Using reverse biased zeners to correct for mosfett turnon voltage in a push pull stage

load

which

parasitic

level

be a

an

more

As it turns out the feedback is simpler than the actual data reconstruction. The simples controller layout is a cascaded PI loop where your controller input from your lock-in or beam deflection signal controls directly the fastest piezo on the stack for a deflection/amplitude setpoint. Every further piezo takes the output of the next faster PID and tries to control for zero voltage there, meaning no deflection of the piezo or in other words keeping the small ranged fast piezos centered around their range. Actually setting up the controller in the lab is still another story ;)

The opamps only see the power from the dc/dc converter, +-12 or some such. They basically float on the output rail.

Of course, the input, from some driver stage, has to swing almost from V+ to V-.

Actually, I like the second circuit better. Its crossover behavior is remarkable.

My experience (limited, since MOSFETs are just that much nicer!) is that, at least without a huge amount of reverse bias, they tend to break down at Vceo, which is a paltry 600V, versus the 1500V Vcbo. This is kind of unusual as BJTs go, because I made that measurement at Vbe = 0 (inductive, pulsed), a condition where the average BJT goes pretty close to Vcbo. (2N3904 goes to 80V or so with a moderate B-E resistor, and also exhibits avalanche behavior which can be of use.)

I'm surprised they snap well, I'd think the capacitance and lead inductance would hamper things greatly.

That 1kV pulser you made, you ended up with a very particular selection of part number, manufacturer and reel, of some common rectifier (1N4007 or some such), right? 50A through a HOT, if it snaps well enough (given its capacitance), should do about as well. Hey, if you can push it to 1500V in a split second, that's even more into 50 ohms than the 1kV you had!

Tim

This is still common practice in audio with discrete bipolar pass transistors, where the Vbe multiplier is a thermally coupled small signal transistor. From the sixties.

No reason why the concept couldn't be used with mosfets. You'd have to have the same part number for the signal as for power though, just to ensure that the tempco's are similar - too much process variation otherwise.

RL

We got about 2KV into a small capacitive load (a tomographic atom probe electrode) at the end of a coax, about 2 ns wide, 100 KHz or so. The best device we found was the c-b junction of a HOT transistor, which was nice to heat sink, too. The coax doubled the voltage, for free.

With FETs it isn't a BE junction where a distinct voltage drop develops, you'd have to design a wee bit of electronics around that. It won't work to well even then because your sense FET will not see tens of watts of dissipation slamming into its die but the FET to be protected and bias-regulated does.

You might try a mojito instead. Very tasty. =20

?-)

a dc

to

statically

the load

way which

parasitic

might be

a level

there to be a

there an

overlooked?

too

distort

current.

SOA

opamp

diode

vary more

and

tons

and

opamps

swing

I would be rather careful with the idea as the final output swings quite = a bit farther than the opamp rails.

?-)

load

which

parasitic

level

be a

It forms a voltage divider, which is what one needs to control a HV output from a LV amplifier.

The general scheme works great.

Cheers

Phil Hobbs

load

which

parasitic

level

be a

an

more

Somebody showed me an AFM a while back. The tip assembly flexed from the local tip force, which they sensed with a laser bounced off the back of the flexure into a differential photodiode. No interferance or anything like that. The flexures are so soft, that's all you need. You can buy such a tip+flexure for around $10 now.

That'll be a lot more difficult than with BJTs. Another major impediment will be those lot-to-lot variations. With FETs they are huge. With BJTs its mainly the beta that varies a lot but that has little impact on the quiescent current if it's all done right.

In this case I'd prefer clamping. Stop the microscope every so often, current drops down to Iq, regulate Iq back to where it should be, re-start the microscope.

That's why the thermal coupling is used.

The issue in a basic schematic, as drawn by the OP, is the unavoidable process variation between P and N-chanel parts. This wasn't an issue with bipolars, as the Vbe multiplication could be trimmed, and all had the same tempco.

A Vgs multiplier also has an interesting impedance, when imbedded in a signal train.

RL

RL

dc

statically

load

which

parasitic

be

level

to be a

there an

overlooked?

distort

current.

opamp

more

Yup, that's the optical lever approach. It was invented by Cavendish or somebody like that, back in the 19th Century, and perfected by R. V. Jones in the 1950s. It works better with lasers, but not as much better as you might think--you get more light and arguably a better beam profile, but with lasers you suffer badly from etalon fringes and spatial side-modes, which are absent with broadband light. Running the laser light through a single-mode fibre before use helps a lot with side modes and pointing stability, but doesn't help the etalon problem.

My gizmo worked mostly in attractive-force mode, and was able to sense force gradients of around 1-10 pN/m. (I didn't invent the technique, but I did put some new wrinkles on it, e.g. using a phase sensitive detector tuned to nearly 90 degrees. That got rid of the major instability of the cantilever devices, which was as follows:

1. Cantilever vibrates at a frequency near but slightly above its first bending resonance, with an amplitude servo detecting detuning of the cantilever due to an external force gradient. Force gradients look just like a change in the spring constant, so that moves the resonance, which changes the response of the cantilever to the piezo's excitation.
2. The tip touches down on some asperity on the surface, and sticks due to capillary force. The oscillation amplitude goes to zilch. The servo interprets this as the tip being way, way too close, and starts pulling the piezo away from the surface.
3. At some point, the tip lets go, and the vigorous plucking action makes it vibrate like absolute mad at its free resonance.
4. The servo interprets this as the tip being way, way too far out, and crashes the tip into the surface again
5. The cycle repeats until somebody notices or something breaks.

Using phase sensitivity makes the servo almost insensitive to the oscillation amplitude, and ringing at the free resonance is outside the loop BW anyway, as opposed to an AM-only system. That trick let me servo much closer to the surface, and improved resolution and sensitivity by a lot. (I published an Applied Physics Letter back in about 1988 showing 25-nm resolution in magnetic force microscopy, which I was pretty proud of at the time.)

A little later, my friend Tom Albrecht came up with a very slick scheme, which was to run the tip as an oscillator. The reason this is good is that when you're using external excitation, you're limited by the Q of the resonance, whereas since the mechanical resonance itself has no memory, it responds instantly to changes in the spring constant. The result is that you can go much faster with a self-oscillating tip.

It's very hard to get the same sort of crash resistance that way, though, so my original phase sensitive approach still has its uses even though it's slowish.

I haven't done any AFM stuff since 1989, though, so I'm sadly uninformed on later developments.

Cheers

Phil Hobbs

turns

an

sort

that

quite a

I was discussing this particular circuit, not one of Adrians whose supply is servoed.

?-)

Yes, but you will have a "fun" time matching the TCs, plus each part is in its own thermal universe = = NO thermal coupling.