<josuaH[m]>
So many orders in total! Wow! Are the total number of orders public? It looks like this goes above 1982 based on the queue ID?
<whitequark[cis]>
it's over 2000
jstein has quit [Quit: quit]
<josuaH[m]>
Hats off!
redstarcomrade has quit [Read error: Connection reset by peer]
redstarcomrade has joined #glasgow
notgull has joined #glasgow
notgull has quit [Ping timeout: 255 seconds]
JoshEnders[m] has quit [Quit: Idle timeout reached: 172800s]
esden[cis] has quit [Quit: Idle timeout reached: 172800s]
redstarcomrade has quit [Read error: Connection reset by peer]
tom has joined #glasgow
Eli2_ has joined #glasgow
Eli2| has quit [Ping timeout: 240 seconds]
g5pw has quit [Quit: leaving...]
as_g5pw has joined #glasgow
fibmod has quit [Ping timeout: 264 seconds]
<whitequark[cis]>
I just got another Glasgow from Piotr with a 2x512 Mbit RAM-Pak!
<galibert[m]>
Looks like Cat girl is gonna have a lot of fun :-)
<whitequark[cis]>
it's kind of mostly work to be honest, I actually wanted to discuss that bit with the folks in the channel but mainly esden
<whitequark[cis]>
I'm thinking about the representation for the addon
<whitequark[cis]>
sorry, that's incomprehensible. I'm thinking about how to represent the installed addon in the onboard EEPROM
<galibert[m]>
Ahhhhhh
<sorear>
is "addon" a class of things of which there can only be exactly one?
<whitequark[cis]>
I don't know
<whitequark[cis]>
basically, by far the easiest solution here is for Piotr to say "we will have no more than 6 SKUs of these" and then I have six spare flag bits left
<whitequark[cis]>
but I suspect I'll regret this
<galibert[m]>
How big is the eeprom ?
<whitequark[cis]>
I have six free bits left in the current configuration area
<galibert[m]>
Ouch
<whitequark[cis]>
I have at least 4KB more elsewhere, but that requires... complex engineering work
<galibert[m]>
One bit, dynamic auto-sizing ?
<whitequark[cis]>
note that the configuration data format is necessarily parsed by the FX2 boot state machine
<galibert[m]>
Not sure how hard it is to size hyperram
<whitequark[cis]>
uh, that's a non-starter
<galibert[m]>
Heh ok
<whitequark[cis]>
hyperram does have the equivalent of SPD but you almost certainly want to modify your gateware to account for the topology of any particular ram-pak
<whitequark[cis]>
since ice40 doesn't have a lot of LUTs to accommodate this
<galibert[m]>
And topology is not in the spd?
<whitequark[cis]>
it is, but by the time you can read it you
<whitequark[cis]>
* it you've already programmed the bitstream
<whitequark[cis]>
because you need a bitstream to talk to the RAM
<whitequark[cis]>
also the spd-type metadata is not exceedingly well specified or documented, so i'm not sure how generic you can be in first place
<whitequark[cis]>
i guess the other concern is that having a zillion SKUs of ram-paks is going to make testing a major headache
fibmod has joined #glasgow
<galibert[m]>
I guess you need a bitstream to reach the extra 4K?
<whitequark[cis]>
no
<whitequark[cis]>
it's just laid out really weirdly in the FX2_MEM EEPROM
<galibert[m]>
Oh
dne has quit [Remote host closed the connection]
dne has joined #glasgow
theorbtwo[m] has quit [Quit: Idle timeout reached: 172800s]
theorbtwo[m] has joined #glasgow
<theorbtwo[m]>
Can you not have a 6-bit integer for 2**6 different possibly things on lvds connector? Then you can have one value for nothing, one for flagging to use that extra 4kb to store a better id, and still have a reasonable number left over?
<whitequark[cis]>
true
<esden[m]>
So currently there are just two types. 128x2 and 512x2. I want to keep things simple and I will make and sell the 512x2 only. But that already shows that there could be 256x2. Additionally the HyperRAM chips come and go. I Saw a bunch of the parts get disconnected over the last few years. So eventually I might be forced to create a new sku due to chip availability.
<galibert[m]>
discontinued?
<esden[m]>
If we constrain that the LVDS connector can only have one add on. (I think it would be reasonable) then using a number to encode adding makes sense.
<esden[m]>
Yes discontinued, as in not available for purchase.
<galibert[m]>
Yeah, sorry, you wrote disconnected :-)
<esden[m]>
s/disconnected/discontinued/
vegard_e[m] has joined #glasgow
<vegard_e[m]>
disconnected from reality 🙂
<galibert[m]>
indeed
<whitequark[cis]>
okay, let's reserve three bits for this?
<esden[m]>
🤦♂️ indeed I did sorry
<esden[m]>
whitequark (@_discord_182174208896401419:catircservices.org) yeah I guess we will have to reengineer things if we need more data in the future
<sorear>
would anything attached via a/b/aux/sync ever be considered an "addon"?
<whitequark[cis]>
these four things are not a part of one group
helene has left #glasgow [client parted]
<whitequark[cis]>
basically, as far as i'm concerned aux isn't a thing
<whitequark[cis]>
it's actually ESD-protected unlike the LVDS port, but it still requires soldering, and I assume (esden what do you think?) it would void your warranty
<whitequark[cis]>
it's just two pins we happened to have a small amount of PCB area to expose
<whitequark[cis]>
I don't really want to think about it or support it, and it's definitely going away on future revisions
notgull has joined #glasgow
<esden[m]>
Yeah soldering on the board would definitely void the warranty. But to be fair, I never had anyone insist to replace a board after they tinkered with it.
<esden[m]>
Also the ESD protection diodes are actually DNP... I had them mounted on the early batch of Glasgow by accident.
<whitequark[cis]>
oh
<esden[m]>
(I mean the ESD diodes on the AUX connector)
<whitequark[cis]>
ok, then aux is just not a thing as far as I'm concerned
<esden[m]>
I was at some point calling the modules plugged into LVDS as "addin" as it fits inside the case... and those that plug into the regular GPIO as addons. I think we also called them briefly kilts. It might be all bit too silly and convoluted though.
<whitequark[cis]>
we didn't
<whitequark[cis]>
someone on the channel suggested it before being swiftly turned down because it's a poor name
<whitequark[cis]>
also "addon"/"addin" distinction is incomprehensible
<whitequark[cis]>
or I guess more specifically unreadable
<whitequark[cis]>
every dyslexic will hate you for picking it
<esden[m]>
I think anyone with even minor dislexia will have trouble figuring out which one is which... and that includes yours truly 😅
<whitequark[cis]>
heh
<esden[m]>
I just wanted to mention the naming thoughts that were floating in the past. I am fine if any kind of addon is called addon. It is pretty clear where they go at the end.
<whitequark[cis]>
I think I should probably just make ram-pak a special-cased thing or something
<esden[m]>
That said, when the RAM-Pak is in full production I have to include a mechanical assembly drawing with it, as I can imagine people plugging it in 180 to how it is intended.
_whitelogger has joined #glasgow
<whitequark[cis]>
I see, yeah
notgull has quit [Ping timeout: 268 seconds]
icb[m] has joined #glasgow
<icb[m]>
Is there enough room on the bottom to say "This side down"?
<whitequark[cis]>
I think "THIS SIDE TOWARD ENEMY" is either a different munuition or just a popular corruption of it
<SnoopJ>
mixture with "THIS SIDE UP" perhaps
<galibert[m]>
The latter I suspect
<whitequark[cis]>
but i also feel like it's maybe not the perfect reference to put on every unit shipped, given the broader context
<galibert[m]>
Yeah, it’s fun here but not in reality
<whitequark[cis]>
(mind you, I'm working on something in Amaranth right now and it takes all of the restraint I have to not name the states of a state machine that has a trigger to refer to operation of a machine gun. it fits perfectly, technically, but would be in poor taste within context)
<tpw_rules>
demolish the state
<whitequark[cis]>
that happens every time you delete the simulator I think
<josHua[m]>
hm this sounds like an application for interference fit pin headers
<josHua[m]>
I mean, like, or claymore mines, also okay
<josHua[m]>
but one is definitely a little more drastic of a solution
duskwuff[m] has joined #glasgow
<duskwuff[m]>
if the user doesn't have a case for their Glasgow there's another two plausible ways the RAM-Pak can be plugged in wrong
<whitequark[cis]>
it has to have an assembly diagram either way... and a real big "read this or lose your warranty" lettering on it
<whitequark[cis]>
actually, what does happen if you plug it in wrong? can someone flip the schematic?
<duskwuff[m]>
(i.e. rotated 180° in plane with the PCB, with the Pak sticking off the edge of the board instead of overhanging it)
Eli2_ has quit [Ping timeout: 268 seconds]
<Darius>
whitequark[cis]: 180 degrees will connect GND to 3.3V and vice versa
<Darius>
and 3.3V to VIO_AUX
<Darius>
apart from that, nothing 🙃
<whitequark[cis]>
GND to 3V3 should trip overcurrent or something, not too bad