Re: Oscillator circuit for solid-state TC

To: tesla-at-pupman-dot-com
Subject: Re: Oscillator circuit for solid-state TC
From: "Harri \"Haba\" Suomalainen" <haba-at-cc.hut.fi>
Date: Wed, 9 Oct 1996 21:09:06 +0300 (EET DST)
On Tue, 8 Oct 1996, Tesla List wrote:
> > a circuit built around the "SG3524" IC which has a maximum operating
> > frequency of 150 khz.
> Yes I've used the 3524 chip - I'm sure I pushed it to 200kHz.

It is specified up to around 250kHz (500kHz osc. frequency) if my
memory server me well.

> The big advantage of these chips is that they are designed
> to provide all the control for a switching power supply.

Indeed! An oscillator will need awfull lot stuff aound unless you
just add a couple of flipp-flopps and think you need nothing
else like protection circuitry :)

> I'm going back to the CA3524 to replace the 555 as the low
> frequency switch. Running continuosly draws serious power
> so I switch the oscillator on/off at about 60Hz with a varying
> on/off ratio - more spark- less smoke :)

..and lot less heat (lot less effects that with 100% duty cycle? :)
I used half wawe rectifying to lower power. At 1kW the caps were
around 0V during the other cycle. It did also look quite different
than the normal dc drive.

> Now that I'm working at 100kHz I may go back to the 3524 for
> that side too.

..or perhaps not. 3524 has a very low output current rating. It means
the mosfets will be charged and discharged slowly. You may also
experience some of the stuff I did: parasitic drain source capasitance
did cause spurious turn-ons ie. the driver was at low state and
still during the switching fets did turn on! It helped a lot when
I moved to a system where gate had -3.3V when both fets should have
been off.

Anyway, 3524 will give something like 100mA absolute maxinum gate drive.
3825 on the other hand is specified for 2A output (0.5ms short pulse).
That means that turn-on or turn-off of a mosfet will take about 20
times as much time with 3524! It means 20 times as much dissipation.

At high power levels and high frequency 3524 just will not work any
more. The situation is another when you add some mosfet drivers but
the cheapest way is to stick to good driver like 3825 in the first place.
Mosfet drivers are not cheap.

Did someone say "what the heck, put a bigger heat sink"? :) It will
not work always. I had a situation where dissipation was so large
the junction-case transfer was limiting. I had practically an infinite
heat sink...
--
"99% of the pain in the world is due to misunderstanding" -M. Jarvelainen

Harri.Suomalainen-at-hut.fi - PGP key available by fingering haba-at-alpha.hut.fi