On 11/13/2013 7:34 PM, luke.leighton wrote:

On Wed, Nov 13, 2013 at 11:16 PM, luke.leighton luke.leighton@gmail.com wrote:

...

http://rhombus-tech.net/community_ideas/router/news/

uh... am i reading this right - that this router design could be up to 20 watts?? 4x USB2 @ 500mA (10W), 1x RTL8366 (3W) and an EOMA68 CPU Card (4W). That's.... a hell of a lot.

anyone any experience at dealing with that much power on small boards (4 amp @ 5V) ?

LM1084? any guidance on PCB layout anyone aware of?

l.

---

arm-netbook mailing list arm-netbook@lists.phcomp.co.uk http://lists.phcomp.co.uk/mailman/listinfo/arm-netbook Send large attachments to arm-netbook@files.phcomp.co.uk

Hey guys... why not just use a 5v 4a external laptop-style brick...? Specify a 5.5x2.1mm barrel plug for the input connector, slap a jack down on the MEB (or whatever other carrier board we want) and you're good to go. Replacements, at least in the US, are $10ish on eBay when the time comes ;) Guessing the factory direct option would be even cheaper, so the initial cost is a brick and a barrel jack... not too much, then!

On 11/13/2013 8:10 PM, luke.leighton wrote:

On Thu, Nov 14, 2013 at 12:54 AM, Christopher Havel laserhawk64@gmail.com wrote:

...

On 11/13/2013 7:34 PM, luke.leighton wrote:

...

On Wed, Nov 13, 2013 at 11:16 PM, luke.leighton luke.leighton@gmail.com wrote:

...

http://rhombus-tech.net/community_ideas/router/news/

uh... am i reading this right - that this router design could be up to 20 watts?? 4x USB2 @ 500mA (10W), 1x RTL8366 (3W) and an EOMA68 CPU Card (4W). That's.... a hell of a lot.

anyone any experience at dealing with that much power on small boards (4 amp @ 5V) ?

LM1084? any guidance on PCB layout anyone aware of?

l.

---

arm-netbook mailing list arm-netbook@lists.phcomp.co.uk http://lists.phcomp.co.uk/mailman/listinfo/arm-netbook Send large attachments to arm-netbook@files.phcomp.co.uk

Hey guys... why not just use a 5v 4a external laptop-style brick...?

still have to have a power regulator chris. yes definitely the 5.5-2.1mm jack, the smaller connector i was thinking of won't take the current.

l.

---

arm-netbook mailing list arm-netbook@lists.phcomp.co.uk http://lists.phcomp.co.uk/mailman/listinfo/arm-netbook Send large attachments to arm-netbook@files.phcomp.co.uk

Thought those supplies were regulated already...? (insert confusion here :P )

On 11/13/2013 8:31 PM, luke.leighton wrote:

absolutely irrelevant. cannot be trusted, cannot provide stable current. wires too long. and they actually supply 5.5V.. and, what happens if some idiot plugs in a 6V PSU? what happens if there's a small spike? so... no.

That makes sense. I should know better than to trust eBay crunk to be exactly what it says on the tin, considering how much of it is in my house :P

btw can you please cut irrelevant text, read up on mailing list etiquette, at least you're not top-posting but it's causing me to spend considerable time each and every time replying to you by having to cut out a whole stack of crap. the more time you force me to spend doing that the less time i will have to reply to you. l.

Is this better...? I found this FAQ particularly illuminating --> https://www.freebsd.org/doc/en/articles/mailing-list-faq/etiquette.html

...and many profound apologies, to the entire list, for not looking at that sooner!

I should probably fess up at this point -- I have Asperger's Syndrome, an Autism Spectrum Disorder. (Hey, at least I'm at the high-functioning end of things....) As a result, at 27 years of age, I'm not yet quite fluent in proper manners, and rather a bit obtuse when it comes to the subject... I'm trying to learn, it's just one of a few things that's extremely hard for me. But I try to be a nice, kind fellow, so when I goof I'm more than eager to make up for my mistake. I'll try to be more careful (and considerate) in the future, for sure!

...

I should probably fess up at this point -- I have Asperger's Syndrome,

you and me both. yaay!

Lucky!!

I got short changed, and all I suffer from is ass berger syndrome.

"luke.leighton" luke.leighton@gmail.com writes:

On Thu, Nov 14, 2013 at 12:54 AM, Christopher Havel laserhawk64@gmail.com wrote:

...

On 11/13/2013 7:34 PM, luke.leighton wrote:

...

On Wed, Nov 13, 2013 at 11:16 PM, luke.leighton luke.leighton@gmail.com wrote:

...

http://rhombus-tech.net/community_ideas/router/news/

uh... am i reading this right - that this router design could be up to 20 watts?? 4x USB2 @ 500mA (10W), 1x RTL8366 (3W) and an EOMA68 CPU Card (4W). That's.... a hell of a lot.

anyone any experience at dealing with that much power on small boards (4 amp @ 5V) ?

LM1084? any guidance on PCB layout anyone aware of?

l.

---

arm-netbook mailing list arm-netbook@lists.phcomp.co.uk http://lists.phcomp.co.uk/mailman/listinfo/arm-netbook Send large attachments to arm-netbook@files.phcomp.co.uk

Hey guys... why not just use a 5v 4a external laptop-style brick...?

still have to have a power regulator chris. yes definitely the 5.5-2.1mm jack, the smaller connector i was thinking of won't take the current.

Possibly a foolish thought, but how about allowing for multiple power supplies, preferably with some way of detecting which one broke so one could then send an email about it before the other one packs up?

Also, or perhaps instead, PoE? (although you'll struggle to get enough power through that -- 802.3at (Type 2) is able to deliver up to 22.5W)

Cheers, Phil.

On Wed, Nov 13, 2013 at 6:34 PM, luke.leighton luke.leighton@gmail.comwrote:

On Wed, Nov 13, 2013 at 11:16 PM, luke.leighton luke.leighton@gmail.com wrote:

...

http://rhombus-tech.net/community_ideas/router/news/

uh... am i reading this right - that this router design could be up to 20 watts?? 4x USB2 @ 500mA (10W), 1x RTL8366 (3W) and an EOMA68 CPU Card (4W). That's.... a hell of a lot.

anyone any experience at dealing with that much power on small boards (4 amp @ 5V) ?

LM1084? any guidance on PCB layout anyone aware of?

l.

Well, this will obviously be a 4 layer board, but I can say with the MEBv2, we had tremendous problems with thermal dissipation of the LM2576 at 9-12vdc, at 6vdc is was manageable, but the MEBv2++ are 2 layer boards, and with the addition of a secondary power plane, it would probably have been fine. For the most recent versions, we are using an AP6503a, and so far at 6, 9,12 vdc the heat is non-existent. Not to mention, the local price of the AP6503 was .60c cheaper than the LM2576, and is 85-95% more heat efficient.

On 11/13/2013 08:41 PM, Christopher Thomas wrote:

On Wed, Nov 13, 2013 at 6:34 PM, luke.leighton
<luke.leighton@gmail.com <mailto:luke.leighton@gmail.com>> wrote:
     > anyone any experience at dealing with that much power on
    small boards
     > (4 amp @ 5V) ?

      LM1084?  any guidance on PCB layout anyone aware of?
Well, this will obviously be a 4 layer board, but I can say with the
MEBv2, we had tremendous problems with thermal dissipation of the
LM2576 at 9-12vdc, at 6vdc is was manageable, but the MEBv2++ are 2
layer boards, and with the addition of a secondary power plane, it
would probably have been fine. For the most recent versions, we are
using an AP6503a, and so far at 6, 9,12 vdc the heat is
non-existent. Not to mention, the local price of the AP6503 was .60c
cheaper than the LM2576, and is 85-95% more heat efficient.

My point in mentioning the LM2576, is that they are functionally similar to the LM1084, and both Linear.

CUI does board mount switching regulators that generate a lot less heat.

http://www.cui.com/product/power/dc-dc-converters/isolated/board-mount/20-w/...

PYB20-Q24-S5 is an isolated supply that will output 5V at 4A. But that digikey pricing suggests their not economical yet.

http://www.digikey.com/product-search/en/power-supplies-board-mount/dc-dc-co...

Christopher Thomas <christopher <at> firemothindustries.com> writes:

PYB20-Q24-S5 is an isolated supply that will output 5V at 4A. But that digikey pricing suggests their not economical

yet.http://www.digikey.com/product-search/en/power-supplies-board-mount/dc-dc-co...

While not the best for the job, it's certainly appealing at .22c versus

$39. But taps out at 3A. if there's a way to cull some wattage from the 20w point.

What about running the voltage regulators in parallel? 2 * 3A = 5.5A, or so.

http://electronicdesign.com/power/double-your-output-current-parallel-voltag...

On Thu, Nov 14, 2013 at 3:22 AM, Christopher Thomas christopher@firemothindustries.com wrote:

While not the best for the job, it's certainly appealing at .22c versus $39. But taps out at 3A. if there's a way to cull some wattage from the 20w point.

http://www.digikey.com/scripts/DKSearch/dksus.dll?Detail&itemSeq=1389362...

that's a 404 not found chris.

anyway pending no complete howlers i'll probably go for the TPS56428 as it's $1.10 in 1k volumes (TDK inductor is $1 as well, eek!) - the TPS56428 looks like quite an awesome and well-designed bit of kit, no completely mad thermal thingies like the LM1084, good advice from TI in the datasheet so... yeah.

l.

On Thu, Nov 14, 2013 at 9:34 PM, Christopher Thomas christopher@firemothindustries.com wrote:

On Thu, Nov 14, 2013 at 3:17 PM, luke.leighton luke.leighton@gmail.com wrote:

...

On Thu, Nov 14, 2013 at 3:22 AM, Christopher Thomas christopher@firemothindustries.com wrote:

...

While not the best for the job, it's certainly appealing at .22c versus $39. But taps out at 3A. if there's a way to cull some wattage from the 20w point.

http://www.digikey.com/scripts/DKSearch/dksus.dll?Detail&itemSeq=1389362...

that's a 404 not found chris.

When will I learn to stop posting links from my Shopping Cart?

:)

Here's the correct link. http://www.digikey.com/product-detail/en/AP6503SP-13/AP6503SP-13DITR-ND/2767...

oo, oo - found one: http://www.digikey.com/product-detail/en/RT8288AZSP/1028-1148-2-ND/3078371

i like that one. much lower-cost than the TI part. about the same as the one you found chris. good datasheet, good advice on layout. yeah.

l.

right - bit of a marathon session last night and most of the routing is done. nothing like as complex as the flying squirrel - no embedded controller, plenty of space (double-sided), it's pretty much a breeze.

questions:

1) joe as this is GPLv3+ open hardware does your offer to PCB-print and populate it still stand, even though i've had to use PADS? (i don't have the expertise yet to not not use KiCAD because it lacks comprehensive design rules).

2) is there anyone who would like to have a go at ordering a rev1 board just for the sheer heck of it?

3) if not, and there is something missing, what would encourage you to splash out on a completely experimental PCB in the way of functions or interfaces? (yes i have to look up POE, see if i can fit it on somewhere).

l.

On Sat, Nov 16, 2013 at 10:38 PM, Scott Sullivan scott@ss.org wrote:

On 11/16/2013 08:29 AM, luke.leighton wrote:

...

3. if not, and there is something missing, what would encourage you to

splash out on a completely experimental PCB in the way of functions or interfaces? (yes i have to look up POE, see if i can fit it on somewhere).

So, the option I've seen taken in many projects is the use of Silvertel's ready made PoE modules.

http://www.silvertel.com/poe_products.htm

ah. awesome.

PoE seems to be a popular request, but is not cost effective for mass production. Currently the Arduino Ethernet addresses this by designing the board to forward up-gradable with one of these modules. This can also make sense from manufacturing and sales, use the same main PCB, but just pre-fit a smaller run of the boards with the module.

http://arduino.cc/en/Main/ArduinoBoardEthernet

ok. that's a huge help - not least because they'll have schematics available.

thanks scott. i'll put that in.

l.

On Sun, Nov 17, 2013 at 11:12 AM, luke.leighton luke.leighton@gmail.com wrote:

On Sun, Nov 17, 2013 at 10:21 AM, luke.leighton luke.leighton@gmail.com wrote:

...

On Sat, Nov 16, 2013 at 10:38 PM, Scott Sullivan scott@ss.org wrote:

...

...

http://www.silvertel.com/poe_products.htm

...

...

http://arduino.cc/en/Main/ArduinoBoardEthernet

ok. that's a huge help - not least because they'll have schematics available.

thanks scott. i'll put that in.

hmmm... it's a 10/100 PoE circuit (and module). need to investigate and/or cut down one of the GbE interfaces to only 10/100... hmmm..

http://www.silvertel.com/AppsNotes/ANX-1000BASE-T-CONNECTIONS-v1-1.pdf

ah ha! o poo. have to use a separate magnetics module. which solves a problem of lack of datasheets on 1x4 RJ45 modules with built-in magnetics, but still poo :)

l.

ok i went for the Ag5300 in the end. the choices are 12W or 30W. 12W is definitely not enough, ruling out the 9000 series. but, 8-211at needs mode-selection (to a micro-processor). there's not enough spare pins, so.. y'know what? sod it. 15W is good enough for POE input. meaning that the Ag5300 won't be pushed hard, overspec'd components means it won't get massively hot.

although i didn't increase the PCB size too much to get this all on i did need to make the connectors overhang somewhat - that's ok too. i pub on some drill-holes as well, might as well, and made them plated so they're earth-points as well. one of them's running through the POE GND heatsink area which is good news, too.

in the end also i went with a TDK magnetics 1-in, 1-out module and put 24 (!!) of them down. they're very small. the advantage is: i could shuffle them around until i found a good combination for placement.

all very odd, but hey it's a good spec board now. power input voltage anywhere between 6 and 21V. POE option which can *output* 12V via the power connector (i wired them straight up. maybe i shouldn't do that - use diodes, eh?). found that app note which shows how to do gigabit PoE...

... whew :)

On Sun, Nov 17, 2013 at 1:14 AM, joem joem@martindale-electric.co.uk wrote:

right - bit of a marathon session last night and most of the routing is done. nothing like as complex as the flying squirrel - no embedded controller, plenty of space (double-sided), it's pretty much a breeze.

questions:

1. joe as this is GPLv3+ open hardware does your offer to PCB-print

and populate it still stand, even though i've had to use PADS? (i don't have the expertise yet to not not use KiCAD because it lacks comprehensive design rules).

2. is there anyone who would like to have a go at ordering a rev1

board just for the sheer heck of it?

3. if not, and there is something missing, what would encourage you to

splash out on a completely experimental PCB in the way of functions or interfaces? (yes i have to look up POE, see if i can fit it on somewhere).

-- If its open, then upload somewhere and post link with all the datasheets,

all that has been available since the beginning of the project. i've mentioned them in almost every single post.

cad files

these have also been available since the beginning of the project.

and some PDFs of the board and circuits

these have also been available since the beginning of the project.

• if it looks

doable, I'll take a shot at it in parallel with any others.

everything has always, always, always been here, mentioned in almost 100% of all posts i've made on this router project: http://hands.com/~lkcl/eoma/router/

Most likely a kicad version will get drawn

there is already a kicad version. it has always been here: http://git.hands.com/?p=eoma.git;a=tree;f=pcb/router

and laid out

... and that's the bit that i don't have the expertise to do (PCB layout without decent and comprehensive design rules) hence why i use PADS. KiCAD doesn't even have differential-pair layout so you have to do this odd "thing" of creating a part which has two pins at the exact width you want, then dragging that around and deleting it once you're done. of course, if you get the impedance wrong you have to start again, but given that KiCAD doesn't have impedance calculation *at all* let alone impedance design rules you can't even check you got it right in KiCAD... hence why i'm using PADs.

as well to try and get it all into kicad format so that everyone can use it.

if i can find out how to use PADS's COM interface i might at some point have a go at automatic conversion. KiCAD's _pcbnew.so is fairly straightforward to use (albeit highly undocumented and missing some key functionality). at least that would save vast amounts of peoples' time, and if i made it a two-way converter it would be possible to use PADS autorouter then import back. which would be cool.

l.

...

If its open, then upload somewhere and post link with all the datasheets,

all that has been available since the beginning of the project. i've mentioned them in almost every single post.

Thanks - what I really meant is I got too many links and difficult to sift through to get to correct information in a timely manner.

everything has always, always, always been here, mentioned in almost 100% of all posts i've made on this router project: http://hands.com/~lkcl/eoma/router/

A lot of the important files there are not accessible - no permission.

...

Most likely a kicad version will get drawn

there is already a kicad version. it has always been here: http://git.hands.com/?p=eoma.git;a=tree;f=pcb/router

...

and laid out

... and that's the bit that i don't have the expertise to do (PCB layout without decent and comprehensive design rules) hence why i use PADS. KiCAD doesn't even have differential-pair layout so you have to do this odd "thing" of creating a part which has two pins at the exact width you want, then dragging that around and deleting it once you're done. of course, if you get the impedance wrong you have to start again, but given that KiCAD doesn't have impedance calculation *at all* let alone impedance design rules you can't even check you got it right in KiCAD... hence why i'm using PADs.

I gave up on proprietary a while back... It has damaged too many projects of mine to ever want to go back.

...

as well to try and get it all into kicad format so that everyone can use it.

if i can find out how to use PADS's COM interface i might at some point have a go at automatic conversion. KiCAD's _pcbnew.so is fairly straightforward to use (albeit highly undocumented and missing some key functionality). at least that would save vast amounts of peoples' time, and if i made it a two-way converter it would be possible to use PADS autorouter then import back. which would be cool.

KiCAD does support external routers.

In other news: Christmas coming up, and things start winding down. If board needs to be built more urgently, need to try parallel paths. I'm looking at other side of Feb to get this board finished.

...

http://hands.com/~lkcl/eoma/router/

A lot of the important files there are not accessible - no permission.

| quick chmod: sorted.

Thanks

I gave up on proprietary a while back... It has damaged too many projects of mine to ever want to go back.

| ... by contrast i simply don't have the required knowledge to use |tools - any tools - that don't have extensive built-in knowledge, and |i also don't have the requirement that you have to support customers |for 20+ years right down to the source code level of the tools that |were used to make the hardware.

Not religious about proprietary - certain things just can't be trusted to proprietary - hence positive action to eliminate proprietary where trust is broken. PCBs in particular - I have need of them years down the road for pasting on to new designs. If the data is not going to be available, then I can't copy paste, so I must do it again. Pointless delays damage projects. e.g. power supplies - I must have laid out the same power supply designs umpteenth times in different packages. Could they get together and agree to copy paste between packages to save the user and their projects? Don't think so. Therefore they are totally untrustworthy and should not be trusted with any PCB data.

On Wed, 2013-11-13 at 19:38 -0600, Christopher Thomas wrote:

we had tremendous problems with thermal dissipation

In the past I have made PIC product with half a dozen switch mode power supplies, charger and fast charger all rolled into one device that needs so many different power supplies for different functions.

The one PIC chip was operating all the power supplies through software rather than hardware.

PIC and ARM Cortex chips are fast enough these days to to generate the signals on the fly in software so that it is possible to get 1.2V, 1.25V, 1.8V, 3.3V, 5V, trickle charger and fast charger functions all into one chip with an optically isolated RS232 or an I2C comms link and built in EEPROM to control internal functions. It should universally wipe out the numerous power solutions that EEs are doing up for their boards and replace it with just the one that is GPL'd, cheap and free to use, and supported by a Linux driver.

May be wrong mailing list to talk up such things, or may be a new mailing list needed to talk up all the infrastructure projects that go into making EOMAs, Cubies, Olixinos, netbooks, IoT devices because these are infrastructure projects with common benefit. The infrastructure projects are fully GPL'd and generate a lot of hard to keep up noise (EE noise and embedded software coding noise) whilst being sorted out.

On Thu, 2013-11-14 at 12:32 +0100, Aaron J. Seigo wrote:

On Thursday, November 14, 2013 09:19:34 joem wrote:

...

It should universally wipe out the numerous power solutions that EEs are doing up for their boards and replace it with just the one that is GPL'd, cheap and free to use, and supported by a Linux driver.

yes, please. :)

...

May be wrong mailing list to talk up such things, or may be a new mailing list needed to talk up all the

imho it needs more than just a mailing list (here or elsewhere). it really calls for a cpan / github for hardware type thing where these ideas can be pooled, shared and publicized together.

what that would look like would need to be driven by the workflow it should enable. this is really the kind of thing that is so much easier / most pleasant / faster to do with a small-ish group of involved people in a room together with a whiteboard ...

Hi Aaron,

I'm least qualified to make comments on how the collaborative workflow will get managed.

There is a whole bunch of fully GPL'd tough nuts to be cracked involving both hardware, test results (which involve a lot of DSO traces), maths, graphing, Linux driver software, embedded controllers and their software. Not everything can be successful, and some things will drag on for years while the issues are identified and sorted. No information can be hidden from view while such work is in progress.

To solve all this for my work environment, over the years I do many things and finally I found HTML.

I work with everything in HTML format so that I get access to EVERYTHING within 3 or 4 clicks. So if its a graph, circuit diagram, scope trace, highlighted code fragment, CAD drawings, calibration procedures, relevant bits of datasheets, photos, video, animations, MP3, etc, its ALWAYS converted into HTML format and either GIF, JPEGs, video or MP3, so that I don't need to start up secondary applications like pdfviewer to view the relevant information. No historical information is hidden from view by more than 3 or 4 clicks.

When life gets more complex with extreme detail, I opt to pump everything through the speech synthesiser / verbose project manager described at http://www.gplsquared.com/SoM2/SoM2.html to march the project and its issues at a fast pace by clicking on the speech buttons until each item coming out through the verbose project manager speech synthesiser is drilled into your head and you do something to address every issue. This can happen for example when there is software and hardware bugs that interact together making it extremely difficult to get to the bottom of an issue without 'something' talking through each and every issue, and repeating it patiently until something makes sense.

How all that can work in distributed collaborative environment without sucking up a lot of time managing it all, I would not begin to comprehend.