MC34063 with small inductors

[Joerg]

It's s'posed to cut off during ramp-down--that's when the flop gets
reset, unconditionally. So says TI, anyhow.

http://focus.ti.com/lit/an/slva252b/slva252b.pdf

the operative paragraph being the last one pg. 5. And, the last
comment in Table 1 confirms:

"Output switch conduction is always
terminated when Ct is ramping down."

I know but as shown in my sim plot if is terminated before Ct is ramping
down. That not s'posed to happen. But does, at least according to SPICE.

 

[Joerg]

What does it do in the lab?

I don't know yet. The package just arrived but I have a much more
pressing problem from an other client to deal with right now. Also a
chip that seems to not exactly do what I think it should. For that one
we have lab bench scope plots and the manufacturer is investigating.
This one simulates alright but chokes in real life.

 

[Hammy]

I don't know yet. The package just arrived but I have a much more
pressing problem from an other client to deal with right now. Also a
chip that seems to not exactly do what I think it should. For that one
we have lab bench scope plots and the manufacturer is investigating.
This one simulates alright but chokes in real life.

I have a NCP3063 Buck on a Breadboard now. I'm using it to test
cross-regulation on a coupled inductor (+/- 5V out). It only
terminates on the falling edge of Ct.

See scope shot. Ct ramp and switch node same as Jorge's sim.

http://i42.tinypic.com/34nlitw.jpg

You can see the glitch on the Ct ramp when the switch changes states.

Joerg I used the MC33063 for a 15W DCM buck before, it was the first
SMPS I did on a PCB in college and it ran fine. Mind you it needed 3 x
1mF of output capacitance to get the ripple down to something
reasonable. The inductor was 12uH for 5V out at 3A; 15 to 20Vin,
60kHz. The peak current was a little over 6A ;-) using an external
P-Fet.

I think the only reason they don't show or recommend DCM is because
the already high and modulated ripple you get with Hysteretic type
converters is much worse in DCM particularly at higher power levels.
This isn't flattering for their IC so they just show CCM examples.
Just my thoughts.

Here is the output ripple 12Vin, 5V out, 0.4A. This is with a 220uH
inductor and a 330uF PSA POLYMER output cap. Doing a proper PCB and a
small LC filter (bead MLCC) would get rid of those HF oscillations

http://i41.tinypic.com/2mebbya.jpg

This is using the NCP3063 at 175kHz

 

[Joerg]

I have a NCP3063 Buck on a Breadboard now. I'm using it to test
cross-regulation on a coupled inductor (+/- 5V out). It only
terminates on the falling edge of Ct.

See scope shot. Ct ramp and switch node same as Jorge's sim.

http://i42.tinypic.com/34nlitw.jpg

You can see the glitch on the Ct ramp when the switch changes states.

Thanks, Hammy. If you never see it turn off before the peak ramp
reversal then it behaves according to the datasheet. That would mean the
SPICE model isn't quite up to snuff or LTSpice misinterprets something
in there.

Joerg I used the MC33063 for a 15W DCM buck before, it was the first
SMPS I did on a PCB in college and it ran fine. Mind you it needed 3 x
1mF of output capacitance to get the ripple down to something
reasonable. The inductor was 12uH for 5V out at 3A; 15 to 20Vin,
60kHz. The peak current was a little over 6A ;-) using an external
P-Fet.

3000uF on the output? Yikes! I don't even have that much on the
rectifier after the mains transformer.

I think the only reason they don't show or recommend DCM is because
the already high and modulated ripple you get with Hysteretic type
converters is much worse in DCM particularly at higher power levels.
This isn't flattering for their IC so they just show CCM examples.
Just my thoughts.

Here is the output ripple 12Vin, 5V out, 0.4A. This is with a 220uH
inductor and a 330uF PSA POLYMER output cap. Doing a proper PCB and a
small LC filter (bead MLCC) would get rid of those HF oscillations

http://i41.tinypic.com/2mebbya.jpg

This is using the NCP3063 at 175kHz

The RF is fairly easy to lowpass out but the lower frequency ripple from
the slightly hysteretic behavior of this chip drops in spectrum a lot
the higher the inductance and the larger the output cap. But in this app
we can live with it since that rail is only used for relay drivers.
There's another regulator downstream for sensitive stuff.

I don't think I'll be able to keep it in DCM considering the cramped
space I've got on the board.

Thanks for posting the scope shots, looks encouraging

 

[Joerg]

Is it happening simultaneous with _on-set_ of ramp-down?

On the 34063? I have simulated the dickens out of that and the
comparator never shuts off the switch at onset, just occasionally in the
middle of a ramp-up which it ain't supposed to be able to do. And in
Hammy's case doesn't do.

 

[Joerg]

But I can't convince any of these companies to hire me to do high
quality modeling... they want me to submit a quote for about the
equivalent of one day's work, when it would generally take a week to
do a real quality job :-(

Sometimes I have those request as well but never from exisiting clients.
If the conversation starts with "We've got something that should be
rather trivial for you ..." I can smell it, and usually must decline.

I guess they really don't care.

Have you tried other models for the 34063? ...

Grand total of four, same problem.

... So far I've seen two, but
haven't had time to try either one yet.... I Usenet while running sims
that might pay off

Same here. But now I've got a li'l RF project for dessert, solder iron
time

And the last Weller ETS tip just bit the dust, have to see if I can
revive that with some sandpaper for now so it'll hang on until Saturday.
Did I already say that I hate RoHS solder?

 

[Spehro Pefhany]

Did I already say that I hate RoHS solder?

You use it even for test circuits?

 

[Joerg]

You use it even for test circuits?

Oh no, never. But I do have clients who market internationally so
there's the occasional RoHS board crossing my lab bench.

 

[Joerg]

Propagation delay? When the ramp comparator trips, it'll take a
minute or two to turn the timing ramp current source around. So to
speak.

That was a good one. Maybe there's a bureaucrat in the chip and the
comparator needs request a permit every time it trips

 

[Joerg]

Geez these threads get messy. First of all, have you called apps?

Not yet, another more pressing problem came up. Other client, other
chip. That one behaves ok in the sims but doesn't behave like the
datasheet says in real life. Hurumph ...

Generally if there is a limit, there is science behind that limit. But
in the chip biz, you have to pick limits that WILL work for the tens
of millions of parts you crank out. Probably more for this part. So
you get it to work in spice, but they fab a "slow" boat of wafer that
is still within spec, and the thing doesn't work. Now if your volume
is low and you temperature test your boards, I wouldn't be too
concerned with using an external component a bit out of the spec
limit.

Again, I am _not_ operating the thing outside its spec limits.

 

[Joerg]

Are you trying to run it faster than spec? Faster rise times
capacitively tripping a comparator?

No, in that mode it's actually running way slower than spec, around
30kHz while spec says 100kHz possible.

 

[John Devereux]

The compensation is that you need very little copper and ferrite at 3
MHz.

Much easier to filter the ripple too.

 

[Tim Williams]

I've done 300KHz, discrete, but 50KHz-ish is usually more efficient.

I've done 1MHz discrete. Didn't come out all that efficient though, needs
work...
http://webpages.charter.net/dawill/Images/Fast_Forward.gif
(note the output filter cap specifically *isn't* tantalum... ESR compensates
the loop)

Tim

 

[Tim Williams]

Hmmm. I try to keep total switching time under roughly 10% of the PWM
period. At 3MHz that's 17nS per transition. Pretty do-able at 5v,
actually. But that's supply's so generic you might as well just buy
it.

Seems to me, D from BC was working on some 1MHz offline switcher. Something
about big loads of EMC from the transistor-to-heatsink capacitance.

I'm going to try an induction heater at 1MHz sooner or later.
C0G tank cap:
http://myweb.msoe.edu/williamstm/Images/1MHz_Ind_Cap.jpg
Inverter board:
http://myweb.msoe.edu/williamstm/Images/1MHz_Ind_Top.jpg
(the whole high side is top and bottom ground plane floating at the 1MHz
output)

Tim