How To: High Power Boost Converter - an ongoing journey. / High Voltage / Forums

Forums

4hv.org :: Forums :: High Voltage

« Previous topic | Next topic »

How To: High Power Boost Converter - an ongoing journey.

Move Thread

LAN_403

Sigurthr

Wed Feb 18 2015, 03:45AM

Registered Member #4463 Joined: Wed Apr 18 2012, 08:08AM
Location: MI's Upper Peninsula
Posts: 597

Hello everyone!

I recently built my first high voltage boost converter, and if you saw my other thread, I did so using a schmitt trigger comparator instead of a uC like I had done in the past. This is the most efficient converter yet, and also the only one to operate at over 12V I/O. I've measured efficiency at around 75%. The previous ones I had made used arduino for feedback and drive, and were all horribly inefficient, but were all very low power, less than 5W, and only output 12V. This one takes in 10-24Vdc and outputs as high as 450V, but I usually top off operation at around 400V because I want to leave headroom for my 500V FET. This one also operates at much more power, over 30 Watts output!

Well, I'd like to scale up to around 200 watts output at 400V. What's stopping me? Well, I don't think I can find a core large enough given my current topology / implementation! Inductor peak current is REALLY high....

I've tried a few inductors to date on this ~30W version:

33uH (+-20%):
1) an 8A peak through hole inductor - was first I tried, worked well, but only about 65% efficient and the core got HOT (140C) at only ~18W output. Maximum voltage also wouldn't rise over about 250V with an appreciable (2Kohm) load even at max input voltage.
2) an air-cored inductor I made with 22ga wire; well, the heating problem was gone! Haha, and efficiency went up to around 70%, but it was VERY bulky, had a relatively large DC resistance (~100mOhm), and voltage would only rise up to about to about 325V on the same load and supply as above.
3) an excessively large gapped core made from an old flyback core; this is working the best so far, 75% efficiency, and I can reach the entire range of output on the same 2Kohm load and 24V supply (actually I can hit 400V as low as 19V in). It stays cool (35C) too. I tried removing the gap in the core but inductance skyrocketed, current dropped, the load severely weighed down output voltage, and the FET started to run hot for the first time. I put the gap back in.

~18uH:
I wound a smaller air-cored inductor, and figured I could just raise the operating frequency to compensate, but current draw was higher and output current/voltage was lower, there was also increased FET heating.

~15uH:
I wound a smaller coil on the gapped flyback core listed above, hoping I could run at a higher frequency to compensate the lower inductance while keeping similar I/O ratios. It increased FET dissipation, chewed up efficiency, and still output less power than the 33uH winding.

Here's the rundown of the circuit I'm using:
It's a standard boost converter power section, the controller is a schmitt trigger comparator which monitors output voltage via a voltage divider and gates the output of a 50% duty wide range VCO via the enable pin. I.e. essentially pulse skipping. Feedback works great and regulation seems solid with less than 500mV ripple, which is all I need for my application. The VCO is set to run at 23KHz, as low as it will go stably on the maximum range divider. I don't want to run lower either as I can hear 22KHz and below. In practice I have a pot in place of the Vset voltage divider shown in the schematic, so I can vary frequency from 23KHz to 1MHz linearly while keeping 50% duty cycle and tune within 1KHz of a target.

see schematic here:

Here's some scope shots of it in operation, set to 250V output on the 2Kohm load:
WT3UAm9

I've tried searching boost converter design information, calculators, and application notes, but all I can find are converters designed to operate in constant conduction mode (which I'm fairly sure mine is NOT), with monolithic IC solutions, or with output control via PWM and not FM/PRF or oscillator gating like I can and am using, respectively.

Clearly the feedback works well, but I think the discontinuous nature might be saturating my inductors? This big flyback core doesn't appear to be saturating on the 2Kohm load I am using, but something is clearly preventing me from scaling up, and I know that I shouldn't need such a HUGE core for ~30Watts!

Goals:
Allow for a higher operation frequency.
Reduce inductor core size.
Increase output power > fivefold.
Increase efficiency if possible, even if I retain 75% efficiency, at 200W, I'd rather not have to dissipate 50W for no reason.

I'm open to all suggestions or tips.

Thanks!

EDIT: If anyone was wondering the current trace was inverted, d'oh!

Inducktion

Thu Feb 19 2015, 06:50PM

Registered Member #3637 Joined: Fri Jan 21 2011, 11:07PM
Location: Buffalo, NY
Posts: 1068

I want to say that with boost converters the gap in the core is a huge recommendation. I can't remember the exact reason for it but it's the same behind why with flyback topology converters you're supposed to use a gap. I want to say it somehow reduces peak currents but I don't want to be giving bad information.

Which leads me to ask, why not just go with a flyback converter? You'll get isolation between the logic and the high voltage side and you'll probably be able to scale it up much much easier.

Or why a boost converter in the first place? You can use a plethora of different topologies, and each have their own advantages and disadvantages; I think a boost converter's advantages are really just ease of use and small part count. Push pull, forward, the aforementioned flyback, etc might give you better results.

What also might be dragging your converter down is the LACK of inductance; My personal rule of thumb is 50 - 100 khz asks for 200 - 100 uH respectively. With higher frequencies the smaller the inductor you need, so maybe try fooling around with inductance values and frequency?

Also what kind of gate drive are you using?

Sigurthr

Fri Feb 20 2015, 01:47AM

Registered Member #4463 Joined: Wed Apr 18 2012, 08:08AM
Location: MI's Upper Peninsula
Posts: 597

Details of drive are all in the above schematic I posted. It's just a cascaded common emitter BJT amplifier.

Reason for boost converter? Application is a 3S LiPo powered capacitor charger to be used as a one-shot "burst mode" power source for portable SSTC. Even with the 200W output it would take 13.2sec to charge the bus capacitance, giving a fire rate of only 4.5 bangs per minute. That already is asking ~16A from the LiPo, which is reasonable given it is only ~6.66C and most LiPos tolerate 10C at least.

[hit enter too soon]

Using my current boost converter it takes about 90seconds to charge the bus capacitance, which is far too long. I can live with > 2 bangs per minute.

Also, I tried increasing inductance while keeping the frequency the same and increasing frequency but keeping inductance the same, it all works out equally; the current draw, output voltage, output power, and efficiency all drop to next to nothing. The series inductive impedance winds up limiting the power level in this mode of drive.

I'm open to using an off the shelf controller IC, I just don't know any good ones for this kind of application. Of course going discrete is always a bonus to me, but sometimes it isn't worth the trade off.

Patrick

Fri Feb 20 2015, 06:58AM

Registered Member #2431 Joined: Tue Oct 13 2009, 09:47PM
Location: Chico, CA. USA
Posts: 5639

Just in case you haven't done so yet, id recommend looking over the TI / Unitrode documents. they've explained alot to myself and met my needs every time.

Shrad

Fri Feb 20 2015, 08:24AM

Registered Member #3215 Joined: Sun Sept 19 2010, 08:42PM
Location:
Posts: 780

there are also good topics at

plenty of appnotes actually

Sigurthr

Sat Feb 21 2015, 03:03AM

Registered Member #4463 Joined: Wed Apr 18 2012, 08:08AM
Location: MI's Upper Peninsula
Posts: 597

Thanks for the links guys! Tons of info there, and I'd never heard of either of them prior.

I spent the better part of yesterday reading what was probably a hundred or more datasheets and application notes. It is now clear to me that the reason I was having trouble is that my initial circuit operates totally in the discontinuous conduction mode (dcm). While pulse skipping or oscillator/fet-driver gating works perfectly in DCM, it doesn't work very well at all in continuous conduction mode. Given that efficiency and power capacity is so very much better in CCM it is clear that I'll have to abandon the entire scheme.

Here's a video of something I was trying last night. I switched back to the air-cored inductor as I was curious what would happen:

One little gem of a formula I found was for determining the required duty cycle to achieve an input:output ratio. It is from a coiltronics app note;

Duty = 1 - (Vin / Vout)

Simply applying this and rigging up a simple circuit to apply a similar waveform to the gate I watched as it switched from 30W output in DCM to over 100W output in CCM with the same amount of waste heat. I was able to crudely vary the pulse width without affecting frequency, and could see the rough threshold of where DCM meets CCM on a given load.

In the same document I found the formula for determining required (but not optimal) inductance:

L = Vind * dT/dI
where Vind is the voltage applied by the psu across the inductor during fet ON state, dT is the pulse width ON time in seconds, and dI is the allowable inductor ripple current.

Elsewhere I found that the ripple current should be no more than 50% of the inductor's saturation current, and should be less than the rated current.

Now that I can determine the approximate inductance value, inductor current ratings, and required minimum pulse width I feel like I am getting somewhere.

I still can't find a formula for determining the input or output currents but it makes sense to me that if I know an approximate or worse-case efficiency, I can look at my output requirements, apply the efficiency, and get an estimate input current. How this correlates with a value of chosen inductor ripple current I don't know. I know from my physical circuit experiments that my core can handle 20A ripple current without overheating or saturating, so I'm using this as a rough guideline as at 10A my circuit in CCM was outputting about half of what I require.

Moving forward I'm going to try the MAX1771 IC with a modified feedback network (both the current sense resistor and the potential divider will be changed). According to the datasheet the current sense threshold is 100mV, and the Rsense should be chosen so that the current through the inductor doesn't exceed ratings or saturate the core when 100mV across is reached. This would translate to 0.005 Ohms for my 20A limit. It also states that the current feedback controls the pulse width maximum value up to an internal limiter. This would mean that if Rsense is too small for the inductance and maximum pulse width 100mV will never be reached, so I'll also have 0.01 and 0.015 ohm resistors available for swapping in and out. In that spirit, I'll have the capacity of trying out different values of inductance with the same core, since it is DIY.

The datasheet provides this formula for minimum inductance with the MAX1771:

L = (Vin * 2uS) / Ilim
where Ilim is the maximum allowable ripple current.

This yields very low inductance values for my parameters; 1.2uH for 20A ripple, and 2.4uH for 10A ripple. So, given that my current inductor is around 55uH and easily adjustable, I shouldn't have much if any issues there.

The datasheet says operation is via two one-shots, one for on time, and one for off time, where the on time continues to maximum unless stopped by the current threshold being reached. It doesn't say what kind of black magic is used to keep the chip running in CCM; i.e. what sets the lower bound of inductor current. If anyone knows, I'm all ears.

It gives a formula for determining the potential divider values for feedback of output voltage, and buried in there you can find the threshold where it sets output to be 1.5V at FB pin. I'll use a pot for the bottom resistor, with a limiting resistor in series.

As always, any and all info or tips are greatly appreciated, and I hope someone else finds all this useful too.

Moderator(s): Chris Russell, Noelle, Alex, Tesladownunder, Dave Marshall, Dave Billington, Bjørn, Steve Conner, Wolfram, Kizmo, Mads Barnkob