Constant current sources

[raivo]

How are current sources made if they have to supply 50A or 500A or more? Are there such things as switchmode constant current sources? Any reading (or links) on this subject would be of great help.

 

[Barrie66]

Dynapower

http://www//dynapower.com

provide switchmode RECTIFIERS up to 4000A. They are available in constant current sources.

 

[ScottyUK]

A thyristor controlled rectifier with a large source inductor is the traditional way of making a constant-current source, with the rectifier being controlled in a closed loop from a current transducer in the DC output. This arrangement forms the front end of a current-source inverter. They aren't so common these days except in really large sizes.

How 'constant' do you want the current to be? And what impedance do you want to be able to drive the constant current in to? These are both pretty important parameters.


----------------------------------

One day my ship will come in.
But with my luck, I'll be at the airport!

 

[elfgriper]

raivo,
The simple answer is that almost any supply can be a current source, although different load types may favor different output types. The main feature is the feedback. Measuring the current thru the load (not the Voltage)by a series resister (shunt) or with a hall effect probe or a current xformer will give you a signal to feed back to the control circuity. We use a switcher (200 Amp DC) with a current xformer in the primary of the output xformer to feedback but there are lots of way to do the same thing. Let your mind do the walking.
-elf

 

[Warpspeed]

What is the application Raivo ?

Yes, switchmode supplies or SCRs will do it.

 

[Skogsgurra]

There are even linear ones. Bruker used to put about a hundred parallel 2N3055 in their high current supplies for NMR and accelerator beam deflection units. And a big heat-sink and fan to cool them. No noise and pretty fast, which is needed in such applications.

Gunnar Englund

http://www.gke.org

 

[raivo]

Thanks guys, some more questions. How paralleling of linear current sources is done? Just diode to the output of every current source before connecting it to the general output line... is it that easy?

Output voltage of supply can change between 0.5 and 24V, current has to be constant if voltage range premits, 10% ripple is acceptable. Can i do this with Buck converter schematics and suitable feedback circuit? What conversion efficiency i can expect from simple Buck converter considering wide output voltage range and rather mad downsteping.

 

[itsmoked]

Yes you probably can.. Spend some time on Linears' site find a "controller" rather than a regulator. Your voltage is well within their offerings.

You will need a custom inductor. 4-sure-Rocky.

 

[Warpspeed]

The picture is now beginning to clear.

Possibly the easiest way to do this is with a suitable transformer and a pair of SCRs. That will act very much like a PWM buck regulator, but it will operate at mains frequency making everything much simpler.

The load absolutely must be made inductive. If it is not already inductive, a suitable choke will be required to maintain current flow through the load during SCR "off" time. A flywheel diode providing the current path (exactly as in a buck regulator).

If very high load currents are required, it is going to be more economical to place the SCRs back to back in the transformer primary, and use a conventional rectifier after the transformer. Very high current diodes are a lot cheaper than very high current SCRs.

If three phases are available, ripple current can be made far lower, and response to load changes made approximately three times faster doing it with three phases.

It is then just a case of building a control system to monitor load current and adjusting the SCR firing point with negative feedback to hold output current at the desired level.

When designing the transformer, realise that a buck regulator averages the voltage waveform, and it is the average voltage (not the RMS voltage) that determines the available final maximum available dc output voltage, after all the other losses and voltage drops have been accounted for.

The only real problem with all of this is the response to load changes will be fairly slow. Possibly hundreds of milliseconds for acceptably low ripple and stable feedback.

If you need more speed, high frequency PWM, or a linear regulator would be better, but at a much higher cost and complexity.

One last observation. If your dc load is highly inductive, that may limit the response time, regardless of what method you use to control the power going to it. There may not be any practical advantage of using anything faster than SCRs.