azonenberg changed the topic of #scopehal to: libscopehal, libscopeprotocols, and glscopeclient development and testing | | Logs:
<GenTooMan> perhaps if I get more data tossed on my network something like a DS920+ from Synology would be good. I have 10T worth of USB drives which is becoming a bit rediculous.
<azonenberg> Yeah my setup right now is one 6TB SATA drive + 2x 1.92 TB NVMe per node
<azonenberg> Three nodes but i have three replicas of each piece of data so the effective capacity is still 6 + 3.84 TB
<azonenberg> The nodes are capable of scaling to 8x SATA + 9x NVME each, which would give me 48 TB SATA + 34.56 TB NVMe assuming I used the same drives. But realistically by the time I need anywhere *near* that kind of storage
<azonenberg> bigger drives will be affordable :p
<azonenberg> Right now i'm only using about half of the capacity so i'm in no hurry to expand
<Degi> What kinda box is it that you have so much NVMe?
<Degi> If you only need N+2 drives, couldn't you do 6 TB * (9 * 3 - 2)?
<azonenberg> Degi: no it's 3N
<azonenberg> it's not ECC like raid6
<azonenberg> every piece of data is replicated 3x
<Degi> Ah
<azonenberg> this gives you redundancy against loss of two servers without data loss, or one while maintaining full operation and writability
<azonenberg> also means you get better bandwidth for bulk reads because you can read from all three at once
<azonenberg> Ceph can of course scale to >3 nodes, 3 is just the minimum for reasonable usage
<azonenberg> And i don't see outgrowing these servers any time soon
<azonenberg> It's a supermicro motherboard that has something like seven pcie slots and four m.2 connectors
<Degi> Ooh neat
<azonenberg> Right now one m.2 is used for the boot drive, two for the 2x 1.92 TB NVMe, one unused
<azonenberg> then one pcie for a 10G NIC and the rest unused
<azonenberg> So i have lots of room to expand if I want. I just see no reason to spend the money without a good reason
<azonenberg> I built these boxes with the intent of scaling them up as my needs grew
<azonenberg> planning to keep them until the 2030s probably
<azonenberg> with drive replacements of course ,but keeping the same basic hardware
Degi_ has joined #scopehal
Degi has quit [Ping timeout: 252 seconds]
Degi_ is now known as Degi
<_whitenotifier-3> [scopehal] azonenberg pushed 1 commit to master [+0/-0/±2]
<_whitenotifier-3> [scopehal] azonenberg 30e5e9c - Finished initial MIL-STD-1553 decode. No protocol analyzer mode support yet. See #286.
<xzcvczx> azonenberg: awake still?
<azonenberg> Yeah, it's early evening for me here - 19:00
<xzcvczx> is threshold meant to go off measured data points rather htan interpolated values?
<azonenberg> As of now, the threshold filter does not interpolate. The few filters that need sub-sample precision, like the CDR PLL, normally work directly on the analog signal and do thresholding internally
<azonenberg> Longer term i will probably make a mode switch in the filter to interpolate or sample
<azonenberg> allowing you to trade performance against accuracy to fit your use case
<xzcvczx> ok cool
<d1b2> <mubes> Yeah, didn't lose anything from my disk crash....the zfs, hosted on the same server, maintains the backups for the flash drive that is used as the active primary (At the time this was configured I didn't trust zfs on Linux enough to integrate the flash into the zfs). Flash blew out but the zfs kept on chugging and didn't miss a beat. It was hardly stressed to be fair (it didn't lose any drives etc) but still, I'm pleased with its performance
<d1b2> and I will look at putting some flash in there to use the array as a primary disk set.
<azonenberg> Yay
<d1b2> <mubes> Found a spare spacewire-usb bridge. Guy who has got it is off till weds, but will follow up then.
<azonenberg> Great. Yeah just a few waveforms should be fine for initial development
Tost has joined #scopehal
esden has quit [Read error: Connection reset by peer]
elms has quit [Ping timeout: 245 seconds]
esden has joined #scopehal
elms has joined #scopehal
bvernoux has joined #scopehal
sam210723 has quit [Quit: Leaving]
<xzcvczx> azonenberg: i wonder when some mfgs will start encrypting/drming their waveforms
<ericonr> xzcvczx: stop giving them ideas
<xzcvczx> no i refuse, you can't tell me what (not) to do
<GenTooMan> xzcvczx, you do realize that the alternative might be permanent. I mean people who refuse to behave often end up in conditions worse than prison.
<xzcvczx> :(
<GenTooMan> xzcvczx, in any case let me clue you in on how bad ideas can lead to big trouble. You ever hear of email spamming? Well seems some idiot in the late 1980's and early 1990's had the idea that it would be a great way to "make money" through advertisement. He did prison time by the way because he used public assets to conduct business, he had to serve a full 90 days in prison. Undetered by that he got some idiot BMs (Business
<GenTooMan> Men) to invest in his scheme. Thus began email spamming.
<tnt> xzcvczx: Already the case in RF signal generators, they have encrypted waveforms format so that if you use their software to generate like ... a 5G carrier, you can't copy and use that waveform on a competitor hardware.
<xzcvczx> >_<
<xzcvczx> how anyone can claim that crap is good for innovation and doesn't stifle it
<GenTooMan> xzcvczx, some people find it easy to say conflicting things apparently.
<GenTooMan> xzcvczx, what scope do you have? Not a siglent I assume?
<xzcvczx> nah i am too cheap
<xzcvczx> i just have the hantek 6022bl and a salae clone
<xzcvczx> the bl does logic as well, but can only be analog or digital at a time
<xzcvczx> so the salae clone helps hwen i need both
<GenTooMan> hantek in the US is more expensive than siglent.
<xzcvczx> but my current analysis thing is like 22kHz so its not exactly taxing
<xzcvczx> well 22kbps
<xzcvczx> the 6022 is like sub $100 usd
<GenTooMan> I know a hantek dealer site let me see what they are selling it for IF they are selling it.
<GenTooMan> the problem is not everything available here is available elsewhere and visa versa.
<xzcvczx> ummm which siglent do you think is cheaper htan hantek?
<GenTooMan> I think I avoided hantek to be fair I believe I was wavering between Siglent and Rigol but went with Siglent because the didn't charge "extra" for decoding serial signals.
* xzcvczx still thinks the new-agg oscopes look weird not being absolute huge bricks
<xzcvczx> GenTooMan: can you not "hack" the unlock for those on the rigol?
<xzcvczx> hmmm the siglent isn't terrible priced here surprisingly
<GenTooMan> I read the reviews and went through all the video on both units. The reality is I prefer not poking things like that because you don't know when they will pull a FTDI stunt and brick the unit.
<xzcvczx> ture
<xzcvczx> what siglent you got?
<xzcvczx> >_< i should probably wait til my sister is coming over next time and get her to bring me one, save $300
<GenTooMan> SDS1104X-E it works surprisingly well I have yet to make use of the 100MHz BW then again I'm mostly doing things like SPI and QSPI
<GenTooMan> I need the digital interface now because of QSPI :D
<azonenberg> GenTooMan: flash or something else?
<azonenberg> If you can get the digital interface working in scopehal that would definitely be good progress
<azonenberg> We have a QSPI flash decoder already
<GenTooMan> azonenberg, I am using QSPI flash memory mapping for IO into an FPGA
<azonenberg> Yeah so we actually have a multi-layered decode for this. you run SPI decodes on the two x1 lines then a QSPI decode on all four
<azonenberg> so you can see the individual bytes in either x1 or x4 mode
<azonenberg> then the serial flash decode takes those three data streams as input and decodes them to commands, addresses, and data bytes
<xzcvczx> GenTooMan: does the -E or -U mean anything?
<GenTooMan> xzcvczx, + means built in MSO -E means external MSO -U means NEVER an MSO
<xzcvczx> ohhhh ok, thanks
<GenTooMan> I don't know if you can get the SDS1XXX+ units anymore.
<GenTooMan> azonenberg, good to know thanks for informing me.
juli9610 has joined #scopehal
<xzcvczx> GenTooMan: is siglent well interfaced with glscopeclient?
<GenTooMan> xzcvczx, not the SDS1000* stuff just the SDS2000 and SDS5000 units.
<xzcvczx> azonenberg: is there any thoughts about getting snap to data points for the cursor option?
<azonenberg> File a ticket. It's something that i could see being useful as an alternative to the interpolation mode we use now
<azonenberg> we'd have to add a pref or menu setting for it
<xzcvczx> could just be a checkbox in the curson context menu
<xzcvczx> cursor
<azonenberg> Yeah something like that. File the ticket and we'll figure out how to implement it later
<xzcvczx> but yeah i will create a ticket
<xzcvczx> in -apps i assume?
<azonenberg> Yeah
<_whitenotifier-3> [scopehal-apps] xzcvczx opened issue #320: snap to data points -
<_whitenotifier-3> [scopehal-apps] xzcvczx edited issue #320: snap to data points option -
<xzcvczx> it makes it a fair bit easier for measuring stuff in digital signals
<xzcvczx> i mean sure you can use threshold and snap to the edges i guess
<azonenberg> i mean i'm used to much more oversampled waveforms
<azonenberg> with 20/40 Gsps scopes
<azonenberg> so single points are usually something that get blurred into the background
<azonenberg> But yeah on lower hardware you push closer to nyquist a lot
<azonenberg> so it makes sense it could be useful
<xzcvczx> azonenberg: well feel free to send one of htem scopes this way :P
<azonenberg> lol
<azonenberg> I don't want to overlook features that make glscopeclient useful for lower end work by any means
<xzcvczx> indeed
<azonenberg> it's just not my focus for stuff i use on a daily basis
<xzcvczx> it does remind me of bunnies fun with hypertransport back in the xbox days though
<xzcvczx> or whatever that bus was called
<azonenberg> yeah and dont get me wrong not everything i do with glscopeclient is in the GHz lol
<azonenberg> I'm working on some stuff on a 20 MHz serial bus for work now
<azonenberg> I actually had to make some optimizations to glscopeclient because of it lol
<azonenberg> because while it was slow, i was recording over four seconds of bus activity
<azonenberg> the protocol analyzer started bogging down when i gave it 300000 transactions to scroll through
<xzcvczx> haha yeah one of my traces currently stuck in sr/pulseview creates a 1.3GB csv file
<xzcvczx> (stuck in sr/pulseview as sr's csv export is broke)
<azonenberg> I have some glscopeclient datasets that are double digit GB
<azonenberg> admittedly some of that is bloat from the sparse file format, which i'm looking to work on eliminating in the future
<azonenberg> But that kind of situation is one of the reasons that I built that big ceph cluster i was talking about earlier
<azonenberg> my old NFS was large enough to hold big files fine, but it was slooow because it only had 1GbE and the drives were all spinning rust
<azonenberg> the new cluster has 10GbE to each node for a total of 30 Gbps of aggregate network bandwidth, and most of the actively used data lives on NVMe
<xzcvczx> very nice
<_whitenotifier-3> [scopehal-apps] azonenberg pushed 1 commit to master [+0/-0/±2]
<_whitenotifier-3> [scopehal-apps] azonenberg 94c9273 - Updated submodules
<xzcvczx> i sorta wish it would show what had changed in the submodules there
<xzcvczx> azonenberg: i still want to say thank you muchly for making it bsd licenced though, i am getting really sick of all the crap going from permissively licenced to gpl/agpl/other weird left-licences lately
<xzcvczx> (dont get me started on qt)
* GenTooMan remains mute about the "qt" thing.
Tost has quit [Ping timeout: 240 seconds]
<xzcvczx> GenTooMan: very wise
tnt has quit [Ping timeout: 240 seconds]
tnt has joined #scopehal
Stephanie is now known as Stephie
Tost has joined #scopehal
<GenTooMan> gah installing ghidra is worse than installing eclipse manually.
<xzcvczx> haha yup
<xzcvczx> welcome to java
<monochroma> is it? iirc i just installed whatever debian's "open" java package and away I went
<Degi> Yeah, like my package manager seems to have it
<xzcvczx> GenTooMan: what are you getting stuck on
Tost has quit [Ping timeout: 252 seconds]
<xzcvczx> monochroma: its the building of the large packages in java that is sometimes not so easy
<monochroma> ah yeah if you're building it from scratch vs just installing the binary
juli9610 has quit [Quit: Nettalk6 -]
<GenTooMan> xzcvczx, someone wrote a nice install script that requires sudo I can't see why it needs sudo except it's trying to install it in /opt <- which is odd
<GenTooMan> thus I am manually installing it in .local
bvernoux has quit [Quit: Leaving]
<xzcvczx> gah its better than cube updates
<xzcvczx> they require sudo even when its installed in a userdir
<_whitenotifier-3> [scopehal-apps] azonenberg pushed 1 commit to master [+0/-0/±1]
<_whitenotifier-3> [scopehal-apps] azonenberg d998107 - WaveformArea: when creating a FFT in a new window, use sane X axis units rather than converting fs to Hz and zooming way too far out
<_whitenotifier-3> [scopehal-apps] azonenberg opened issue #321: Filter graph editor crashes when the last reference to a filter is removed -
<_whitenotifier-3> [scopehal-apps] azonenberg labeled issue #321: Filter graph editor crashes when the last reference to a filter is removed -
<_whitenotifier-3> [scopehal] azonenberg opened issue #430: Add "spectrogram" filter -
<_whitenotifier-3> [scopehal] azonenberg labeled issue #430: Add "spectrogram" filter -