Forums

CPU fan buck converter, miniature

Started by Winfield Hill January 26, 2020
https://www.dropbox.com/s/wttdn9r8np3o0t6/buck_CPU-fan.JPG?dl=0

 I had finished a new PCB version of my compact
 250A analog pulser, with a newly-added CPU fan.
 The fan was powered with an 78L12, hey, there
 wasn't room for much else.  But it was obvious
 its 300mA would overheat the TO-92 regulator,
 plus what a waste of power!  I needed to fit a
 buck converter.  This called for a SOT-23 part;
 no room for a SO-8, and SON or LFCSP packages
 not allowed, too hard to solder.  There are lots
 of low-voltage candidates, but RIS-796A is used
 up to at least 40 volts.  Only one part fits the
 bill: LM2842Y.  And it runs at 1.25MHz, allowing
 for a small inductor.  The PCB circuit is only
 0.4 x 0.5 inches, fits on a corner of the PCB.
 (The AOZ1282CI is a close alternate 36V 450kHz.)


-- 
 Thanks,
    - Win
Winfield Hill wrote...
> >https://www.dropbox.com/s/wttdn9r8np3o0t6/buck_CPU-fan.JPG?dl=0 > > I had finished a new PCB version of my compact > 250A analog pulser, with a newly-added CPU fan. > The fan was powered with an 78L12, hey, there > wasn't room for much else. But it was obvious > its 300mA would overheat the TO-92 regulator, > plus what a waste of power! I needed to fit a > buck converter. This called for a SOT-23 part; > no room for a SO-8, and SON or LFCSP packages > not allowed, too hard to solder. There are lots > of low-voltage candidates, but RIS-796A is used > up to at least 40 volts. Only one part fits the > bill: LM2842Y. And it runs at 1.25MHz, allowing > for a small inductor. The PCB circuit is only > 0.4 x 0.5 inches, fits on a corner of the PCB. > (The AOZ1282CI is a close alternate 36V 450kHz.)
Here's why this is a miniature buck converter. It's in the little green circle at the lower right corner, of a fairly-small 3x3-inch PCB. https://www.dropbox.com/s/iqxegp4he8d6ast/buck_in-796A.JPG?dl=0 Save yourself a copy, you might need it oneday. I imagine it's got about 5W of capability. -- Thanks, - Win
On 26 Jan 2020 05:44:44 -0800, Winfield Hill <winfieldhill@yahoo.com>
wrote:

>https://www.dropbox.com/s/wttdn9r8np3o0t6/buck_CPU-fan.JPG?dl=0 > > I had finished a new PCB version of my compact > 250A analog pulser, with a newly-added CPU fan. > The fan was powered with an 78L12, hey, there > wasn't room for much else. But it was obvious > its 300mA would overheat the TO-92 regulator, > plus what a waste of power! I needed to fit a > buck converter. This called for a SOT-23 part; > no room for a SO-8, and SON or LFCSP packages > not allowed, too hard to solder. There are lots > of low-voltage candidates, but RIS-796A is used > up to at least 40 volts. Only one part fits the > bill: LM2842Y. And it runs at 1.25MHz, allowing > for a small inductor. The PCB circuit is only > 0.4 x 0.5 inches, fits on a corner of the PCB. > (The AOZ1282CI is a close alternate 36V 450kHz.)
We did one box recently that has a temperature sensor on the main PCB, which was a good idea anyhow, so now the uP can control fan speed. That is less noisy usually, and to some extent reduces the delay tempcos. The uP output is PWM, lowpass filtered into an LM317. The algorithm is a simple slow up/down counter based on temp high/low compared to the setpoint. That's simple and stable and has no acoustic drama. On this one, each amplifier board has its own fan, plus one for the box. All run full blast. https://www.dropbox.com/s/5nwlbkep2y97baq/3d_4.jpg?raw=1 On your little board, in one place two traces make a V where they hit a pad. One of our high-PITA customers wants us to change a board layout to eliminate any not-90-degree pad entries, mumbling something about "acid traps", which I think may have happened a few times around 1960. -- John Larkin Highland Technology, Inc The cork popped merrily, and Lord Peter rose to his feet. "Bunter", he said, "I give you a toast. The triumph of Instinct over Reason"
On 2020-01-26 08:44, Winfield Hill wrote:
> https://www.dropbox.com/s/wttdn9r8np3o0t6/buck_CPU-fan.JPG?dl=0 > > I had finished a new PCB version of my compact > 250A analog pulser, with a newly-added CPU fan. > The fan was powered with an 78L12, hey, there > wasn't room for much else. But it was obvious > its 300mA would overheat the TO-92 regulator, > plus what a waste of power! I needed to fit a > buck converter. This called for a SOT-23 part; > no room for a SO-8, and SON or LFCSP packages > not allowed, too hard to solder. There are lots > of low-voltage candidates, but RIS-796A is used > up to at least 40 volts. Only one part fits the > bill: LM2842Y. And it runs at 1.25MHz, allowing > for a small inductor. The PCB circuit is only > 0.4 x 0.5 inches, fits on a corner of the PCB. > (The AOZ1282CI is a close alternate 36V 450kHz.) > >
I use a fair number of those AOZ parts--they're cheap and they just work. There's also the AOZ1282CI-1, which gives you half the current at twice the frequency, but is hard to get. Cheers Phil Hobbs -- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC / Hobbs ElectroOptics Optics, Electro-optics, Photonics, Analog Electronics Briarcliff Manor NY 10510 http://electrooptical.net http://hobbs-eo.com
Phil Hobbs wrote...
> > I use a fair number of those AOZ parts--they're cheap > and they just work.
Alpha & Omega makes lots of interesting parts. And they seem determined to sell them at very low prices.
> There's also the AOZ1282CI-1, which gives you half > the current at twice the frequency, but is hard to get.
They want you to purchase full reels? Twice the frequency, that would have put it in good contention with the LM2842Y part. But the extra 6V also helped in this case. Most of the 20V parts are synchronous, save the space of a sod-123 diode. -- Thanks, - Win
jlarkin@highlandsniptechnology.com wrote...
> > On your little board, in one place two traces make a > V where they hit a pad. One of our high-PITA customers > wants us to change a board layout to eliminate any > not-90-degree pad entries, mumbling something about > "acid traps", which I think may have happened a few > times around 1960.
Do you follow the principle on your layouts? Or only for that customer? Most of the time a 90-degree rule isn't a problem. But there are times I prefer a mecca approach, where all traces must meet only at the pads, despite whatever room around the pad is available. A few years ago I starting worrying about flaring any narrow traces as they reached a large pad, to reduce the possibility of trace breakage or lifting. If you start worrying about outlier cases, is there any end? -- Thanks, - Win
On Sunday, January 26, 2020 at 9:16:56 PM UTC-5, Winfield Hill wrote:
> jlarkin@highlandsniptechnology.com wrote... > > > > On your little board, in one place two traces make a > > V where they hit a pad. One of our high-PITA customers > > wants us to change a board layout to eliminate any > > not-90-degree pad entries, mumbling something about > > "acid traps", which I think may have happened a few > > times around 1960. > > Do you follow the principle on your layouts? Or only > for that customer? Most of the time a 90-degree rule > isn't a problem. But there are times I prefer a mecca > approach, where all traces must meet only at the pads, > despite whatever room around the pad is available.
A 90 degree entry rule is not the right solution. The right rule would be to have no internal corners less than 90 degrees. I often use traces entering at a corner which the 90 degree entry rule would exclude even though it has larger than 90 degree internal corners. I have no idea if acid traps are a real issue or not, but this is easy enough to deal within general, so I design as if it were a real issue. It would be even easier if the software had a design rule to check for that, which mine doesn't. Manual inspection is not so easy when your boards have hundreds of pads.
> A few years ago I starting worrying about flaring any > narrow traces as they reached a large pad, to reduce > the possibility of trace breakage or lifting. If you > start worrying about outlier cases, is there any end?
I don't typically worry with that, but then my traces are typically not much smaller than the pad. Sometimes I have to reduce a trace width to match the pad. Pad for digital components are often 0.5 or even 0.4 mm pitch. The pad is less than 8 mils, so a 6 mil trace matches pretty well. -- Rick C. - Get 1,000 miles of free Supercharging - Tesla referral code - https://ts.la/richard11209
On 26 Jan 2020 18:20:23 -0800, Winfield Hill <winfieldhill@yahoo.com>
wrote:

>Winfield Hill wrote... >> >> A few years ago I starting worrying about flaring any >> narrow traces as they reached a large pad, to reduce >> the possibility of trace breakage or lifting. If you >> start worrying about outlier cases, is there any end? > > This is an example. I manually added the flaring. > >https://www.dropbox.com/s/x6a82iq0nz218w2/flared-traces.JPG?dl=0
PADS will do teardrops automatically, but I think they're ugly so don't do them on my boards. We very rarely have a board problem. We buy from known good vendors and specify bare-board testing. We have had some bad cheapie-quick-proto Chinese boards. -- John Larkin Highland Technology, Inc The cork popped merrily, and Lord Peter rose to his feet. "Bunter", he said, "I give you a toast. The triumph of Instinct over Reason"
Rick C wrote...
> > A 90 degree entry rule is not the right solution. > The right rule would be to have no internal corners > less than 90 degrees.
I'm confused, no 45-degree direction changes in traces? -- Thanks, - Win
On 27/01/2020 09:50, Winfield Hill wrote:
> Rick C wrote... >> >> A 90 degree entry rule is not the right solution. >> The right rule would be to have no internal corners >> less than 90 degrees. > > I'm confused, no 45-degree direction changes in traces? > >
Perhaps they'd permit 135 degree instead :) piglet