RE: How to power a PCI card from either 3.3V OR 5V..

[Ivor Bowden <ivor@peritek.com>]

At 05:02 PM 3/28/00 -0500, Rick Collins wrote:
>A 3.3v regulator only dumps 8.5 watts if you are using a linear regulator.
>A switching regulator can be as high as 90% efficient dissipating as little
>as 1.8 watts. So using a regulator is not a real burden. 

like I said:
>>use a switching regulator, which still wastes heat, takes up
>>board space, and makes noise

It's a lot more burden then just connecting the 3.3V pins
(if one could count on them to have 3.3V).

>But the original question of why not use the regulator at all times is
>valid in any case. If you have to use the regulator when you don't have
>3.3v available and you can get away with the extra heat dissipation, then
>why do you need to switch it out when you do have a 3.3v supply available?
>You will have already designed the board to handle the higher dissipation,
>so what will the lower power accomplish, longer battery life?  ;)

If you have a high performance product that is already on the edge,
heat dissipation wise, then the extra heat of using a regulator (even
"only" 1.8 Watts) could cause chip operation to be derated (e.g. run at
a lower clock frequency). In general, you will get longer component
life and more reliable performance if you can run at a cooler temperature.

And if you don't think 1.8 Watts is hot, try putting your finger on it
for a few seconds (OUCH!).

Ivor

>Rick Collins
>
>rick.collins@arius.com
>
>
>Arius - A Signal Processing Solutions Company
>Specializing in DSP and FPGA design
>
>Arius                         http://www.arius.com
>4 King Ave                    301-682-7772 Voice
>Frederick, MD 21701-3110      301-682-7666 FAX
>
>
>

-----------original thread------------------------------------------
oops, sorry, my mistake. 25W is for PCI, not PMC.
please excuse..
I think you are right, 7.5W total was the correct number.
All the more reason to use 3.3V instead of regulating 5,
if at all possible (why oh why wasn't 3.3V mandatory?).
Ivor

At 01:42 PM 3/28/00 -0700, you wrote:
>Where are you getting the "25W max" number?  I believe an earlier version of
>the PMC/CMC spec called out a maximum of 7.5W, but this requirement was
>removed.  25W on a PMC could be difficult to cool at all and could add up to
>a lot of heat in a system using several PMCs, especially if they are mounted
>on intelligent baseboards that are already dissipating considerable heat.
>(This is particularly true if you ever have to convert the PMC to a
>conduction cooled version.)
>
>John
>
>> -----Original Message-----
>> From:	Ivor Bowden [SMTP:ivor@peritek.com]
>> Sent:	Tuesday, March 28, 2000 12:19 PM
>> To:	pci-sig@znyx.com
>> Subject:	RE: How to power a PCI card from either 3.3V OR 5V..
>> 
>> The same reason it was stupid to not require 3.3V supply on PMC
>> slots in the first place, 3.3 Volt regulators dump a lot of heat.
>> Like, if you need 5A @ 3.3V (16.5W, well less than the 25W max)
>> then the regulator will be dissipating 8.5W, then the total is
>> 25W. If the 3.3V can be used instead of regulating the 5V down
>> that would be a tremendous thermal savings, increasing the range
>> of operation and reliability greatly. Putting a switch on a PMC
>> board to use 3.3V if available makes sense. Yes, it is a hassle
>> and can have sequencing, etc. problems. Probably easier just to
>> use a switching regulator, which still wastes heat, takes up
>> board space, and makes noise, but given that one can not rely on
>> 3.3V supply on PMC slots, the 3.3V pins are pretty much worthless
>> (except for AC bypassing ;-)
>> 
>> Ivor
>> 
>> At 10:19 AM 3/28/00 -0500, you wrote:
>> >>  On the board there is going to be a 5V to 3.3V regulator, but if the
>> >>host is providing 3.3V power in the system, then I want to turn off the
>> >>output of the 5V-3.3V regulator and power the board from the hosts 3.3V
>> >>power.
>> >
>> >Why not just use your on-board regulator all the time?
>> >
>> >There is some advantage to using a common supply if your card will be in
>> the
>> >3.3V signaling environment, since both output and input threshold levels
>> >track the supply voltage in that environment.  However, not in the 5V
>> >signaling environment.
>> >
>> >A common supply also avoids power sequencing problems, but if your card
>> is
>> >going to sense and switch supply voltages, it may have the same problems
>> >anyway.
>> >
>> >Regards,
>> >Andy
>> >
>> >
>> >
>
>
>