<azonenberg>
i havent written any control s/w for the PSUs yet but they work fine if controlled from the front panel
<rqou>
what part is new?
<awygle>
Dang, I want that lecroy just if only for dark mode lol
<azonenberg>
Lol
<azonenberg>
i find myself using it about 50/50 in local console and remote mode
<awygle>
All my test equipment is ugly white or uglier gray
<azonenberg>
on the left hand 24" monitor
<azonenberg>
you can see my own analysis doing the eye
<awygle>
Pretty sure I asked before and you got the rack on ebay?
<rqou>
hmm azonenberg your unlisted/ doesn't have listing enabled
<azonenberg>
waaay better looking than lecroy's persistence :p
<rqou>
security through obscurity?
<rqou>
:P
<azonenberg>
rqou: actually i think i just have a blank index.html
<azonenberg>
and its kinda like my private pastebin for images and stuff
<azonenberg>
nothing on there is that secret
<azonenberg>
i just dont want it getting spidered and mass downloaded and using all of my host's b/w
<rqou>
on this note, i really should get a second vps
<rqou>
then i can have a staging/testing server
<azonenberg>
That would be nice
<rqou>
and not one giant clusterf*ck that OOMs occasionally when minecraft leaks too much memory
<azonenberg>
:P
<azonenberg>
awygle: also my next lecroy will probably be a 2U headless
<rqou>
i still need to split out the "router" part of my router+nas+buildslave-thingy
<azonenberg>
i can buy my own LCD and keyboard/mouse if i need to do local stuff
<azonenberg>
the rest will be remote
<awygle>
I set up a swap file on my crappy vps because it was oom-ing my test jenkins install
<rqou>
jenkins now runs on my router+nas-thingy
<rqou>
much to the displeasure of housemates :P
<awygle>
I am thinking of getting away from hosting stuff at home. It's just a pain. I can afford to pay linode or somebody 10$/mo.
<rqou>
anyways azonenberg last time this was brought up you mentioned bandwidth routing between subnets?
<rqou>
this can be fixed by having multiple routers on the same segment
<azonenberg>
awygle: i mostly need high bandwidth *locally*
<azonenberg>
i dont host public services from home
<azonenberg>
i have VPSes for public facins stuff
<rqou>
with the 10g router only doing cross-internal-subnet traffic and the nuc doing to-internet traffic
<azonenberg>
then internal hosting for things that don't leave the house
<rqou>
i'm sure that multiple routers on a segment is definitely a well-tested and common setup /s
<rqou>
every time i touch networking stuff i get an increasing temptation to write my own "routerd" that just unifies dnsmasq+radvd+dynamic-dns+misc.
<rqou>
instead of a bajillion poorly-interacting programs
<awygle>
A NAS would be fun but I don't really have a good use case. My apartment is not that big.
<awygle>
rqou: easy there, Poettering
<azonenberg>
lol
<azonenberg>
ooh ooh i know
<azonenberg>
systemd-tcpipstack :D
<awygle>
systemd-routerd is a year away at best
<azonenberg>
wait
<azonenberg>
thats actually a thing?
* azonenberg
cries
<rqou>
at least my proposed program doesn't actually hijack anything in the tcp/ip stack
<rqou>
it only handles management packets
<awygle>
Oh sorry, it's "networkd" and "resolved" and already exists.
<azonenberg>
...
<awygle>
Systemd is one of those things that I really wish I could get on board with
<awygle>
The previous solutions were terrible, and systemd is not even that bad, they're just huge dicks to everyone gratuitously
<azonenberg>
my problem is having that much stuff shoved in pid 1
<azonenberg>
especially running as root
<azonenberg>
The existing stuff all needed a rewrite, i'm not denying that
<azonenberg>
but the rewrite should have been more modular, not more monolithic
<rqou>
my experience with systemd was unbelievably bad
<balrog>
much of it doesn't run under pid 1
<balrog>
my experience has been all right btw
<awygle>
Yeah they've gone way overboard. I kind of feel like somebody said to Lennart "you're eating everything" and he was like "muthaf**ka you ain't seen nothing yet"
<rqou>
"herp derp, let's upgrade from ubuntu 14.04 lts to 16.04 lts" "wait, nothing works now, why?" "wtf is this whole systemd thing that they added?"
[X-Scale] has joined ##openfpga
<azonenberg>
So how long until poettering decides to write his own kernel
<awygle>
Like with the "if you fuck up your password I now have root" thing - he probably would have fixed it except he felt attacked and he's too much of a child to handle that
<rqou>
i mostly just hated the complete lack of a migration path as i went from 14.04 to 16.04
<rqou>
not looking forward to seeing 18.04 lts
<balrog>
on this system, systemd has six auxiliary processes
X-Scale has quit [Ping timeout: 240 seconds]
[X-Scale] is now known as X-Scale
<balrog>
rqou: you had a lot of init scripts?
waylon531 has quit [Ping timeout: 250 seconds]
<rqou>
not "a lot" but definitely a few
<rqou>
and the 16.04 release notes don't say anything about "oh btw, in this lts we rewrote the entire bootup sequence"
<awygle>
14.04 was just bad imo
<awygle>
I ended up skipping it entirely
<rqou>
my laptop actually ended up stranded on 14.10 for not-exactly-great reasons
waylon531 has joined ##openfpga
<rqou>
iirc i reinstalled with debian sid after that
<rqou>
but i still run ubuntu lts releases on server machines
<azonenberg>
i run debian stable on everything but my router and nas right now, which are oldstable
<azonenberg>
those won't get upgraded except for critical security patches until i shut things down to move
<awygle>
Between 12.04 and 16.04 I bounced around a lot because of hardware support and the init transition
<azonenberg>
i got fed up with canonical pretending to be apple
<azonenberg>
completely reinventing the UI every release, etc
<rqou>
i never even heard about the init transition until i tried to update my vps
<azonenberg>
So now i run XFCE because i like the windows 2000 look :p
<awygle>
None of my Linux systems have a GUI
<Zorix>
im thinking about going debian testing with mate
<rqou>
i heard that testing is the worse of all worlds
<azonenberg>
with just a touch more polish
<azonenberg>
actually i guess its closer to win7
<azonenberg>
in terms of being clean but not overly fluffy and fancy
<Zorix>
i heard it was a balance between old and new
<rqou>
it's not actually much less buggy than stable, but it gets constipated every time debian actually tries to make a release
<Zorix>
hmm
<azonenberg>
Zorix: personally, i run stable then i build packages from source on the rare occasions i need something bleeding edge
<Zorix>
i just want newer kernels.. more rolling release without problems
<azonenberg>
right now the only things i have installed from source are solvespace and kicad i think
<awygle>
I need to set up an apt repo
<azonenberg>
everything else is distro packages
<Zorix>
and newer drivers and newer x.. etc
<azonenberg>
and of course my own projects
<rqou>
i'm thinking of moving to arch actually
<azonenberg>
I just run the LTS kernel
<rqou>
supposedly it is better at being rolling release than debian
<azonenberg>
thats one of the things i avoid changing the most
<awygle>
I like arch
<azonenberg>
i dont want a rolling release
<azonenberg>
i want something that stays put
<rqou>
plus arch is the weeb distro :P
<azonenberg>
and wont break anything for five years
<rqou>
debian is the angry german distro :P
<Zorix>
im kind of used to ubuntu so debian isnt much of a new playground for me
<awygle>
rqou: thought that was gentoo
<rqou>
i thought gentoo was the neckbeard/meme distro
<awygle>
I'd be like azonenberg if I had 5 year hardware cycles but all my gear is too crappy
<awygle>
And I end up with "why is your cmake so old" too much
<azonenberg>
every time a new debian comes out, i do a rolling upgrade a few machines/VMs at a time
<azonenberg>
my router and nas are the last to get upgraded
<azonenberg>
native machines are before that
<azonenberg>
and VMs are first
<rqou>
ubuntu has imho the worst release cycles (at least for my usage patterns)
<azonenberg>
So usually the router gets upgraded maybe a year or so after a new stable comes out
<rqou>
because it turns out that october and april are both "right in the middle of the semester"
<azonenberg>
rqou: so run a LTS and upgrade in june or july
<azonenberg>
or december/january
<rqou>
but LTSs always don't have a new enough <foobar>
<rqou>
for different values of <foobar> depending on what i'm working on
<awygle>
Lol yeah, wait for June so the release bugs are gone
<azonenberg>
awygle: exactly
<azonenberg>
thats what i do
<azonenberg>
i start upgrading a few months after the release goes out
<azonenberg>
then it takes me a year or so to roll the upgrade out everywhere
<rqou>
so now i just run a rolling release and live with everything being slightly broken constantly rather than majorly broken every six months
<Zorix>
what kinda things break
<azonenberg>
Zorix: let's see, video drivers routinely dont upgrade correctly when you change kernel revs
<rqou>
in my experience it's usually random desktop bits
<azonenberg>
if you upgrade to a new kernel version
<azonenberg>
VMware often breaks too
<Zorix>
ah wonderful heh
<rqou>
like "oh yeah, now everything is unreadable/4x the correct size no matter how you adjust the dpi"
<azonenberg>
i had to buy a new vmware version after switching from debian 8 to 9
<Zorix>
i use virtualbox anyways
<azonenberg>
and even then, i had to apply a custom patch to the kernel modules to make it work with the debian kernel
<azonenberg>
which was apparently newer than what the latest vmware was built for :p
<rqou>
i just use qemu/kvm+virt-manager
<rqou>
no problems here :P
<azonenberg>
I needed one last vmware license
<Zorix>
yea i use that on my server
<azonenberg>
The plan is to move to qemu+KVM after i move
<Zorix>
i like virtmanager
<azonenberg>
and buy a dedicated VM server
<rqou>
even osx on qemu/kvm works
<azonenberg>
and a native gaming+photoshop machine
<azonenberg>
at that point i wont need a gaming VM, i wont need browsing VMs on my desktop
<azonenberg>
i can push all of that out to the server or the gaming/photoshop box
<azonenberg>
and basically not run a hypervisor natively at all
<azonenberg>
on my host system i meant
<Zorix>
i have been looking at qubes.. i like the concept so far
<azonenberg>
i'll continue using vmware at work because it has better support for weird quirks of OSes and peripherals
<azonenberg>
In any case, when vmware killed unity mode on linux
<azonenberg>
they lost my business
<azonenberg>
i bought one more license to cover me during the migration
<azonenberg>
and that's it
<Zorix>
not worth buying that when virtualbox or kvm/qemu can do almost everything really
<azonenberg>
The issue is i had existing vmware machines i was using
<azonenberg>
sure, i could convert or rebuild them
<rqou>
afaik virt-manager can import some of them
<azonenberg>
but i couldnt afford the downtime
<azonenberg>
This vmware license will stay with this dekstop
<azonenberg>
which will probably stay on debian 9 for the rest of its life
<azonenberg>
i bought it in 2014-15 and it's overdue for replacement
<azonenberg>
so after the move i'll buy that dual xeon workstation i've always wanted
<Zorix>
i assembled this machine in early 2010
<azonenberg>
and move this motherboard into a 2U case and call it my vm server
<awygle>
azonenberg: god your hardware setup depresses me.
<azonenberg>
awygle: why?
<azonenberg>
Because it's less sucky than yours?
<awygle>
2 year old desktop ready for replacement
<awygle>
Yes!
<awygle>
Lol
<Zorix>
im on the 8 year schedule heh
<azonenberg>
awygle: Well, it's a single socket i5
<azonenberg>
with 32GB of RAM
<azonenberg>
which i OOM frequently
<azonenberg>
i saturate the cpu all the time
<azonenberg>
and it has no more free pcie slots
<azonenberg>
so i cant put in a 10g NIC which is going to be a critical feature once i start doing 10G stuff with my FPGAs
<Zorix>
yea well most new board dont come with many pcie slots anymore
<awygle>
My pc is a single socket i7 with 8gb of not very good ram
<azonenberg>
So my new build is going to be dual socket xeons, probably 12 cores per socket
<azonenberg>
128GB of ECC DDR4
<azonenberg>
then either my current GPU or a then-current nvidia card, plus a dual port 10g SFP+ NIC
<azonenberg>
then probably a raid1 of PCIe SSDs
<Zorix>
amd phenom ii 965, 16gb ram, geforce 760.. 1tb hd heh
<rqou>
azonenberg you should run a silly setup like me with ecc ram on a pentium :P
<azonenberg>
sure, it'll cost something like $4K, but for something i use professionally on a heavy basis? worth it
<azonenberg>
and THAT box shouldn't be ready for replacement for a while
<azonenberg>
i normally have a 3-5 year cycle for stuff but i derped and didnt get enough pcie slots on this mobo
<azonenberg>
and at this point it doesnt make sense to get a new mobo for a haswell i5 with ddr3
<rqou>
azonenberg: the special keyword you want is "HEDT" :P
<azonenberg>
So rather than retiring the box i'll push it to VM server duty
<azonenberg>
and keep it in active service
<azonenberg>
the last machines i *retired* were in 2015, purchased in 2009-2010
<azonenberg>
core2quad's and 1st gen i7s
<Zorix>
my box is in such bad shape if it powers off it takes like 30 min of dicking around with ram and pressing various parts of the motherboard to get it to post or pass memory tests.. its such a pain i rarely turn it off heh
<azonenberg>
ouch
<lain>
D:
<Zorix>
hehe yea its sad
<Zorix>
but when its up its stable.. no issues
<qu1j0t3>
most likely caps
<qu1j0t3>
sadly, on motherboard, but still replaceable
<rqou>
so i heard you like committing industrial espionage and stealing electrolyte formulas :P :P
<qu1j0t3>
eh no caps have finite lifespans.
<azonenberg>
qu1j0t3: electrolytics especially
<qu1j0t3>
this 'bad formula' stuff affects almost nothing.
<azonenberg>
This is one of the reasons why i
<qu1j0t3>
azonenberg: Yes, that's exactly what i mean.
<azonenberg>
why i prefer solid polymer caps, or ideally ceramics
<azonenberg>
All of my own PCB designs are 100% ceramic lately
<rqou>
yeah i'm aware
<qu1j0t3>
Zorix: replace the decoupling caps
<rqou>
but the ESR characteristics are different
<Zorix>
i recap a lot of things that need it
<awygle>
Polymer is definitely better
<azonenberg>
if i ever build that FPGA backplane i've been planning, the bulk caps there will be polymer
<rqou>
but $$$
<Zorix>
i have a couple of amiga 4000 motherboards im planning to recap here shortly
<azonenberg>
wet electrolytic and tantalum have no place in my designs, period
<awygle>
Got into that habit in spaceland and just never changed
<qu1j0t3>
Zorix: :D:D
<azonenberg>
I dont build high volume consumer stuff
<awygle>
Tantalum isn't so bad
<azonenberg>
so i'm not going to penny pinch
<rqou>
hook it up backwards :P
<azonenberg>
tantalum has catastrophic failure modes
<azonenberg>
more so than other cap types
<Zorix>
id repair the board i have in this machine but its not really worth it when im killing the cpus fairly frequently.. i need more power
<qu1j0t3>
i can't count the numebr of CapXon's i've replaced
<azonenberg>
why are you killing cpus?
<Zorix>
i have seen failed tantalums.. they crater the board
<qu1j0t3>
Zorix: yeah......
<qu1j0t3>
i haven't seen it myself but i get frequent reports from fellow retrocomputing people
<Zorix>
i have an old fluke network meter at work that it happened to.. i wanted to repair it but i dont have any idea of the value of the cap that was in it
<awygle>
Sure, but like... Don't do that lol
<qu1j0t3>
Zorix: does fluke publish schematics?
<qu1j0t3>
Zorix: they used to
<Zorix>
no and i asked them for it anyways and they basically told me to get a new one heh
<qu1j0t3>
ugh burn it down
<qu1j0t3>
that's so unacceptable
<Zorix>
i agree
<qu1j0t3>
Zorix: ask on eevblog/test equipment forum?
<azonenberg>
(suggestions re the list are welcome btw)
<Zorix>
im not really a developer... yet.. i just have repair experience
<awygle>
I'd pay like 20% more for an aluminum polymer over a tantalum
<awygle>
And I always derate caps by 50% for voltage
<awygle>
So 25v minimum for a 12v rail
<qu1j0t3>
Zorix | responsible for so much e-waste too oh god yes this is a crime
<Zorix>
i didnt even know they existed until this conversation, i knew of the dipped polymer ones of course.. but the aluminum can ones are like electrolytic but without all the issues
<Zorix>
yea
<awygle>
Tbh I prefer the dipped ones just mechanically
<Zorix>
yea they make sense for a lot of things except for space considerations
<Zorix>
i dont know the failure modes of polymer though, i havent seen any fail.. i really only test capacitance but it doesnt have a rising esr like electrolytic
<Zorix>
and i think their characteristics can be unsuitable to some circuits that electrolytics are used in but i dont know too much there
<lain>
ceramic caps don't need voltage derating :3
<lain>
well ok -- you don't have to derate the applied voltage, but you *do* have to account for the reduced capacitance when a voltage is applied, because polarization
* lain
shakes fist at magnetism
<lain>
hm
<Zorix>
good to know heh
<lain>
when you reflow a ceramic cap, you take it beyond the curie point, which resets the cap to its highest design capacitance. over time it will polarize due to applied DC voltage, and reduce capacitance (though the permanent polarization isn't a big deal iirc)
<lain>
iirc it's logarithmic
<azonenberg>
lain: yeah i always look at the C/V curves
<azonenberg>
when speccing a ceramic cap
<lain>
so you'll see a stark reduction in capacitance within the first few days of burn-in after a reflow
<lain>
after that it's effectively stable out to decades of time
<azonenberg>
i'm a lot less familiar with electrolytic behavior
<azonenberg>
And yeah i just saw i linked you a 16V cap
<azonenberg>
i'd probably use a 25 for 12V if it was an electrolytic
<azonenberg>
awygle: yes some higher end gaming/server mobos do actually use all solid caps
<awygle>
C/V on ceramics is brutal, especially the small ones
<awygle>
lain: I still derate ceramics 50% because there's usually no reason not to, and I'm used to an environment where anything on the board could randomly eat a huge quantum of energy
<lain>
haha
<azonenberg>
awygle: lol
<azonenberg>
Well package size is more important for me
<awygle>
Also it lets me be lazy with C/V curves most of the time lol
<azonenberg>
a 25V 0402 with high capacitance
<azonenberg>
may be at 10% of rated cap at 12V
<lain>
awygle: fair enough, but note that ceramics can typically survive 2-10x their rated voltage indefinitely with no damage
<azonenberg>
lain: i think the 25V rated parts are just up-specced 12V
<azonenberg>
:p
<lain>
haha
<lain>
likely
<awygle>
azonenberg: yeah OK. For high cap I almost always just go up a size
<azonenberg>
the c/v curves look the same
<azonenberg>
awygle: I just buy Samsung and look at the curve
<lain>
I've run into problems finding higher C/V caps, where every time I find a higher-V rated cap it's got the same C/V curve so is still useless at the desired V :P
<lain>
yeah
<azonenberg>
typically i target 20% or so max cap loss at my rated voltage
<awygle>
10uF 0402s may exist but it's not worth the headache
<azonenberg>
samsung is good at publishing curves in easy-to-find places
<azonenberg>
exactly
<azonenberg>
i usually do 0402 up to 1 uF
<awygle>
I use kemet
<azonenberg>
0603 up to 10ish
<awygle>
And murata
<awygle>
Usually
<azonenberg>
then 0805 and larger for big stuff
<azonenberg>
Those are not bad brands but the datasheets are harder to find
<awygle>
I usually skip 0603
<azonenberg>
samsung c/v curves are right on the digikey page
<awygle>
I have a big Murata stock already. Same with Panasonic resistors
<azonenberg>
I have a lot of Stackpole resistors, i use them for 5% stuff when i dont really GAF
<azonenberg>
like pullups and led series resistors
<azonenberg>
i usually do panasonic for when i care about tolerance
<azonenberg>
idk if stackpole even has 1% parts? :p
<awygle>
azonenberg: do you have a kitting substitution procedure?
<azonenberg>
?
<azonenberg>
I dont keep inventories of parts per se
<awygle>
I put one together a while back when someone else was doing kitting for me
<azonenberg>
usually what i do is, if i need a value i check what i have in stock
<awygle>
But I can't find mine
<azonenberg>
if i have something that meets spec i use it
<azonenberg>
if not i buy either the required amount plus 10% for small volumes of expensive parts (like big caps)
<azonenberg>
or, if its something cheap like a 1% resistor
<azonenberg>
i buy a tape of a thousand
<azonenberg>
and my stock gradually increases
<azonenberg>
i dont have what i would consider a formal "standard lab inventory" of parts
<awygle>
My overage rules are 2x for passives, 10 percent up to 2$, 1 extra up to 5$, and exact quantity after
<azonenberg>
simply because most of the time when i'm doing a new design i'm ordering a PCbs
<azonenberg>
a PCB*
<azonenberg>
and that takes time
<azonenberg>
so i do a digikey order for the board
<azonenberg>
and well, since this is my lab and not a production run
<awygle>
Yeah kbaker keeps trying to convert me to breadboards but I basically never bother
<azonenberg>
some things i intentionally order way extra because i expect to use them eventually
<azonenberg>
Like, i have a whole reel of 0.1 uF caps
<azonenberg>
When my supply of a thousand 0.47 uF's runs out, i'll buy a reel of those too
<awygle>
I have a reel of 10k resistors because they went on an insane sale lol
<awygle>
I have a tray of my standard microcontroller too which will probably never get much more than the half used it is now
<azonenberg>
Lol
<azonenberg>
what mcu?
<awygle>
MSP430FR5969
<azonenberg>
ah
<azonenberg>
i dont actually have a standard mcu at this point
<azonenberg>
i liked pic32 back when i did mcu stuff more
<azonenberg>
Been meaning to play with stm32
<awygle>
My last two mcu designes used the fr2111 instead
<awygle>
I need to do more stm32 stuff, arm has clearly won
<azonenberg>
that's like my standard decoupler now
<awygle>
I usually do 0.1 and 1
waylon531 has quit [Ping timeout: 248 seconds]
<awygle>
I was sketching a design today for an ingestible microphone and wondering how much decoupling that would actually need
<awygle>
(note this will never be built so nobody start telling me about biocompatibility or anything lol)
<Zorix>
thats a little creepy
<awygle>
A friend is at a conference and was asking how I'd approach some of the tech being presented
<Zorix>
i occasionally ask people questions like.. how would you go about designing a computer that can last 1000 years of usage.. in somewhat decent conditions.. like in a normal home environment
<awygle>
I'd ask cr1901_modern, probably :-P
<Zorix>
hehe
<azonenberg>
Zorix: redundancy
<azonenberg>
:p
<Zorix>
thats the normal answer i get
<Zorix>
heh
<Zorix>
but even with redundancy i think you'd be hard pressed to make it past 100 years
waylon531 has joined ##openfpga
<azonenberg>
you'd probably want cold spares
<azonenberg>
i.e. some cores are not powered up until others have failed
<awygle>
Well the home the environment is in will be knocked down by that time probably
<azonenberg>
so the time under bias is kept to a minimum
<Zorix>
interesting
<Zorix>
what about passives
<lain>
I often wonder how long a modern desktop cpu would last under full load conditions, assuming it was kept within thermal limits
<Zorix>
yea
<azonenberg>
Zorix: all ceramic caps
<azonenberg>
the MTBF on those is pretty good
<Zorix>
yea?
<azonenberg>
multiple power domains with fuses between them
<azonenberg>
So if a cap fails short it only takes out that core
<Zorix>
thing would have to be massive
<azonenberg>
Yes, it would
<azonenberg>
i dont think making single cores that reliable is tractable
<azonenberg>
i mean we're basically talking the google approach
<azonenberg>
dont harden each node to a crazy extent
<azonenberg>
make the network reliable and able to withstand failure of single servers
<Zorix>
yea
<Zorix>
but what about storage
<Zorix>
data storage
<azonenberg>
if you want max reliability storage
<lain>
climate-controlled tape.
<lain>
:P
<azonenberg>
I was going to say NOR flash made on an older process tech
<azonenberg>
Do your own ECC on top of it
<lain>
that too
<Zorix>
ahh
<azonenberg>
Mirror writes to multiple dies
<azonenberg>
say eight dies with ECC on top of them
<azonenberg>
As dies reach a given number of writes and/or fail, rotate new dies in
<lain>
modern flash can start degrading when left in an unpowered state for a few years
<azonenberg>
with a working set of eight good dies active at any time
<Zorix>
think that would be stable over 1000 years though?
<azonenberg>
Zorix: i dont know what the shelf life of an unpowered flash is
<Zorix>
ah
<azonenberg>
the *retention* lifetime is limited, usually 20-100 years
<lain>
Zorix: if you're mostly reading, use those "stone" discs :P
<azonenberg>
But if you didnt care about the data surviving
<Zorix>
m-disc?
<lain>
M-disc
<lain>
yeah
<Zorix>
i heard they were mostly bs
<azonenberg>
i.e. you just soldered a chip to a board and bring it up as needed
<lain>
they claim it lasts 1000 years :P
<lain>
no idea if it's accurate
<lain>
sure takes ages to burn though
<azonenberg>
oh wait, you want read only? not read-write?
<azonenberg>
mask rom :p
<Zorix>
well read only, read/write, etc
<Zorix>
what you'd expect a computer to have
<azonenberg>
For read-write i'd do older nor flash
<azonenberg>
Every year or so do a scrub, read and rewrite every address
<azonenberg>
correct ECC errors as necessary
<Zorix>
interesting
<azonenberg>
and swap out any dies with more than X uncorrectable errors
<azonenberg>
or with more than X write/erase cycles
<azonenberg>
so if you dont scrub eventually it'll degrade
<Zorix>
ahh
<azonenberg>
plus, if one bit cell fails
<azonenberg>
you want to detect and correct that while you still have enough intact ECC bits to recover
<Zorix>
right
waylon531 has quit [Ping timeout: 248 seconds]
<azonenberg>
i actually thought about this exact problem but for smaller volumes of data, crypto key storage
<Zorix>
nor in any large capacity is kind of unfeasible from what i have read
<azonenberg>
Define large
<Zorix>
ah
<Zorix>
something on order of current ssd storage drives
<azonenberg>
i mean if you have a thousand-year target lifetime
<azonenberg>
you probably have some budget
<Zorix>
yea
<azonenberg>
let's see, how big is nor flash on digikey now?
* azonenberg
checks
waylon531 has joined ##openfpga
<Zorix>
i like to propose the question once in a while because its an interesting thought experiment.. i think the current trend of disposable technology is just bad thinking and the path to newer technology tends to be at the expense to longevity.. so i like when designs try to keep longevity in mind
<azonenberg>
i dont like the disposable trend either
<azonenberg>
but at the same time, trying to get longevity beyond say 10-20 years requires serious engineering with that as the goal
<Zorix>
too much human intervention to keep technology operating.. it really is an issue.. you got massive datacenters with equipment coming and going just to keep up with short lifetimes
<Zorix>
yea
<Zorix>
i have a phone system at work that has been online and operational since 1990.. its computerized.. essentially a 68030 based system
<Zorix>
but i look at it to see the design choices because its reliable
<azonenberg>
So it looks like 256Mx8 is the biggest nor flash on digikey and that's a $26 BGA package
<azonenberg>
(in single units)
<Zorix>
daang heh
<azonenberg>
So you're looking at about $100 per GB before ECC and RAID overhead
<azonenberg>
But again thats for single chips
<azonenberg>
not sure how much it'd drop in bulk (digikey only lists single chip price)
<Zorix>
ah
<Zorix>
usually they have 10, 100, 1000 etc
<azonenberg>
yeah not for this part for whatever reason
<Zorix>
low volume im sure
<Bike>
i think chronic implant stuff is neat for that. unfortunately i don't know a ton beyond that one time they used a betavoltaic battery
<azonenberg>
So say you wanted to do a 20GB volume, that'd be $2K for one copy of the disk
<Zorix>
betavoltaic is not something i have heard about before
<azonenberg>
now say you do a nice bulky FEC code like, say, BCH(31,16)
<azonenberg>
with nearly 100% overhead
<azonenberg>
So say $4K for one copy
<azonenberg>
then eight mirrors of the disk to handle sector or even full chip failures
<azonenberg>
that's $32K per 20 GB
<Bike>
yeah, it's cool, right?
<Zorix>
getting up there
<Bike>
not what i would have thought of for an artificial heart, admittedly
<azonenberg>
then say add another 50-100% on top of THAT
<azonenberg>
to allow you to rotate chips in and out of the pool as they fail
<Zorix>
even with 8 mirrors would you expect it to last more than a few hundred years?
<azonenberg>
so $50K ish
<azonenberg>
Zorix: well i was thinking you'd have a pool of chips that were kept turned off
<azonenberg>
and apply power as other ones fail
<azonenberg>
(this of course requires reliable switching fets)
<Zorix>
ah
ZipCPU|Laptop has quit [Ping timeout: 255 seconds]
<azonenberg>
it would be a nontrivial project
<Zorix>
and waht about fet lifetimes?
<azonenberg>
That i dont know enough about to comment
<Zorix>
ok
<Zorix>
well thats some new ideas i havent heard before
<Zorix>
really fascinating
<Zorix>
what about good offline storage for something like a time capsule that would remain powered off
<Zorix>
magnetic media would oxidize and fall apart within about 20-30 years..
<Zorix>
optical media is known to degrade pretty fast, even pressed discs.. m-disc's reliability is questionable.. even when you consider that there might not be an optical drive anymore at some point
<Bike>
for a thousand years? carve it into a lump of platinum?
<azonenberg>
Zorix: glass plate with sputtered chromium
<azonenberg>
FIB patterned into it
<azonenberg>
by "glass" i mean "fused quartz"
<Zorix>
oh?
<azonenberg>
Basically an IC mask
<azonenberg>
choose your feature size as a tradeoff between durability/longevity and data density
<azonenberg>
You can do either binary coding, for machine reading
<azonenberg>
or actually etch text into it
<Zorix>
wow
<Zorix>
never heard of that idea
<azonenberg>
then just seal it inside a 2-part clamshell case made of PTFE
<azonenberg>
(pretty inert to almost any chemical i can think of)
<azonenberg>
put that inside a copper tube and solder a cap onto either end
<azonenberg>
encase that in concrete
<azonenberg>
i suspect your data would be retrivable after 10k or more years
<azonenberg>
Assuming anyone can find it and knows how to use a microscope / recover the data itself
<azonenberg>
you might have to etch instructions for opening and reading into the concrete or something
<lain>
haha
<Bike>
where would you put it such that people could find it again?
<azonenberg>
middle of a desert?
<azonenberg>
on a salt flat
<azonenberg>
it'll be the only large object for miles in any direction
<Bike>
do they last that long?
<lain>
at 1000 years geologic process becomes relevant :D
<azonenberg>
pretty sure a big salt flat will stay desert for a century
<Bike>
well, century sure, but you said ten thousand
<azonenberg>
sorry i meant a thousand years
<azonenberg>
Ten thousand i'm less sure of
<azonenberg>
but if you encase it in a giant mountain of concrete and have it hermetically sealed inside that
<azonenberg>
and are in a seismically stable area
<azonenberg>
it might get buried by blowing sand/dirt
<Bike>
salar de uyuni is apparently only a couple thousand years old...
<azonenberg>
but if you could find an excavate
<azonenberg>
it'd still be intact
<Bike>
well, that's what i'd be worried about. think how long it took civilization to even find easy mode shit like troy again
<azonenberg>
i'm thinking something closer to a pyramid
<azonenberg>
maybe make a giant pyramid of concrete around it
<azonenberg>
To make it easier to find
<azonenberg>
that should last a few centuries easily
<Bike>
sensible
<azonenberg>
and if you use good concrete, a few thousand
<azonenberg>
if its in the middle of a desert, it should be easy to find too
<azonenberg>
literally the only big thing there
<Bike>
just convert amundsen scott into a big ol death pyramid. the pole will move, but it'll probably be in the middle of ice hell for a while longer
<azonenberg>
That would work too lol
<Bike>
that bust of lenin is still hanging out there, so why not, right
DocScrutinizer05 has quit [Disconnected by services]
DocScrutinizer05 has joined ##openfpga
DocScrutinizer05 has quit [Disconnected by services]
DocScrutinizer05 has joined ##openfpga
azonenberg_work has quit [Ping timeout: 240 seconds]
pie_ has joined ##openfpga
cr1901_modern has quit [Ping timeout: 248 seconds]
cr1901_modern has joined ##openfpga
linux_user has joined ##openfpga
<linux_user>
Hello all!
<linux_user>
I am interested to know if anyone is aware of any free software that will give me a "Tangent Piano" sound using a USB Midi Keyboard with Linux?
<linux_user>
I am interested to know if anyone is aware of any free software that will give me a "Tangent Piano" sound using a USB Midi Keyboard with Linux?
<shapr>
I think this channel is about reprogrammable CPUs
<linux_user>
I have no idea
<linux_user>
I saw synthesis and figured it was a music synthesizer channel
<linux_user>
sorry, I will ask elsewhere.
<Ultrasauce>
zynaddsubfx
<shapr>
linux_user: I like the intuitive leap :-)
<shapr>
linux_user: my only connection with music+linux is Tidal Cycles for live coding
<pie_>
ive experienced tunnel vision before heh
<awygle>
Dell has a black Friday offer for a Xeon server for 300$. I find it strangely tempting even though I'm sure it's terrible
digshadow1 has quit [Ping timeout: 248 seconds]
eduardo has joined ##openfpga
azonenberg_work has joined ##openfpga
digshadow has joined ##openfpga
nrossi has quit [Quit: Connection closed for inactivity]