current limiting for power supply

[akis]

Originally Posted by macboy

constant at 100 mA? Current limiting is a dynamic linear process just like voltage regulation. Hysteresis is neither needed nor desired in a current limiter. Oscillations or instability of the control loop are no more (or less) of an issue than in any voltage regulator.

OK this is how it (does not) work in my case:

A sense resistor raises approx 150mV at the desired current limit level.
A current monitor chip (INA170) senses that voltage and produces a ground reference voltage of 1.20V at its output pin.
This voltage is tied to the positive input of an op-amp.
The negative input is attached to a very stable 1.20V.
When the positive input reaches and exceeds 1.20V, which is what happens when we reach and exceed the current limit level, the op-amp swings wildely to positive rail. The op-amp is OPA340 which works off single supply and can swing from ground to rail.
This voltage (supply rail) is then fed to the base of an NPN transistor (the current limit action) which shorts the voltage reference which drives the voltage regulator transistors.
This has the effect of bringing the output voltrage to almost ground, and as the output voltage falls the error condition (cuurent limit breached) is removed!
At that point the op-amp abrupty and suddenly swings to ground making the NPN limit action transistor go open collector, which allows the voltage reference to drive the voltage regulators to full voltage.
The output rises until it breaches the current limit, and the cycle repeats.

[megajocke]

Such a design has way too many poles at relatively low frequency to be stable without some kind of compensation. It should be doable though. Could you post your schematic?

Otherwise, a common trick that is often used is to use two opamps, one for current regulation and one for voltage regulation. The outputs of these then go to diodes so that the drive to the pass transistor is determined by the amplifier currently demanding the lowest drive.

To be able to use standard opamps the positive output (if using NPN/NMOS pass transistors) is often made the ground reference of the control circuit which has its own supply of say +-12V. Here is an example from HP/Agilent, also including a 4-level transformer tap changer using thyristors:

http://cp.literature.agilent.com/lit.../5959-5310.pdf

[akis]

Here is the circuit follwing Andrew's advice to put a constant current source before the voltage regulator.

The voltage reference is derived from a separate supply (not sure if I also need a separate winding), and I divide the 30 Volts into 3 equal ranges so then I can use a rotary switch to select 0-10, 10-20 and 20-30. Via this rotary switch a potentiometer then swipes from 0 to 10 volts, which is added on the base voltage.

This very reliable voltage reference (excluding such exotic considerations as choise of zener and temperature coeffiicient) is then used to drive the base of the voltage regulator transistors (the main transistor needs to be a beefier one to withstand at least 90 Watts, but the simulation software does not have a model for it).

The PNP constant current source limits the current to a predefined level determined by R24, R25 and the two diodes at the base of Q6. This has been put together in a hurry following Andrew's suggestion and I have not yet experimented with the correct way to wire it up.

Outside of your general comments, my question would be what is the correct voltage to drop on R24 ?

[akis]

This is the current limiter based on a sense resistor. The output drives an NPN transistor that shorts the voltage reference from the previous diagram (base of Q5, and an addition of a series resistor past the voltage control pot to allows to drop the voltage across) thereby driving the output voltage down.

[cliffforrest]

Originally Posted by akis

This is the current limiter based on a sense resistor. The output drives an NPN transistgor that shorts the voltage reference from the previous diagram thereby drivign the output voltage down.

Can you put an extra switch bank for the op-amp input? Otherwise the contact resistance is going to dominate on the higher current ranges.

[AndrewT]

Originally Posted by macboy

It's quite obvious that you've never actually designed a current limiter. Your thought experiment above is interesting but impractical.

yes, my thought experiment.
I have assembled many current limiters but none using a comparator.
Maybe my thoughts are worthless.

[akis]

Originally Posted by cliffforrest

Can you put an extra switch bank for the op-amp input? Otherwise the contact resistance is going to dominate on the higher current ranges.

Do not understand that

[akis]

Originally Posted by AndrewT

yes, my thought experiment.
I have assembled many current limiters but none using a comparator.
Maybe my thoughts are worthless.

As I see it, the comparator has not got a linear range between "on" and "off", at the point of the theshold it will abruptly go on or off. But to maintain the current at a limit it would need to float gently somewhere in between "on" and "off" depending on the level of the excess current it would be trying to cover for.

What we need is something quite fast, but not as fast as a comparator.

I used the current sense monitor (INA170) with a factor of approximately 10 to 1 at its output (100mV sense in, 1000mV ground referenced voltage out). I fed it to the OPA340 op-amp set a voltage gain between 10 and 22 (currently 22). The output of the op-amp then drives the current limiting transistor and it seems to work well, quite tight around the current limit and no oscillations even when you short the output.

Now it is a matter of tieing it all together and to decide whether the constant current source with its simplicity (1 transistor) is better than the current sense solution which requires 2 chips, 1 transistor and a very accurate voltage reference.

[cliffforrest]

Originally Posted by akis

Do not understand that

Only that you have a switch contact in series with a 50mR resistor, and are sensing the voltage across both in series.

A switch contact can easily have 100mR of contact resitance, and be very noisy and erratic in the bargain.

[akis]

Originally Posted by cliffforrest

Only that you have a switch contact in series with a 50mR resistor, and are sensing the voltage across both in series.

A switch contact can easily have 100mR of contact resitance, and be very noisy and erratic in the bargain.

Oh yes how true. Hmm I need to think more about this.

Edit: ok so a "standard" 2 x 6 switch will do, as you originally suggested.