Openbot has quit [Quit: jmIrc destroyed by the OS]
<DocScrutinizer05>
~seen timriker
<infobot>
timriker <~TimRiker@bzflag/projectlead/TimRiker> was last seen on IRC in channel #infobot, 22h 44m 9s ago, saying: 'yay. another outage. this one for 10 hours. sheesh I hate phone companies'.
<DocScrutinizer05>
[2015-01-02 Fri 19:45:39] <mvaenskae> DocScrutinizer05: how was my link btw?
<DocScrutinizer05>
great, thanks! :-) I guess ebay should have the item as "sold" now
<DocScrutinizer05>
I wish there were more like those offers, possibly 10 or 100 times the "size" resp volume
<DocScrutinizer05>
(rationale: the Neo900 is depending on a system that's tailored towards embedded systems with limited resources. Systemd is considered toxic for that purpose)
<kerio>
(second rationale: fuck lennart poettering, really)
<DocScrutinizer05>
:-)
<DocScrutinizer05>
not a good rationale but adding to the satisfaction felt by supporting devuan, yes
<DocScrutinizer05>
basically it's all about freedom
<DocScrutinizer05>
basically the systemd cabal by refusing to discuss ways to allow linux without systemd inevitably forced that fork
<DocScrutinizer05>
it been plain obvious since almost a year
<Steven__>
IMO, sounds like a good thing. Linux should be modular enough that a few people can work together to replace any given component.
NIN101 has quit [Quit: No Ping reply in 180 seconds.]
NIN101 has joined #neo900
<Steven__>
Some of the desktop infrastructure is like that too. A program that requires a small dependency can pull a large number of other dependencies with no good alternatives.
SylvieLorxu has joined #neo900
<kerio>
on my last linux desktop, that was also used by my family, i ran a really lean xfce/rox desktop
<kerio>
while they ran gnome but with qt/kde loaded, just to run the cd burning program
<kerio>
and it was on a system with less than 1gb of ram
<jake42>
DocScrutinizer05: do you know if what you call "zeroclocking" is part of TI's smartreflex?
vakkov has joined #neo900
<kerio>
no, smartreflex reduces voltage of parts of the chip, iirc
Wizzup has quit [Quit: brb]
Wizzup has joined #neo900
Wizzup has quit [Client Quit]
<Steven__>
Question about systemd, etc.: One of the threads on the Debian mailing list mentioned that a significant number of packages depend on the choice of init system. That sounds crazy to me. Why should a package care about how the init system works unless it is somehow involved in booting the computer?
<kerio>
daemons
Wizzup has joined #neo900
<Steven__>
kerio: What about them?
<Steven__>
So a daemon gets started by init. Why should it care how it got started?
<kerio>
also, less about depending on PID 1 and more hard depending on systemd: stuff that needs some particular udev things, stuff that depends on logind
<kerio>
Steven__: init scripts
<Steven__>
But init scripts are usually simple stuff. It could be handled numerous ways so that the daemon/package/maintainer doesn't need to care what init system was used.
<kerio>
or they can lobby for systemd
<kerio>
so it becomes the de-facto default init system for linux
<Steven__>
You mean that package maintainers have political motives in specifying their dependencies?
<Steven__>
If I were writing a daemon I would look for a simple/easy way of abstracting the differences between systems so I wouldn't need to care. I could just call some "install init script" function and it would handle things for me.
<Steven__>
Well, at least there are people working on a solution...
<DocScrutinizer51>
jake42: yes, it's "part of" that bundle of measures TI iirc calls "SmartReflex"
<DocScrutinizer05>
kerio: what you think was SM is actually just Adaptive Voltage Scaling, one of several aspects of SR
<DocScrutinizer05>
s/SM/SR/
<kerio>
ayy lmao
<jake42>
DocScrutinizer05: tnx, gave me an overview
<DocScrutinizer05>
however zeroclock is prolly waht TI calls clock gating and predates SmartReflex, but SR offers a more finegrained way to shut down subparts of the SoC
<DocScrutinizer05>
actually zeroclock is only a part of Dynamic Frequency Scaling and that is part of SR
<DocScrutinizer05>
I wonder if 0MHz CPU clock is OPP0
<jake42>
what's OPP0?
<DocScrutinizer05>
OPerationPoint - there are like 5 OPPs OPP1 ... OPP5, with OPP5 being 600MHz for OMAP3530
<DocScrutinizer05>
sth along that line
<bencoh>
like C0->C4 ?
<DocScrutinizer05>
nah, that's not exactly similar
<DocScrutinizer05>
aiui
<DocScrutinizer05>
C0 .. Cn denominate states of "idleness"
<DocScrutinizer05>
while OPPs denominate valid CPU clock frequencies basically
<DocScrutinizer05>
the OPP0 remark was with tongue in cheek
<Steven__>
OPPsqrt(-1)
* DocScrutinizer05
just wonders if the General Purpose Memory Controller (GPMC) meant to manage the NAND could actually also manage another 1GB of **RAM**
<DocScrutinizer05>
it allegedly could handle SRAM as well
<DocScrutinizer05>
>> Glueless Interface to NOR Flash, NAND Flash (With ECC Hamming Code Calculation), SRAM and Pseudo-SRAM <<
<DocScrutinizer05>
;-D
<Steven__>
That would be very nice.
<DocScrutinizer05>
I doubt we will find a POP chip with 1GB LPDDR and SRAM on top
<Steven__>
What is the communication difference between dynamic and static ram? I didn't think the CPU was too involved in the refresh cycle.
<DocScrutinizer05>
RAM nowadays runs interface protocols almost as complex as IDE/ATA
<DocScrutinizer05>
at least in some cases
<Steven__>
Is that DDR3 your referring to?
<DocScrutinizer05>
sort of, yeah
<Steven__>
PATA isnt that complicated though, if you are just accessing data on a drive.
<Steven__>
(still more than dumb memory accesses though)
<Steven__>
Maybe some sort of interface could be hacked together.
<DocScrutinizer05>
sorry, we have no space for such interface, nor the technology to implement one in reasonable volume and cost frame
<DocScrutinizer05>
we need a PoP chip providing all the RAM of the system
<DocScrutinizer05>
I already pondered a piggyback PCB in place of such PoP chip, but that didn't lead anywhere
<Steven__>
You could have a no assemble spot with pads for a high density connector for these sorts of extra things.
<DocScrutinizer05>
LOL
<Steven__>
=)
<DocScrutinizer05>
show me a FBGA 0.4mm pitch connector
<DocScrutinizer05>
and I tel you why we cannot route the traces for such critter
<Steven__>
I know. All of my ideas are terrible. Sorry.
<DocScrutinizer05>
np :-)
<Steven__>
If you did put it on though, it would drive someone crazy enough to figure out how to use it by hooking up some Rube Goldberg machine to a rats nest of wires hanging out of the phone.
<DocScrutinizer05>
won't fly. The signal integrity would go south on such setup
<Steven__>
Rube Goldberg machines don't need signal integrity. You just add on more gizmos until it works!
<DocScrutinizer05>
also we already have some component connected to the bus: the NAND in PoP chip
<Steven__>
I'm curious: are there any chips that implement flash backed RAM? Sort of like what you would get if you put a ramdisk on top of a read-only partition using unionfs.
<Steven__>
I think that would be an interesting setup.
<DocScrutinizer05>
FPBGA168 PoP chips: definitely none
<Steven__>
Last time I checked there didn't seem to be much interest in ramdisks, even though they should be much faster than SSDs.
<DocScrutinizer05>
well, seems to me linux comes with a number of ramdisks per default
<Steven__>
I mean hardware ramdisks. Sorry ramdrive I think is the typical term.
<DocScrutinizer05>
they are probably obsolete as long as you can have real fast RAM attached to CPU's RAM bus
<Steven__>
Well, with a ramdrive you can have a large array of ram without needing an attachment to the bus for each chip.
<DocScrutinizer05>
downside: magnitudes slower than real RAM
<Steven__>
Yes, but much faster than SSD, and much cheaper than denser ram.
<DocScrutinizer05>
well
<DocScrutinizer05>
extreme niche requirement. Usually peeps get a larger server
<DocScrutinizer05>
throw in more cheap fast standard RAM into that
<Steven__>
Yeah. When you are talking about server setups the price difference probably isn't that big.
<DocScrutinizer05>
and then, a server farm often is preferable to one server with huge amounts of RAM
<DocScrutinizer05>
cheaper, redundant, and more performing
<DocScrutinizer05>
I mean, even super computers nowadays are large arrays of "inexpensive" server nodes, with a fast network thrown over it
<Steven__>
I guess. Still, I would think that somebody would have stuffed a bunch of ram in a 2.5" SATA SSD package, or something, but all of the ramdrives I have seen are pretty big and use older buses.
<DocScrutinizer05>
tinkering with a SSD to beef up the buffer RAM from a few MB to - say - 64GB would be a funny project
<Steven__>
Yes, exactly. =)
<Steven__>
What do you think of encrypting RAM? There has been some work on that to prevent cold-boot attacks.
<DocScrutinizer05>
ugh
<DocScrutinizer05>
never even tought about it
<DocScrutinizer05>
I guess coldboot attacks can get defeated otherwise
<DocScrutinizer05>
when somebody has sufficient control over the system to run a coldboot attack, I don't hope for RAM encryption to save me
<bencoh>
DocScrutinizer05: you just need to steal it
<bencoh>
nothing more
<DocScrutinizer05>
RAM encryption only would help when the system isn't tainted before reboot already
<bencoh>
(doesnt sound like uncommon situation)
<DocScrutinizer05>
errr
<DocScrutinizer05>
I'd rather use stuff like TrustZone or UEFI or whatever comes to mind, to not even allow attacks during boot
<Steven__>
Well, if your machine has removable ram, then it could just be moved to a machine set up to dump it as soon as it is plugged in.
<DocScrutinizer05>
mind you, the RAM only keeps its content when you boot from a running system into a tampered bootloader
<DocScrutinizer05>
and a secure first stage bootloader can easily clear all RAM
<Steven__>
Wouldn't that be pretty slow?
<DocScrutinizer05>
honestly, a too exotic bizarre attack vector, in my book
<Steven__>
To zero everything.
<DocScrutinizer05>
there's no such thing like perfect safety. You always need to evaluate the value of the info to get protected against the effort needed to steal it
<Steven__>
Really? It is fairly standard for when an educated attacker has seized a device for the data. Granted that wouldn't be an ordinary thief.
<DocScrutinizer05>
RAM encryption is the XY answer to that
<DocScrutinizer05>
~xy
<infobot>
[xy] The XY problem: You want to do X, but don't know how. You think you can solve it using Y, but don't know how to do that, either. You ask about Y, which is a strange thing to want to do. Just ask about X. http://www.catb.org/~esr/faqs/smart-questions.html#goal
<bencoh>
he did mention the context (cold-boot attacks) though ;)
<Steven__>
I don't think that is applicable, since it is known how to do the Y and it would solve X. It might not be a very good solution, depending on the situation.
<Steven__>
But that should be taken into account by the person deploying the security.
<DocScrutinizer05>
coldboot is a fine method to read out a life snapshot of an unprotected system. On a system with decently protected bootloader (think "TrustZone"), there are other way simpler attack vectors than coldboot
<bencoh>
Steven__: where would you keep the key btw ?
<Steven__>
bencoh: In a protected CPU register.
<DocScrutinizer05>
based on a RNG?
<Steven__>
Or some other protected storage location.
<bencoh>
oh, fair enough :)
<Steven__>
It could be on a crypto processor too.
<DocScrutinizer05>
meh
<DocScrutinizer05>
when the system refuses to *boot* when the bootloader got tampered with, you need no protection for RAM
<bencoh>
DocScrutinizer05: what if you steal a powered device ?
<DocScrutinizer05>
what then?
<Steven__>
I assumed that was what we were talking about.
<DocScrutinizer05>
what if you steal a powered N9? could you attack it by booting it? No!
<bencoh>
Steven__: me too
<Steven__>
Cold-boot attacks are against running systems.
<Steven__>
Yes. RAM rememberence allows a fast-booting kernel to dump the memory from the previously running system.
<Steven__>
Or, on a system with removable ram it can be moved to another system.
<DocScrutinizer05>
you'd need to solder chips in the running powered system. When you can do this, you also can attach an in-circuit-emulator and analyze the data before it even gets encrypted
<DocScrutinizer05>
sorry, this is too bizarre a topic for me right now
<Steven__>
?
<Steven__>
I agree that for RAM that is soldered to a board, a secure boot sequence is probably suffencient.
<Steven__>
But not all ram is soldered.
<DocScrutinizer05>
on N900 all RAM is soldered
<Steven__>
Yes.
<Steven__>
I think we were both confused because I was referring to the general case and you were referring to embedded systems specifically.
<DocScrutinizer05>
for devices with plug-in RAM there are watchdogs that bite your leg and vaults and IR motion sensors
<Steven__>
Lol.
<kerio>
watchdogs as in literal dogs
<DocScrutinizer05>
all way simpler, more effective, and cheaper than RAM encryption
<kerio>
?
<kerio>
like, the ones that go woof
<Steven__>
I think a dog is more expensive than encryption, lol.
<DocScrutinizer05>
yes, those
<kerio>
actually no, the ones that DON'T go woof cuz otherwise they'd be warning the intruder
<DocScrutinizer05>
haha, RAM encryption needs to get done inside the chip that manages the RAM bus
<Steven__>
kerio: Intimidation is sometimes a pretty good security tactic.
<DocScrutinizer05>
since... *all* your data is in RAM
<DocScrutinizer05>
you cannot encrypt RAM by a CPU that runs a program in... RAM
<DocScrutinizer05>
and then store the decrypted RAM data to... RAM
<Steven__>
+DocScrutinizer05: You can run AES without using RAM.
<kerio>
DocScrutinizer05: this chip will need some volatile memory to store the key
<kerio>
:>
<Steven__>
Yes, or just a hardware implementation with a few registers.
<Steven__>
That is probably how you would want to do it, so that it is not too expensive resourcewise
<DocScrutinizer05>
could we stop that topic now, please?
<DocScrutinizer05>
it has potential to worry users
<DocScrutinizer05>
and add workload to my "hotline"
<nox->
he probably thinks of how ps3 etc do it, tho i dunno if that uses a fixed key
<Steven__>
PS3 encrypts RAM?
<nox->
hm or maybe it only encypted the bus, dont remember
<Humpelstilzchen>
AES without RAM? in HW?
<Steven__>
(playstation 3?)
<DocScrutinizer05>
Neo900 doesn't need any of this
<nox->
indeed
<Steven__>
There are linux patches to do AES in the CPU registers, though I don't think I would want to try that without acceleration.
<DocScrutinizer05>
please!!!
<Steven__>
Well, this started with a hypothetical, not a specific suggestion.
<kerio>
DocScrutinizer05: what's your problem?
<Steven__>
I am not asking you to develop anything.
paulk-collins has quit [Ping timeout: 256 seconds]
<DocScrutinizer05>
my problem?? "Hey dudes, why don't you add RAM encryption to Neo900? Don't you care about user's safety and privacy?"
<DocScrutinizer05>
so pretty please take that elsewhere!
<Steven__>
Did I ever ask that?
<Steven__>
No.
<bencoh>
he didnt say that, you're just assuming some guy will eventually ask for it :)
<DocScrutinizer05>
I'm not willing to say "Toldya" and quote the mails I receive, in two weeks
<bencoh>
(you might be right in assuming some guy will eventually do, though ... but that cant be helped anyway)
<DocScrutinizer05>
anyway it's massively off topic in this channel. So a last time, it's not appreciated
<Steven__>
Lol. So the accusation is that I am granting knowldge without wisdom to potentially annoying people.
<Steven__>
Well, sorry.
<DocScrutinizer05>
we can discuss here why Neo900 doesn't need RAM encryption. everything beyond is confusing in this channel which is meant as a source of info for Neo900 prospects
Oxyd76 has quit [Ping timeout: 250 seconds]
xes has joined #neo900
<DocScrutinizer05>
and there's no accusiation of any sorts
<DocScrutinizer05>
just a firm request to please not discuss it any further here
Oxyd76 has joined #neo900
<Steven__>
Well, here is a question about the Neo900 if you like. How likely is the addition of the humidity sensor at this point? I am interested in tracking temp/pressure/humidity.
paulk-collins has joined #neo900
illwieckz has quit [Ping timeout: 245 seconds]
<DocScrutinizer05>
Highly likely, unless that particular nice chip becomes unavailable
<DocScrutinizer05>
pressure will be in anyway, no matter what (unless another sourcing disaster throws sticks at our legs)
<Humpelstilzchen>
nice so I'll finally know the huminity in my pocket
<Steven__>
That's cool.
<Steven__>
Lol.
<DocScrutinizer05>
ambient temperature yet up to evaluation, it might not be as accurate and fast as you would like it to be
<bencoh>
hm what would you guys use it for ?
<bencoh>
(out of curiosity)
<DocScrutinizer05>
the chip?
<bencoh>
yeah, or rather the result ;)
<Steven__>
I have sleep apnea that is not responding to treatment and I have been considering tracking conditions in my room to try and correlate symptoms with causes.
<bencoh>
oh, okay
illwieckz has joined #neo900
<Humpelstilzchen>
Steven__: uhm I don't think the neo900 will be a medical device..
<Steven__>
Speaking of pockets: Having that stuffed in and out of a pocket won't mess up the calibration on that kind of sensor?
<DocScrutinizer05>
for that it's definitely fast and (after calibration) exact enough
<DocScrutinizer05>
bencoh: THE ultimate source of latest data :-)
<Steven__>
Humpelstilzchen: This data would not form the basis for medical diagnosis or treatment, so that does not apply.
<bencoh>
DocScrutinizer05: I was actually asking for a user point of view, but thx for the link anyway :-)
<DocScrutinizer05>
bencoh: see Si7020
<DocScrutinizer05>
bencoh: sorry, you lost me
<DocScrutinizer05>
ooh, I missread
<Humpelstilzchen>
Steven__: just make sure you do not trust any sensor data
<bencoh>
DocScrutinizer05: :)
<DocScrutinizer05>
I'm not planning any particular usecase. That's in line with the openmoko philosophy that making usecases is up to users. We provide hardware
<Steven__>
Humpelstilzchen: Sure. The use case is that a possible cause is checked by a stricter method.
<Humpelstilzchen>
Steven__: and btw how did you find out that you have this problem?
<Steven__>
Humpelstilzchen: Polysomnogram.
<DocScrutinizer05>
generally the Neo900 design is not based on usecases much. Rather on general technical specs that are guided by feasibility
<Steven__>
Or "sleep study".
<Humpelstilzchen>
Steven__: Gesundheit
<Steven__>
Lol.
<DocScrutinizer05>
the idea is that users come up with new innovative usecases that were made possible by the hacker friendly open design of the device
<bencoh>
DocScrutinizer05: haha, yeah :)
<Humpelstilzchen>
Steven__: but you must have been a bit suspicious before doing such a test
<Steven__>
I actually forget why I had a sleep study in the first place (I think I had some symptoms). Of course, the reason I have forgotten is that memory loss is a symptom of sleep problems. =/
<bencoh>
DocScrutinizer05: I guess you would embed a class4 laser if you could ? :p
<DocScrutinizer05>
sure thing
<Steven__>
If you had a small laser with a frequency that the camera could pick up, it could possibly be used for reflective photoplethysogram.
<Steven__>
You could get heartbeat data and possibly blood oxygenation from that.
<DocScrutinizer05>
Steven__: (calibration) shouldn't get impaired by carrying the device in pocket. Drowning it into water would, but then... you get other problems than calibration of humidity sensor
<Humpelstilzchen>
Steven__: that doesn't sound very funny..
<Humpelstilzchen>
can we get a rfid zapper in addition to the class4 laser?
<DocScrutinizer05>
Steven__: blood oxygen saturation and heartbeat rate can get acquired by simple VGA camera and software
<Steven__>
What is an RFID zapper? A small EMP device?
<Humpelstilzchen>
yes
<bencoh>
yay
<Humpelstilzchen>
as in Matrix 2 ;)
<Steven__>
DocScrutinizer05: Using a laser is much much more reliable and accurate.
<DocScrutinizer05>
bencoh: don't you know my age old whitepaper (or product design draft) for GTA04? It has all sorts of such stuff :-)
<bencoh>
you can also measure heartbeat with a simple piezometer
<bencoh>
DocScrutinizer05: I think I read the neo900 one, dont remember about the gta04
<Steven__>
That would also be very inaccurate.
<Steven__>
With a laser you could detect arythmias.
<DocScrutinizer05>
yeah sure. But measuring heartbeat rate from dozens or 100s of meters distance with a simple camera is sort of smart, no?
<Steven__>
How would you do that? I thought you were talking about in close proximity?
<DocScrutinizer05>
and I'm quite sure you can detect arythmias via VGA cam and sw too
<DocScrutinizer05>
no, it works by simply analyzing the color of your face. a camera is waaaay better on that than your eye
<Steven__>
I am skeptical. Maybe with direct contact with the heart, but not plethysogram.
<Steven__>
Really? Interesting.
<DocScrutinizer05>
add a few filters on top... done
astr has quit [Ping timeout: 245 seconds]
<DocScrutinizer05>
on each heartbeat the color of your face changes to the red, due to blood vessels expanding
<Steven__>
I will have to look into that, it sounds promising.
<Steven__>
I have some benign arrythmias from my apnea.
<DocScrutinizer05>
it's on the edge of detection with a common dynamic range and precision of a camera, but you got a lot of pixels all showing your face, so you can inprove precision
<Steven__>
Right, right. Think you could do that in nearIR?
<DocScrutinizer05>
it's done in common visible range
<DocScrutinizer05>
in that paper I've seen
<Steven__>
I know it usually is, but I like to be able to sleep with the lights off. =)
<DocScrutinizer05>
(thanks to speedevil who brought it to my notice, iirc)
<DocScrutinizer05>
hehe, good point
<Steven__>
How are the camera LEDs done on the Neo900? Similar to the N900?
<Steven__>
There could be potential for a nearIR LED there.
<Steven__>
Hmm, or maybe the IR communication could be adapted...
<Steven__>
You would need a camera with the filter removed though.
<DocScrutinizer05>
too many "would need"
<Steven__>
Hypotheticals, not requests.
<DocScrutinizer05>
the flash LEDs are pretty identical to N900
<DocScrutinizer05>
maybe a tad stronger
<DocScrutinizer05>
but still white, and still two, and on the same location for the same purpose
<kerio>
DocScrutinizer05: i hope they're not strong enough to not be turned on for long periods
* Humpelstilzchen
only uses the led as torch light, the leds never did work for fine for camera..
<DocScrutinizer05>
when you mod the camera module to remove filter, you could replace them by IR LEDs
<DocScrutinizer05>
kerio: sorry?
<kerio>
flashlight
<Humpelstilzchen>
on the n900 the LEDs always added a blue touch to the photos
<Steven__>
How strong are the IR LEDs?
<DocScrutinizer05>
which IR LEDs?
<Steven__>
For data.
<DocScrutinizer05>
err, some 100mA continuous current iirc
<DocScrutinizer05>
as strong as we were able to make it
<Steven__>
Ah, ok, so the data LEDs could potentially be used for illumination if the IR filter is removed from the camera.
<DocScrutinizer05>
after all we want a better CIR range than N900 ;-)
<DocScrutinizer05>
yes
<Steven__>
They are not quite pointed the same direction, but they might be okay.
<DocScrutinizer05>
when you attach a mirror to redirect the beam
<Steven__>
Yeah.
<DocScrutinizer05>
but I guess that's a bit weak illumination for anything more than 1 or 2 m distant
<Steven__>
Let me see if the spec sheet on the camera says anything about the filter.
<DocScrutinizer05>
no, it doesn't
<Steven__>
Oh, okay.
<Humpelstilzchen>
while we are on the wishlist, can we get a RGB-D Sensor ;)
<DocScrutinizer05>
but iirc some guys actually removed the filter
<DocScrutinizer05>
Humpelstilzchen: what's that?
<Steven__>
So it is doable without breaking it? Some cameras have them fused in or something, as I recall.
<Humpelstilzchen>
DocScrutinizer05: just kidding, D is for distance, its basicly what Kinect (Primesense based sensor is), for each pixel you get its distance
<Steven__>
Humpelstilzchen: Is this based on the phase of the returning photons or something?
astr has joined #neo900
<Humpelstilzchen>
Steven__: Kinect v1/Primsende on the base of a known IR image that is projected, from the deviation of the pixel you get the pixel
<Steven__>
Ahh.
<Humpelstilzchen>
*get the distance
<Steven__>
You might be able to do that with the N900 with some fancy algorithms.
<Steven__>
Use the lens distortion of the camera LEDs and their lenses as your pattern.
<Steven__>
Pulse them at high speed so any change in what is being measured is low, and then sync your detection with the pulses picked up from the camera.
<Steven__>
It would be crude and inaccurate though.
<Humpelstilzchen>
DocScrutinizer05: (youtube video) afaik you can do that with any digital camera
<DocScrutinizer05>
es
<DocScrutinizer05>
yes
<DocScrutinizer05>
I'm not sure if somebody really removed the IR filter of N900. Maybe I'm mistaken
<Steven__>
Thanks for the links. I think getting at least pulse data is feasible for this setup. NearIR is already used in pulse oxymeters for measurement.
<Humpelstilzchen>
I guess someday we will have RGB-D Sensor on little devices like n[eo]900 but not yet
<Steven__>
Cool. So it gives a purplish tinge to normal photos.
<Humpelstilzchen>
when removing the filter downside is of course that you can not take any RGB Image (aka selfie) pictures anymore ;)
<Steven__>
We do have two cameras to work with...
<Humpelstilzchen>
might be me, but I never was able to get a _good_ picture out of my ne900 front camera
<Steven__>
That might be the one to remove the lens on then.
<Humpelstilzchen>
:)
<Steven__>
I don't see why it is needed anyway. The only major use case I can think of is video calls.
<Humpelstilzchen>
selfies ;)
<Steven__>
Flip the phone around?
<Humpelstilzchen>
or well as mirror
<Steven__>
A very very expensive mirror!
<Humpelstilzchen>
to check your makeup
<Steven__>
Mirror900: for when looking at a simple reflection just isn't stylish enough. Buy today!
Oxyd76 has quit [Ping timeout: 250 seconds]
Oxyd76 has joined #neo900
paulk-collins has quit [Quit: Quitte]
<Steven__>
I wish thermal cameras were cheaper. A couple of phone sized thermal cameras came out in the last few years IIRC, but they are still a few hundred dollars.
<Steven__>
(though that is much better than a few thousand dollars)
<DocScrutinizer05>
the true mirror app is the one that switches the screen to black ;-P
<Humpelstilzchen>
uhm I usually want to see my face in the display, not my soul
<Steven__>
Because... you are a vampire?
<DocScrutinizer05>
I get better mirror with batery out than with that mirror app
<DocScrutinizer05>
YMMV, depending on screen protector you might use
<DocScrutinizer05>
some do an excellent job at deglaring the display
* Humpelstilzchen
is not using any screen protector
<Humpelstilzchen>
I just get a very dark image, unless its in the sun, and summer
illwieckz has quit [Remote host closed the connection]
<DocScrutinizer05>
((it gives a purplish tinge to normal photos)) sure, since the IR light adds to the red component of picture
<Humpelstilzchen>
I just never cared enough to debug it
<DocScrutinizer05>
when you get an environment without any near-IR light sources, your pictures will still look normal
<DocScrutinizer05>
hardly anything to debig. The front cam of N900 is crap
<Humpelstilzchen>
^
<DocScrutinizer05>
debug*
<Humpelstilzchen>
uhm..find some light source without any IR components..
<DocScrutinizer05>
well, that's the purpose of the IR filzter in the camera
<Steven__>
Does the IR receiver let you get analog input? You could try using it measure the amount of IR in general and filter some of it out.
<DocScrutinizer05>
they wouldn't add any if it wasn't for the purple tint issue
<DocScrutinizer05>
Steven__: the Neo900 IR receiver? yes it has analog coupling, though AC
<DocScrutinizer05>
however for camera it's useless
<Steven__>
The tinge is probably DC mainly.
<Steven__>
Well, I was thinking you could use the IR receiver to measure the ambient IR in general and then use a software filter to remove that much red from the image.
<Steven__>
It would only improve it slightly under the best conditions though.
<DocScrutinizer05>
you can adjust the picture of a cam with IR filter removed, by simply reducing the red gamma. But it will result in somewhat incorrect colors
<Humpelstilzchen>
IR is on ADC? lol, I guess we have lots of fun with the neos sensors
<DocScrutinizer05>
sure :-)
<DocScrutinizer05>
actually it's on audio-in
<Humpelstilzchen>
there are quite a few algorithms to play with
<Steven__>
Haha.
<Steven__>
Can't the DC filter on it just be removed then? Or perhaps switched with a GPIO.
<DocScrutinizer05>
so you can sample up to maybe 40kHz carrier frequencies via audio ADC
<Steven__>
IIRC most audio systems remove DC just be putting a cap in series with the ADC.
<Humpelstilzchen>
Just think of a compass app, combining the magnetometer with gyro and acceleromater, bayes filter & sensor fusion anyone?
<DocScrutinizer05>
DC ADC is useless due to the design of the IR sensor, which is very low sensitivity
<Steven__>
Or that is what I have seen when I have looked at audio circuits anyway.
<Steven__>
Ah.
<Steven__>
How tolerant will the audio in from the connector be? I assume that will be filtered for DC?
<DocScrutinizer05>
you got a proper IR sensor at front of N900 anyway, integrated in the usual ALS
<Steven__>
There are some projects floating around for converting cheap USB DAC/ADCs into multimeters.
<DocScrutinizer05>
high accuracy, decent sensitivity, low sampling rate
<Steven__>
I bought a $10 one, but I lost input when I removed the cap. Propably fried it.
<Steven__>
probably*
<DocScrutinizer05>
audio-in on AV is HiFi and blocks DC, yes
<Steven__>
Is that by a cap, or by the audio chip?
<kerio>
can we get optical audio? :>
<DocScrutinizer05>
I pondered to not block DC and use it for DC ADC and impedance testing as well
<DocScrutinizer05>
kerio: I pondered that but the components are too large and generally not compatible with N900 AV jack
<Humpelstilzchen>
just remember as long as there is usb host (otg) you can attach an usb sound card
<DocScrutinizer05>
:nod:
<Steven__>
I've never been quite sure why you would want to use short-range optical for audio rather than say... twisted pair.
<DocScrutinizer05>
indeed
<Humpelstilzchen>
because its known to be even better then golden contacts!
<kerio>
isn't toslink the standard
* kerio
has no hifi system at all
<DocScrutinizer05>
there's an electric complementary to toslink
<kerio>
mh apparently my laptop can output optical audio at 192000Hz
<Steven__>
Toslink is proprietary AFAIK. S/PDIF is the standard, and it is usually carried on electical line.
<Steven__>
Uhm, I forget which connector.
<DocScrutinizer05>
S/PDIF was the name, right
<Steven__>
Toslink is just S/PDIF with an optical connector.
<DocScrutinizer05>
Neo900 doesn't support either ;-)
<Humpelstilzchen>
Steven__: srsly? I thought s/pdif is optical?
<kerio>
what about ethernet
<kerio>
rj45
<DocScrutinizer05>
via USB adapter, no problem
<kerio>
that's cheating!
<kerio>
is the n900 tall enough
<Humpelstilzchen>
lol
<Steven__>
According to Wikipedia: S/PDIF (Sony/Philips Digital Interface Format) is a type of digital audio interconnect cable used in consumer audio equipment to output audio over reasonably short distances. The signal is transmitted over either a coaxial cable with RCA connectors or a fibre optic cable with TOSLINK connectors.
<kerio>
my laptop doesn't have an internal ethernet card because it's not tall enough
<Humpelstilzchen>
Steven__: aah ok the toslink is the mechanical component to s/pdif ;)
<Steven__>
Only for optical.
<Humpelstilzchen>
kerio: bitten wodfall?
<kerio>
?
<DocScrutinizer05>
peeps, I need dinner
<DocScrutinizer05>
o/
<kerio>
DocScrutinizer05: it's half past midnight
<kerio>
the fuck
<DocScrutinizer05>
yeah, a tad early
<Steven__>
I ate cold pizza at the computer earlier today.
<kerio>
it's early for breakfast
<kerio>
it's way late for dinner
<Steven__>
It is 6:33 here.
<Steven__>
pm
<Humpelstilzchen>
its greeting time!
<kerio>
it's not my fault you're living in the wrong place
<DocScrutinizer05>
~ugt
<infobot>
well, ugt is Universal Greeting Time. Created in #mipslinux, it is a rule that states that whenever somebody enters an IRC channel it is always morning, and it is always late when the person leaves. The local time of any other people in the channel, including the greeter, is irrelevant. http://www.total-knowledge.com/~ilya/mips/ugt.html
<svetlana>
was it really created there
<svetlana>
:)
<Steven__>
Lol.
<DocScrutinizer05>
hi svetlana
<svetlana>
hi
<kerio>
meh, greeting time
<DocScrutinizer05>
no more freenode staff cloak?
<kerio>
only useful for people who disconnect from irc :s
<svetlana>
no more, I left
<Humpelstilzchen>
just curious, I thought DocScrutinizer05 was a the same timezone as me which is EST (gmt+2)..which is still LATE!
<DocScrutinizer05>
cmd: So 4. Jan 00:36:46 CET 2015
<Humpelstilzchen>
that sums it up
<Steven__>
What do you guys think of UTC and TAI? Do most programs really do okay with Unix time has a leap second?
<kerio>
aren't leap seconds POSIX
<Steven__>
As I recall, there are.
<Steven__>
Which surprised me.
<kerio>
then programs really have no excuse, do they
<Steven__>
I think after getting rid of time zones and daylight savings and days and months and years and so on, it kind of weakens your purpose to allow leap seconds.
<Humpelstilzchen>
The time difference is probably not big enough to to some real damage, just my two cents ;)
<Humpelstilzchen>
(yet)
<Steven__>
It is a whole second though!
<kerio>
Humpelstilzchen: seconds=61 can fuck some shit up i guess
<Steven__>
For most applications it would be fine, but you could have some critical thing that fails unexpectedly.
<kerio>
i like the google solution
<Humpelstilzchen>
kerio: too low
<kerio>
slow or speed up the next 3600 seconds by 1/3600th of a second
<Humpelstilzchen>
kerio: I have no idea what the hard coded value for e.g. ssl verification is...
<Humpelstilzchen>
(and I still hope I'll never find out)
<Steven__>
Simple solution: Use TAI to derive Unix Time, where every second is exactly the same and you never have to think about the position of the Earth wrt Sol.
<kerio>
hard coded value for what? :o
phre4k has joined #neo900
illwieckz has joined #neo900
<Humpelstilzchen>
guys lets just hope we won't have any problem before 2039
<Humpelstilzchen>
*8
<Steven__>
Lol. 64bit numbers for seconds?
<kerio>
who the hell thought that a 32-bit time_t was good enough? :s
<kerio>
Steven__: yes
<kerio>
because the alternative is jumping back to the '70s after 2038
<Humpelstilzchen>
kerio: signed...
<kerio>
even an unsigned one wouldn't help much
<Humpelstilzchen>
sure it would help a few years
<kerio>
besides, signed time_t makes sense
<Humpelstilzchen>
kerio: not really
<Steven__>
Here is a possibly stupid question: Why can't things like time, addresses, identifiers, etc, be sort of like UTF-8 or strings or something, where you have something to tell you how long the number is?
<Humpelstilzchen>
(note I was born after 1970 ;)
<kerio>
lol
<kerio>
C strings are null-terminated
<kerio>
which is not something you really like in a number
<Humpelstilzchen>
Steven__: less processing power
<kerio>
unless you want to store numbers in base 255
<kerio>
which is AWKWARD AS FUCK
<Steven__>
I guess, but how many calculations are you doing on an address?
<Steven__>
Sorry, afk for a bit.
<kerio>
bignum in C is really awkward
<Humpelstilzchen>
of course length is only one problem, 10e50+1-10e50 is the other..
<bencoh>
yeah, they fucked up with the 32b choice
<bencoh>
just like in ipv4 ;)
<Humpelstilzchen>
well still nice for the age
<kerio>
is ipv6 enough for a billion addresses for every atom in the universe?
<Humpelstilzchen>
actually thats a myth
<bencoh>
I stopped counting ;p
<kerio>
no that's a question
<Humpelstilzchen>
I learned that ipv6 only has a much as MAC-Addres
<Humpelstilzchen>
rest is prefix
<kerio>
wut
<bencoh>
well, not really ipv6 addresses are stored as 128b