Power-down emergency power source.

[tALSit de CoD]

G'day people of wisdom!

I have a project that writes to an SD card, and is externally powered. It's based on the STM32F4, so the current draw is about 200mA.

Since it's externally powered and I don't trust for a second that the users will do a proper shut down, I have the problem that my SD card will get corrupted by getting powered off mid - write.

So I was thinking of having an "electron reservoir" somewhere in the power path.

My current set up is thus:

Vin (8-34V) -> Switch Mode Power Supply (SMPS) -> Diode (BAT4JFILM) -> Linear Voltage Regulator (LDO).

The SMPS outputs 5V, the linear 3.3V.

I was thinking of putting a largish capacitor after the diode, where the voltage is about 4.5V (due to the drop of the diode under load). I monitor the input voltage before the SMPS, so I can find out when the thing is unplugged.

I've done a basic simulation and a 1000uF capacitor will give me about 5ms response time, which should be enough for me to issue an fclose() command.

I have 2 problems with this set up: massive inrush current, about 5A, which my SMPS won't be able to supply, which may work out - the capacitor will charge up more slowly, but the SMPS may not like it.

The other problem is that this large capacity would probably have to be. An electrolytic, which I'm trying to avoid due to the exploding nature :-)

I have looked at super cap, but their high ESR rules then out. Also, having a tiny LiPo is also ruled out due to the complexity (and, again, fire).

Anyone have any ideas / suggestions?

[Andrew]

Tap off the SMPS via a 10ohm resistor to charge the capacitor (deals with inrush current).

 

Connect the capacitor via a second diode to the LDO - the two diodes give power supply OR'ing.

--
You received this message because you are subscribed to the Google Groups "Robots & Dinosaurs" group.
To unsubscribe from this group and stop receiving emails from it, send an email to sydney-hackspa...@googlegroups.com.
To post to this group, send email to sydney-h...@googlegroups.com.
Visit this group at http://groups.google.com/group/sydney-hackspace?hl=en.
For more options, visit https://groups.google.com/groups/opt_out.
 
 

[Arik Baratz]

One thing I've done in the past is, charge the capacitors through a diode and get a power-off signal from before the diode into the MCU (by adding a pull-down resistor), to a high priority interrupt. This way the MCU has enough time to shut down whatever it needs to do before power drains. I used it to trigger the low power consumption mode on mine, the one that halts the oscilator.

-- Arik

[Andrew]

Yeah, charging the cap via a diode + resistor is better - stops the capacitor discharging back through the power supply in shutdown.  Also lowers the capacitor voltage to guarantee that there is no current passing through the OR'ing diode.

 

The two paths are:

SMPS -> diode -> resistor -> cap -> diode -> LDO

SMPS -> diode -> LDO

[tALSit de CoD]

So that's TWO diodes in the path between the SMPS & the linear... Time to do some actual tests, me thinks!

But thanks for the advice. I am monitoring in input voltage, before the SMPS and my MCU has an analogue watchdog when its input falls below a set threshold, and it triggers and interrupt.

I may try that out...

[Kean Maizels]

One way to improve this is to replace one or more of the diodes with small P-MOSFETs – wired “backwards” so the internal diode is in the correct direction, and when active you get almost no voltage drop.

 

Some more info at http://www.eevblog.com/forum/projects/mosfet-as-blocking-diode-how/

 

Kean

[Luke Emrose]

"An electrolytic, which I'm trying to avoid due to the exploding nature :-)"

In my experience it's Tantalum's you want to watch out for.

Electros should be just fine, no? The only thing that will upset them is reverse voltage, other than that, you should be just fine.

Low ESR should mean more current required to charge. In fact, it's entirely possible that you WANT high esr for power supply filters. A few links I found online corroborate this theory.

i.e. http://tinyurl.com/bpnppw2

--

You received this message because you are subscribed to the Google Groups "Robots & Dinosaurs" group.
To unsubscribe from this group and stop receiving emails from it, send an email to sydney-hackspa...@googlegroups.com.
To post to this group, send email to sydney-h...@googlegroups.com.
Visit this group at http://groups.google.com/group/sydney-hackspace?hl=en.
For more options, visit https://groups.google.com/groups/opt_out.
 
 

--
Luke Emrose
aka evolutionary theory
www.evolutionarytheory.com
www.soundcloud.com/evolutionarytheory
www.reverbnation.com/evolutionarytheory
http://www.facebook.com/pages/evolutionary-theory/110717958952413

[and...@arcadius.com.au]

Yep, +1 on that! 

 

[Arik Baratz]

You could use a low voltage drop diode such as a Germanium or Schottky (0.3v and 0.2v respectively). Limiting the current to the capacitor is important so you won't exceed the potential current through it.

-- Arik

[Nick Johnson]

Why is the ESR of a supercap an issue? It oughtn't to be so high that you can't draw 200mA from it, should it?

-Nick

--

[tALSit de CoD]

The ESR is around 50Ohms or more... I would have thought that the voltage drop would have be significant to rule out, but I'll give it another look...

[tALSit de CoD]

From a simulation, I get this:

Though I don't know if I'm simulating it totally right. I just plugged the numbers into LTSpice and that's what it shows.

// talsit.org

[tALSit de CoD]

Though, I just did a search, and found some swanky new supercaps with sub-1Ohm ESR... Like this one, a 5.5V 16mF, with 500mOhm

http://www.mouser.com/ds/2/67/cellergy_spec-209.pdf

And if I plug that into LTSpice, it gives me over 10ms of response time before the voltage falls below 4V (I'm being pessimistic with my LDO).

I'll have a further look into modelling all the diode/p-ch mosfet and the rest in a bit.

// talsit.org

[Kean Maizels]

Attached are some quickly drawn up LTspice models – diode OR-ing vs. P-Ch MOSFET OR-ing.  Note the specific way the FETs are wired – the FETs internal body diode (not shown) is in the same direction as the normal diodes.

 

When the difference in voltage between the two supplies is more than the gate threshold the P-FET turns on, otherwise you get the diode drop of the internal body diode.  MOSFET selection will need to be based on review of Rds(on), Vgs(th), Vgs(max), and Vds(max).

 

Of course using comparators to drive the FETs would give better results.

 

 

Kean

 

From: sydney-h...@googlegroups.com [mailto:sydney-h...@googlegroups.com] On Behalf Of tALSit de CoD
Sent: Monday, 15 April 2013 7:41 PM
To: Robots&Dinosaurs
Subject: Re: [RnD] Power-down emergency power source.

 

Though, I just did a search, and found some swanky new supercaps with sub-1Ohm ESR... Like this one, a 5.5V 16mF, with 500mOhm

[Arik Baratz]

On 16 April 2013 04:54, Kean Maizels <Ke...@kean.com.au> wrote:

Attached are some quickly drawn up LTspice models – diode OR-ing vs. P-Ch MOSFET OR-ing.  Note the specific way the FETs are wired – the FETs internal body diode (not shown) is in the same direction as the normal diodes.

 

When the difference in voltage between the two supplies is more than the gate threshold the P-FET turns on, otherwise you get the diode drop of the internal body diode.  MOSFET selection will need to be based on review of Rds(on), Vgs(th), Vgs(max), and Vds(max).

 

Of course using comparators to drive the FETs would give better results.

I don't like the look of the arrangement with the FETs. You can see that there is a strong current draw from your v2 source at just under 0.2s when both FETs are conducting. Graph the current and I bet there will be a spike there.

-- Arik

[ada]

On Apr 15, 11:32 am, Luke Emrose <evolutionarythe...@gmail.com> wrote:

> Low ESR should mean more current required to charge. In fact, it's entirely
> possible that you WANT high esr for power supply filters. A few links I
> found online corroborate this theory.

> i.e. *http://tinyurl.com/bpnppw2*<http://tinyurl.com/bpnppw2>

How exactly does that link corroborate this theory? Did you actually
read it?

(for those who engaged in a case of TL;DR: the ESR changes the RC
time constant and thus the complex frequency response of the output
filter. the suggested fix is putting another cap in the compensation
circuit if you can, for externally compensated SMPS chips.)

I don't see how high (as in, above 1ohm) ESR helps anyone.

[ada]

On Apr 16, 10:43 am, Arik Baratz <goo...@arik.baratz.org> wrote:

> I don't like the look of the arrangement with the FETs. You can see that
> there is a strong current draw from your v2 source at just under 0.2s when
> both FETs are conducting. Graph the current and I bet there will be a spike
> there.

I think that's stray capacitance rather than the FETs conducting. In
that region, both FETs are acting as diodes and M2 isn't turned on
yet.

You can see other capacitance effects at 0.5s in the diode graph where
the output is higher than both inputs - ie neither diode should be
conducting in that region.

[Luke Emrose]

You are right, it didn't show anything useful.
I clicked send too quickly, and thought it was something else I'd read the previous day.

My mistake.

Here is a simpler source:
http://www.electronics-tutorials.ws/capacitor/cap_3.html

And the relevant quote:
"Capacitors with a relatively high ESR have less ability to pass current to and from its plates to the external circuit because of their longer charging and discharging RC time constant."

--
You received this message because you are subscribed to the Google Groups "Robots & Dinosaurs" group.
To unsubscribe from this group and stop receiving emails from it, send an email to sydney-hackspa...@googlegroups.com.
To post to this group, send email to sydney-h...@googlegroups.com.
Visit this group at http://groups.google.com/group/sydney-hackspace?hl=en.
For more options, visit https://groups.google.com/groups/opt_out.

[ada]

On Apr 16, 2:54 pm, Luke Emrose <evolutionarythe...@gmail.com> wrote:
> And the relevant quote:
> "Capacitors with a relatively high ESR have less ability to pass current to
> and from its plates to the external circuit because of their longer
> charging and discharging RC time constant."

Why is this a good thing?

" Capacitors with very low ESR ratings are available and are best
suited when using the capacitor as a filter."

[Kean Maizels]

Arik,

 

There is no current spike.  The FETs won’t ever be conducting at the same time, but their body diodes will be.  The voltage sources are specified with 1R series resistance, and there is a 20R load, so you’re seeing the sources droop a little at those cross over points.

 

Kean

 

 

From: sydney-h...@googlegroups.com [mailto:sydney-h...@googlegroups.com] On Behalf Of Arik Baratz
Sent: Tuesday, 16 April 2013 10:43 AM
To: sydney-h...@googlegroups.com
Subject: Re: [RnD] Power-down emergency power source.

 

 

On 16 April 2013 04:54, Kean Maizels <Ke...@kean.com.au> wrote:

--

[Arik Baratz]

On 16 April 2013 16:08, Kean Maizels <Ke...@kean.com.au> wrote:

Arik,

 

There is no current spike.  The FETs won’t ever be conducting at the same time, but their body diodes will be.  The voltage sources are specified with 1R series resistance, and there is a 20R load, so you’re seeing the sources droop a little at those cross over points.

 

Hmm that makes more sense, now that I think about it.

-- Arik

[tALSit de CoD]

I've done a couple more simulations (*), and this seems to be something that would work. 

Note, R1 would be "my circuit", linear LDO and all.

Andrew, you said putting a third diode between the positive output of C1 and its path to R1. Do I really need it? The cap charges at about 60mA in about 1ms with that 50Ohm resistor, and then, when V1 is cut off, it supplies all its voltage nicely, right?

Now, the question remains, to use electrolytics or to avoid them at all costs? I keep hearing of doomsday scenarios with electrolytics, with them having a definite use life. At least, when they go, they don't short out, like tantalums. I can get a suitably high-reliability electrolytic for about $2, whereas a tantalum would be about $10, and the biggest ceramic I can find is about 100uF, so I would need 10 of them, making that venue not too good.

(*) Since it's been impossible for me to do some actual tests as Osaka, Japan, doesn't have any P-Channel MOSFETS sitting on shelves that weren't obsoleted 20 years ago...

Oh, and btw, thanks heaps for all great info!

// talsit.org

--

[tALSit de CoD]

Isn't ESR just a by - product of the manufacturing technique? As in, supercaps tend to have high ESR (~50Ohm), because they are made for long lasting (months), low current (uA's) supply to things like RTC, where low ESR isn't important. And electrolytic are typically used for very short stabilisation of high current, so their ESR has to be low or else they won't be able to supply the current?

That's my understanding on caps.

Now, in my case, a 1000uF electrolytic cap with 0.08Ohm ESR will draw around 5A until it charges, which is in microseconds, but still way too high. Something with more ESR would be better for charging, but then I get into the problem of discharge. If it has 50Ohm ESR, the voltage drop of the car's supply is non-good!

I'm sure that a 0Ohm (if possible) cap would cause all sorts of other issues too...

[Andrew]

tALSit, the problem is that, as shown, the circuit won't kill the inrush current you said you were worried about.  C1 will just charge via D1.  The additional diode prevents this but at a cost of voltage in the powering-down scenario, which will affect your run time.

 

[aaron...@exemail.com.au]

I recommend the top floor of Spa World in Osaka if your want to have some
non-techie fun :)

Also there was a small bakery near Osaka station that kept quite a few of
my Yen - but then there are nice bakeries just about everywhere in Japan.

Aaron.


> I've done a couple more simulations (*), and this *seems* to be something

> that would work.
> Note, R1 would be "my circuit", linear LDO and all.
> Andrew, you said putting a third diode between the positive output of C1
> and its path to R1. Do I really need it? The cap charges at about 60mA in
> about 1ms with that 50Ohm resistor, and then, when V1 is cut off, it
> supplies all its voltage nicely, right?
>
> [image: Inline images 1]
>
> Now, the question remains, to use electrolytics or to avoid them at all
> costs? I keep hearing of doomsday scenarios with electrolytics, with them
> having a definite use life. At least, when they go, they don't short out,
> like tantalums. I can get a suitably high-reliability electrolytic for
> about $2, whereas a tantalum would be about $10, and the biggest ceramic I
> can find is about 100uF, so I would need 10 of them, making that venue not
> too good.
>
> (*) Since it's been impossible for me to do some actual tests as Osaka,
> Japan, doesn't have any P-Channel MOSFETS sitting on shelves that weren't
> obsoleted 20 years ago...
>
> Oh, and btw, thanks heaps for all great info!
>
> // talsit.org
>
>
> On 16 April 2013 16:32, Arik Baratz <goo...@arik.baratz.org> wrote:
>
>>
>> On 16 April 2013 16:08, Kean Maizels <Ke...@kean.com.au> wrote:
>>

>>> Arik,****
>>>
>>> ** **

>>>
>>> There is no current spike. The FETs won�t ever be conducting at the
>>> same
>>> time, but their body diodes will be. The voltage sources are specified
>>> with 1R series resistance, and there is a 20R load, so you�re seeing
>>> the

>>> sources droop a little at those cross over points.****

[tALSit de CoD]

I knew I was forgetting something!!!

// talsit.org

On 16 April 2013 19:30, Andrew <and...@arcadius.com.au> wrote:

tALSit, the problem is that, as shown, the circuit won't kill the inrush current you said you were worried about.  C1 will just charge via D1.  The additional diode prevents this but at a cost of voltage in the powering-down scenario, which will affect your run time.

 

--

[Andrew]

Oh, and I think for this application, electrolytics should be fine.

 

The usual conditions that age these components do not appear to apply in the situation.

 

PC motherboards these days are using a "solid" electrolytic for long-life.

[Andrew]

What I really think is that the inrush current shouldn't be a problem.  The SMPS already will have some substantial output capacitors in play and a simple diode/capacitor arrangement should be sufficient.  Even then, the additional capacitance will only be extended (by some small amount) the inrush current of the SMPS charging its own capacitors.

 

I suspect the situation you will more likely encounter is that when input power to the SMPS is removed, the output voltage will decay at a relatively slow rate due to the output caps in play.  Unless you have a particularly sensitive method of detecting the voltage decline, you may find that the voltage in your secondary cap as already started to droop significantly before you notice, possibly giving less reaction time than you had planned.

 

If the failure case you are protecting against is the sudden removal of the connection between the SMPS and the rest of the circuit, then things will likely work as you expect as the charge in the output caps won't be in play.  It may be worth putting a 10k resistor across the SMPS connection to the circuit to quickly drain any parasitic capacitances hanging around.

 

[tALSit de CoD]

I am monitoring the voltage to the input of the SMPS, and I can set a analog watchdog on that - though I haven't tested it. At the least, I can sample it at 1kHz, which would give me at least some warning.

// talsit.org

--

[ada]

On Apr 16, 2:55 pm, tALSit de CoD <tal...@talsit.org> wrote:
> Isn't ESR just a by - product of the manufacturing technique? As in,
> supercaps tend to have high ESR (~50Ohm), because they are made for long
> lasting (months), low current (uA's) supply to things like RTC, where low
> ESR isn't important. And electrolytic are typically used for very short
> stabilisation of high current, so their ESR has to be low or else they
> won't be able to supply the current?

Yes.

What Luke is trying to say is that high ESR is better than low ESR.
that's crazy.

> Now, in my case, a 1000uF electrolytic cap with 0.08Ohm ESR will draw
> around 5A until it charges, which is in microseconds, but still way too
> high. Something with more ESR would be better for charging, but then I get
> into the problem of discharge. If it has 50Ohm ESR, the voltage drop of the
> car's supply is non-good!

Your 1mF electrolytic with 80mOhm ESR will draw more like 75A when it
starts to charge (if charging from a voltage source). It's likely to
be limited by the output impedance of your (nonideal) voltage source.

> I'm sure that a 0Ohm (if possible) cap would cause all sorts of other
> issues too...

Given that ESR is specced as a maximum, anybody whose circuit runs
into trouble with a perfect capacitor deserves everything they get.

Get one of these:
http://www.digikey.com/product-detail/en/PM-5R0V105-R/283-3015-ND/2185065

Stick it after the linear reg. Their droop after draining 200mA for
10ms is only 20mV.

[tALSit de CoD]

And how do I softly charge that 1F beast?

// talsit.org

[ada]

On Apr 17, 4:00 am, tALSit de CoD <tal...@talsit.org> wrote:
> And how do I softly charge that 1F beast?

LDO in currentlimiting mode.

[kaindub]

I've been following this for a few days and have to add some commentary

there seems to be a lot of discussions about ESR (Equivalent Series resistance) of capacitors. the key word here is equivalent>

ESR is measured at 100 or 120Hz (look at the datasheets) It is a measure of the RC time constant of the capacitor. It is NOT the real value of the capacitors resistance. this figure is important for switch mode power supplies and the like.

There is another parameter called tand. This is the measure of the lossiness of the capacitor and is what relates to the heating of the capacitor.

For a simple circuit using a capacitor as a filter the ESR can be ignored.

I think the solution to your problem is to 

a) have a power supply (the main supply to your circuit, designed for a certain hold up time. It probably only needs to be 1-2 ms - the time it will take your processor to process a shut down command.

b) A signal to your circuit from Before the power supply signalling power good.

When your processor loses the power good signal, go straight into an interrupt routine to stop or complete any memory writes. This will take only a few instructions and less than 1-2 ms

Your going to need to do the above step in any case for an orderly shutdown of the memory.

This solution alos does away with backup supplies supercaps batteries etc.

Robert

[ivanjh]

I think a few 2600F ultra-caps will do the job.

http://afrotechmods.com/videos/ultracapacitors.htm

[Luke Emrose]

After reading some more, apparently ESR is irrelevant at DC, which is what talsit I believe is dealing with.

Sorry for accidentally trolling this thread. I really was just trying to help. Can we stop talking about ESR now please?
I simply brought it up as one possible point amongst the list of things to consider, but clearly in talsit's case it's completely irrelevant.

As you were. I learnt something. Great. Thanks.
L

--

You received this message because you are subscribed to the Google Groups "Robots & Dinosaurs" group.
To unsubscribe from this group and stop receiving emails from it, send an email to sydney-hackspa...@googlegroups.com.
To post to this group, send email to sydney-h...@googlegroups.com.
Visit this group at http://groups.google.com/group/sydney-hackspace?hl=en.
For more options, visit https://groups.google.com/groups/opt_out.
 
 

[tALSit de CoD]

You mean in OVERcurrent mode where I'm trying to push the component way past its specifications?

[tALSit de CoD]

Hey Robert,

Yep, that's the kinda of idea that I was thinking of. Have the PSU front end last long enough for an orderly shutdown, triggered by the input voltage monitor. 

The problem that I have is my way-too-basic-knowledge of these things, hence me thinking and asking of how to do it.

I had the notion that ESR also limited how much instantaneous current you could draw from it. 

But I guess I should do more reading...

// talsit.org

[ada]

On Apr 17, 9:36 am, tALSit de CoD <tal...@talsit.org> wrote:
> On Apr 17, 2013 3:38 AM, "ada" <a...@panda2.net> wrote:
> > On Apr 17, 4:00 am, tALSit de CoD <tal...@talsit.org> wrote:
> > > And how do I softly charge that 1F beast?
> > LDO in currentlimiting mode.
> You mean in OVERcurrent mode where I'm trying to push the component way
> past its specifications?

See, for example, MIC5156, which has a provision for a current sensing
shunt so that it can do both voltage and current limiting
simultaneously.

So you program that with say a 25mOhm resistor and it will limit the
current to 1.5A.

[Andrew]

Same physical size (credit card), interesting additional functionality,
similar price.

http://www.marsboard.com/

[Patrick Barnes]

Looks pretty cool.

Does anyone know whether the A10 chip actually supports gigabit
ethernet? (on this board, or anywhere)

-Patrick

[Madox]

Look at the UDOO guys, iMX.6. Looks pretty nice.

[Luke Weston]

"Now, the question remains, to use electrolytics or to avoid them at all costs? I keep hearing of doomsday scenarios with electrolytics, with them having a definite use life."

I don't see why you need to fear aluminium electrolytics. Everybody uses aluminium electrolytics every day with no problems.
Choose an electrolytic with the right voltage rating for your circuit, install it with the right polarity, don't get the dodgy Taiwanese nasty ones with incorrect electrolyte (not that that issue really exists any more today), don't run the capacitor in an environment close to its maximum temperature rating and you should never have a problem within a sensible duration of time.

What's the typical temperature inside the environment where the capacitors are operating? If it's less than 70C it should never even come into play as a significant factor that even starts to derate the capacitor's MTBF or necessitate the choice of high-temperature caps.

Design it properly and it should just run happily without any intervention for many years.

[Angus Gratton]

On Wed, Apr 17, 2013 at 09:35:37PM +1000, Patrick Barnes wrote:
> Looks pretty cool.
>
> Does anyone know whether the A10 chip actually supports gigabit
> ethernet? (on this board, or anywhere)
>
> -Patrick

Hi Patrick,

No gigabit on the Allwinner A10, 10/100 only. I think the A31 (new
quad core Allwinner chip) has gigabit, but devices (and GNU/Linux support
if you're after that) are still a bit scarce.

You might consider Freescale i.mx6 if you want gigabit but don't mind
the form factor being a little bigger. For instance:

Wandboard: http://www.wandboard.org/

MarS Board: http://www.cnx-software.com/2013/04/02/99-mars-board-features-freescale-i-mx6-dual-4gb-emmc/
(Not to be confused with the A10-based Marsboard Andrew posted. Naming confusion ahoy!)

Same chip as the UDOO that Madox linked to. Only difference being
shipping times. I think Wandboard is available now, MarS Board says
next month. UDOO says September.


As an aside, I've messed with Allwinner devices a little bit now. For
running Android or other people's precanned Linux images they are
fine, but the internals are kinda hacky - multi-stage bootloader with
binary config files that preload values into registers before they
chain off to the next stage.

I ordered an i.mx6 based device (GK802) myself the other day,
supposedly they have much better vendor kernel and bootloader support
- a proper uboot without any weird hacks. It's probably only a
significant difference if you want to mess with installing OSes from
scratch, but I thought I'd mention it.

- Angus

>
> On 17/04/2013 9:07 PM, Andrew wrote:
> >Same physical size (credit card), interesting additional functionality,
> >similar price.
> >
> >http://www.marsboard.com/
> >
> >
> >
>

[miloska]

> Does anyone know whether the A10 chip actually supports gigabit
> ethernet? (on this board, or anywhere)

What is the use-case for a small device like these + Gb eth?

As I'm thinking about it one option can be firewall, but I doubt if these small boxes can do  connection tracking at the Gb level, so I run out of ideas.

  

Cheers,

[Terry Dawson]

Fileserving might be one.

[Jake Anderson]

check out the http://cubieboard.org/ as well
lots of i/o i think

[miloska]

My bad - I was thinking about what could possible fill the 1Gb, but the real question is what can fill the 100Mb.

I need an other coffee :)

Thanks!

[Madox]

On that line of thought (file serving), I am very happy with my HP N40L I purchased recently, very good value for money for what it is.

I'm not sure the use case for a Gb Eth is unless you are just switching it...at which point you should use a switch...

Many of those boards don't have the grunt to do anything productive with that much bandwidth...

A High-Res IP camera maybe?

//Note - Should have a word with Andrew about 'replying' to threads :)  Looks weird on the groups web interface, even with the subject edit it is still tagged against Talsit's thread.

[Madox]

On Thursday, April 18, 2013 8:24:15 AM UTC+10, Angus Gratton wrote:

I ordered an i.mx6 based device (GK802) myself the other day,
supposedly they have much better vendor kernel and bootloader support
- a proper uboot without any weird hacks. It's probably only a
significant difference if you want to mess with installing OSes from
scratch, but I thought I'd mention it.

- Angus

>

Hi Angus,

What are you doing with it?  I got my GK802 earlier in the year and have only used it to play YouTube Videos on my TV :)

They run a bit warm, and something I found odd when I used mine that I should tell you about... if you're powering it up from a PC, then plug in an USB device to the GK802 - the GK802 acts as a pass through and the USB gets enumerated on the PC...

I also think my TV didn't provide enough power so it would brown out on boot.  So you might have to use a plugback/wallwart/whatever they're called. 

[Andrew]

Didn't know about the reply-to-thread issue. I do the reply thing for efficiency (laziness) reasons rather than trying to remember the group address.

 

GB Ethernet is not only about the bandwidth/volume of traffic - it is also about the latency of traffic.  While these devices may not handle the volume, they will possible benefit  from the latency.  There are other considerations as well.

[Madox]

On Thursday, April 18, 2013 9:24:19 AM UTC+10, Andrew wrote:

Didn't know about the reply-to-thread issue. I do the reply thing for efficiency (laziness) reasons rather than trying to remember the group address.

:) I assumed so, just always looks as a thread hijack since I only use the web interface and it all gets confusing.

 

GB Ethernet is not only about the bandwidth/volume of traffic - it is also about the latency of traffic.  While these devices may not handle the volume, they will possible benefit  from the latency.  There are other considerations as well.

GB Ethernet might give you all sorts of 'issues' too, is the small latency at the PHY level really worth it, will it make a difference if you factor in processing?  

----

Did anyone mention the upcoming Beagle Bone update?  It was meant to be April...which is running out...

[Angus Gratton]

Hi Madox,

On Wed, Apr 17, 2013 at 04:21:23PM -0700, Madox wrote:
> What are you doing with it? I got my GK802 earlier in the year and have
> only used it to play YouTube Videos on my TV :)

Headless server, I'm hoping. Want to migrate my VPS (runs my mail,
tt-rss install, other stuff) to run locally, plus some other periodic
CPU-heavy tasks that I used to run on MHV's server until I left
Canberra. With a USB cell phone battery pack acting as a "UPS".

That's the plan anyhow, assuming it runs stably enough. :)

I was going to do this with an MK802 but I lifted the TX UART pad off
one (derp) and then the second MK802 I bought had an A10s in it, which
doesn't support GNU/Linux (or even a UART on that particular board, as
far as I can tell they routed non-UART pads to the two test points.)


> They run a bit warm

I noticed some of the newer kernels have CPU thermal throttling turned
on. Was considering cutting a hole and mounting a heatsink on it if it
comes to that.

> , and something I found odd when I used mine that I
> should tell you about... if you're powering it up from a PC, then plug in
> an USB device to the GK802 - the GK802 acts as a pass through and the USB
> gets enumerated on the PC...

Weird! So they just wired one port to two sockets? Can you use the
GK802 as an OTG device?

Maybe I'm being a bit optimistic "This $60 device will be different!
I've finally found the one good one!" :P.

> I also think my TV didn't provide enough power so it would brown out on
> boot. So you might have to use a plugback/wallwart/whatever they're
> called.

Yeah, this seems pretty common with these things. 2-2.5A is a
bit of an ask for a standard USB port. :)

- Angus

[Jaz]

I wonder if mars board + VR goggles + bluetooth keyboard would make an acceptable netbook replacement? With battery and SSD it could be a bar of soap form factor. A true pocket PC.

[Jaz]

Actually possibly scratch that. What is youtube playing performance like on the raspbian/midori? I've seen tutorials on getting it working but reviewer stop short of actually showing it in action.

[Kris]

Ooo if we're plugging file severs:

I really recommend open media vault
Its a debian based file server with a focus on home media storage. It has lots of plugins and has been really stable since version 2 (now 4). I run it off a fanless via mini itx board and two drives in raid1.
If you have a clunker sitting about, its perfect.

Don't try and run the os from a USB thumb drive though, its eats them in about a week.

--
Sent from my Android phone with K-9 Mail. Please excuse my brevity.

[Madox]

Hi Kris, I am running FreeNAS 8.3.1 on an USB stick. Installation was painless (dd an image) and a decent web GUI.

The only thing I don't like is that it is BSD... Can't easily run say Crashplan on it...

Otherwise it is great, designed to run off USB I presume... The HP N40L has an internal USB socket so that was cool :)

[James Laird]

On Wed, Apr 17, 2013 at 04:21:23PM -0700, Madox wrote:

> On Thursday, April 18, 2013 8:24:15 AM UTC+10, Angus Gratton wrote:
> > I ordered an i.mx6 based device (GK802) myself the other day,
> > supposedly they have much better vendor kernel and bootloader support
>

> They run a bit warm, and something I found odd when I used mine that I
> should tell you about... if you're powering it up from a PC, then plug in
> an USB device to the GK802 - the GK802 acts as a pass through and the USB
> gets enumerated on the PC...

They don't run warm with the latest kernel. I'm using the stock heat
spreader, with a little paste on the bottom as I mashed it up a lot
while I was developing the u-boot port.

Now that we have cpufreq working stably, they also throttle reasonably
well. The GPU clocks down too, with attendantly abysmal performance.

The USB design is a bit disappointing though; I was really hoping to use
both ports. I ended up making a grotty cable with a switch to choose between
the two modes while working on u-boot... You could take out the crappy
internal wifi module if you're desperate for another USB port.

- James

[tALSit de CoD]

Ok, so op delivers.

I've done some actual tests with actual components, and here are my results:

https://plus.google.com/u/0/116808312983714776242/posts/Lx5qJUHg955

Basically, I get about 20ms with a 1000uF capacitor. I ended up going with one of those nice Panasonic "high reliability" electrolytic ones. I may, for the sake of science, try with a 5.5V 0.10F super cap I have...

// talsit.org

[tALSit de CoD]

Ok, with that supercap, I get absolutely no time increase from having no cap there at all! The output voltage drops off very sharply about 6ms after the 7V trigger, which is exactly the same as if I remove the whole capacitor part.

// talsit.org