Tunnel Diode Voltage Reference

[CaptnYellowShirt]

Check out this patent:

http://www.google.com/patents/US5099191

I've never even heard of a tunnel diode. Does anyone here have experience with such things?

[mathias]

As a physicist, this was one "partical" example of quantum mechanics in one of my courses. I've played around with them with a curve tracer a while back, but apart from that, I haven't done anything practical with them. You can buy them relatively cheap on eBay from Russian sources, usually as amplifier, detector or pulse generator. Tek apparently also used them in their triggering circuitry as fast switches.

[CaptnYellowShirt]

The zener diode also using is a tunneling -- albeit in the reverse bias direction. The zener effect allows electrons to tunnel from the valence band to the conduction band. Do you know the mechanism in the tunneling diode? What type of tunneling it is?

[jimmc]

Never heard of them as a voltage reference, they're more of a negative resistance device.
Useful as high frequency oscillator or switch.
High cost meant that they never really caught on, most common use was in the trigger circuits of early Tek 'scopes e.g. 475

Plenty of information here
https://archive.org/details/GeTunnelDiodeManual
or here
https://archive.org/details/RcaTunnelDiodeManual

Jim

[G7PSK]

I first heard of tunnel diodes in the late 60's when I read about someone in New Zealand getting a long distance QSO record for a quarter watt tunnel diode transmitter, NZ to US. I never really saw any mention of their use other than for that purpose.

[muvideo]

Interesting, looks like a funny thing to play with.
If I understand the circuit well the Diode is used as a
constant current source that feeds the 1kohm resistance.
Output voltage depends on the 1kohm stability, input offset
voltage of the opamp, and mostly on current trough the Diode at
Voltage Setpoint.
Working near Vpeak the current will be Ipeak and will depend
only slightly on Vp. A resistance in parallel to the diode can even
flatten out Vp vs Ip curve.
AFAIK Ip variation over temperature has a minimum at a specified
forward voltage.
Also Vp variation over temperature is not very high.
Ip is nicely charachterized.
For example for 1N2939/2940 Ge TD, Delta Ip minimum
can be obtained for Vp around 60mV, Ip will be around
1mA +-0.1 at around 25°C, deltaVp can be 0.1% per °C:
see http://w140.com/Ge1961TunnelDiodeManual.pdf

Reading the patent, the authors seem to try to
exploit the Radiation Resistance of Tunnel Diodes
to create a simple yet stable and RadHard voltage reference.

Maybe not high precision, but esotic enough

[saturation]

I knew them as Esaki diodes.  To build oscillators around them of high performance was easier, Tektronix used them quite a bit.  I have a Tek pulse generator 284, which back in the 1960s generated 1 GHz.

http://w140.com/tekwiki/wiki/284

Tunnel diodes are used today is the GHz sphere, such as in radar and radio communications.

http://www.aeroflex.com/ams/metelics/micro-metelics-prods-TD-MTD.cfm

[mathias]

The zener diode also using is a tunneling -- albeit in the reverse bias direction. The zener effect allows electrons to tunnel from the valence band to the conduction band. Do you know the mechanism in the tunneling diode? What type of tunneling it is?

They are heavily doped diodes. This means that at certain voltages, the conduction band in one half of the diode is at the same energy as the valence band in the other half. That's when tunneling is possible. The electrons tunnel through the depletion area. The fact that they are constructed from heavily doped semi's, means that they are quite insensitive to damage (e.g. radiation).

Apparently, people were able to measure on one of the original tunnel diodes, more than fourty years after construction (from Wikipedia).

[T3sl4co1l]

I don't recall the effect being that the conduction and valence bands overlap (when projected across the junction), but that the junction is small enough (which does come awfully close, due to the high doping level).  Thus the quantum tunneling is the usual manner, crossing a physical width (junction thickness) with some barrier height (junction potential - applied voltage).

It's been years since I had a class on semiconductor physics, so I don't even remember how to do those problems...

I also remember separately that there was an experiment to build a laser diode in the same way, using a direct bandgap semiconductor of course; a heavily doped junction won't fully overlap (ah, I think I'm remembering something about this!), but given enough bias current, one can force a gradient over the junction and thus achieve an excess of conduction band electrons, a population inversion -- which dutifully recombines producing coherent light.  But they sucked because the efficiency was crap.   (The introduction of heterojunctions made that completely irrelevant of course, and we enjoy high efficiency laser diodes today.)

Tim

[RatherBeHiking]

Greetings.

On the topic of deriving reference voltages/currents from negative resistance devices, see: 
US patent 5,384,530
and
"A Bootstrap Voltage Reference Based Upon an N-Type Negative  Resistance Device", IEEE Transactions on Instrumentation and Measurement, vol. 42, no. 3, JUne 1993, 719-725.
For a proof-of-concept and an alternative application of this type of circuit see , for example,
 A.D. Martin, M.L.F Lerch, P.E. Simmonds, and L. Eaves,  Appl. Phys. Lett. 64 (1994) 1248, along with subsequent publications by  this  group. The reference works like a charm, but requires startup circuitry.

For another whack at accomplishing the same result, see
US patent 5,422,563,
and
"New RTD-Bootstrapped Current and Voltage References", IEEE Trans. Circuits and Systems I. Fundamental Theory and Applications, vol 41, no. 11 November 1994, 740-743.
Note Fig. 3 and Fig.  4 for estimates of output stability.

[CaptnYellowShirt]

Greetings.

Welcome to the forums!

Thanks for the helpful information. Did you run across this tread on google?

[RatherBeHiking]

Yes, via Google.

I don't think that these techniques have been exploited commercially. They may, however, offer some advantages compared with Zener-based references. The group out of Wollongong and Nottingham (Eaves et al.) is the only one, to my knowledge, exploiting NOR biasing of RTDs in the region of negative differential resistance (NDR) w/o oscillation.

While the patents are easily had from the internet, the technical papers are not, in general, available online and require some effort to obtain. However, it is interesting that the idea (for the op-amp bootstrap circuit) evolved at two places at almost the same time, motivated by two entirely different applications.

To access the dc I-V curve of an RTD or TD in the NDR a stability criteria must be met. To my knowledge this requires some plumbing.

[edavid]

Tunnel diode and voltage reference just don't go together... conventional tunnel diodes are sloppy, ill-defined, drifty devices, the opposite of what you want for a precision circuit.  The only reason you would try is to make this work is the special circumstance of high radiation, but it's not as if you can get space qualified tunnel diodes.

[RatherBeHiking]

First, thanks to the members of this forum for the opportunity to discuss this topic.

TDs and RTDs certainly have not been engineered to the degree that reverse-biased p-n junction diodes have. However, in defense of these devices I'd suggest that the glass is half full rather than half empty.

Voltage regulation due to a diode in reverse breakdown, which (for reference diodes) is typically in a state of mixed conduction wherein the reverse current is a mixture of tunneling and avalanche current, is achieved by via a steep but NEVER vertical current-voltage characteristic. Pick your operating point and take your chances.

Now, consider that (virtually - there are extreme counter-examples of lightly-doped devices) EVERY tunnel diode exhibits a local maximum and hence a zero in the I-V  function. This suggests an opportunity for current regulation by confining V_op to the region of V_p. However, it gets better. For voltages slightly larger than V_p the curvature (second derivative) of the I-V characteristic decreases, so that by using a parallel resistor the negative resistance (which for the composite device just became less negative) operation at V_op slightly > V_p (of the uncompensated TD) requires less regulation of the voltage than for uncompesated devices.  Perhaps there is an opportunity for some interesting engineering of the type that yielded such fine Zener devices.

In certain semiconducting materials it is not possible to realize high-quality Zener diodes (as is possible in silicon). The second patent and paper mentioned above is a first attempt at addressing that problem.

There has been no mention of RTDs so far. These are fabricated via epitaxial growth, a somewhat more controlled process than degenerative doping via diffusion. I doubt that there is sufficient data on engineered devices to determine if there is an opportunity to achieve regulation for industrial/space applications.

[RatherBeHiking]

Forgot  something.

Regarding drift, stability, et cetera and regulation - these can be two different issues.depending upon certain time frames over which the circuit is sampled. So, in the short term a device with a local I-V  maximum achieves near-perfect regulation in I_p for small fluctuations in V_op for approximately equal to V_p. However, TD or RTD device properties may drift in the long term (that time frame thing again) so that  overall circuit regulation degrades.

[RatherBeHiking]

Errata: Two messages prior a sentence  should read "... a zero in the first derivative of the I-V function" rather than "... a zero in the I-V function."

[edavid]

Voltage regulation due to a diode in reverse breakdown, which (for reference diodes) is typically in a state of mixed conduction wherein the reverse current is a mixture of tunneling and avalanche current, is achieved by via a steep but NEVER vertical current-voltage characteristic. Pick your operating point and take your chances.

Not a problem in practice, since you can cascade as many references as you like to get the desired line regulation.  On the other hand, there's nothing you can do to make tunnel diodes stable or predictable.

In certain semiconducting materials it is not possible to realize high-quality Zener diodes (as is possible in silicon). The second patent and paper mentioned above is a first attempt at addressing that problem.

What is this bizarre material that people want to use for voltage references, where you can build good tunnel diodes, but not zener diodes?

There has been no mention of RTDs so far.

Yes, because there is no sign that they will ever be more than a PhD thesis generator.

[CaptnYellowShirt]

Yes, because there is no sign that they will ever be more than a PhD thesis generator.

How much for this thesis generator... where can I buy one?!

[edavid]

Yes, because there is no sign that they will ever be more than a PhD thesis generator.

How much for this thesis generator... where can I buy one?!

OK, I should have have said thesis topic generator... "RTD based voltage reference", boom!

[Vgkid]

By rtd are yiu referring to reverse tunnel diode or (back diode)

[RatherBeHiking]

In this discussion, RTD = resonant tunneling diode.

[RatherBeHiking]

Another circuit that does the job. A prototype was fabricated and performed well:

Negative resistance device local extremum seeking circuit
US 4839832 A
Abstract
A negative resistance device local extremum seeking circuit including a negative resistance device (NRD) and a biasing device for providing a bias signal to operate the NRD. The output of the NRD is detected by a detector that provides a signal which indicates the position of the bias signal in relation to the peak of the NRD. A bias adjust circuit responsive to the detector adjusts the bias signal toward the peak of the NRD to operate it at the peak.

[edavid]

Another circuit that does the job. A prototype was fabricated and performed well:

Negative resistance device local extremum seeking circuit
US 4839832 A
Abstract
A negative resistance device local extremum seeking circuit including a negative resistance device (NRD) and a biasing device for providing a bias signal to operate the NRD. The output of the NRD is detected by a detector that provides a signal which indicates the position of the bias signal in relation to the peak of the NRD. A bias adjust circuit responsive to the detector adjusts the bias signal toward the peak of the NRD to operate it at the peak.

That's useless for a reference, since the peak point of a tunnel diode is not stable.

[RatherBeHiking]

eDavid - This is interesting. What data and sources do you rely on to state that the peak voltage is not stable, and that the devices can not be fabricated to tight tolerances? The statement about fabrication is a substantial assertion. Do you know if this been tried? Are you a device physicist?