Need the topology for a current to current SMPS, not voltage to voltage SMPS

So SMPS SHUNT regulator ?:-} ...Jim Thompson

--
| James E.Thompson                                 |    mens     | 
| Analog Innovations                               |     et      | 
| Analog/Mixed-Signal ASIC's and Discrete Systems  |    manus    | 
| San Tan Valley, AZ 85142   Skype: Contacts Only  |             | 
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  | 
| E-mail Icon at http://www.analog-innovations.com |    1962     | 
              
I love to cook with wine.     Sometimes I even put it in the food.

--
So you want to pump, say, 0->1mA into an input with an impedance of 
zero ohms and output, say, 0->1A into a load from an infinite source 
impedance? 

What's your: 
1. Supply voltage? 
2. Input current range? 
3. Output current range? 
4. Load resistance range? 
5. What do you mean by "Current-time profile?

So you've got a current source driving it? Do you need more current/ less current or both? (I'm taking a smps on the web... which means I know enough to say something stupid.) My first thought was to switch around the L and C and then chang e shunt and series... so a shunt cap on the input side and a series L on th e output. (I'd have to play with it some though...)

George H.

You didn't give much data here. Assuming that measuring the load current is iffy (because it usually it) one could instead measure the waste current. This has to include the current used by the converter itself. You could have a sense resistor and run a sync buck into that, which except for this sense resistor would be a dead short situation. The ground of the converter would also connect to the top of that current sense resistor so it includes its own consumption. Now set the loop so it sinks away the portion of the input current you don't want to reach the output into ground. The load would hang on it's input and would see the balance of the current.

Use a converter that will not go into foldback mode or goose it into regular mode, for example by fooling the voltage control (which is otherwise disabled).

--
Regards, Joerg 

http://www.analogconsultants.com/

In the current domain, the conventional topologies (boost, buck, etc.) work just fine with capacitors. Waveforms are an exercise for the student :)

I'm not so sure about isolation (flyback, forward), that might look like a signal coupled with two capacitors, one for power and one for ground return.

How ideal is the source it's attached to? If it's got a honking big bypass cap, it's not very constant for SMPS purposes. If it's got a massive inductor (or an open collector or something like that), chances are it's a pretty good current source.

Tim

-- Seven Transistor Labs Electrical Engineering Consultation Website:

Bandwidth? Allowed current ripple? Must the load be ground-referenced, or can it float? Can it _include_ a transformer? Expected load reactance?

If the bandwidth is well below the switching frequency, and the load is not strongly capacitive, then why not a 'plain old' voltage SMPS wrapped with a current loop?

If you can float the load all the better -- sense the load current with a resistor to ground.

If you can't float the load but you can have a transformer, then do something that's transformer isolated and sense the current with a resistor from ground to the output section's "ground".

--

Tim Wescott 
Wescott Design Services 
http://www.wescottdesign.com

I scribbled this at lunch... which I think works as a current buck.

(Fun question) George H.

Current output = some sort of flyback topology.

Done that. Stability of the control loop could be tricky.

It could be actually simpler to take classic voltage controlled buck or boost as a base for design; and implement whatever current control loop around it using MCU. This would be easy to do using standard SMPS ICs.

Vladimir Vassilevsky DSP and Mixed Signal Designs

A synchronous buck running at fixed duty cycle is the electronic equivalent of a pair of gears, or an ideal autotransformer. Current out is current in times 1/n where n is duty cycle.

--

John Larkin         Highland Technology, Inc 

jlarkin at highlandtechnology dot com 
http://www.highlandtechnology.com 

Precision electronic instrumentation 
Picosecond-resolution Digital Delay and Pulse generators 
Custom laser drivers and controllers 
Photonics and fiberoptic TTL data links 
VME thermocouple, LVDT, synchro   acquisition and simulation

But you need to take into account losses and its own power consumption because Robert likely has to feed it from the input current source as well. In forced non-burst mode a sync buck can burn off an incredible amount of the input energy, easily 5% or more just by idling. Like a car with the A/C running.

--
Regards, Joerg 

http://www.analogconsultants.com/

Yup :)

I think one of the strangest V-I transformations is the current mode push-pull inverter. The first thing that should be apparent is, it looks like a half bridge (switches go from side-by-side to in series), but what the CT transformer winding does is tricky to synthesize.

Tim

--
Seven Transistor Labs 
Electrical Engineering Consultation 
Website: http://seventransistorlabs.com

By push-pull do you mean the buck-boost? I had to draw all three out after work today. The buck-boost took the longest, (I had the current going the wrong way.)

I ended up with this,

+---+----+---+--LLL--+ | | | | | ^ FET | _ R I | C ^ R | | | | | +---+----+---+-------+

^ where I is the current source.

George H.

No! I think it needs another diode... maybe there is a better way?

+---+->|-+--+-LLL--+ | | | | | ^ FET C _ R I | C ^ R | | | | | +---+----+--+------+

George H.

Nah, PP, as typically used in low voltage inverters (DC input, balanced AC output), or tube amplifiers. Should really be called "pull-pull" (often was, back in the day), but that doesn't roll off the tongue so well.

Hmm, buck-boost is an inverting topology, needs to produce negative current (L and R should be "below" the ground wire :) ).

Tim

--
Seven Transistor Labs 
Electrical Engineering Consultation 
Website: http://seventransistorlabs.com

If the shunt is linear, it will eat power; but if it is a switched shunt..., got to think about that.

However, I forgot to mention something really important here. The input is AC, which [I think] I do NOT want to allow to go maximum compliance. ARRRGG! Usually right after I post a question the answer is obvious, but not this time. I must not even have a handle on the problem yet. My question is coming out like drivel. It's just I have this sense that the solution is an 'invrsion' of a normal solution, and I must turn my thinking 'wrong-side' out.

I could allow full compliance and then tap off what's necessary, but allowing the full compliance voltage and only tapping off to what is needed raises the spectre of what-ifs. One should ALWAYS allow for potential operation ANYWHERE within a supply loop. I remember the days when power supply designers neglected to consider 'brown outs'.

.....snip...

Thanks for the schematic, I see the philosophy in it.

I neglected to mention the MOST important aspect, the input is AC, not DC. There is a potential full compliance could 'pop' to over 1kV, which is obviously destructive. But, with a dead short voltage is not much, because current is not much. Thus, my question here.

I neglected to mention that the input is AC! Sorry. I can't take full copliance voltage, so a dead short protects all the ocmponents.

Why would stability of the control loop be a problem? Let me reword that. Why would trying to do this type of control loop be more difficult?

I'm not sure how to do what you suggest.

Picture 1kV AC in and I either short it out or use it, that way the voltages on all the components keeps low.

What kind of source? Huge assed resistor? Analogous to series street lighting (massive AC CCS transformers, the kind with the secondary on a hinge and counterbalance!), etc.?? Any idea of parasitic capacitance or inductance near the load?

AC input isn't *quite* a big deal, because you can use a synchronous switching design, where MOSFET and diode are both replaced with antiseries MOSFETs. (A bidirectional synchronous buck, at fixed duty cycle, is pretty cool -- it's a VARIAC without the iron core!)

Also... EMI concerns? Can the line be switched directly, or does it need filtering? (You might recall from the current-to-current buck circuit, the line will see full switching voltage -- just as a conventional voltage-to-voltage circuit sees full switching current. Both kinds are nasty to nearby radios without bypass and EMI filtering.)

Tim

--
Seven Transistor Labs 
Electrical Engineering Consultation 
Website: http://seventransistorlabs.com

OK... We haven't done many AC circuit yet.. but I'm guessing the same sort of thing can be done. For completeness here's the way a current source buck-boost should look.

(I figured out in the shower that the capacitor was in wrong.)

George H.