3-5V High Current 150A Power Supply

[FvM]

If you are going to design a transformer current source, which is surely possible, you need to calculate load impedance, connection wire resistance/inductance and transformer properties exactly instead of guessing about it.

For the intended low frequency, an iron core transformer is surely better suited. You may even use an industry standard current transformer with sufficient VA rating, reverse connected.

In case of a PWM source driving the transformer, you should have sufficient filtering after the H-bridge.

I don't understand your leakage inductance considerations, would you mind to sketch an equivalent circuit?

 

[BradtheRad]

My soldering gun contains a step-down transformer. I plug it into 120VAC. It creates 0.1 VAC at sufficient amperes to heat a thick copper wire red hot.

I would not be surprised if it's over 100A. I don't have an easy way to measure it.

 

[Warpspeed]

While the multiple wire thing is OK, let us say I want to generate 100Amps RMS.

I like the short circuiting the secondary of a transformer and using it as a reverese CT thing but please help me with the math a bit.

Aim: To get 100A rms pure sinusoidal current through a conductor.

Ferrite core transformer: 100 turns primary and 1 turn secondary. So 12Volts applied to the primary will give 0.12V or 120mV output.

We must provide a very small impedance to the output (basically close to a short circuit) such that it draws 100A when provided with 120mV. i.e. 1.2mohm impedance. Most probably even a short circuit will not allow such a small impedance and so Primary voltage will have to be increased till we get 100A output.

Problem: As the primary will be excited by SPWM, to get a sinusoidal current at the output, the equivalent impedance of the secondary, primary, leakage reactances etc must have a dominating inductive component to filter out the switching frequency which I'm planning to have around 10-15kHz at the moment.

Do you guys thing this will naturally be the case? If not will providing small air gap which would increase leakage reactances X1 and X2' make the equivalent inductive without leading to other issues?

The SPWM will be bipolar (half bridge or push pull) so saturation should not be an issue.

This current transformer only needs to work at the power line frequency, so a tape wound silicon steel toroidal mains transformer core would be a much better choice than ferrite. And you could use the existing 220v primary.

This is a current transformer, so all the nasties such as copper winding resistance and leakage inductance all appear in series, and can be discounted from the actual current transforming process.
All these losses just increase the impedance at the primary, and require an increased driving voltage and some excess driving power.
The basic current multiplication function (due to the turns ratio) will not be effected.

If its driven from a true current source, all the transformer nasties can be ignored.

If I were doing this myself, I would prefer to scan a sinewave stored in EPROM fed through a DAC.
If the DAC is supplied from a regulated 5v source, the sine amplitude will remain regulated too. A simple low pass filter, even just a shunt capacitor will readily smooth out the very small steps into a nice clean sine wave voltage.

An audio power amplifier (using current feedback) would then drive the primary of the current transformer.

A relatively small toroidal mains transformer with 220v primary might have in the region of 1,000 turns as a wild guess, the smaller it is, the more turns it will have.
The power amplifier may only need to supply 100mA for 100A output if that were the case.
Its not difficult to work out the number of turns on an unknown transformer.

The overall power requirement is not high, and your idea of a portable battery powered 100A current calibrator instrument would be quite practical.

 

[chinuhark]

The overall power requirement is not high, and your idea of a portable battery powered 100A current calibrator instrument would be quite practical.

First of all Thank You for this. I needed that as I was afraid people were going to call the idea unfeasible and silly. I just had it when I was actually standing in the middle of nowhere (sort of;-)) and knew my visit was a waste as I couldn't even measure the currents there. Even if we have such a device and check meters before leaving the office, it should make sense.

Just woke up to 4 replies (India).

@FvM: What you are suggesting is PWM Power stage--->Filter--->Off the shelf CT--->LEM sensor feedback back to to the power stage.

I'm thinking something similar, just hoping that maybe I could skip the filter if the self/leakage inductances of the transformer are large enough. (A hope/guess simply, and best way to find out will be to actually try it on a scope, which would take lesser time then going into the math).

Since I want to skip the filter and PWM output will directly be applied to it, hence the ferrite core Xmer.

https://blog.oureducation.in/wp-content/uploads/2013/05/hh.png

Here's a Xmer equivalent circuit. The equivalent Inductive reactance of both the coils and as seen by the source is X'eq as shown.

I did some math:

For 125mV open loop secondary voltage, about 0.5ohms of inductive reactance (@10kHz) would draw about 0.25A , i.e. for L = ~8uH
This is about 2*pi*50*8uH or 0.0025ohms @50Hz so I = 50A.

Worst case, I put a small 20 turn 8uH air core inductance at the output.

So even if the leakage reactance (+external if required) is greater than or equal to 8uH, the current at PWM frequency is 250mA and that at power frequency is 50A.

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

Does any of the above math make any sense whatsoever or am I making some fundamental silly mistake in all this. I have no idea how sine LC filters are designed, and there is no C here btw as we are only interested in current but will the current get filtered by above logic?

 

[Easy peasy]

The only problem with a battery powered approach is power, say you want to generate 50A rms, if the total of all the R in your ckt (mosfets, ALL wiring, DC choke res, Cap losses (ESR)..etc) is 10 m-ohm then you are looking at 25 watts from your battery.
A more likely figure is 50 m-ohm giving 125 watt.
So it is critical to select your max test current and then see if you can design to a minimum total R to meet a decent amount of battery time...
(and add in likely switching losses too if you are PWm modulating the o/p, and any connector losses and the output cable R)

 

[Warpspeed]

An off the shelf CT may be problematic, there are two potential problems that I can see.

All standard off the shelf mains type CTs are designed for a very limited burden voltage, usually just enough to operate a meter movement. The ones I have tested here often saturate the core at typically one to two volts peak which is fine for many 50/60 Hz metering applications.

If you plan to turn it around and use it as a high current source, it may fall on its face with insufficient output voltage. Its just not big enough...

Another problem is the core material may either be of high permeability ferrite if its a high frequency switch mode type CT, it may be very unhappy with the very high volt microseconds at 50/60 Hz.

If its a traditional 50/60 Hz mains monitoring type CT, it will probably be a wound steel tape type core which may have very high eddy current losses when you try to drive it directly with high frequency and high voltage PWM.

I still think a mains transformer of possibly about 25 Va size and a 220 volt primary might be your best bet.
That size primary will continuously carry 114 mA, so the wire size will be o/k.
The number of primary turns might turn out to be fairly reasonable too, although you will need to test that.

As it will be operating well below 220v the flux in the core will be absolutely minimal, and you can drive it with as many volts as you like without any fears of core saturation.

At 17 Khz or whatever switching frequency you decide on, the flux swing will be microscopic and both eddy current loss and magnetising current will be very low.

The power transfer efficiency will be extremely high, the only down side is the physical size and weight, although a 25vA mains transformer is probably not that large.

 

[FvM]

An off the shelf CT may be problematic...

It seems to me that you are disproving a number of points that won't be actually considered for a serious design.

I was talking about a CT with sufficient VA rating for a reason. 50 Hz current transformers have clear specifications, you can well calculate the saturation margin in this application. Obviously I'm not suggesting a high frequency CT.

Similarly, I won't load the transformer with a large PWM current. A small ripple won't hurt however.

It's a different question if low voltage, possibly battery powering of the tester makes much sense. Provided you want/need it for some reason, you should know how to.

 

[Warpspeed]

I was talking about a CT with sufficient VA rating for a reason. 50 Hz current transformers have clear specifications, you can well calculate the saturation margin in this application. Obviously I'm not suggesting a high frequency CT.

That's all very true.
Most common current monitoring transformers are designed to operate a rectifier plus moving coil or moving iron meter or even a digital readout, and are certainly not usually designed to transmit much power.

The winding will obviously be up to the current, but its the voltage and core flux limit that needs some care.

The original poster mentioned using a ferrite core, which we both agree will fall far short at 50 Hz.

I am sure a suitable commercial CT could be nailed down, but as I believe this is to be a one off design project, a possibly free junk box toroidal power transformer might do quite nicely.

 

[chinuhark]

OK I realized yesterday that a ferrite core is pretty much useless as I kept thinking about the 20kHz carrier and forgot the main part, the 50Hz modulated signal, the very reason all of this is being done. (I do forget such obvious things sometimes:lol

OK so it's either a 100/5 CT or a 25VA toroidal Xmer with a welding cable through either. This part is now final.

What I'm worried of now is the filtering. Do I need to filter both the current and voltage or just the current. I will obviously have to add a series inductor for current filtering, but I'm afraid how the control loop will react if I add the C to complete the LC filter. The filter's overall phase lag and the sluggishness it will introduce to the system resonse, surely can't be good for the control system.

Would the PWM voltage fed to the CT primary cause heating issues? The current is going to be sinusoidal but I think the voltage would be PWM. Will this cause issues? Didn't older sinewave inverters just directly feed the SPWM output to step up transformers and use the transformer leakage reactance for filtering? Or am I mistaken?

 

[Easy peasy]

Any L in the system will cause a filtering effect on the current and hence voltage too, if you want accurate o/p current control you have no choice but to have a CT (or shunt) on the actual o/p i.e. to measure the 100A rms, sure filtering (on the HV side of the mains Tx - as easier) will cause delays etc but given you are interested in say 50Hz output, a control loop can tolerate these filtering delays (to remove the ~20kHz).

 

[FvM]

The amount of filtering has to be chosen according to your ripple current specification, might be either simple L or a LC filter. The achieveable control loop bandwidth will be of course affected by the output filter, the ability to suppress harmonics can be reduced. All-in-all it's a compromise. Precise regulation of 50 Hz output magnitude respectively RMS current should be possible.

 

[Warpspeed]

One sneaky way to overcome the feedback phase shift issue would be to run it as a class D amplifier, where the switching waveform itself provides voltage feedback before any subsequent low pass filtering.

As only a very few watts of ac power are involved, I cannot see any real advantage in using a switching power amplifier as opposed to using a common off the shelf modular audio power amplifier with direct current feedback around it.

This could drive your output transformer directly from a low distortion sine wave current source. It would be a lot simpler and have far fewer potential problems.

If the output transformer is to act as a true current transformer, it MUST be driven from a current source, and that would be a lot easier to do with direct current feedback around a linear amplifier.

 

[Easy peasy]

We have used high power class AB amps to run into 50Hz transformers to generate various outputs for test gear with good success (at all sorts of frequencies) - and its easily done - these days you can get a 150W class D amp board add +/- 35V and you have an amp without the big heatsink, easy to drive and to set up a current feedback system to control the o/p current of your step down Tx to 100A rms... or 100A average, sine wave...

 

[Warpspeed]

The first step might well be to start with testing the proposed current transformer.

Hook it up to a variac and measure the required peak to peak drive voltage (and current) for 100 A RMS circulating current in the output coupling link.

If the transformer has low losses, as it should have, you may be surprised how little actual drive power is required. You can then work backwards and spec your amplifier and dc power supply rails to suit.

Its real red neck engineering at its finest, but its entirely practical, especially if you can recycle an existing small junk toroidal mains transformer..

 

[Easy peasy]

A 1000:1 CT with a 5VA allowable burden (5 ohm) would give you 100A rms in one pass thru the middle with 100mA in the 1000 turns, also it would work down to 10Hz.

However if you do something transient to the "output" current, i.e. the 100A like breaking it suddenly, then the reflected volts back to the driven side (1000 T) will be in the kilo -volts region and may upset your driving amp without some big TVS (>30V say) across the 1000T wdg.

 

[Warpspeed]

It could be simply clamped to the two dc supply rails with diodes.
This is good design practice anyway, and many modular amplifiers already have such diode protection.

If there are enough microfarads of energy storage on both supply rails, the resulting voltage bounce need not be destructive.

Its a case of weighing the max possible mJ of energy storage in the transformer core, to the total mJ energy storage capacity in the decoupling capacitors. It's not a difficult situation to anticipate.

There is no real need to be able to disconnect the output current loop anyway, as we are testing clamp meters. The two ends can be permanently crimped very solidly together.

 

[BradtheRad]

As long as we're brainstorming ideas here, a battery might suit the purpose. Two large lead plates in acid. Similar makeup as a car battery (except this is one cell, 1 or 2V). It is messy and hard to calibrate. I suspect you'd need to experiment with plate immersion, acid concentration, etc.

Nevertheless such a battery may be a substitute for a 50 Farad capacitor (since you're probably thinking of a large smoothing capacitor for your power supply).

I have read that large 2V batteries (rather, single cells) are used in forklifts, telephone exchanges, submarines. They may be a second-hand market for the power source you're looking for.

 

[Warpspeed]

Let's do some wild off the wall guessing with numbers.

Assume the secondary link uses 25mm^2 wire (to carry 100 amps)
And its only 300mm long.
The resistance of that will be close to 0.2 milliohms.
If the transformer has a ratio of 1,000:1 that would be 0.2 ohms when reflected back into the primary.

Now I grabbed a 240v 30VA EI core transformer and measured a 66 ohm primary.
The additional reflected secondary impedance into that is so low it can be ignored.

So our 1,000:1 transformer requires 100mA primary current flowing through maybe about 66 ohms, plus leakage inductance which we can probably ignore too.

So a wild off the wall guess might be about 6.6 volts RMS to drive the thing and about 660 mW of audio power.

We can probably ignore magnetising current too.
A 240v transformer driven with 6.6v is not going to draw much zero load current !

Probably +/- 12v dc rails at perhaps 50mA average does not look too tough to get straight from a couple of batteries.... for this example.

If it has more than 1,000 primary turns, (which it probably does) the drive voltage may need to be a lot higher, but also requiring less drive current.

A larger transformer will have fewer primary turns, and much less winding resistance and so be even more efficient.


So its a case of testing a few odd transformers of different VA ratings to find one that gives us nice suitable convenient supply rail voltages.

As long as the wire in the primary can safely carry sufficient current, that is probably the only real limit to minimum usable size.

 

[Easy peasy]

impedance is transformed by turns ratio squared....

- - - Updated - - -

100A at 200u ohm is only 2W so yes a small Tx is feasible,

 

[Warpspeed]

impedance is transformed by turns ratio squared....

Oops, you are absolutely right.
(Warpspeed slaps himself in the head)

Reflected impedance back into the primary is about 200 ohms for this example.
Thicker secondary link wire could easily halve that, 50mm^2 welding cable perhaps ?

Total impedance now maybe 266 ohms for this example.
At 100mA that is 26.6 volts rms, and 2.66 watts.

The power is still nice and low, but the drive voltage would be better if lower.
A larger more meaty transformer with fewer and thicker primary turns would solve that, along with fatter secondary wire.

A bit of quick and dirty bench testing with a variac will quickly resolve all of this.