High current boost-buck circuit

[arclight]

Hi all,

I got brought in to work out the digital side of this project, and now that its working, they're asking me to help with the power electronics bit. Here is some background:

This project is part of a grant to test the durability of supercaps for a hybrid power project. They want to run a few thousand 1-minute charge/discharge cycles and measure how hot the capacitors do as well as testing them for capacity afterwards. Originally, they were going to just charge them up from a power supply and then dump them through a load bank. To save power, they got the idea of having two banks of supercaps, and just moving power from one to the other using a boost-buck circuit and a large IGBT. The dummy load would only get used to discharge the banks completely when needed.

Here's where we're at: They have a test program in Labview that is successfully generating 5Khz square waves at varying duty cycles to drive the two IGBT channels via a high-current buffer chip (74HC04) and the Powerex BG2B driver board. This works just fine - the TTL signals from labview are driving the 15V gate voltage on the IGBTs, they are moving 100A+ with good heat sinking, and the driver board prevents both from firing at once.

Now the problem: I don't think they sized their inductor correctly. There is a 500F, 24V string of supercaps on each side, and a ~720uH inductor wound on a large ferrite torroid core between them. Schematic is attached.



I went ahead and ran some tests in buck mode. At 5Khz, 50% duty cycle, I was able to run a 12V, 20 power source at about 20A into one of the capacitor banks, with the 1.5Ohm dummy load attached. It got an average voltage of 20-26V, with 50-120V peak pulses.



The pulses are 5Khz apart and don't seem to overlap much and I'm not sure that high-voltage spike is OK.



The target for this is to be able to handle up to 200A peak current, 0-24V, <=7% current ripple and be able to boost or buck between the two capacitor strings.

Can anyone sanity check the inductor calculations? The guy who designed this used a calculator from the ferrite vendor:

http://www.micrometals.com/software_index.html


I'm assuming that saturation is also a possible problem here.


Thanks!


John

[Odysseus]

Generally speaking, If you want to boost and buck in either direction you would need an H-bridge configuration.  The way it's shown now, you can boost from left to right and buck from right to left, but not the other two ways.

At 5kHz, boosting from 12V to 24V the inductor will experience a peak-peak ripple current of 167mA 1.67A.  (V=L*dI/dt => dI = V*dt/L = 12V*100us/720uH)

What were you probing when you observed the ringing on the scope?

[Marco]

All you've given is the inductance, that calculator is for calculating the required core and windings ... we can't double check that because you haven't said what core and windings you're using.

You've not given enough information to determine if the inductance is correct either, it's going to depend on allowable ripple at minimum current ... we have the ripple but not the minimum current.

PS. Odysseus, shouldn't that be 100us? (and 1.67A)

[Odysseus]

PS. Odysseus, shouldn't that be 100us? (and 1.67A)

Oops, it should.

[megajocke]

I went ahead and ran some tests in buck mode. At 5Khz, 50% duty cycle, I was able to run a 12V, 20 power source at about 20A into one of the capacitor banks, with the 1.5Ohm dummy load attached. It got an average voltage of 20-26V, with 50-120V peak pulses.

Whoa, slow down. You are suddenly talking about "the" dummy load without having introduced it before. We can't read your mind. 

Which string are you probing?
Which string is your load connected to?
Which string is your power source connected to?

The schematic you showed seems to be missing local energy storage across the half-bridge positive and negative terminals. The higher-voltage side has high current ripple and the supercap string probably has too much inductance to give reasonably performance.

As others already calculated the current ripple on the low-voltage side should be 1.67 A pk-pk maximum if the high-voltage side of the converter is at 24 V. That seems like great overkill if the aim was for 7 % of 200 A as current ripple. A 200 A 720 µH inductor is going to be enormous.

IGBTs aren't really very good in this voltage range either, but sure it will work if you can tolerate the conduction losses.

[SeanB]

At that current it is likely to be air cored and made out of copper bar, or at least 0AWG wire. Probably would be best wound as a toroid to keep the stray magnetic field down so you do not cook the housing it is in or anybody close by just from induced current.

[c4757p]

Is it really worth this much effort to save the power? It's not like you're building this into a product, it's just a test rig, right? I say make it double as a toaster!

[arclight]

So I checked it out. The inductor cores are the T650-52 from Micrometals:

http://www.micrometals.com/parts_index.html

Each core has two 36 turn windings, made of 10 turns of #16 magnet wire, wrapped up to be sort of like one #1 wire.

There are two of these in series.

I was measuring the ringing at the terminals of the supercap string on the output of the boost circuit.


Arclight

[Marco]

That's strange, if I throw a PFC boost inductor of 300uH, 200 ampere, 6 to 12 volt at that calculator it tells me it needs 2 stacked T650-8s with 38 turns of 10x#9 wires ... oh and it will dissipate 80 Watts from copper losses.

The turns match closely enough (2 stacked cores have double the core area so need half the windings) but I don't see how the design you have can be good for 200 Ampere. If I go down to 100 amperes the calculator gives solutions for a non-stacked core, but it is still recommending FAR more copper (75*10*#12) than you used.

Any way, the system you want can work in theory (with the right inductor, which I don't think you have at the moment) but I don't see the reason to experiment at this point. "Lets throw a 5 KHz square wave at this" is not really a useful experiment ... it's not really behaving like it should, but trying to find out why not will only tell you things you don't need to know.

Design a real controller, with a current shunt and a PWM IC which really does what you want it to do (at least in the buck direction) and then start experimenting. You can throw that together on a breadboard at these frequencies.

PS. you should of course use a current limited power supply (and one able to supply 200 Ampere, the storage bank won't be able to deliver that on the first few cycles until it's voltage gets high enough after the boost cycles) and put a (big) diode between the power supply and the storage bank.

PPS. you probably want some kind of overvoltage protection on the DUT as well, just in case your circuit doesn't operate correctly ... 6x3000F of venting capacitors is going to make a mess (probably not going to explode at these voltages, but still a mess).

[arclight]

Marco,

I appreciate the advice.  This is not my area of expertise, and I'm pretty sure the guy who spec'd that out does not have the expertise either.

Here is a picture of that inductor, FYI:



He's been testing either one or two of these in series. For the current, I think the 200A was meant to be peak and not sustained. Now, how long that "peak current" could heat up the copper is something to find out.

On the ringing, I went ahead and put some 250V, 0.1uF caps across the power terminals on the IGBT, and got that down.  I've got a working hall effect current sensor on there, and his labview program is supposed to be able to track current and adjust the duty cycle.

I'd much rather have a dedicated controller chip from LT or similar running this, but it isn't my project. I might still breadboard that up anyway.


Arclight

[peufeu]

The left cap bank voltage never goes above the right bank voltage, so you don't need a real 4 switch buck-boost, simply a bidirectional buck (it bucks in one direction and boosts in the other).

Here's how I'd do this : build a cheap 10 amp bidirectional current source, and put 10 of them in parallel.

The current source is simple. I have used this to drive LEDs and charge/discharge LiIons and also supercapacitors (much smaller than yours). You got a pair of MOSFETs, a driver (the likes of ADP3120) and a fast comparator with hysteresis whose common mode input includes ground. And of course shielded inductor the size of your thumb.

The comparator is set to control the current, centering it around the target value plus or minus hysteresis. When the inductor current drops below the lower threshold, it turns the top MOS ON. When the current crosses the top threshold, it turns the top MOS OFF and the bottom MOS ON. To change the direction of current or its value, the DC offset added to the comparator input via a few resistors is controlled.

That's it, one channel will set you back about 10 bucks, build 10 and you're done.

Pros :

- divide and conquer approach spreads dissipation, I2R losses etc, so it becomes manageable
- 10 amp module can be built and tested without risk or expensive gear
- blown parts cost $1, not like an IGBT brick
- hysteretic buck is foolproof since it regulates the current
- no risk of inductor saturation, overcurrent, etc
- it is a current source, doesn't care about short circuits, can be paralleled etc
- no need for precision on anything
- frequency self-adjusts, minimizing switching losses ; frequency is easy to calculate depending on in/out voltages

Cons :

- You need to check that the MOS driver can startup with a non-zero output voltage, ADP3120 cannot, some of its cousins from IRF or OnSemi (NCVxxx) can.

That's about it... it just works...