SMPS inductor calculation confusion

[Simon]

I've desined a simple SMPs for my circuit but i'm going around in circles in trying to calculate the inductor. I've used the formulas on the MC34063 datasheet but then looking around the internet I am finding many other formula's and everyone seems to contradict each other, some want me to select using the max current or peak current, others using the ripple current that is 10% the output current, the result is difference between the various methods of various orders of magnitude !

I know it's supposed to be a black art, but there is black art and there is people contradicting each other.

[jahonen]

If we ignore the obvious saturation current of the inductor, there is no single correct answer. If the loop compensation is good, you'll probably found that your power supply works fine with several values of inductance just fine

Exact value of the inductor depends on the ripple specifications and transient response you want. Lower inductance results higher ripple current in the inductor which in turn causes higher ripple voltage to output voltage. But then, that can be compensated by increasing capacitance. Increasing either output capacitance or inductance, slows down the transient response, because loop bandwidth will go down. Depending on the application, transient response may not matter at all (if load is pretty constant), as long there is no massive nasty turn on-transient.

Generally, you'll want the inductor current ripple to be in some sane limits, so having it something between 10 to 30% of maximum current is reasonable starting point. Also, it is usually beneficial to have enough inductance so that your SMPS will run in continuous conduction mode, i.e. inductor current never goes to zero.

Regards,
Janne

[Simon]

ok it's just that if I use the MC34063 formula that looks at peak current (2x Io) and i get 822uH, but other formula's devide by the ripple current which is 10% Io so 20 times smaller than the other formula, I jyst can't see how such huge difference in calculation can exist for the same thing.

I have calculated on the MC34063 formula's for a Vinmin of 20V, Iomax of 25mA and frequency of 100KHz 822uH and 15uF.

LTspice is a further confusion because trying to model the circuit or something close gets me a total fail with any value of inductor which clearly means that the program won't work properly but then it's so limited and a paiun in the arse.

All thin on the back of the fact that I recently threw together with no calculations a SMPS that worked in "bang bang" mode used a LM293 comparator, directly driving a P mosfet with 280uH and 2000uF and worked just fine, I've not taken any measurements yet and only used it to power a 300mA LED but it worked great, and now I try to actually do it properly it all goes tits up apparently.

[rob77]

"eyeball" the values -> measure -> adjust values
that works for me (hobby projects , not production stuff)
sometimes this method produces some blown parts... but well... that's life

btw.. for such a low current with 34063 basically "any" inductor will work , i'm using mc34063 quite often and i'm using only 100,150,330,470 uH inductors and never ran into issues... don't know how critical your project is, but probably it would be sufficient to just chose one , eyeball the capacitor value and then just quickly check the output ripple and inductor current under min and max load. if it's continuous and output ripple within spec... then job well done if not - adjust the values and measure again

[owiecc]

I am finding many other formula's and everyone seems to contradict each other, some want me to select using the max current or peak current, others using the ripple current that is 10% the output current, the result is difference between the various methods of various orders of magnitude !

Different formulas probably describe different ways of converter operation. It all depends whether you want you converter to run in CCM or DCM. Do you want big ripple but fast transient response or small ripple and slower response.

If you share what type of SMPS it is and what is the application we can probably suggest a good method to choose the inductor.

Ultimately it all comes down to U_{L} = L*di/dt

U_{L} depends on the converter topology
di/dt is connected with current ripple and switching frequency

[Simon]

its a buck from 20-30v down to 5v, the formulas i am comparing are for same type of converter.

[Kremmen]

ok it's just that if I use the MC34063 formula that looks at peak current (2x Io) and i get 822uH, but other formula's devide by the ripple current which is 10% Io so 20 times smaller than the other formula, I jyst can't see how such huge difference in calculation can exist for the same thing.

I have calculated on the MC34063 formula's for a Vinmin of 20V, Iomax of 25mA and frequency of 100KHz 822uH and 15uF.

LTspice is a further confusion because trying to model the circuit or something close gets me a total fail with any value of inductor which clearly means that the program won't work properly but then it's so limited and a paiun in the arse.

All thin on the back of the fact that I recently threw together with no calculations a SMPS that worked in "bang bang" mode used a LM293 comparator, directly driving a P mosfet with 280uH and 2000uF and worked just fine, I've not taken any measurements yet and only used it to power a 300mA LED but it worked great, and now I try to actually do it properly it all goes tits up apparently.

You can use almost any arbitrary values for L and C and such a simple circuit will most likely "work" although the dynamics are likely to be - hmm - disappointing may be the word. Sometimes a cookbook approach works just fine but then you tend to end up with exactly this type of questions when/if not everything is crystal.

Usually there are a number of design criteria/targets for the performance of a SMPS supply. At least these are usually interesting:
- the maximum allowable voltage ripple in the output  --> sufficient filtering of the switching waveform
- compensation of load current variations to output voltage --> max limit of converter output impedance
- compensation of input voltage variations to output voltage --> max value for line-to-output transfer function
- transient response time --> placement of loop gain crossover frequency of the transfer function
- output overshoot and ringing --> sufficient phase margin of the loop gain

Now the component dimensioning is impacted by the relative priority of these design targets. You say the design rules contradict each other; i say they just march to different tunes. You can't have everything at the same time, there will be priorities and compromises. And only you can decide which they will be in each particular case. (Think of bread: there is wheat baguettes and sourdough dark loaves and a million recipes between. Contradicting? - no, just different. But all are bread).

The MC34063 is meant for simple circuits because it won't let you access the compensator circuit. That makes it impossible to work with the feedback loop transfer function and you are left with the output filter component values. Should you aim for both fast dynamics, high stability as well as low ripple then this chip most likely won't cut it. But for a constant-ish load it can be OK and the component values are not critical at all. You will get varying behavior but if it doesn't matter then who cares.

[nctnico]

ok it's just that if I use the MC34063 formula that looks at peak current (2x Io) and i get 822uH, but other formula's devide by the ripple current which is 10% Io so 20 times smaller than the other formula, I jyst can't see how such huge difference in calculation can exist for the same thing.

There are two driving forces for choosing an inductor:
- The saturation current which is Vinmax * Ton * L
- The ripple current which causes core losses
So it depends on what you want more: a small size inductor or low losses. The tolerances on the choice of the inductor are large.

[Simon]

overall efficiency is not a concern anything will be better than a linear regulator, the input voltage will vary slowly as it's vehicle batteries so it's just based on engine on or off or varying states of charge. The load is fairly constant although I'll be sending out 50Hz pulses of up to 10mA but then I'm working at 100KHz so that should not be a problem in terms of recovery. My capacitors will be 66uF made up of 3x22uF so Plenty of overhead there. I've designed for 25mA continuous but I'm not planning to use quite that much.

[Precipice]

the input voltage will vary slowly as it's vehicle batteries so it's just based on engine on or off or varying states of charge.

Errr....
Alternator ripple can be up to a few volts of variable frequency. (It's normally less than a volt, though - are you planning to sell this as a product, or just use it locally?)
Then there's the awful automotive load-dump transients, surviving variously hamfisted jumpstart attempts, making sure that you don't catch fire as the battery goes flat, and, and, and...
If this is something you care about, then I would start by using an IC designed to survive such things. It doesn't necessarily need to be automotive qualified, but I'd say you were making life hard for yourself here.
(If you're only making one unit for yourself, then knock yourself out, it'll be educational. Even if you miss a few potential failure modes, it'll still work fine and you can repair it when it fails.)

Edit: From further up the thread - "Vinmin of 20V, Iomax of 25mA". Does this mean you're running on a 24V system? If so, the abs max Vin of 40V of the MC34063A is waaaay to close for comfort. The amount of protection you'll end up wrapping around it will outweigh the rest of the power supply design, even with your tiny input current (which works very much in your favour!)

[HackedFridgeMagnet]

Post your LTSpice sim, I'm sure someone could get it working.

I find some of the FET models on it a bit borked, and changing the FET models fixes things, or setting initial conditions helps too.

[Simon]

That is why I'm not actually using an MC34063 just using the formula's as I could not find any anywhere else and now that I have it's still a nightmare.

Yes 24V automotive, which is why i ended up using comparators to make my own. I need -46/+83C ambient (very few parts go under -40 the voltage reference alone was hard enough to find and costs £4.90), to withstand up to 50V input as that is the maximum clamping voltage of my TVS's, I have put my comparators that only go up to 36V behind a 15V zener regulator actual supply is not critical and They interface back to the power switch at up to 30V (50 in case of a spike) with a mosfet that can take that. The main switch transistor handles 200V but should never see more than 50V. Bearing in mind that all of these parts need to be in near sensible packages that I can solder (no BGA's or other odd balls, I stopped at the SOT23-5 voltage reference package as I had no option)

[Simon]

Post your LTSpice sim, I'm sure someone could get it working.

I find some of the FET models on it a bit borked, and changing the FET models fixes things, or setting initial conditions helps too.

I got fed up and ditched it, all i got out was Vin as the switch just stayed on but not being able to find the correct models for the parts makes it pointless anyhow. In theory it will work, Done it before, it's the practice that needs looking at.

[Precipice]

Isn't Sulphuric acid safely frozen by -46?
http://www.sulphuric-acid.com/techmanual/Properties/properties_acid_freezingpt.htm
If there's a battery heater and a block heater (and probably a cabin heater, if these temperatures are real), can you nestle in there and reduce your required temperature spec?
I take it you're very, very sure that you're not just beating yourself up for no reason on this spec? It all seems exceedingly unusual. If you're enjoying it, fine, but it doesn't entirely look that way.

(Also, if it's that cold, you can't begin to rely on 20V from the battery. Shortish burst of 3V might be more reasonable on cold cranking.)

[Simon]

Well the battery temperature is not going to be my problem and it is probably being heated with a fuel fired heater before the vehicle starts. I'm nowhere near any heating and I've been asked for those temperatures. We will test to those temperatures anyway. As far as i know my gear will be working in a vehicle with a running engine so at least 20V supply, if the engine is not running my kit is pretty pointless.

[owiecc]

I need -46/+83C ambient

Well, maybe add a heating element so you can find more ICs.

[Niklas]

24 V automotive and TVS diodes clamping at about 50 V. Have you checked the ratings of those TVS diodes, even the 5kW rated ones? If you are supposed to handle the standard set of pulses (load dump, 600 V transients etc) you might have a problem with most diodes. A transistor disconnect circuit or time limited clamping circuit could be options that works. Google "radiocad load dump"

Also check Murata's datasheet catalogue for ceramic capacitors, specifically the derating chart on one of the first pages. Make sure you have enough voltage rating for high temperature operation. Don't forget the power dissipation on the PCB as that will increase the temperature above ambient. At some input/output voltage and switching frequency combinations a buck converter can be unefficient.

Check the technical requirements before thinking about adding heating elements. Standby current? Full/limited operation over the complete temp range? When the vehicle is left for a week in freezing climate, a heater will drain the battery fast. Can you make sure that the heater always works? Safety related equipment? Better to design something that works without a heater if possible.

[Simon]

I've got 2 TVS diodes that do 3KW each, they peak at 48.5V and that's good, many hit something like 76-80V at full pulse. I'm only using 10W of heating, the device won't be left on if the vehicle is unused for a week.