On 2/11/2012 3:18 PM, Jim Thompson wrote:
> On Sat, 11 Feb 2012 13:59:29 -0600, Tim Wescott
> <t...@seemywebsite.please> wrote:
>
>> On Sat, 11 Feb 2012 11:22:38 -0700, Jim Thompson wrote:
>>
>>> In the past I have designed boost converters in burst mode, with peak
>>> current control, but never with a PWM-controlled loop.
>>>
>>> Outputs: +5 @ 25mA, -5V @ 20mA
>>>
>>> Can someone point me to a tutorial?
>>>
>>> Thanks!
>>>
>>> [Don't point me to an off-the-shelf part. This has to go into a
>>> _custom_ chip... just a little thing... 1mm x 25mm :-]
>>
>> Actually, I'd go looking at off-the-shelf parts to snag ideas from.
>>
>> Efficiency?
>
> As high as practical.
>
>> Regulation bounds?
>
> Loose. Lots of compensating innards for power supply variation.
>
>> % of total product cost that'll be taken
>> up by the converter?
>
> They've already committed to external switch(es)
>
>>
>> Unless the above answers are "low, loose, and low", I think your biggest
>> challenge won't be the silicon, it'll be finding small, economical& high
>> performance inductors, designing the thing to take advantage of them, and
>> dancing around the inevitable unstable zero in the boost converter.
>
> The inductor _is_ a problem. They're already at their height limit
> and their present implementation (at lower current) is just pushing
> into saturation :-(
>
Has the value of inductor been selected properly? The rule of thumb for
a boost converter in continuous conduction mode is L = (V *
D)/(R*I_l*f), where the parameters are evaluated at the _lowest_
specified (worst case for a boost) input voltage. V is the input
voltage, I_l is the DC inductor current (I_o/(1-D)), D is the duty
cycle, f is the switching frequency. R is the "ripple current ratio"
and has an optimum value of 0.4 for continuous conduction mode in most
cases.
At low voltages the inductor can be sized solely based on the DC load
current requirements - if the current limiting is fast enough it doesn't
matter if the inductor saturates under abnormal conditions.
If the inductor is still too big maybe use discontinuous mode instead?
It'll be harder to stabilize, unfortunately.