Low voltage negative resistance oscillator design, lessons learned

They do have a negative slope of C vs V. Sort of a negative reactance. Sort of.

I bet a step-recovery diode could be an amplifier. A PIN diode certainly can.

A MASER is a pumped amplifier.

Paramps have to be pumped. I vaguely recall that this one was pumped with a klystron, but it would have had to be stabilized somehow--the fifth harmonic of an ammonia maser (120 GHz) would have been in the right ballpark. A TD could probably be run as a self-oscillating converter as John says, but I expect it would be even rattier than a hexagrid converter tube.

Cheers

Phil Hobbs

I've appended the imgur post with a scope capture of the oscillator. It runs at about 200kHz.

I do still wonder how they found the 2N5962. The datasheet says they are general-purpose amplifier transistors. The ones in the varactor amp are marked with red paint, too (a bit hard to see in the pictures), so AD must've handselected them. Maybe they were just cheap? I guess the main important thing is the matching.

Nice info regarding the color codes and the capacitors "disguised" as resistors, I didn't know about that.

Can't find the scope photo. So you have a url?

What is the waveform and amplitude into the bridge?

Thanks

Sure, here's the image:

I found the waveform. Thanks. It's a plain multivibrator. I guess that makes sense. It's about the simplest oscillator you could make at these frequencies.

I'm a bit confused by the amplitude reading. The photo shows the 10X scope probe is on, and the readout along the bottom shows Vpp=6.56 V. You give the amplitude as 80mVpk-pk. None of these numbers agree!

The datasheet for the 2N5962 says see the 2N5088 for characteristics. The sN5088 has the SPICE model. Here's the links:

Thanks for your help.

Ah yes I see it now. Geez, I drew it pretty much in the worst way possible :)

The scope capture is somwhere in the oscillator. Must've been the collector of the side not going into the bridge. The signal into the bridge is tapped off on the 33 Ohm resistor that together with the 3300 Ohm resistor forms the load in the other collector.

The 80mVpk-pk was measured across the junctions/bridge windings. That's pretty close to the 2 * 7/9 * 1/100 of the two 7-turn bridge windings in series on the core with the 9-turn primary winding after a 1/100 divider.

Thanks for the spice model, by the way. I might model it sometime.

I drew the schematic pretty quickly, there might be some errors. The blended picture should help pick those out.

I'm glad to share it.

Thanks for the readings. That is a strange bridge. The 2N5088 datasheet shows the Cob is around 4.8pf at 0 Volts. I had a hard time finding the meaning of Cob, but it shows up in the Toshiba Bipolar Transistors Application Note at

Cob is the capacitance between collector and base at the specified collector-base voltage and frequency when the emitter terminal is open-circuited.

The description says "Amplifier Ein varies varactor capacitances, unbalances bridge and develops pump voltage output proportional to bridge unbalance. [...] Matched low leakage varactors inherently eliminate 1/f noise, give excellent offset and drift specs, and provide 3 x 10^11 Ohms differential input impedance."

So we are getting some more pieces to the puzzle.

Thanks!

No problem, this is fun :)

Interesting. Nothing in the datasheet jumped out at me as a reason for this device. Perhaps it has/had a high dC/dV, low leakage or both. I always thought the reason for this bridge is that only 2 "varactors" need to be matched, unlike for the P2s fully differential bridge.

There's an AD patent on the concept:

I think I have an AD article on the design somewhere, but I haven't been able to find it yet. It might have been an analog dialog issue.

What do you think of the feedback to the JFET source? My best guess is that it's there to bootstrap the capacitance.

There are teardowns and schematics available of the Philbrick P2 over at the Philbrick archive. Pease describes the design in "Analog Circuit Design Art, Science and Personalities" by J. Williams et al. He mentions its bridge getting "(...) perhaps 100mV of RF drive (...)".

Yes, 60-80mV seems about right. The RF modulation has to be very small to keep well away from the onset of picoamps forward conduction at higher temperatures.

piglet

True, it will bootstrap capacitance somewhat but I think more simply it is there to stabilize the open-loop gain.

piglet

Wow! Schematics! Description! Priceless! Thanks!

I can't find the jfet source you are talking about. Which file is it in?

Thanks. The paper is at

Your help has been invaluable. The patent especially. Now we have enough information to spend hours studying this amazing design. It still has merit for the unmatched common mode range.

It is the input FET in the OP's handrawn schematic, feedback is from the emitter of the second BJT stage.

The AD310 common mode range is only good at DC/low-AF since they penny-pinched and used capacitive coupling from the bridge to the AC amplifier. A fully transformer isolated design like the Philbrick P2 would be good to much higher volatges and/or frequencies.

piglet

The 310 bridge is transformer coupled. Essentially unlimited common mode range.

Why do you need wide bandwidth in a circuit intended for pA applications?

No they have negative noise, it's the negative of the problematic noise you'd expect :). Shame that doesn't help any.

NT

Hi Steve,

The AD310 has transformer coupling from carrier oscillator to bridge, but from bridge to AC amplifier is capacitor coupled. A glass 22pF ultra low leakage cap in one leg and a garden variety ceramic 1nF in the other (cold) leg. I suggest that a balanced transformer design like P2 or P2A or the AD patent would have better AC CMRR.

Wide signal bandwidth is not the issue but AC common mode interference could be a problem. Imagine the AC CMRR needed for a worst case dc-like pA signal riding on 120V 60Hz !

piglet

applications?

You are not going to find pA signals riding on 120V 60Hz. The 310 was intended for electrometer applications. These are static, DC applications.

In a related topic, whit3rd stated on Fri, 30 Aug :

"The AD310 usually sat in a teflon-bushed socket and might have to be hand-wired,"

Page 1 of the AD 310 description talks about the small size, high input impedance and low bias current needed for use in electrometer applications:

Yet the pcb photo and schematic posted earlier shows a circuit that would not fit in a teflon socket.

Which one is the real AD 310?

Thanks that's interesting. I made this 'audio' double balanced mixer with four diodes in a ring. I found that diode connected transistors (B-C shorted) worked 'better' than regular signal diodes.

George H.