Voltage controlled 10V->100V DC-DC converter

Output current is going to be in the low mA range, though I don't have an exact number (the load is almost entirely capacitive, with a very high R in parallel). I was hoping that the physical size would be something like 1"x1" on a PC board, if possible. And as far as efficiency, I assume you mean

Power out / Power in ?

If so, I can deal with very low efficiency. I'd have no problem having a standard rack HP power supply, or a few of them, powering the whole thing. As far as a control input voltage, I'll be controlling these DC-DC convters via CPU (with a National Instruments analog I/O board) that can supply -10V-10V analog (I can do digital too, but sounds more complicated).

I think you want a linear high-voltage opamp, such as those made by Apex, powered from a small 150V dc-dc switching power supply. Setup the opamp for a gain of 10x. Yawn.

--
 Thanks,
    - Win

--
How small is "kinda small"?
How much output current and what kind of efficiency do you need?

I don't know much about switching supplies, but how about something like this:

+12V>--+-----------------------------------+ | | Vc Vout [L1] | 0-10 0-100 | | | / +---[CR1>]--------------------------|---|---+--->>---+ | | | | | | | | | | C A---+ /+|--+ [R3] | Q1 B--[R1]-+-Y NAND | A-------< | | +---+ E | B---+-+-Y NAND \\-|------+ | | | | | B---+ | | [C] [R] | | | | | | | | | | +--[R2]--+ | [R4] +---+ | | | | | | | +--------[C1]-------+ | | | | | | | GND>---+-----------------------------------+-------+--->>---+

Basically a boost converter where the two NANDs form a gated oscillator which turns Q1 on and off, forcing current through L1 and then shutting it off abruptly which makes a spike that gets integrated by C (your capacitor). As the stream of spikes charges C, the voltage at the junction of R3 and R4 (a 10:1 voltage divider) rises until it equals the voltage on the comparator's + input, at which time the output of the comparator goes low, stopping the oscillator and turning Q1 off. Then, when R (your load resistance) discharges C to the point where the voltage on the - input of the comparator goes lower than the voltage on the + input, the comparator's output will go high again, starting the oscillator and recharging C.

--
John Fields
Professional Circuit Designer

I didn't notice you were going to make 100 of these, powered from a single power supply. The APEX amplifiers are easy to use, but they are expensive. I still think a linear solution bests a switching supply solution, but with 100 units something simple and cheap is in order. I suggest you use a circuit like fig 3.75 in our book, except with transistors instead of FETs. Here's one that should work well:

.. ---+-------+-- 130 to 150V HV supply .. | | 100 amplifiers requires .. R4 10k R5 70mA maximum (all at 0V) .. Simple, Cheap 220k | .. Precision Slow | |/ Q2 .. HV Amplifier +-----| mpsA42 .. | |\\v .. G = 1 + R2/R1 | | R3 .. +--|--+-- 3.3k --' 1.00M 1% .. ,--|-_/ | | .. | === 330pF | .. | | C1 | .. '---------+---------------------+ .. | R1 .. +/-15v opamp supply 100k 1% .. | .. gnd

I've used a common-base non-inverting high-voltage stage to allow simple one-cap C1 feedback-loop compensation. R5 and R6 provide short-circuit limiting. R3 isolates the feedback loop from large capacitive loads, but can be eliminated if you increase C1. Note, the mpsA42 transistors can handle 300 volts, so the output range could easily be increased.

You haven't mentioned the accuracy or speed you need. If these are very relaxed specs, you can eliminate the opamp and drive the level- shifting stage directly from your D-A stage, using the -10 to 0V part of your D-A range. Although a regulated HV supply is required, this is a dramatically simplified circuit!

.. ---+-------+------- Vcc = 120V .. | | regulated HV supply .. R4 27k R5 60mA max for 100 amps .. 200k 1% | .. Super Simple | |/ Q2 .. no-feedback +-----| mpsA42 .. HV Amplifier | |\\v .. | | .. +--|

A few gotchas to watch for in that boost converter are that 1) the 12V feeds through L+CR1 to output when the transistor is off so that output is never less than 12V and 2) the steady state Vout/Vin= 1/(1-Duty) where Duty is transistor switch duty cycle- so you would want that to be about 0.1 or so when Vin=12V. A 50% duty cycle gives you a maximum of

24V for Vout.

Okay, thanks, gotta sort all this info out.

And sorry, what is "our book"?

Voops->(1-Duty) is about 0.1 which makes Duty=0.9 for a x10 boost...

Jesse,

Are you going to have HV generators with 100 separate outputs and 100 separate (0-10v) inputs, all controlled independently?

I thought you needed 100 high voltage supplies running independently and that you needed to control the outputs based on an independent analog input, hence my suggestion of the low power EL inverter chip which could be controlled with through the dimming input.

I'm wondering, can you have a single (relatively large and powerful)

100 volt source with 100 outputs that are (effectively) electronic variable resistors???? I don't think there are any cheap low power high voltage op amps out there, they usually tend to be high voltage and high power (and expensive).

If you have a single big 100 volt power supply, I'd think that a PWM chip hooked to simple switching transistor would do the job. There are some low cost low power pwm chips available.

M

--
OK, thanks.

The Art of Electronics, by Horowitz and Hill,

--
 Thanks,
    - Win

Its cheap for its size and content. We'll have a new edition in a few years, if we can pick up the writing pace. But, respectfully, it looks like Jesse needs help now. He should get a copy. You can get one too, Mebart. :>)

--
 Thanks,
    - Win

It looks like an absolutely outstanding book, but an expensive one. Are there any plans for an updated version??

M

Oh, I DEFINITELY have a copy. We use piezos on all of our lasers actually (referring to figure 3.75 -- 1kV low-power piezo driver) though I shamefully say that all of our laser driving hardware is bought, and not built (except for some of the PID locks). I just didn't exactly know where to look for a power supply circuit like this

-- It should have hit me that driving a piezo is basically what I'm trying to do (big capacitive load, no current).

Incidentally, in case anyone is interested, we're a neutrino physics group at Stanford

I'm on a team building a linear ion trap (for Barium actually), which is basically a set of 4 segmented rods (15 segments each) that are driven at about 1MHz@100Vpk-pk with a DC offset that needs to be controllable. By changing the DC offset, you can move on ion around wherever you need it. After we trap an ion, we hit it with 493nm and 650nm lasers (650nm is an off the shelf diode, 493 is a doubled diode using KTP) to drive it's levels, and we look for the fluoresence. And, though I'm a physics grad, I unfortunately have WAY less electronics experience than I should (other than the super basics -- opamps and stuff).

Excellent!

Did you say exactly what capacitive load you were driving?

Then I assume it's the dc-voltage on these electrodes you're driving, with, I suppose, some sort of LC network to allow the RF drive onto the electrodes, which I assume are resonantly driven (that's how we do our traps and quadrupole guides). How many electrodes do you have? Is there a paper describing this setup, or one that's similar?

--
 Thanks,
    - Win

Then just use the deflection amps from a 'scope. (well, maybe a couple of 'scopes, or maybe a couple of dozen - do you have to drive each pair of each segment individually?) A couple of hundred Vpp at almost no current is what those guys were designed for. And the "position" control gives you your DC offset. ;-)

After we trap an ion, we hit it with 493nm and 650nm lasers

Good Luck! Rich

Sounds interesting, but y'all have to aim at something that's pretty simple, more simple than the stripped-down linear circuits I posted. Let's see some drawings and parts counts.

--
 Thanks,
    - Win

To be more detailed, the trap is 4 "rods", where each "rod" is actually

15 rod segments, each electrically isolated from the other, but very close together.

Cross section:

o o o o

Side view: ___ ___ ___ |___| |___| |___| ___ ___ ___ ... |___| |___| |___|

each rod segment is ~ 3cm long, with a diameter of ~6mm. All rods have to be driven at 1MHz, 100Vpkpk, and each set of 4 rods (e.g. the 4 in the cross section view) are going to be at the same DC potential.

In order to drive the RF, I just built the following extremely simple circuit

G1|\\ +--|/--- ------------------- | ( ( | | Vac ) ) --- --- | ( ( ---Cmatch ---Ctrap | ) ) | | -------- ------------------- | N1 N2 | GND L1 L2 GND (wound toroid)

Where (1) Vac is just an HP fuction generator (needs 50 Ohm output) (2) G1 is a 50dB honkin RF amplifier (50 Ohm output) (3) N1/N2 = 20:5 wound micrometals toroid transformer (4) Cmatch is a variable cap for making sure that Ztot=50Ohm as seen by G1 (5) Ctrap is the total capacitance of the rod segments (~500pF max)

I'll probably have to add a small resistor (10 Ohm?) somewhere in the right hand side to lower the Q of the transformer, which is too high right now. The tranformer does 2 things for me. First, it matches the impedance so that the right hand side looks like 50Ohm to G1 (tweakable by Cmatch) and it gives me voltage gain of ~ 4, so that if I want to run Ctrap at 100Vpk-pk, then the power dissipated in this circuit is only 12.5^2/50=3W.

Now, the obvious question is how to insert my DC supply (I'm building it right now...) so that Ctrap sees the DC potential, but also so that the RF doesn't go back into the DC circuit to kill it. I'm not really sure how to do this yet... Possibly a bias-tee, but that's my only idea.