Bugzilla – Attachment 717 Details for
Bug 1487
hantek-6xxx driver loses samples on 12MB boundary
Home
|
New
|
Browse
|
Search
|
[?]
|
Reports
|
Help
|
Log In
[x]
|
Forgot Password
Login:
[x]
irc log
irc.log (text/x-log), 12.86 KB, created by
L29Ah
on 2021-01-18 17:55:56 CET
(
hide
)
Description:
irc log
Filename:
MIME Type:
Creator:
L29Ah
Created:
2021-01-18 17:55:56 CET
Size:
12.86 KB
patch
obsolete
>[07.01.2020 20:57:50] uwe_: /me has set the subject to: Portable, cross-platform, Open-Source signal analysis software (logic analyzers, scopes, multimeters, and more) | www.sigrok.org | Questions? Just ask and wait, we're not always around >[07.01.2020 20:58:50] L29Ah: have anyone succeeded running a fx2lafw oscilloscope w/o missing data at >=8MHz? >[07.01.2020 21:00:01] uwe_: which one specifically? >[07.01.2020 21:00:16] uwe_: (and for how long, I guess :) >[07.01.2020 21:03:34] sigbot: 1 commit(s) pushed to pulseview on branch 'master': >[07.01.2020 21:03:35] sigbot: https://sigrok.org/commits/pulseview/0672b79..feda6c6 >[07.01.2020 21:03:36] sigbot: Soeren Apel: clang-tidy and clazy proposals >[07.01.2020 21:03:40] uwe_: abraxa: merged, thanks! >[07.01.2020 21:04:11] L29Ah: uwe_: for whatever, but i'm mostly interested in dds120; for one second at least >morri_od has left >morri_od (morri_od%irc.freenode.net@irc.l29ah.blasux.ru/5495a034@p5495A034.dip0.t-ipconnect.de) has entered as none/participant >morri_od has left >[07.01.2020 21:12:40] uwe_: sounds like it should work, can give it a try if you like >[07.01.2020 21:12:56] uwe_: as per usual with FX2 it depends on all kinds of things, though >[07.01.2020 21:23:16] L29Ah: i've checked all the ports of my laptop and also gave pulseview realtime priority via chrt, no luck >[07.01.2020 21:23:48] L29Ah: tested by poking my fingertip with a probe and then looking for sudden jumps in the sampled signal >[07.01.2020 21:24:35] L29Ah: https://pstbin.ru/f/41599/ >Yamakaja has left: Quit: Bye >[07.01.2020 21:25:28] L29Ah: also no usb devices plugged in except the embedded webcam that is not used >[07.01.2020 21:26:38] L29Ah: PV consumes <10% CPU during acquisition >[07.01.2020 21:27:58] abraxa: L29Ah, what's the period of that signal >[07.01.2020 21:28:07] abraxa: is it 20 ms, aka 50 hz >[07.01.2020 21:28:24] abraxa: and omg my keyboard doesn't work right, can't type question marks >[07.01.2020 21:28:29] L29Ah: yea 50Hz >Yamakaja (yamakaja%irc.freenode.net@irc.l29ah.blasux.ru/~yamakaja@vps.pub.yamakaja.me) has entered as none/participant >[07.01.2020 21:29:16] abraxa: a 50 hz signal of 2 volt amplitude is suspicious >[07.01.2020 21:29:17] L29Ah: a finger is the most convenient signal source for oscilloscope testing >[07.01.2020 21:29:28] abraxa: i'd say the preamp stage is shot, if there is one >[07.01.2020 21:29:30] L29Ah: i don't think it's suspicious >[07.01.2020 21:29:45] L29Ah: lemme check with my soviet analog 'scope >[07.01.2020 21:30:11] abraxa: well, 4 volts peak-to-peak is a lot for coming out of your finger, where do you live >[07.01.2020 21:30:12] abraxa: lol >[07.01.2020 21:30:49] abraxa: you could also attach it to a constant voltage of, say, 1 volt >[07.01.2020 21:31:16] abraxa: heck, even a battery would do >[07.01.2020 21:32:29] L29Ah: the analog scope tells the amplitude is 2V2 >[07.01.2020 21:32:40] L29Ah: 4V4 peak-to-peak >[07.01.2020 21:32:51] abraxa: wow >[07.01.2020 21:32:53] L29Ah: i live in a flat in Moscow >[07.01.2020 21:33:04] abraxa: apparently with really strong ac around you >[07.01.2020 21:33:18] abraxa: or you're just a really good antenna >[07.01.2020 21:34:04] L29Ah: don't you get the similar results if you stick the probe in ur thumb? >[07.01.2020 21:34:31] abraxa: xD >[07.01.2020 21:34:33] L29Ah: anyway the problem is that the damn scope loses samples so sometimes i can't measure stuff properly >[07.01.2020 21:34:58] abraxa: does acquisition stop when that happens >[07.01.2020 21:35:02] L29Ah: no >[07.01.2020 21:35:03] abraxa: i thought it was supposed to >[07.01.2020 21:35:05] L29Ah: as you can see >[07.01.2020 21:35:24] abraxa: does it have two channels >[07.01.2020 21:35:36] L29Ah: what >[07.01.2020 21:35:46] abraxa: the dds120 >[07.01.2020 21:35:52] L29Ah: the scope has two channels, but i've disabled the second one in the hope it will improve things >[07.01.2020 21:36:37] abraxa: that's the question, isn't it... maybe you still have the usb bus load of 2x8 bit, meaning 2x8mhz in reality >[07.01.2020 21:37:12] abraxa: either way, a fifo underrun should be detected and acquisition stopped >[07.01.2020 21:37:19] abraxa: uwe_, do you know why that doesn't happen >[07.01.2020 21:37:37] abraxa: does only fx2la support that maybe >[07.01.2020 21:37:51] abraxa: (the driver, not the firmware) >[07.01.2020 21:37:51] L29Ah: if i set it to 48MHz the skips appear a lot more frequently >[07.01.2020 21:38:02] L29Ah: but i think even 2x8MHz should fit into an USB2 bus >[07.01.2020 21:39:02] abraxa: with 24mhz being hard to achieve for a lot of people, i'd say 16mhz could already be an issue for some configurations >[07.01.2020 21:39:41] abraxa: did you check lsusb -t to see if there's anything else connected to that root hub >[07.01.2020 21:40:09] L29Ah: set to 4MHz, see a couple of skips as well >[07.01.2020 21:40:11] uwe_: let me give the dds120 a try real quick >[07.01.2020 21:41:35] L29Ah: abraxa: http://dpaste.com/1W11Y9Q >[07.01.2020 21:42:03] uwe_: L29Ah: you're using git sigrok and the fx2lafw-based firmware, I assume? and the hantek-6xxx libsigrok driver? >[07.01.2020 21:42:18] L29Ah: yes, yes >[07.01.2020 21:42:20] L29Ah: eys >[07.01.2020 21:42:30] abraxa: L29Ah, dev 24 is the dds120- >[07.01.2020 21:42:34] abraxa: isn't it >[07.01.2020 21:42:35] L29Ah: yes >[07.01.2020 21:42:58] abraxa: so it shares the bus with other usb devices, can you move it to the other root hub maybe >[07.01.2020 21:43:20] L29Ah: even if i plug it into the usb3 port it doesn't leave the bus01 >[07.01.2020 21:43:42] abraxa: wtf >[07.01.2020 21:43:48] abraxa: that's not even usb3 >[07.01.2020 21:43:57] L29Ah: should i probably ask linux to forget about all the other devices? can they be re-identified later? >[07.01.2020 21:44:19] abraxa: L29Ah, unloading the kernel modules should do >[07.01.2020 21:44:27] L29Ah: what modules i have no modules >[07.01.2020 21:44:43] abraxa: you have a completely static kernel then, do you >[07.01.2020 21:44:45] L29Ah: oh turns out i do >[07.01.2020 21:45:24] L29Ah: removed the modules, no luck at 8MHz >[07.01.2020 21:45:57] L29Ah: 4MHz too >[07.01.2020 21:46:25] L29Ah: 1MHz has a glitch too >[07.01.2020 21:46:25] uwe_: 8MHz works fine for me length-wise, I can easily get 15 seconds or so without the sampling stopping >[07.01.2020 21:46:48] L29Ah: i don't care about the length, i care about my signal being silently skipped in the middle >[07.01.2020 21:47:02] L29Ah: like a few times in the screenshot i've posted >[07.01.2020 21:47:45] uwe_: do you know if this happens regularly? I'm seeing one or two such items when I'm hooked up to a function generator, but it could also be an issue with the fngen, not sure >gruetzkopf has left: Remote host closed the connection >[07.01.2020 21:49:17] L29Ah: dunno, looks quite random, as i don't have a proper slow triangle gen to spot it >gruetzkopf (gruetzkopf%irc.freenode.net@irc.l29ah.blasux.ru/~quassel@daemon.gruetzkopf.org) has entered as none/participant >[07.01.2020 21:49:36] L29Ah: so it can silently throw out a whole period w/o me noticing >[07.01.2020 21:50:11] uwe_: I'll grab a few .sr files with triangle, sec >mrec has left: Ping timeout: 240 seconds >mrec (mrec%irc.freenode.net@irc.l29ah.blasux.ru/~markus@sundtek.de) has entered as none/participant >[07.01.2020 21:51:00] abraxa: L29Ah, lsusb -t shows the dds120 is alone on the bus then >[07.01.2020 21:51:04] abraxa: does it >[07.01.2020 21:51:11] L29Ah: abraxa: ofc no >[07.01.2020 21:53:33] abraxa: Finally have my full keyboard back :D >[07.01.2020 21:54:32] L29Ah: https://pstbin.ru/f/46019/ meanwhile look at this cutie :3 >niska has left: Quit: Leaving >martling-win (martling-win%irc.freenode.net@irc.l29ah.blasux.ru/~Martin@32.136.187.81.in-addr.arpa) has entered as none/participant >[07.01.2020 21:57:05] martling-win: uwe_: working on an msbuild project file for libserialport. One issue is that config.h and all the package/lib version stuff comes from autotools, so there's going to be a certain amount of duplication if we want to do this. >[07.01.2020 21:59:08] benimautner: the scan() function works properly already >[07.01.2020 21:59:11] martling-win: also, we don't even have libserialport.h to include until libserialport.h.in is processed >[07.01.2020 21:59:22] L29Ah: ok i did /sys/bus/usb/devices â echo > 1-8/remove >[07.01.2020 21:59:23] L29Ah: now i have zero clue on bringing it back; usb host doesn't seem to have a `scan` node like ata does >[07.01.2020 22:00:08] benimautner: (to avoid confusion) from my driver >niska (niska%irc.freenode.net@irc.l29ah.blasux.ru/~niska@static.38.6.217.95.clients.your-server.de) has entered as none/participant >[07.01.2020 22:00:22] L29Ah: after removing both devices like that now my 'scope is alone but still losing samples >[07.01.2020 22:01:46] abraxa: L29Ah, sysfs/procfs aren't persistent, so a reboot will bring them back if nothing else will >[07.01.2020 22:02:00] abraxa: Still odd your device is losing samples like crazy, though >[07.01.2020 22:02:09] abraxa: Did you try a different USB cable, too? >[07.01.2020 22:05:11] L29Ah: relaced the cable, no effect, but i've noticed a peculiar thing: >[07.01.2020 22:05:45] L29Ah: the data comes into pulseview in a couple of packets around 5.5Ms long >[07.01.2020 22:06:04] L29Ah: and the glitch occurs between the "packets" >[07.01.2020 22:06:10] uwe_: martling-win: the latter would only apply to git, tarballs ship the generated libserialport.h, I think. but I see what you mean. >[07.01.2020 22:06:28] L29Ah: as if something buffers it improperly before drawing on screen >[07.01.2020 22:07:28] L29Ah: observed on 4 and 8MHz >[07.01.2020 22:07:58] martling-win: uwe_: i think we need it to work from a git version, really. >[07.01.2020 22:10:43] abraxa: L29Ah: Hm, no idea what that could be, honestly... >blight (blight%irc.freenode.net@irc.l29ah.blasux.ru/~greg@80-109-10-222.cable.dynamic.surfer.at) has entered as none/participant >blight has left: Changing host >blight (blight%irc.freenode.net@irc.l29ah.blasux.ru/~greg@reactos/developer/blight) has entered as none/participant >bvernoux has left: Quit: Leaving >[07.01.2020 22:30:23] Depili: hmm, my crystal ball says that I might end up writing a IBIS (VDV 300 tram/bus/whatever information bus for ticket validators, line number displays etc) PD in the future... assuming that I get a working implementation on the display units that recently fell into my lap... >[07.01.2020 22:31:17] Depili: quite simple protocol on top of uart or tcp/ip, just finding documentation is little PITA >[07.01.2020 22:36:24] Dar1us: Depili: neat! >[07.01.2020 22:37:42] uwe_: L29Ah: ok, I think I have some reproducible test setup where I'm seeing issues every 786ms or so at 8MHz sampling rate >[07.01.2020 22:38:01] Depili: So far the best protocol documentation I have found: https://web.archive.org/web/20170427164053/http://cybox.ib-luehning.de/ibis/ibisdata.html >[07.01.2020 22:38:38] Depili: but I might have a way of getting my hands on the VDV 300 standard specification for the protocol... >[07.01.2020 22:38:39] L29Ah: every 6.288Msa i'd say! >[07.01.2020 22:38:45] uwe_: L29Ah: when it happens, whatever it is, it affects both channels >[07.01.2020 22:40:09] L29Ah: that's almost 6 * 1024 * 1024 >[07.01.2020 22:40:35] L29Ah: 5.9967041015625 in fact >[07.01.2020 22:42:22] L29Ah: src/hardware/hantek-6xxx/protocol.h:#define MAX_PACKET_SIZE (12 * 1024 * 1024) >[07.01.2020 22:43:07] L29Ah: no idea what to do with this peculiar coincidence tho >andy-burns has left: Ping timeout: 260 seconds >[07.01.2020 22:47:17] uwe_: that "/* Weird flushing needed for filtering glitch away. */" also looks weird, guess someone needs to properly analyze what the firmware and the hantek-6xxx driver do *exactly* >[07.01.2020 22:54:42] uwe_: "sr: session: bus: Received SR_DF_ANALOG packet (6291456 samples).", so yeah. one of those per channel yields the 12*1024*1024 in total >c4017w (c4017w%irc.freenode.net@irc.l29ah.blasux.ru/~c4017@209.52.68.8) has entered as none/participant >[07.01.2020 23:04:45] L29Ah: it's interesting that hantek driver initiates one bulk transfer while fx2lafw driver does multiple at once >[07.01.2020 23:06:21] L29Ah: src/hardware/fx2lafw/protocol.h:#define NUM_SIMUL_TRANSFERS 32 >[07.01.2020 23:06:21] L29Ah: in fact >[07.01.2020 23:06:21] uwe_: if that's the case it's probably part of the problem. for FX2 you want to always queue multiple transfers in the OS at the same time >[07.01.2020 23:06:22] Dar1us: probably because the fx2lafw driver has been optimised more for performance.. >c4017w has left: Client Quit >[07.01.2020 23:06:37] Dar1us: issuing multiple transfers lets the OS know so it can schedule actual transfers >[07.01.2020 23:06:39] uwe_: send_chunk() takes 70-90ms, with sr_session_send() commented out it's 14-19ms. >c4017w (c4017w%irc.freenode.net@irc.l29ah.blasux.ru/~c4017@209.52.68.8) has entered as none/participant >[07.01.2020 23:06:50] L29Ah: why are fx2lafw and hantek drivers separate btw? >[07.01.2020 23:06:51] uwe_: (for the DD120 on my setup) >c4017w (c4017w%irc.freenode.net@irc.l29ah.blasux.ru/~c4017@209.52.68.8) is now known as c4017 >[07.01.2020 23:07:48] uwe_: L29Ah: different devices and features etc., but yeah, various similarities as well
You cannot view the attachment while viewing its details because your browser does not support IFRAMEs.
View the attachment on a separate page
.
View Attachment As Raw
Actions:
View
Attachments on
bug 1487
: 717