Searching \ for '[EE] Re: : voltage regulator' in subject line. ()
Help us get a faster server
FAQ page: www.piclist.com/techref/power/regulators.htm?key=voltage
Search entire site for: 'Re: : voltage regulator'.

Exact match. Not showing close matches.
'[EE] Re: : voltage regulator'
2005\09\18@101121 by

On 9/18/05, Tom Sefranek <tcscmcorp.com> wrote:
{Quote hidden}

20A x (48-12)V = 720W

Assuming 80W/transistor it means he need 9 transistors and a lot of
water + a pump recirculation on the heatsink.
with that ugly thing. One great dummy design I saw, it was than one
transistor collapse it broke all supply.

There is no other solution than a switching supply. With a very good
dimensioned filter you may expect -40dBm noise at this load.

cheers
Vasile

{Quote hidden}

> --

On Sep 18, 2005, at 7:11 AM, Vasile Surducan wrote:

>>> Input range is 20-48 and required
>>> output is 12vdc @ 10Amps with a max of say 15-20A.

Can you tighten up that spec?  15-20A for how long?  With what
sort of behavior on the output voltage?  Otherwise, you're looking
at a 12V@20A power supply, which is twice the nominal 12V@10A...

>> A classic Linear regulator is the multi 2N3055 pass transistors feed
>> by
>> a single 3055 driven by a 723.
>
>   20A x (48-12)V = 720W
>
Ouch! (of course.)  The nominal 240W switcher is no piece of cake
either,
especially since the specs miss the 36-72V input range used in telco
equipment that circuit designers tend to aim for...

BillW

>>>>Input range is 20-48 and required
>>>>output is 12vdc @ 10Amps with a max of say 15-20A.
>
>Can you tighten up that spec?  15-20A for how long?  With what
>sort of behavior on the output voltage?  Otherwise, you're looking
>at a 12V@20A power supply, which is twice the nominal 12V@10A...

I need 10 amps (prob 99%)  but it should be capable of higher peaks
without letting the smoke out. The input voltage is going to be from 20-48
and not steady. I have no control over that one. I did look at telcom stuff
but the input seems pretty fixed around 48.

Switching type is the only way to go on this as far as I can tell. It would be
nice to keep the efficiency pretty high and the heat low. I caught myself
looking
at Romans circuit the other night wondering if a few different bits might make
it work. Anyway just was hoping someone knew of something so I wouldn't
have to slowly reinvent the wheel.

Dave

On 9/19/05, Dave King <KingDWSshaw.ca> wrote:
>
> >>>>Input range is 20-48 and required
> >>>>output is 12vdc @ 10Amps with a max of say 15-20A.
> >
> >Can you tighten up that spec?  15-20A for how long?  With what
> >sort of behavior on the output voltage?  Otherwise, you're looking
> >at a 12V@20A power supply, which is twice the nominal 12V@10A...
>
> I need 10 amps (prob 99%)  but it should be capable of higher peaks
> without letting the smoke out. The input voltage is going to be from 20-48
> and not steady. I have no control over that one. I did look at telcom stuff
> but the input seems pretty fixed around 48.
>
> Switching type is the only way to go on this as far as I can tell. It would
> be
> nice to keep the efficiency pretty high

this will produce a high output noise, of course

> and the heat low.

360W of switching will not be low heat at all, so think to forced air cooling.

I caught myself
> looking
> at Romans circuit the other night wondering if a few different bits might
> make
> it work.

Nothing much appreciated than a good joke. Take a look on some
schematics about how are designed PC power supplies. Then try to
compute the switching circuitry for
much higher current and lower input voltage. You need at least two
power switching devices working in push-pull fashion. 20V and 400W
means 20A switched. So you need some low Rds devices, else will be
0.1ohm Rds x 20Ax20A = 40W dissipated by one device. Also rectifiers
must be carefully chosen, as well as coil designs. Avoiding core
saturation is must at this power. If it's the first time for you, be
prepared for a long and difficult project.

success,
Vasile

part 1 3273 bytes content-type:text/plain; charset=ISO-8859-1; format=flowed (decoded 7bit)

Having just designed and built a 0-50VDC 0-5A fully regulated LINEAR
power supply I can quite rightly say that any work in this power range
and above, serious consideration has to be given to getting the wasted
heat out of the system.

The power supply in question was designed for a magnet ramp supply, so
noise was a serious consideration and therefore a switching supply was
out of the question. The DC input voltage was 65VDC.

The final design used an OPA541 pwr opamp as the series pass regulator,
being driven by an ultra-stable, low-noise reference. Three MJ15003's in
parallel were used as the pass elements. Current regulation was achieved
using a fully floating/high side P-Channel MOSFET regulator. The unit
was tested in abuse conditions to destruction several times. The two
major concerns were the amount of latent heat that had to be removed and
load dump due to the switching and disconnection of two very large
electro magnets. This necessitated forced air cooling, both feed and
drag fans at either end of the system and active surge clamping on the
output to protect the power supply.

For your application a similar configuration with an appropriate number
of pass elements should be do-able.

Nino.

Vasile Surducan wrote:
{Quote hidden}

part 2 373 bytes content-type:text/x-vcard; charset=utf-8; name=Nino.Benci.vcf
(decoded 7bit)

begin:vcard
fn:Antonio L. Benci
n:Benci;Antonio L.
org:Monash University;School of Physics
email;internet:electronic.servicesspme.monash.edu.au
title:Professional Officer
tel;work:+613 9905 3649
tel;fax:+613 9905 3637
x-mozilla-html:FALSE
url:http://spme.monash.edu.au
version:2.1
end:vcard

part 3 35 bytes content-type:text/plain; charset="us-ascii"
(decoded 7bit)

Dave King wrote:
> Switching type is the only way to go on this as far as I can tell. It
> would be nice to keep the efficiency pretty high and the heat low. I
> caught myself looking
> at Romans circuit the other night wondering if a few different bits
> might make it work.

Roman's circuit is a cheap and simple switcher with relatively poor
performance.  It's OK for powering a PIC from a 20V source.

However you are looking for a 240W power supply.  This is going to require
some careful design and the right parts.  And yeah, it's going to cost more
than a few transistors from the junk box and an RF choke.  This is also not
a beginner's switching power supply.  Frankly, if you have to ask here, you
probably shouldn't be attempting this by yourself.

The wide input voltage range and high current suggest to me a dsPIC
controlled multi-phase approach.  We have experience with this sort of
thing.  Let me know if you want professional help.

*****************************************************************
Embed Inc, embedded system specialists in Littleton Massachusetts
(978) 742-9014, http://www.embedinc.com
This is not necessary a dsPIC application. Analog based SMPS is still
much more popular and more prevalent in the market. Digital PWM
ASIC is relative new. Of course DSP/MCU based power supply
has been in the market for long. Still the control is not so simple
since the basic of the PWM control is still lack of rigorous theory
to back up the analysis. Cuk's averaging theory is quite okay
for practical analysis and PID control is still the most important
control method and is proved to be quite effective in practical applications.
However it is still quite an approximation. The major problem with
DSP/MCU based control is the speed of the ADC. It is still not
up to the requirement of the control application in quite a few cases.

Anyway I agree with you that this is not a beginner's switching regulator
and professional help is indeed needed if the OP does not have
the relavant background.

Regards,
Xiaofan

On 9/19/05, Olin Lathrop <olin_piclistembedinc.com> wrote:
{Quote hidden}

Xiaofan Chen wrote:
> This is not necessary a dsPIC application.  Analog based SMPS is still
> much more popular and more prevalent in the market.

Yeah, so?  Pet rocks were popular at one time too.

I think digital control is useful in this case due to the large input
voltage dynamic range.  At this power level, it will be important to get the
maximum out of the components.  At the highest power level, you will want to
run the inductors as close as possible to saturation without actually
saturating.  This will be very difficult to do with analog control.  A dsPIC
can measure the instantaneous input and output voltage each pulse and
calculate the maximum on time accordingly.  This will allow use of smaller
inductors because less margin is required.

Digital control is also useful for intelligent startup, shutdown, power good
output, etc.  These things are not impossible with analog chips, but fall
out nicely from a digital design allowing you to get exactly the behavior
you want and not be stuck with whatever the chip vendor decided.

> Still the control is not so simple
> since the basic of the PWM control is still lack of rigorous theory
> to back up the analysis.

Huh?  Exactly what theory do you think is missing?

> Cuk's averaging theory is quite okay
> for practical analysis and PID control is still the most important
> control method and is proved to be quite effective in practical
> applications.

PID is certainly one control scheme.  It is readily implemented digitally so

> The major problem with
> DSP/MCU based control is the speed of the ADC. It is still not
> up to the requirement of the control application in quite a few cases.

That is a consideration, and microcontroller control can't go as fast today
as some of the top analog chips.  However, the dsPIC 10 bit A/D can sample
at 500Khz on the first parts and over 1MHz on newer parts.  Switching
frequencies of a few 100KHz are quite achievable with a dsPIC.  At 250KHz
you get 120 instructions per pulse.  That's a lot.  If you go too fast the
switching losses will start to be significant.  100-250KHz is a nice range
for many practical switching power supplies.

*****************************************************************
Embed Inc, embedded system specialists in Littleton Massachusetts
(978) 742-9014, http://www.embedinc.com
On 9/19/05, Olin Lathrop <olin_piclistembedinc.com> wrote:
>
> I think digital control is useful in this case due to the large input
> voltage dynamic range.  At this power level, it will be important to get the
> maximum out of the components.  At the highest power level, you will want to
> run the inductors as close as possible to saturation without actually
> saturating.  This will be very difficult to do with analog control.  A dsPIC
> can measure the instantaneous input and output voltage each pulse and
> calculate the maximum on time accordingly.  This will allow use of smaller
> inductors because less margin is required.
>
> Digital control is also useful for intelligent startup, shutdown, power good
> output, etc.  These things are not impossible with analog chips, but fall
> out nicely from a digital design allowing you to get exactly the behavior
> you want and not be stuck with whatever the chip vendor decided.

Oh yes that is what TI / AD are pushing. And since the cost of DSP/FPGA
are getting lower and lower, there will be more and more power supply
application which uses DSP/FPGA/MCU. Still the majority of the application
now is still analog based.

{Quote hidden}

Exactly how many theories have you learned? There are so many PhD thesis
on this topic. Anyway, never mind. PWM looks very simple yet it is
actually very difficult to understand. I am certainly not an expert but I
know that the theory of PWM control is not fully understood until now.

{Quote hidden}

I higher power application, one will be very happy with 50Khz with digital
control.

Simply mentioning digital control does not solve all the problems. Anyway
this is still not a very difficult task so both digital control and analog
control can be used.

Regards,
Xiaofan

Sorry I should have not ask the question "Exactly how many
theories have you learned?". I believe you must have learned
quite some so that you can offer professional help to others.

Still you quote my first email  wrongly. What I have said is
that even averaging theory enabled people to use simple PID
control method for switching converters and proves to be good for
lot of practical application, averaging theory itself is an
approximation. There are still a lot of unknowns in the
PWM switching converter modeling and control. PWM converters
are highly non-linear system and right now the theory to
tackle the problem is linear approximation. Still for some
applications like resonant converters, averaging theory simple
is not working. Even for simple control method like current
mode control, there are quite some theories trying to explain
what is really under the hood.

As for digital control versus analog control, I believe that
digital control will be one direction to go. However this
is not to say that analog control (ASICs) will lose its leading
role in such applications. The good point of digital control
is the adding capabilities of diagnosis/communication/...
Still there are many difficulties in the real applications.

As for the inductors, I do not think this will be too difficult
for those below 1kW. If it is 10kW, then things will be totally
different and an inductor can cost a lot if the current is really
high.

Regards,
Xiaofan

{Original Message removed}

What I have been seriously looking at is using a LT3800 based regulator
design. That gives me my requirements quite nicely as well as the result
being fairly compact which is also going to be
quite critical. This is going on a small generator engine with a built in
alternator. It will have plenty of cooling air so that is well covered. I
am wondering about the noise on it now that someone
mentioned it. I might just take care of that by filtering on the subsystems
(5, 4.1, 3.7).

I was going to suggest something based on the bag of 2222a's I have sitting
here based on Roman's stuff but the humor impaired do seem to have missed
that bit ;-]
Nothing like 40-60 of the little buggers on a board to warm things up a bit.

Dave

{Quote hidden}

>
The LT3800 is indeed an interesting product. The example application
in the datasheet and the LT journal is only for lower power
(12V, 75W and 3.3V/33W) so you need to adapt the design to suit
your needs. 20A to 30A (or slightly higher) are practical limit
of per phase in this kind of application. If more current
is required, more interleaved channel will be required. For example,
good Pentium 4 supply will be three to four channel and each channel
will provide 20A to 30A to the CPU.

Regards,
Xiaofan

{Original Message removed}
I'm sorry I should have looked at the part. It is a LTC3703EG
not the 3800. As well I have a LT1270ACT to play with. As I mentioned
this is going to be engine generator based application so if the total
cost is too high then I might as well just grab something from Napa
for
\$20.

The actual generator is rated (marginally) for 10 amps so the
suggestion
to use a DSP is actually quite interesting though realistically
somewhat
techie overkill ;-]It would be nice to have but I'd rather just
current
limit it to a safe level instead of pushing it to that last ma of
output.

Thanks for the sggestions.

Dave

{Quote hidden}

>> This is not necessary a dsPIC application.  Analog based SMPS is
>> stillmuch more popular and more prevalent in the market.

> Yeah, so?  Pet rocks were popular at one time too.

Pet rocks, and analog SMPS ICs both share one truly endearing trait -
once their makers have sorted out any early design problems (the lava
state can be hard to deal with) neither crash or take occasional time
out to get their act together after being subject to certain
conditions within their specs
but not anticipated by the designer. Depending on application this can
be a significant factor.

RM

Interesting comment!

It will still crash sometimes. Then again we will say it is a design
problem. The long term stability of some power supply can only be
found out after quite some time. And I will not try to push any
components to its limit in any SMPS design. Margins are very
important in SMPS design.

Regards,
Xiaofan

On 9/20/05, Russell McMahon <apptechparadise.net.nz> wrote:
{Quote hidden}

On 9/20/05, Xiaofan Chen <xiaofancgmail.com> wrote:
> Interesting comment!
>
> It will still crash sometimes. Then again we will say it is a design
> problem. The long term stability of some power supply can only be
> found out after quite some time. And I will not try to push any
> components to its limit in any SMPS design. Margins are very
> important in SMPS design.

And dry filtering capacitors.
:)

cheers,
Vasile

More... (looser matching)
- Last day of these posts
- In 2005 , 2006 only
- Today
- New search...