Switchmode PSU controller?

[Lewin A.R.W. Edwards]

Hi all,

I've got a very simple requirement, which I can't seem to fulfil: I'm
replacing a very inefficient linear regulator with a switchmode DC-DC
converter, and I'm looking for an appropriate controller IC. Input
voltage varies between 11.3-14.0V, output voltage is 5V and minimum
required output current is 3A (3.5A would be better). This is a
prototype, so I would prefer something in a small, easy-to-use package
(no QFP or TSSOP, in other words - either DIP or SO8 would be best for
me).

I'm puzzled by all the options here - I've browsed Maxim, Analog
Devices, and Linear Technology, and haven't found an exactly suitable
product. Is there anything people can recommend? External switching
transistor is fine, my main criteria are ease of hand-assembly and
ready availability of parts at Digi-Key et al. (I'd also like the
total BOM for the power supply to be <$20).

 

[JW]

On 17 Sep 2003 08:32:07 -0700 larwe@larwe.com (Lewin A.R.W. Edwards) wrote
in Message id: <608b6569.0309170732.44fb22c0@posting.google.com>:

Hi all,

I've got a very simple requirement, which I can't seem to fulfil: I'm
replacing a very inefficient linear regulator with a switchmode DC-DC
converter, and I'm looking for an appropriate controller IC. Input
voltage varies between 11.3-14.0V, output voltage is 5V and minimum
required output current is 3A (3.5A would be better). This is a
prototype, so I would prefer something in a small, easy-to-use package
(no QFP or TSSOP, in other words - either DIP or SO8 would be best for
me).

I'm puzzled by all the options here - I've browsed Maxim, Analog
Devices, and Linear Technology, and haven't found an exactly suitable
product. Is there anything people can recommend? External switching
transistor is fine, my main criteria are ease of hand-assembly and
ready availability of parts at Digi-Key et al. (I'd also like the
total BOM for the power supply to be <$20).

I'd go with the LTC1735.
http://www.linear-tech.com/prod/datasheet.html?datasheet=485

Your output current capability would be determined by your FET.

 

[Sřren A.Mřller]

In article <608b6569.0309170732.44fb22c0@posting.google.com>,
larwe@larwe.com says...

Hi all,

I've got a very simple requirement, which I can't seem to fulfil: I'm
replacing a very inefficient linear regulator with a switchmode DC-DC
converter, and I'm looking for an appropriate controller IC. Input
voltage varies between 11.3-14.0V, output voltage is 5V and minimum
required output current is 3A (3.5A would be better). This is a
prototype, so I would prefer something in a small, easy-to-use package
(no QFP or TSSOP, in other words - either DIP or SO8 would be best for
me).

I'm puzzled by all the options here - I've browsed Maxim, Analog
Devices, and Linear Technology, and haven't found an exactly suitable
product. Is there anything people can recommend? External switching
transistor is fine, my main criteria are ease of hand-assembly and
ready availability of parts at Digi-Key et al. (I'd also like the
total BOM for the power supply to be <$20).

The widely used MC34063A from On-semi ( http://www.onsemi.com ) should be
able to do it with an external transistor.

It comes in DIP-8 and SO-8, works with 3-40V input, has a 1.5A internal
switch and costs less than a dollar.
Because it only operates up to 100kHz, the inductor and capacitors need
to be larger than with newer, higher frequency chips.


Sřren A.Mřller

 

[Lewin A.R.W. Edwards]

Hi Sřren,

converter, and I'm looking for an appropriate controller IC. Input
voltage varies between 11.3-14.0V, output voltage is 5V and minimum
required output current is 3A (3.5A would be better). This is a

The widely used MC34063A from On-semi ( http://www.onsemi.com ) should be
able to do it with an external transistor.

Okay, great My records show that I have 1,704 of the ST version of
that part somewhere in my office.

So, how to select the output transistor? Based on a quick check of
Digi-Key's parametric engine, International Rectifier IRL3716 looks
plausible (VDS=20V, nominal ID=180A with a package limitation of 75A,
switching speed way within the 34063's range). Does that sound
reasonable? Will I be making an unnecessarily noisy circuit by using
this rather fast FET? (Sorry to bother you with this. But I am not an
analog guy; I only understand the rudiments of BJTs, and I've never
even built a circuit using a discrete FET).

It comes in DIP-8 and SO-8, works with 3-40V input, has a 1.5A internal
switch and costs less than a dollar.
Because it only operates up to 100kHz, the inductor and capacitors need
to be larger than with newer, higher frequency chips.

Space is no real concern here, so that's OK. I notice that some of the
newer SMPS controllers go up to several megahertz. Besides smaller
inductors, what's the advantage to higher switching rates? Smaller
ripple?

Thanks for the response.

 

[Lewin A.R.W. Edwards]

Hi!

I'm puzzled by all the options here - I've browsed Maxim, Analog
Devices, and Linear Technology, and haven't found an exactly suitable
product. Is there anything people can recommend? External switching

I'd go with the LTC1735.
http://www.linear-tech.com/prod/datasheet.html?datasheet=485

Thanks for the suggestion, but that part is only available in 16-SSOP,
16-TSSOP and 16-SOIC. For this particular project, 8 pins is the
largest SMD I'd like to handle (a DIP of any size would be fine).

The MC34063 looks like a better bet... I've used this chip before as a
black box, but only to supply 750mA (so no ext transistor required). I
need to learn more about SMPSses and analog electronics in general.

 

[Sřren A.Mřller]

In article <608b6569.0309180643.36d5e0a5@posting.google.com>,
larwe@larwe.com says...

Hi Sřren,

converter, and I'm looking for an appropriate controller IC. Input
voltage varies between 11.3-14.0V, output voltage is 5V and minimum
required output current is 3A (3.5A would be better). This is a

The widely used MC34063A from On-semi ( http://www.onsemi.com ) should be
able to do it with an external transistor.

Okay, great My records show that I have 1,704 of the ST version of
that part somewhere in my office.

Then you have a few to blow up

So, how to select the output transistor? Based on a quick check of
Digi-Key's parametric engine, International Rectifier IRL3716 looks
plausible (VDS=20V, nominal ID=180A with a package limitation of 75A,
switching speed way within the 34063's range). Does that sound
reasonable? Will I be making an unnecessarily noisy circuit by using
this rather fast FET? (Sorry to bother you with this. But I am not an
analog guy; I only understand the rudiments of BJTs, and I've never
even built a circuit using a discrete FET).

Well, I am not much of an analog guy either, but here are my thoughts on
this:

I think a P-MOSFET would be better (or a PNP if you can live with the
extra heat).
And 4mOhm/180A is probably a bit overkill. You are probably going to
dissipate at least 1W in the diode anyway.
You could ofcourse do something with a capacitor and a diode to get the
gate voltage above the input voltage.

The easiest would probably be to connect the Switch Emitter(2) to
ground, the Switch Collector(1) to the gate of the P-MOSFET with a
resistor to Ipk Sense(7) and connect the Drive Collector(8) to Switch
Collector(7) - pretty much like "Step-Down With External PNP Switch" in
the data sheet (minus one resistor).
The P-MOSFET is the connected with Source to Ipk Sense(7) and Drain to
the inductor and diode. A schottky diode would be a good choice for
better efficiency.

The transistor should be rated for at least 3.5A peak and 1.5A avg.
The diode should be rated for at least 3.5A peak and 2.5A avg.

If you keep the current paths (input capacitor, switching transistor,
diode, inductor, output capacitor) short and close and the switching
node small you shouldn't get to much noise.

By the way, On Semi has a lot of application notes - these might be
useful:

SWITCHMODE Power Supply Reference Manual:
http://www.onsemi.com/pub/Collateral/SMPSRM-D.PDF

Theory and Applications of the MC34063 and uA78S40 Switching Regulator
Control Circuits:
http://www.onsemi.com/pub/Collateral/AN920-D.PDF

It comes in DIP-8 and SO-8, works with 3-40V input, has a 1.5A internal
switch and costs less than a dollar.
Because it only operates up to 100kHz, the inductor and capacitors need
to be larger than with newer, higher frequency chips.

Space is no real concern here, so that's OK. I notice that some of the
newer SMPS controllers go up to several megahertz. Besides smaller
inductors, what's the advantage to higher switching rates? Smaller
ripple?

Yes, both due to the higher frequency and the fact you can use ceramic
capacitors with very low ESR compared to tantal or electrolytic.

They can also have better response to load transients due to a faster
regulating loop.

And then ofcourse the total size of the power supply is smaller - try
cramming a 34063-based switcher into a modern cell phone


Sřren A.Mřller

 

[Lewin A.R.W. Edwards]

Hi Sřren,

So, how to select the output transistor? Based on a quick check of
Digi-Key's parametric engine, International Rectifier IRL3716 looks

I think a P-MOSFET would be better (or a PNP if you can live with the
extra heat).

Argh, but how to choose it? When I use Digi-Key's parametric engine, I get
results that have max current ratings specified in negative numbers. What
does this mean?
How about IRF9Z10? 50V, 4.7A.

You could ofcourse do something with a capacitor and a diode to get the
gate voltage above the input voltage.

Er... no I couldn't, apart from putting them on the table and shaking until
random action assembles a working circuit

the Switch Collector(1) to the gate of the P-MOSFET with a
resistor to Ipk Sense(7) and connect the Drive Collector(8) to Switch

How do I determine the value of the resistor, though?

The transistor should be rated for at least 3.5A peak and 1.5A avg.
The diode should be rated for at least 3.5A peak and 2.5A avg.

SWITCHMODE Power Supply Reference Manual:
Theory and Applications of the MC34063 and uA78S40 Switching
Regulator Control Circuits:

I did actually read these, but didn't come away with much more
understanding.

And then ofcourse the total size of the power supply is smaller - try
cramming a 34063-based switcher into a modern cell phone

This is going into a model submarine, there is plenty of space I have
been using a "black box" integrated DC/DC converter module, but they're too
expensive.

 

[Sřren A.Mřller]

In article <GP5bb.43250$NM1.39596@newsread2.news.atl.earthlink.net>,
larwe@larwe.com says...

Hi Sřren,

So, how to select the output transistor? Based on a quick check of
Digi-Key's parametric engine, International Rectifier IRL3716 looks

I think a P-MOSFET would be better (or a PNP if you can live with the
extra heat).

Argh, but how to choose it? When I use Digi-Key's parametric engine, I get
results that have max current ratings specified in negative numbers. What
does this mean?

That just means that the device is P-type.

How about IRF9Z10? 50V, 4.7A.

I can't seem to find any datasheet on this, but it looks like it have a
Rds,on of 0.5 Ohm, which is a bit high and would cause apx. 3W of heat
loss.

I would go for a FET with Vds > 20V (< -20V), Ids > 4A (< -4A), Rds <
0.05 Ohm, Qg low, Vgs,th,min < -1V and Vgs,th,max > -5V.
An example in SO-8 is FDS4435: -30V, -8.8A, 20mOhm@-10Vgs.

You could ofcourse do something with a capacitor and a diode to get the
gate voltage above the input voltage.

Er... no I couldn't, apart from putting them on the table and shaking until
random action assembles a working circuit

That could take some time to get working

the Switch Collector(1) to the gate of the P-MOSFET with a
resistor to Ipk Sense(7) and connect the Drive Collector(8) to Switch

How do I determine the value of the resistor, though?

It should be low enough to switch the FET off fast enough (this depends
on the resistor, Vgs,th,min and Ciss) and high enough not to dissipate to
much heat.
For FDS4435, with Ciss = 1.6nF and Vgs,th,min = -1V, this would be
something like 100 Ohm, which would dissipate something like 0.75W and
give a switch-off time of 0.3us - the time it takes to discharge Ciss
from -12V to -1V through 100 Ohm.
It probably would be a good idea to use one resistor from Source to Gate
and one from Gate to pin 1/8 to keep Vgs a bit closer to Vgs,th,min thus
increasing the switch-off speed and reducing dissipation in the
resistors.
A capacitor (comparable in size to Ciss) could be placed from Gate to Pin
1/8 to speed up the switch-on if nessecary.
With two 100 Ohm resistors the dissipation would be less than .2W in each
and turn-off would be reduced to apx. 0.2us.
This should keep the dissipation to less than 1W in the FET, but that is
just my estimate without any warranty.

If you instead choose to use a PNP like D45H8 from On Semiconductor, the
calculations of the resistors become a bit simpler - you just need enough
base current to saturate the transistor when on, an a small enough Base-
Emitter resistor to turn it fast off. But you vill need something like
100mA base current which would dissipate more than 1W in the resistor.

The resistors could be replaced with a buffer with active pull-up, if
faster turn-off and lower dissipation is wanted.

The transistor should be rated for at least 3.5A peak and 1.5A avg.
The diode should be rated for at least 3.5A peak and 2.5A avg.

SWITCHMODE Power Supply Reference Manual:
Theory and Applications of the MC34063 and uA78S40 Switching
Regulator Control Circuits:

I did actually read these, but didn't come away with much more
understanding.

If you build it, it will come

And then ofcourse the total size of the power supply is smaller - try
cramming a 34063-based switcher into a modern cell phone

This is going into a model submarine, there is plenty of space I have
been using a "black box" integrated DC/DC converter module, but they're too
expensive.

It should be easy to get rid of that waste heat with water cooling then.


Sřren A.Mřller
--
Note: Return address is modified due to my mailbox filling up with Swen.