<ds2>
did the need for custom cans ever get solved?
<DocScrutinizer05>
custom cans? RF shielding? GTA04 was so good layout it passed all certs (CE et al) without those. If push comes to pull, there's a factory claiming to be able doing custom RF shielding cans by etching, in prototype quantities, some 30km from my home
<DocScrutinizer05>
alternative: shielding spray to coat the inside of plastic case, should also help a lot for improved ESD prot
<ds2>
okay. you have it undercontrol then
<ds2>
came across a metal etching supplier who could do it in small quantities at a reasonable cost
<DocScrutinizer05>
kerio: ((what happened)) physical breakdown, manifested in a quite recurring/chronic flu (plus side effects) which now is annoying me since 2.5 months
<DocScrutinizer05>
ds2: ta
<ds2>
no time to deal with the video output stuff yet
<DocScrutinizer05>
ds2: you seen the killer fact that DVI etc all need more than the available 8 contacts total we can share to implement that video interface?
<ds2>
yes
<DocScrutinizer05>
k
<ds2>
it also precludes HDMI :D
<DocScrutinizer05>
:nod:
<DocScrutinizer05>
slimfoo etc would work but needs TWO complex chips since none of those chips takes parallel/serial input, they all need digital encoded input already
<ds2>
i wonder if there are nice DSI to that 4 wire display protocol the automotive folks use
<ds2>
nice chips that is
<DocScrutinizer05>
I think - given 8pin restriction and the range of available external adapters and TV/monitor-inputs - VGA is our #1 candidate
<ds2>
the automotive one keeps things digital, IIRC
<ds2>
that way, it can be converted to whatever folks want
<DocScrutinizer05>
"standard" digital video interfaces need more than 8 pins, Non-standard invariably needs a custom complex/active adapter
<DocScrutinizer05>
it would be still acceptable to use a non-standard IF that nevertheless has commercioally available complex adapters, like slimbus(?) or MHL(?)
<ds2>
no this is a standard that uses the USB mini connectors
<ds2>
can't find the proper name for it :(
<DocScrutinizer05>
but those need 2 chips in device, as explained above, plus they still need that active adapter even when such adapter is available
<DocScrutinizer05>
I don't know any monitor or TV that would offer such input
<ds2>
it is used mostly for automotive
<ds2>
but that signal can then be converted without quality loss to whatever digital format dejour
<DocScrutinizer05>
for that it's great for sure
<ds2>
I don't like VGA cuz it is analog and noise sensitive
<DocScrutinizer05>
yeah, 'only' needs a custom made active adapter
<DocScrutinizer05>
Neo900 won't build such adapters
<ds2>
side business... ;)
<DocScrutinizer05>
for the max resolution Neo900 device can provide (800*720p iirc), VGA is just good enough. digital video is basically overkill
<DocScrutinizer05>
on analog VGA noise will just cause a few percent of inferior quality. On digital interfaces, noise will cause 100% failure
<DocScrutinizer05>
actually DM3730 calims to have improved CVBS signal quality compared to OMAP3530
<kerio>
*too much noise
* DocScrutinizer05
ponders to put 'check GTA04 CVBS quality' on todo list
<ds2>
when you show me a 720p or even a 800x480 CVBS display that doesn't look like crap....
<ds2>
and that 800x480 must be visible pixels :P
<DocScrutinizer05>
anyway Neo900 product feature specs: decent standard definition (PAL/NTSC) video, if feasible 800*720 (or higher) workstation static output, at minimum 20fps/RGB565
<kerio>
checkerboard pattern
<DocScrutinizer05>
I think there's no such thing like 720p CVBS. Is there?
<ds2>
is isn't exactly a standard but...
<DocScrutinizer05>
I mean, you can think of a CBS-alike signal that has 720 visoible horizontal lines
<ds2>
its analog, you just toss in a few more lines before you do a vsync each time ;)
<ds2>
but go fit that in the <6MHz BW :D
<DocScrutinizer05>
yep. exactly. At higher H freq, since you want Vsync still be 50Hz
<DocScrutinizer05>
or would you go 720p@25Hz then?
<ds2>
let me put it another way... 800x525 then
<ds2>
or "525p"
<ds2>
show me what
<ds2>
:D
<DocScrutinizer05>
anyway H-freq/Hsync and pixel clock is insane
<ds2>
there is no pixel clock
<DocScrutinizer05>
virtually there is
<ds2>
it is analog
<DocScrutinizer05>
inside device there for sure is
<ds2>
<-- refuses to look inside the magic black box ;)
<ds2>
even the current "320x262" video looks like crap
<DocScrutinizer05>
it doesn't make sense to employ a 30MHz DAC to output 30Mpixel per second on a signal path that has a max bandwidth of only 6MHz
<ds2>
taking that senstance by itself, I'd argue, it does
<ds2>
using a 30MHz DAC would make filtering a lot less nasty
<DocScrutinizer05>
which filtering?
<ds2>
the min. is a 12MHz DAC
<ds2>
reconstruction filtering on the analog side
<DocScrutinizer05>
hmm?
<ds2>
with a 12MHz DAC, you're looking at near brickwall filters to prevent harmonics, etc
<ds2>
think of this as reverse of an ADC
<DocScrutinizer05>
why bother about harmonics?
<ds2>
cuz things (and gov't agencies) get irked by them
<kerio>
fuck the man
<DocScrutinizer05>
meh, when the bus has a bandwidth of only 6MHz, why add a dedicated filter to do same?
<ds2>
that sir, is why I will not be investing my $$
<DocScrutinizer05>
no really, I won't lowpass-filter digital-to-analog video output signal
<ds2>
*shrug* your call
<ds2>
the good thing is this is a moot discussion as the DM3730 is doing this all inside the SoC
<DocScrutinizer05>
the bus drivers themselves do enough of filtering
<ds2>
but in general - the reason you do need to do it is - unless you want to be dealing with nice sine waves, the simpliest way to generate color burst is to use a square wave
<ds2>
so something needs to chop off the harmonics
<ds2>
or you risk things like them beating down and interfering
<DocScrutinizer05>
sure, when the harmonics would reach any active/non-linear component
<DocScrutinizer05>
I'm aware of the problems of beat-down/aliasing
<DocScrutinizer05>
I just say we won't see any of that
<DocScrutinizer05>
anyway, moot dicussion as you say
<ds2>
with the DM3730, probally - if sw is setup right
<ds2>
been spending time figuring out ways to generate a CVBS signal in SW
<DocScrutinizer05>
and since no device I know of would accept a HSYNC@32kHz CVBS signal 720p, it's doubleplus-moot
<DocScrutinizer05>
show me a TV that accepts 720p CVBS and I'll rethink
<ds2>
:)
<DocScrutinizer05>
VGA otoh accepts quite a wide range of different formats regarding VSYNC and HSYNC
<DocScrutinizer05>
and yes, for VGA we will want proper signal shaping to avoid harmonics and overshooting etc
vakkov has quit [Ping timeout: 272 seconds]
<ds2>
strictly speakimg, I'd disagree with that. but having said that, most modern monitors are multisync
<ds2>
remember the X configs that were frying some monitors?
<DocScrutinizer05>
yep, sure
<DocScrutinizer05>
strictly speaking VGA is 640*480p@32kHz_HSYNC
<DocScrutinizer05>
iirc
<ds2>
with a fixed refresh rate
<DocScrutinizer05>
but *all* monitors with "VGA" input nowadays (and since at least 10 years) can do pretty much arbitrary SYNc freq and resolutions
<ds2>
aka the ones with the 13pin D connector for analog signals
<DocScrutinizer05>
I guess the last momitor which escaped blue magic smoke from either deviation solenoid or H-driver died 20 years ago
<DocScrutinizer05>
ugh, "deflection coil"
* DocScrutinizer05
idly wonders how many true CRT monitors are still in use
* ShadowJK
has a CRT on his desk
<DocScrutinizer05>
WOW!
<ShadowJK>
"Hyundai deluxscan 15G"
<DocScrutinizer05>
EIZO? Sony?
<kerio>
"true crt"?
<ShadowJK>
eizo? sony? what?
<DocScrutinizer05>
sort of tautological, kerio. Just most people nowadays know CRT acronym only for "some monitor, whatever technology"
<DocScrutinizer05>
ShadowJK: those two were the most awesome CRT monitors available back when any CRT been available. So the idea been it's likely somebody would keep those, if any
<ShadowJK>
it's mostly because I'm lazy
<DocScrutinizer05>
:-D
<DocScrutinizer05>
as long as it works ;-)
<ShadowJK>
I've got a Sun monitor on floor next to desk
<ShadowJK>
21" or something, CRT
<ShadowJK>
with weird Sun connector
<DocScrutinizer05>
yay
<ShadowJK>
Sun Sparcstation 5
<DocScrutinizer05>
sorry, exhausted, back to curing flu
<DocScrutinizer05>
o/
<ShadowJK>
ANd Sun .. something or other, 800Mhz CPU with 1.5G ram
<ds2>
Mmmmm hypersparcS?
<ShadowJK>
doesn't talk to the Sun monitor though :-(
<ShadowJK>
Yes, the SS5 is Hypersparc
<ds2>
oh the pizza boxes
<ds2>
cg6 video?
<ShadowJK>
Want it? Come get it, it' s free
<ShadowJK>
iirc cg5? is there such a thing? nfi
<ds2>
no, I have an ultra5
<ds2>
cg5 is old, IIRC. cg6 was the standard
<ShadowJK>
I'll throw in a case of beer
<ds2>
hehehehe
<ShadowJK>
Also got a Netra :P
<ds2>
only if you will take the ss2 off my hands in trade ;)
vakkov has joined #neo900
<ShadowJK>
lol
<ShadowJK>
no
<ds2>
netra is a SS5 in shoebox formfactor like the javastations, right?
<ShadowJK>
The SS5 and its monster monitor actually served in modern times
nox- has joined #neo900
<ShadowJK>
The Netra is a 1U thing with a 400Mhz CPU and 2 x 20 GB SCSI hard drives, iirc
<ds2>
is it 19" wide or merely 6"?
<ShadowJK>
hm?
<ShadowJK>
It comes with thumb screw rack mount things
<ds2>
Oh
<ShadowJK>
And fits nicely in rack, if I'd have one
<ds2>
must be thinking of something else
<ds2>
there was something else also in the same shoebox formfactor as a javastation
<ShadowJK>
I'm not familiar with javastation
<ShadowJK>
Though SS5 is slightly narrower than rack, I guess
<ds2>
that's a real SS5. but they moved thigns around to make it shoebox sized
<ShadowJK>
It's rather amusing how SS5, the first time I encountered it, in the day of Pentium 600, was quieter than PCs
<ShadowJK>
Nowadays it's a monster :D
<ds2>
it was??
<ShadowJK>
yeah
<ds2>
you must have some noisy PCs
<ShadowJK>
SS5 was a "wait, computers can be this quiet?!" experience for me
<ShadowJK>
Plus, the 10krpm HD made more impressive noise than the 7.2krpm of PCs
<ds2>
hahahaha
<ds2>
SSA's and netapps!
<ShadowJK>
Walking into the room full of SS5s at Uni was a more awesome experience than walking into the room of 20 Pentium-II 500 :)
<kerio>
SS5?
<ShadowJK>
SparcStation 5
Pali has quit [Remote host closed the connection]
<ds2>
StarShip5
<ShadowJK>
:)
vakkov has quit [Ping timeout: 256 seconds]
<kerio>
according to wikipedia, pizza-box form factor is a thing
<ds2>
the last chance for a peaceful computing
<ShadowJK>
By modern metrics, the SS5 was noisier. By the metrics of the day, the Linux room filled with P2-500 was like stepping onto the deck of an aircraft carrier filled with WW2 planes. The SS5 room was like stepping onto the desck of an aircraft carrier filled with F-18
<ShadowJK>
Also, all the "makes teacher look clueless" coders always sat in the SS5 room
<ds2>
all the SS5's I worked with was on people's desks
<ds2>
SS5's was for the interns and no bodies
<ShadowJK>
There was also an Ultrasparc room, but I don't remember much of it
nicksydney has quit [Quit: No Ping reply in 180 seconds.]
nicksydney has joined #neo900
<kerio>
i notice that my laptop's doing something heavy when the fans become audible :3
<ds2>
gcc seems to have a lot fo dead time otherwise
vakkov has joined #neo900
<DocScrutinizer05>
well, depending on ratio of total computational grunt of all CPU cores vs speed of storage IO you want sufficient overprovision of active processes so there's always a process not waiting for IO for every available core
<DocScrutinizer05>
note however that swap is also storage IO
<ds2>
gcc seems to do IO-Compute-IO....
<DocScrutinizer05>
I think every process does ;-)
<DocScrutinizer05>
that's why you use time sharing multitasking :-)
<DocScrutinizer05>
particularly in -j
<ds2>
yep
<ds2>
hence oversubscribing helps
<ds2>
though, memory is a major constraint. swapping sucks ;)
<DocScrutinizer05>
swapping mostly defeats the purpose of increasing compile speed, when increasing -j beyond the limit that RAM allows. Unless swap is significantly faster than "ordinary" IO
<DocScrutinizer05>
calculating the right parameter for -j option is very complex - otherwise there would already exist a "-j auto-optimum" option ;-)
<ds2>
machine learning to the rescue
<ds2>
-j autolearn ;)
<DocScrutinizer05>
though I think nowadays -j should probably default to "<cores> * 2", not to "1"
<DocScrutinizer05>
but changing defaults is always a nasty thing to do
<kerio>
there's no default for -j
<DocScrutinizer05>
yeah, it defaults to "no -j"
<DocScrutinizer05>
which is similar to "-j1" aiui
<DocScrutinizer05>
aka "no multi-job"
<ds2>
there are poorly done makefiles where anything other then -j1 fails
<kerio>
nethack and variants :c
NIN101 has quit [Remote host closed the connection]