs but now I suppose I should try it out, I think I have them. I had a topic a while back where I messed with lambda diodes to give clean(er) octave effects.

See here

[Taylor]

Hmm, very interesting indeed. Do let us know if you build it.

[puretube]

Hi time to dig out them over 70 years old Benz-patentz... 

[jasperoosthoek]

I'm going to read though all that tomorrow! If the IV curve is smoother then it might sound better. Still you have to be very carefull with connecting devices with negative differential resistance. If the loadline crosses at more than one point it can switch. That is exactly what I tried to avoid in my sketchup above. The parallel resistors remove the instable point. They have to be smaller than then negative resistance. I've seen the IV curve of a real tunnel diode on a scope over a hundred times and I know how tricky it is to make it smooth. But those TDs were clamped, not soldered. The ones I ordered are GaAs btw, 50% rat poison...

The problem is that switching causes a glitch in the signal which, of course, will sound fizzy. Something I hate. Me want full harmonically rich fuzz. If I want that I'll try a Schmitt trigger like circuit.

I'm interested in your lambda diodes. I hope you start tinkering with them. If I get my tunnel diodes working I might try them too. It might just be that the end result would be the same.

[Lurco]

The curve in Fig.50.(b) looks good in this manual

[jasperoosthoek]

Yeah, that's exactly what I'm going to use. I received my tunnel diodes on saturday. But I have to ask a Russian guy to help me interpret the datasheet. My Cyrillic is rather bad :-).

I almost have the circuit planned out. First I'm going to play it safe and copy the circuit I used a student. You can only use the IV in fig 50b if the load line is almost vertical. At least the slope (1/resistance) has to be higher than the negative slope of the tunnel diode. My tunnel diodes have a negative differential resistance of 14 ohms, this means that my load resistance needs to be smaller than 14 ohms. If you have a higher load resistance then the tunnel diode will start to switch between the stable points X and Y. That will generate a fizzy sound.

This is another sketch of what I'm planning to build:

The arrangement with the zener diodes is based directly on the circuit that I'm familiar with. I just doubled the circuit but this could also be done with one polarity.

It is basically a fase splitter and a push pull transistor arrangement driving a transformer. I'm planning on using the Mouser 42TM022 transformer that is also used used in the Tycho octavia: It is readily available and works hopefully. I use a transformer instead of a LM386 amp because it allows me to easily get the signal from the tunnel diode stack. And because it means I don't need opamps. I really favor the simple transistor circuits.
Another two transistors amplify the voltage across the two 2 ohms resistors. I didn't do any circuit similation to check how much gain I need. I will start with getting the zener/tunnel diode part to work with the transformer and see how it works for real.

I didn't include all the necessary bias resistors on most transistors but just assume they are there. Furthermore, the reason I stepped away from having the tunnel diodes back to back is because the forward resistance will probably be too high causing the load line to be more horizontal. This might even damage the tunnel diodes. I don't want any switching in the circuit, just a smooth TD characteristic.

[jasperoosthoek]

Just a small update, nothing special except that this is a hint of an IV curve of one of the russian GaAs tunnel diodes I ordered from ebay! I connected the zeners you can see in the circuit above directly to a signal generator and displayed the voltage across the tunnel diodes (X) and the voltage across the 2 ohms resistor (Y). The problem is that the scope and the signal generator share their ground. But, you can see the negative differential resistance!   I will use an isolation transformer in my design later. Probably a cheap one 100ohms to 100ohms  from mouser. My GaAs tunnel diodes require more voltage and current (power) to work than the usual germanium ones so I'll need either have to change the design (no zeners) or use a bulky transformer (more than 200mW). But I prefer to keep it simple and cheap by using the $1.60 42TM030-RC transformer from Mouser.