sb0 changed the topic of #m-labs to: ARTIQ, Migen, MiSoC, Mixxeo & other M-Labs projects :: fka #milkymist :: Logs http://irclog.whitequark.org/m-labs
fengling has quit [Quit: WeeChat 1.4]
kuldeep has joined #m-labs
evilspirit has joined #m-labs
key2 has joined #m-labs
key2 has quit [Read error: Connection reset by peer]
<sb0>
mithro, no. just that putting something else than ddr3 on the board would require a lot of arguing.
<mithro>
sb0: _florent_'s stuff seems to work pretty well on the Opsis
<sb0>
yes, but the ddr3 is running out of spec
<sb0>
I'm not sure if ddr3 can be done cleanly with the spartan6 ios, unless you use the MCB. unfortunately, our PCB guy seems to love things like the MCB, and flashy new toys...
<mithro>
that worried me a bit initially, but I've learned to live with it
<mithro>
sb0: I was looking at the MCB the other day, it was unclear to me what was actually hardware and what was "wrapper code" stuff they provided in
<mithro>
sb0: It looked like a lot of the MCB features wasn't really hardware?
<mithro>
sb0: Why would you put a spartan 6 on a device?
<sb0>
oh right, it might be easier to argue for artix-7 than ddr<3
<sb0>
hm, though artix-7 doesn't have odelay
<sb0>
mithro, as for training courses, I have none of them planned right now. what do you have in mind?
fengling has joined #m-labs
<mithro>
sb0: Not sure, I'd love to try and get better at FPGA stuff and especially at migen/misoc
<cr1901_modern>
I've been working on a MiSoC tutorial that explains the basics. Unfortunately, it assumes experience w/ Migen
<cr1901_modern>
I do wish it were easier to prototype FPGAs outside dev kits like you can microcontrollers (I'd consider using softcores like those provided by MiSoC instead of ARM cores for my ucon work if that were the case)
<sb0>
and yes, the MCB is hardware
<sb0>
with multiple undocumented and proprietary blocks, and another Java GUI wizard to produce horrid HDL code that puts them together
Gurty has quit [Ping timeout: 244 seconds]
Gurty has joined #m-labs
<mithro>
sb0: We have a whole bunch of people casually interested in FPGAs here at work and there is some chance we could get some training budget to fund a course type thing
mumptai has quit [Remote host closed the connection]
rjo2 has joined #m-labs
evilspirit has quit [Ping timeout: 276 seconds]
<sb0>
mithro, the idea of an ARTIQ workshop has been floating for a while
<sb0>
maybe migen/misoc can be combined
rjo2 has quit [Ping timeout: 276 seconds]
fengling has joined #m-labs
<mithro>
sb0: What would an ARTIQ "workshop" cover? I'm still only really fuzzy what ARTIQ actually does :P
evilspirit has joined #m-labs
evilspirit has quit [Ping timeout: 244 seconds]
<cr1901_modern>
sb0: That MCB file you linked. Looks like an (few?) engineer was paid a full time salary to write that file. Somehow I don't see a hobbyist duplicating it :/
<mithro>
cr1901_modern: 90% of the code looks to be stupid mapping x to y code
<cr1901_modern>
mithro: Hmm, you're right. I wonder if it was autogenerated now that I take a closer look
<mithro>
cr1901_modern: It is - it's generated using the MIG
<cr1901_modern>
Ahhh, I've never used the MIG before
<cr1901_modern>
welp scratch what I said. Was mainly overwhelmed by the size of the file
<sb0>
cr1901_modern, you are aware that, among other things, misoc contains the full functionality of the MCB (minus the idiotic overengineering) from scratch, right?
<sb0>
and yes, what mithro said.
<cr1901_modern>
sb0: No, I was not. I don't have any boards w/ a DDR interface, and tbh, DDR is a little bit overwhelming. Me grokking it has been slower than I'd like.
<sb0>
mithro, processing pulses/signals with the core device and/or PDQ, writing drivers/controllers for various devices
* cr1901_modern
is open to a DDR3 board in the 3-figure range
<sb0>
cr1901_modern, start with old school SDR SDRAM. DDR is the same but with annoying IO problems and a few other details.
<cr1901_modern>
sb0: Does SDRAM require the same transmission-line effect concerns as DDR and friends? Or is the clock speed still sufficiently low as to ignore these?
<sb0>
those concerns with DDR are overrated.
<sb0>
unless you are designing a motherboard with long traces and many SDRAM modules, or you are using particularly fast SDRAM, you can ignore most of them
<cr1901_modern>
sb0: Ack. And the high speed transceivers. I presume that serves a dual role to reduce noise and to reduce the low-pass filtering effect?
<cr1901_modern>
Drepper's "what everyone should know about memory" series is a good article, but it doesn't really give me enough knowledge to actually design my own controller thanks to physical layer stuff.
<sb0>
SDR SDRAM physical layer is pretty straightforward
<cr1901_modern>
_florent_ also gave me a link to your thesis, which is on my "to read" list
<sb0>
it's just D flip-flops
<sb0>
and regular 3.3V single ended IO
<cr1901_modern>
So no special FPGA directives needed? Okay, that's good.
<sb0>
no
<cr1901_modern>
Well, sounds a bit easier than I was making it out to be. Although MiSoC has a fine SDR impl, I should consider making my own as a test for miniSpartan
<cr1901_modern>
IIRC, the ppl who made Papilio got enough complaints about their DRAM that in a new version they replaced it with 2 MB of SRAM. Kinda telling.
sandeepkr_ has joined #m-labs
<rjo>
whitequark: the pythonparser unittests fail here (4 failures, 1 error)
sandeepkr__ has joined #m-labs
sandeepkr_ has quit [Ping timeout: 246 seconds]
kuldeep has quit [Ping timeout: 244 seconds]
kuldeep has joined #m-labs
<sb0>
cr1901_modern, getting SDRAM to work is a rite of passage.
<sb0>
SRAM is small, slow, and expensive.
<sb0>
(well, QDR SRAM isn't slow, but very expensive)
<sb0>
oooh, apparently 375nm leds, which are used in counterfeit money detectors and therefore are not horribly expensive, can be used to ablate *and* ionize calcium
<rjo>
but that is not ablating. it simply isnt. there is ome photodesorption. but no ablation.
balrog_ has joined #m-labs
<rjo>
and you have to think it through. the scatter makes this not much fun.
balrog has quit [*.net *.split]
mindrunner has quit [*.net *.split]
bentley` has quit [*.net *.split]
<sb0>
"When we first tested the LED in place of the 389 nm laser we were surprised to find that it was capable of loading ions into the trap while the 423 nm laser was blocked and the oven was switched off"
gric has quit [*.net *.split]
mindrunner_ has joined #m-labs
<rjo>
they could have just read a bit of literature instead of "being surprised" people have loaded MOTs for ages using photdesorption.
<sb0>
well, was their trap loading or not?
balrog_ is now known as balrog
<rjo>
yes. but there is no ablation.
<sb0>
that might be even better. is photodesorption selective for calcium?
<rjo>
and i wouldn't get too excited about it. if you try it you will whine about it being inefficient, producing scatter, warming, shifting your trap, etc.
<rjo>
no. not selective.
<sb0>
Alek's recipe for producing calcium in-place and avoiding oxidation problems when building/venting would also emit magnesium
<rjo>
which reaction?
<sb0>
inefficient seems totally fine. even 100W LED modules are cheap.
<sb0>
CaCrO4 + Mg
<rjo>
inefficient as in: you might have to wait a long time to load and will kill your trap and your oven before you load.
<rjo>
Cr is a problem. i would think. Mg would be acceptable. iirc it has a lower vapor pressure than ca.
<sb0>
how will it kill the trap?
<rjo>
as i said. he should mix it with one of the NEG powders that do a similar redox reaction.
<sb0>
depleting the oven and shorting things due to running it too long?
<sb0>
you mean damage to the trap electrodes from the UV light? even with a large one (coax cable or razor blades)?
<rjo>
what is large? for 500 µm electrode distances you need many 100 V already. also you mentioned 100W LEDs.
<rjo>
and every bit of change to your trap due to ablation or damage will need readjustment of imaging, voltages, and lasers
<sb0>
voltages? from my experience with the dust trap, it didn't care much about voltage
<sb0>
I'm looking for a simple, AC-only trap
<sb0>
100W LEDs, was just saying. the Oxford team used 500uW
<sb0>
and 100s of volts... as long as dielectric doesn't flash over, I can just get a bigger radio transmitter
<rjo>
you can not extrapolate from a grain of dust to an atom.
<sb0>
actually, if I crank up the voltage enough, the trap should be operable without laser cooling and I might be able to use a less expensive laser for fluorescence, isn't it?
<rjo>
what makes you think that?
fengling_ has quit [Quit: WeeChat 1.4]
<sb0>
which part? trapping without laser cooling, or some fluorescence being easier than laser cooling?
<rjo>
an "AC-only" trap does not magically make things simple.
<rjo>
doppler shifts at photodesoption temperature are ~GHz. do the physics and it will tell you that your ion scattering rate will go down. increasing laserpower makes scatter from the rest of the trap larger.
stekern has quit [Ping timeout: 250 seconds]
<rjo>
you won't be able to see your ion.
stekern has joined #m-labs
sandeepkr__ has quit [Ping timeout: 268 seconds]
<sb0>
what if the fluorescence laser also has a large line width?
<sb0>
and photodesorption produces hotter ions than evaporation?
<rjo>
that does not help. yes. photodesoption will produce hotter ions on average.
<GitHub123>
[artiq] sbourdeauducq pushed 1 new commit to master: https://git.io/vVhz2
<GitHub123>
artiq/master f395a63 dhslichter: Updated qc2 pinouts for SPI and 2x DDS bus, update docs
<GitHub111>
[artiq] sbourdeauducq pushed 1 new commit to release-1: https://git.io/vVhgK
<GitHub111>
artiq/release-1 08e742c dhslichter: Updated qc2 pinouts for SPI and 2x DDS bus, update docs
<sb0>
okay. well in any case I can use some high power UV diode and focus it on the evaporated stream, like Oxford did, which ditches at least one of those expensive lasers
<sb0>
or less expensive but still somewhat difficult vacuum work for the electron gun
sandeepkr__ has joined #m-labs
<rjo>
this is why nobody uses leds: you can't focus them like a laser.
<sb0>
yes, the paper mentions a pinhole.
jaeckel has quit [Ping timeout: 268 seconds]
<sb0>
but I'd rather use a $50 100W diode and turn 99% of its output into heat than a $50k laser...
<rjo>
that only gets you so far.
<sb0>
the latter also requiring months of fiddling around
<rjo>
ha. try getting a pinhole that handles 99 W.
<rjo>
trying an LED for desorption and PI will in all likelyhood take a beginner about the same number of months of fiddling as a laser.
<sb0>
PI only
<sb0>
and it's a large trap, I don't need such a fine focus