Inverting boost converter inductor selection

[Yansi]

Hi!
I'm having a little trouble calculating a reasonable inductor in a inverting boost topology converter (schematic below).

It is a some sort of bizzare converter topology, but is correctly working.  Quite nice solution actualy, if you need to get a +-16V from a lower voltage with two same convertor ICs.

The converter I'm trying to design should be as follows: 12V input, -16V 0.3A output:

By taking the rule of equal voltseconds, ton*Uin = toff*Uout, one can get that for CCM mode the duty cycle s = 1 - Uin/Uout. So the duty cycle required for 16V output should be about 30% (s=0.3)

The switching frequency is 700kHz and current ripple should be selected low enough, due to the core loss at such high frequency. So lets say 10%.

U = L * di/dt   can be modified to such form: L = U * dt / di so we can now calculate the inductor. Inductor average current estimated is about 0.55A (17*0.3 = 5W output means about 6W of input at 12V)

L = U * dt / di  = 12V * 0.3/700kHz / (0.1*0,55A) = 90uH

What I have a problem with, is that the inductance seems to be too high.

Problem 1: the saturation current of such coil should be at least 0.7A or more to be on the safe side. At 100uH this is not a small inductor, this is a solid chunk of core.
Problem 2: SRF of such coil probably too low, wouldn't that cause trouble at those 700kHz of operation?

Can you please check my calculations and/or suggest what I have done wrong or what should be considered different way? Am I too conservative with the 10% current ripple?

I am used to design quite higher power convertors (up to two orders of magnitude more power), where the switching frequencies are mostly well below 100kHz. I don't usually design those funny small converters with rocket science frequencies. So I'd be very happy to hear your suggestions for this stuff.

PS: Also what inductor type would you recommend using? Is something like this okay for 700kHz? http://datasheet.octopart.com/B82472G4224M-EPCOS-datasheet-519061.pdf or do I need something completely different?

Thank you,
Y

[station240]

Have you tried simulating it in OrCad or the like ?
Gives you the ability to view the ripple and experiment with different LC values.

[Yansi]

Thank you for suggestion. DC/DC converters are not (er.. should not) be designed by experimenting with values, but by calculating those values. I have not run any kind of simulation, as they would be pointless, if not having precise models for the inductor and other components. (which I don't have) The circuit physical layout also does play a role here and can't be simulated easily. When talking about such simulations - the best simulation tool is of course to build the circuit and measure - as I've already done (for different in/out parameter setup).

I am kind of confident, that the inductor value is calculated right (my friend has done an independent calculation like me) for the given inputs. (Using formulas in the datasheet also gives same results, as those formulas are the same, only written in different form)

There are not many parameters in the inductance calculation process you could choose or deliberately change - there is only one to be exact: Current ripple.

The "rule of thumb" from the datasheet says the current ripple should be 30 to 50% of input current. In my opinion based on experience with previous designs, this should be fine, but only at low frequencies and for ferrite core inductors. It seems to me that so much current ripple they suggest is waaay too much for any kind of magnetics at 700kHz.

So I'd like to ask you for a suggestion or opinion, if such high current ripple is okay for 700kHz operation. (as I think it is not so much).

The other thing in question is the SRF (self resonant frequency). How much should I be concerned with that for such DC/DC converter? Are those inductors I have presented above okay for such operation frequency? (sadly, the datasheet doesn't provide SRF values, nor stray capacitances).

Thank you
Y

[T3sl4co1l]

This is a bad topology because it effectively connects a capacitor across the switching transistor.  Which means it's a hybrid charge pump and inductive converter, which is fine for steady state but particularly stressful during startup or dynamics.

Tim

[Yansi]

So then replace the D2 by an inductor, done, problem gone.    That should work.

[T3sl4co1l]

Er, the problem is the capacitor into D1...

Now, a nice option would be a Cuk converter, which has the bonus of low supply and output ripple.

Tim

[chris_leyson]

Yansi, that is a bizzare topology, I would go for an inverting Cuk.

Sorry for straying off topic, Tim, I'm going to try a coupled inductor Cuk as soon as I get time, built one years ago and very nearly got zero ripple on input and output, the transformer was a dogs breakfast and I had to hold it together with my fingers, could shim core gaps right, would have cost a small fortune to have it manufactured. Getting zero ripple on the input OR output wasn't too difficult. I just added some extra inductors and went for low ripple continuous conduction. Use a Mosfet for the diode and it works in both directions. The cleanest and quietest converter I've ever built.

[T3sl4co1l]

Nice thing about coupled inductors is, if you don't need any particular ratios (which you don't, Cuk is strictly(?) 1:1), you can buy off-the-shelf parts in vast quantity, because these are fairly normal things now!

Tim

[Yansi]

Er, the problem is the capacitor into D1...

Now, a nice option would be a Cuk converter, which has the bonus of low supply and output ripple.

Tim

What are you talking about? If D1 parasitic capacitance (or Qrr) is a problem, then why you suggesting a ?uk converter? It is the absolutely same circuit.  I am quite confused now... My implementation in the first post differ from ?uk only by placing a diode instead of the second inductor. (to get away with less inductors...)



//The hell!! The forum doesn't like to eat Cuk with the diacritics.

[T3sl4co1l]

Sorry, I did mean D2!

[Yansi]

Okay, misstakes happen. Now it makes sense. So I will not push the limits with saving cost of single inductor.

I am going to design a small prototype board (better than the home-etched I used for initial measurements) and will post some results and also to compare results.

How much % of current ripple can one afford with a common ferrite-core SMD choke at ~700kHz operation? What do you think about those 10% I have used for the calculation above?


Thx,
Yan.

[Buriedcode]

Isn't this a boost converter with a charge pump on the output? Granted, the feedback comes from the output of the charge pump rather than the diode/caps just added to an already regulated output.

I believe this is fairly common for portable equipment that need positive and negative supplies of higher voltage (portable equipment with LCD that need ~-18V drive).  Often a low value resistor is used to feed the charge pump, as to limit the current going into the cap and reduce stress on the converter - at the cost of reducing efficiency of the charge pump part.  I have used simple dual schottky  diodes and ceramic caps to get +/-9V from a 5V boost converter, just used for low current opamp supplies.  Not as efficient as using couple inductors but very low cost and uses little board space

[Yansi]

Yes, it is. Somene named it "a hybrid charge pump" - I like that name.

Yes, the circuit was intended to supply a few tens of miliamps to a bunch of analog stuff in an USB bus powere measurement gizmo. 

I have used MP1542 as the converter IC (haven't written that anywhere so far to prevent possible hate), i.e. 2A switch device. These converters are cheap and provide great spec. (Unlike those LM257x/LM259x series bipolar overpriced shit from NSC/TID with buttloads of fakes on ebay).
It did work, indeed. Supplied me very stable -15V output at 50mA, quite happily. Efficiency 80%+ (peaking 85% at 60mA load).