Continue to Site

Welcome to our site!

Electro Tech is an online community (with over 170,000 members) who enjoy talking about and building electronic circuits, projects and gadgets. To participate you need to register. Registration is free. Click here to register now.

  • Welcome to our site! Electro Tech is an online community (with over 170,000 members) who enjoy talking about and building electronic circuits, projects and gadgets. To participate you need to register. Registration is free. Click here to register now.

Super fast transient linear regulator.

Status
Not open for further replies.
I wil be doing repeated measurements for averaging, but, only with my previous idea of a single current ramp. My main motive for this is power - once energy is put into the inductor on the rising ramp, how can I recover that whilst stll maintaining the same voltage over the inductor? As in, the ramp up and ramp down are symmetrical. With my original idea, I'll be doing say 8 tests for each current threshold, and averaging these, but each time, only charging the inductor, and allowing it to discharge back into a cap bank (which is used to provide ~5V to the linear regulator, so the power input doesn't have to provide more than say 200mA at a time to replenish the cap bank which will deplete due to losses). Allowing it to discharge this way, means that the voltage across it isn't fix, so the 'down ramp' would be steeper, and dependent on the cap banks voltage.

I do see what you're saying, and it does seem a much better way, one could time the entire sequence (basically making an oscillator, but with higher currents), starting from 0, then turning off the voltage once the inductor current reaches a threshold, then one times how long this takes to get back to 0, rinse, repeat etc.. If you can explain in further detail your idea, I would be very interested! My original idea has its issues and many sources of error, its a kind of brute force approach for getting saturation current.

I do like the idea of 'burst tests', as that is what I've been on about, not continuously using the inductor in an oscillator, but a fixed number of tests that are averaged, and then parameters changed, and it does another few tests, eventually giving a fairly good idea of whats happening.

I also like the idea of going through the BH curve for cores. However, I fear we're approaching the sort of test that would require an oscilloscope (to visualise) or heavy processing to extract such information from the test above. There are of course precision test gear for such things, but I was going to strip down the design to very basic, and buildable (using AVR or PIC) whilst still maintaining some accuracy.

I think I have got to the point where I'm going to have to lean less on simulations, and actually build something, get my scope out, and see what happens :D
 
Hi,

Well the basic concept is not too difficult or complex. It would mean switching a positive voltage to the inductor followed by a negative voltage to the inductor, as a square wave. The amplitudes and frequency under your control. However, the offset would also be under control, so that we can apply a positive voltage of say 3v and a negative voltage of say -2v, which means there is a net DC current flowing in the inductor. The net DC is used to test the ability of the inductor to handle DC current as in a buck circuit for example. With +3v and -2v that would mean a net offset voltage of 0.5v and with an inductor series resistance of 0.5 ohms that would mean 1 amp net DC current would be flowing through the inductor. That's the way they work in some converters anyway. In fact, the simplest way is to build a test set that is actually a buck converter, and control the DC current with the output load. This might not allow testing without any DC current though.

The voltages would be applied little by little, starting from 0v and working up and down by small increments until the test point is reached. In fact it may be a good idea to run symmetrical voltages to the choke first and reach some max, then slowly decrease to zero, then increase again for the test reaching up to the test voltages.

I never tried to automate this test procedure but im sure it can be done. Might take a little work.
 
Coulds this be done with a half-bridge (or full bridge) configuration? Perhaps with a current sense resistor in series with the inductor, or for measuring one ramp only, on the leg of one of the switches. I decided to fix the current and measure 'time' because that is much easier to do than measure a single point on a current ramp very quickly (would require a fast ADC and precise timing), but using comparators.. if I understand your idea correctly, could be used to control the switching of the square wave. If the energy built up in the inductor is then dumped back into the power supply cap (like in a half or full bridge) then I guess I could allow this to oscillate and measure the frequency.

I did consider using a buck (or boost) converter, and fixing some parameters whilst measuring others, but as I'm trying to save on power here, I don't' want to be wasting watts in a load resistor as heat, I was unsure how to do it and recover energy - after all the whole point of this was to be able to measure saturation current without a hefty PSU (or an oscilloscope!) but use capacitors, super or otherwise, to supply short high current bursts, then recover the energy. I just realized though, I guess I could use a buck converter to charge a capacitor - similar to how regenerative piezo actuator drivers work, but with inductance instead of capacitance.

I'm still not quite getting what this method is, soo, any chance of a quick diagram? :D sorry to be a hassle, but I'm open to idea's, and I'm sure something good can come out of this.

Back to my simple design, after many simulations it seems the MAX4172 *should* be fast enough, I was just putting it before the output cap, which meant it measured the sudden increase in current as the output cap's voltage fell, and the voltage regulation loop kicking in (<1us) which over loaded it so it took a few us to recover. Putting a higheside current monitor directly in series with the inductor solved this, and it appears now to be really rather accurate - at least as accurate as a low-side with low-offset opamp and the extra opamp required to measure the voltage across the load (because now, the inductor and switch are at ground) just with less components.

Still working on a way to switch out voltage regulation and switch in current regulation, without the output rising, effectively, shutting down the regulator, setting max current, then re-enabling. With only one feedback to the opamp (directly connected to the +V of the inductor) it seems a couple of analogue switches could be used to change the feedback path, and add some heavy compensation to slow it down for current. With only one opamp doing the regulation right now, I'll have one spare as I often by op amps in dual packages when prototyping.
 
Hi,

I supposed you can use an H bridge but then you'd have to use a differential measuring circuit of some type.
With a half bridge and dual supplies, ground would always be ground.

I'll try to draw up a circuit sometime today although i have to do some running around first which will take some time. Food supply, Christmas, etc. ha ha. The only thing i have enough of right now is beer.
 
@BT
I do have the service manual with complete schematics for an SMU. They are available in the paper manuals, but not the online manuals. Here **broken link removed** is a service manual, but you may have to register. On PDF page 89 there is a block diagram.

Mine, in essence stayed within a comfort zone of +-10 V, 100 mA. Voltage was measured with an IA and it was basically an output and was always measuring V. But, it's output could be fed back to the I-V converter biasing and force it's biasing system to bias at that value, so lead lenght could be ignored. Current limit was a function of the driver chosen and it was fixed. So, I had 10-12 V compliance and about 200 mA short circuit protection and the output was analog.

The zeroing stuff wasn't thought out too well, so I ended up with 20-30 pA of offset current and a few uV of offset voltage.
Only the AC accuracy was important, The actual DC value not so much. Part of the auto-zero required zero volts in, but setting a D/A to zero wasn't zero and my zero D/A had to be divided down. Screwdriver adjustments were not in the design either.

So, yea, precision clamping could have controled current and to be a good instrument, you need precision clamping for voltage too. Relays were mercury wetted because of the low currents and voltages.

When you look at the design of an Electrometer amp and zero check, you can see how the amp can amplify Vos (The OP Amp Offset) with the addition of a resistor.

The "output" of this device was two system DVM's. The input, an 8 bit output port and a 4 channel IEEE-488 D/A converter. 2 Channels were supposed to be used. One for the voltage bias and one for zero.

Modes were: 2T/4T, Zero Check, Voc ans Suppress
 
Hi again,

That 'manual' requires logging in to read?

Anyway, here is the drawing i mentioned. As you can see it is just a basic block diagram showing how the components would be connected together. The real work to be done is to devise a method that enables setting the required test parameters, and a method to drive the test set up to that test level. For example, if we want to test with a 10 amp DC current we would want to gradually increase the power supply voltages with one voltage higher than the other (one is negative and one is positive, but "higher" here means larger absolute amplitude of course).
Obviously the two power supplies have to be adjustable too, so that their outputs can be varied as needed. I suppose that would be easy enough to measure however. The timing would be set by the "CONTROL" circuit that also contains appropriate MOSFET driver IC chips not a home made driver or parallel CMOS gates or other cheats.

Granted this isnt the simplest thing to build but then again we want to do a good test on the inductor so it takes a lot more than a simple DVM and 9v battery to test <smile>.

If you wanted to try this out first you could build a knocked down version with smaller power supplies and just test lower current inductors. If that works to your liking then you could beef up the circuit. The control would be the same however, it has to be fast and we have to be able to vary the high going pulse and the low going pulse lengths as needed via some control like with a micro controller or something.
 
Hey KISS,

I'm thinking the precision involved in such an instrument might be more than I require - although I did of course mention <1% - but its impressive engineering!

MrAl,

Can't see a Schem/diagram on your post, but I think I'm starting to understand what you're getting atm but I'd certainly like to see the schem :D

As for my own one, I think I'm getting somewhere. Your mention of MOSFET drivers reminded me I just used a totempole for ease of use in simulation (didn't have any MOSFET driver libraries) and that of course was a source of delay, around ~700ns with the FET I was using. I updated it to a cheap microchip one TC1412 to reduce this down. This didn't affect the inductance measurement so much (when the test 'starts' doesn't really matter as I'm measuring the time difference between two current levels) but it affected the resistance measurement, as the regulator would go into voltage mode before current mode, so the output spiked (current in inductor shot up to 10A+ before settling down to the test current of ~200mA).

Now, along with a spreadsheet of 'predicted inductance' based on timing, voltage, and current peak, I'm getting virtually the same results. Of course so far this is all simulation, but I already have some surplus parts to play around with. Tis the season so I've no doubt this will fall by the wayside.

I've been carefully watching this thread for ideas: https://www.eevblog.com/forum/proje...-alternative-route-to-dump-the-excess-energy/

The half bridge approach seems great, although I would fixed one end of the inductor to a voltage, and the upper switch would allow discharge current to flow 'back in' to the power supply caps that supply the regulator. If loses are small theoretically I could let this run as an oscillator and measure the frequency - basically a boost, but one that sends its output to a cap bank.
 
Hi again,

Oh sorry i must have been in a hurry and didnt upload the schematic. I am uploading it to this post now.
Take a look, and you can see how simple it is.
If you want to modify that's up to you of course.
 

Attachments

  • HalfBridge-1b.gif
    HalfBridge-1b.gif
    13.5 KB · Views: 159
Ahh I see what you're getting at I think. yes a half bridge, I was thinking instead of a bipolar supply, the upper switches to +v, the lower to 0v, and the other side of the inductor is fixed on a low impedance supply (say... 2V), but the principle is still the same.

By varying the power supply, or rather (in your diagram) the difference between VCC, GND, and VEE, we will get varying slopes of current through the inductor, but what are we measuring here? If we fixed the voltages and timing pulses for the switches, that only leaves current to be measured. Perhaps with all voltages fixed, if we use current mode control (turn off top switch when current in inductor reaches a threshold) we could measure the frequency of oscillation? That really will be like making a buck, plugging in an unknown inductor, and letting it run. By increasing the peak current we would see it go into saturation, and the frequency should increase non-linearly. Is this what you're thinking? or have I missed something?

Edit:
I just read your previous post about allowing a net DC current to flow through the inductor. Its DC current rises as we vary voltages. So its a buck converter working in continuous, where the current never falls to 0, we increase average current until it saturates by which time its inductance would have fallen and the timing changes? If I can find a way to recover the energy, or rather have a load that can recover the power and put it back in (not a perpetual motion machine :) ) then I should only have to provide power for losses, drivers, and opamps?

The key things I'm after is, not necessarily measuring the inducance, but rather saturation - so one needs to have a baseline inductance to measure against (to see it drop as it goes into saturation). Also, as noted before I'm attempting to avoid requiring hefty PSU's. Free-running converters require a load, where-as I was going to do lots of single-shot measurements (to average) each time letting the inductor discharge back into the power supply cap (before the voltage regulator), so I can have high peak currents, up to 10A, without having to supply more than 100-200mA, because the cap gets recharged inbetween burst measurements.
 
Last edited:
Hi again,

Well it seems to me that if you want to test for saturation with a DC current (makes sense to do so) then the power required to do that will be lost no matter how you do it unless you use bootstrapping. The reason i say this is because the DC has to be constant and so it wont be able to supply any power back into the power supply line.

So one way to do it would be to build a variable buck circuit with normal output cap and some resistive load, and then another converter to convert the output back to the power supply level. So it would be a buck circuit powered from the original power supply followed by a boost circuit powered by the buck circuit and the boost circuit would in turn power the original power supply to return some of the energy taken from the output of the buck. Might be a little elaborate though as the boost circuit has to be able to handle the power taken from the buck circuit. There would still be some losses of course due to the inefficiencies of both circuits.
 
Status
Not open for further replies.

Latest threads

New Articles From Microcontroller Tips

Back
Top