| wolfspraul | if they are a conservative outfit, they will first have to have an internal meeting about this 'milkymist group' | 00:03 |
|---|---|---|
| wolfspraul | and that will be more like grandpa trying to understand grandson's hiphop than anything else | 00:04 |
| wolfspraul | so I just wait and see :-) | 00:05 |
| wpwrak | yeah, could be a generation issue. who knows :) | 00:05 |
| wpwrak | wolfspraul: http://www.linkedin.com/pub/johnathan-mantey/4/54/192 | 00:07 |
| wpwrak | whatever that means :) | 00:08 |
| wolfspraul | ah you do the spy thing :-) | 00:08 |
| wpwrak | the transparency thing ;-) | 00:08 |
| wolfspraul | looks like an engineering manager, and at lattice only 2.5 years | 00:09 |
| wolfspraul | lemme find my linkedin account ;-) | 00:10 |
| wolfspraul | gotta reach out on all channels | 00:10 |
| wolfspraul | nice, 12 invitations queued up. don't we like to be loved... | 00:12 |
| wolfspraul | now I can just click 'accept' on all of them and feel happier - thank you linkedin! | 00:13 |
| wpwrak | someone should make an antisocial networking site ... "You have 0 friends." ... | 00:17 |
| wolfspraul | message sent, let's see | 00:22 |
| wolfspraul | ah, he just wrote on the list again | 00:22 |
| wolfspraul | "discussion with the uP team" | 00:24 |
| wpwrak | that sounds pretty good | 00:24 |
| wolfspraul | uP = ? | 00:24 |
| wpwrak | microprocessor | 00:25 |
| wpwrak | i.e., lm32 and whatever else they may have or do | 00:25 |
| wolfspraul | yeah. too bad we know so little. | 00:25 |
| wpwrak | looks good. we got the attention of them :) now he can talk to people who can actually do something about the issues | 00:26 |
| wpwrak | interesting. seem that i got myself a soc regression :-( | 00:38 |
| wpwrak | ah well. something to fix later | 00:39 |
| wpwrak | hmm. i update FN and the soc regresses ? strange things are happening ... the funny thing is that the regressions actually make the patch look nice :) | 01:05 |
| cladamwa | good morning | 01:26 |
| wolfspraul | good morning | 01:27 |
| wolfspraul | I can't think of anything better to work on than Milkymist :-) | 01:27 |
| wolfspraul | maybe I should go see a doctor... | 01:28 |
| wpwrak | is wolfgang's water supply downstream from a techno club, with massive consumption of ecstasy ? :) | 01:29 |
| wolfspraul | I'm actually listening to more music during work again now, which is wonderful | 01:31 |
| wolfspraul | I used to do that when my hair wasn't grey yet | 01:31 |
| wpwrak | ;-)) | 01:31 |
| wolfspraul | cladamwa: how's R4 going? | 01:32 |
| wpwrak | i have music running pretty much all the time. helps to focus. | 01:32 |
| wolfspraul | oh wait, I finally want to send the small HDMI snippets from William to the list... | 01:32 |
| wolfspraul | nothing major, but then it's out | 01:32 |
| cladamwa | wolfspraul, need to confirm Q4 circuit & audio again then start to see insertion/detection. | 01:33 |
| wolfspraul | ok | 01:35 |
| wolfspraul | I'm always concerned when I read something like "we are adding protection everywhere" from Sebastien | 01:35 |
| cladamwa | wpwrak, have you seen lastest R4 for Q4 fet? | 01:35 |
| wolfspraul | are we? is our design still simple and easy and cheap? | 01:35 |
| wolfspraul | risk is beautiful | 01:35 |
| wolfspraul | :-) | 01:35 |
| wolfspraul | take risk, enjoy life | 01:36 |
| cladamwa | we added protections on external usbs/audio/ now trying to add in DVI-I's +5V precaution, etc. ;-) | 01:36 |
| wolfspraul | I cannot and don't want to comment on the technical reasoning for adding it or not. | 01:37 |
| wolfspraul | I can just express that I hope we are not driven by fear and 'protecting everything' | 01:37 |
| wolfspraul | we cannot protect everything... | 01:37 |
| wolfspraul | but if we are fixing loose ends now that are clearly not good and were skipped in the intial design rush, then great | 01:38 |
| wolfspraul | cladamwa: you and Werner have to decide :-) | 01:38 |
| wpwrak | fet fet ... looking ... | 01:39 |
| cladamwa | we've known consumer products equipped with each details EMC solutions in design, then DNP them in MP. don't they ? ha..at least i've seen many. especially phone industry, lies everywhere though. | 01:39 |
| wolfspraul | making a good product is hard | 01:40 |
| wolfspraul | saying no is the hardest | 01:40 |
| cladamwa | sure, he, we do our things not related to others. ;-) | 01:40 |
| wolfspraul | I'm not blaming others for 'lies', we just try our best ourselves | 01:40 |
| wpwrak | yeah, if my guesses of how these critters work are right, then this is good | 01:40 |
| wolfspraul | I fully trust you guys, I just read Sebastien's "protection everywhere" comment and remember how much I trust you ;-) | 01:41 |
| cladamwa | wolfspraul, you're right. though. do you remember that we were at SZ, you bought many sch in marketing ? | 01:41 |
| wolfspraul | sure | 01:41 |
| cladamwa | i loved those stuffs/sch. | 01:41 |
| wpwrak | wolfspraul: well, R143 deserves some consideration. if that one gets shorted to ground, it would simply overhead and go up in flames (unless we harden it) | 01:42 |
| wpwrak | s/overhead/overheat/ | 01:42 |
| wpwrak | i use a very similar design to ignite fireworks ;-) | 01:42 |
| cladamwa | i've seen every IO path equipped with all protections, but eventaully not very parts will be monuted. ;-) | 01:42 |
| cladamwa | wpwrak, is that Q4 fet okay from Sebastien suggegsted ? | 01:43 |
| kristianpaul | wolfspraul: do with a M1.., a satellite DBV-S receiver... well just ideas :) | 01:44 |
| cladamwa | wpwrak, protection using purely resistance seems not very good to solve completely. | 01:44 |
| wolfspraul | kristianpaul: hey. not just satellite - *everything*! | 01:44 |
| Action: kristianpaul imagines droping osmo-sdr MCU with a milkymist soc | 01:46 | |
| wpwrak | cladamwa: Q4 looks good to me | 01:51 |
| wpwrak | cladamwa: R143 yeah. i've connected a 47 Ohm to 5 V and let it cook for a while. temperature is about 90 C. | 01:53 |
| wpwrak | should try a 0603 or 0402. watch it burn :) | 01:54 |
| cladamwa | wpwrak, okay. btw, i tried to understand why R232 to be 10K ohm. Does this mean that R143/ 47 ohm in series with R232/10 K ohm, so as voltage picked up on HPD pin is rough 4.98V (i.e. let R232 >> R143) ? | 01:55 |
| cladamwa | wpwrak, when determining a voltage divider, you concern in such way as possible, right ? | 01:56 |
| cladamwa | your R143 may go up in flames when a 0402. ;-) | 01:57 |
| wpwrak | at 0603 it unsolders itself :) | 02:00 |
| wpwrak | (R232) it's just that you don't need a strong pull-down there. do we know the drive strength of the screen when driving HPD ? | 02:01 |
| cladamwa | wpwrak, we probably don't know drive strength from screen side. but we're using fet so that letting fet work is much more determined by voltage not current. | 02:03 |
| cladamwa | also Q4 fet is not transistor, but yes, now we considering R143 consumption. | 02:04 |
| cladamwa | current R143 yes, hard to soldering a 0603. :) | 02:04 |
| wpwrak | (drive strength) yes, the FET needs almost nothing. but the R232 does. | 02:05 |
| wpwrak | 0603 gets 95 C according to my thermometer. but it's probably quite a bit more. | 02:06 |
| cladamwa | wpwrak, let's see page 19 of http://www.ddwg.org/lib/dvi_10.pdf | 02:08 |
| cladamwa | maybe have good answer on current. ;-) | 02:08 |
| wpwrak | 0402 also desolders itself :) doesn't burn, though | 02:11 |
| wpwrak | kinda disappointing | 02:12 |
| wpwrak | well, for my fireworks, i burn something like 20 W in the resistors ... | 02:13 |
| cladamwa | wow 20W | 02:15 |
| wpwrak | ... for about one second ;-) | 02:17 |
| cladamwa | i am thinking that now why we need a R143 ? shall we really need it ? | 02:19 |
| cladamwa | if no R143, how's going to be impacted ? | 02:19 |
| cladamwa | see this http://www.compon-tech.cn/Product%20EMC%20design/DVI%20EMC%20design.pdf | 02:19 |
| cladamwa | it's even without a resistor there, but just a bead connected. | 02:20 |
| wpwrak | so the bead will burn ;-)) | 02:21 |
| cladamwa | ha~ | 02:21 |
| wpwrak | hmm, your DDWG document says that we must be able to provide at least 55 mA | 02:21 |
| cladamwa | now we're worry HPD to connect gnd to avoid, right? how about PTC again? | 02:22 |
| wpwrak | through 47 Ohm, that would result in a "5 V" supply of 2.4 V | 02:22 |
| wpwrak | ptc are pricy but besides that, i have nothing against them | 02:23 |
| wpwrak | at least they won't overheat ;-)) | 02:23 |
| cladamwa | wow. there's expected 2.585V at least on R142 when 55 mA | 02:23 |
| wpwrak | yup | 02:24 |
| cladamwa | s/R142/R143 | 02:24 |
| cladamwa | man~ i like PTC but painful in price. | 02:24 |
| wpwrak | could a BJT do the job of acting as current limiter ? | 02:27 |
| cladamwa | yup, on page 15, monitor draw <= 50 mA, and supply must be min. 55mA, | 02:27 |
| cladamwa | you want BJT? | 02:27 |
| wpwrak | may be cheaper than PTC | 02:28 |
| wpwrak | (and considerably easier to source :) | 02:29 |
| cladamwa | yes, but why not FET again: a FET (~ usd 0.018) v.s. PTC(~usd 0.484) ? | 02:30 |
| cladamwa | man~ i always want to use same p/n if we already have it. hehe... ;-) | 02:31 |
| cladamwa | alright, let's surf a current limiter first. | 02:31 |
| wpwrak | (FET) because BJTs are current-controlled but FETs aren't ? :) | 02:32 |
| wpwrak | something like this: http://www.eevblog.com/forum/projects-designs-and-technical-stuff/simple-current-limiter/ | 02:32 |
| cladamwa | sure, FET it is. but with one or two resistor can do current limiter too. | 02:32 |
| wpwrak | ok. yo know a lot more about FETs than i do :) | 02:33 |
| cladamwa | http://www.vishay.com/docs/70596/70596.pdf | 02:37 |
| wpwrak | "if anyone on the Milkymist team has an oscilloscope" sigh ... | 02:38 |
| cladamwa | no, your knowledge base is much powerfull. ;-) | 02:38 |
| wpwrak | where's a generous investor when you need one ? ... | 02:38 |
| wpwrak | (knowledge) naw, i really don't know much about FETs. need to look at them again. when i tried to figure out how they work, all i came up with is that they don't (needed much larger voltages that we have in modern digital circuits). so there's a bit of a gap between my theory and practice :) | 02:40 |
| wpwrak | (AN103) hmm, they all seem to be low-side limiters. we need high-side. | 02:41 |
| wpwrak | but maybe that's a trivial conversion ? | 02:42 |
| cladamwa | still surfing now... ;-) | 02:45 |
| wpwrak | Vgs = 2.1 V -> Rs = Vgs/Id = 2.1 V/0.05 A = 42 Ohm. that looks familiar :) | 02:50 |
| cladamwa | :) you are right. | 02:52 |
| cladamwa | wpwrak, seems this time I need to do this experiment on my site here to see if 2N7002MTF can limit it. :) | 02:57 |
| wpwrak | yeah :) | 02:59 |
| cladamwa | 2N7002 is continous on 115mA @ 25 C, 73mA @ 100 C, this should be okay. but Rs is still need 105mW | 03:01 |
| cladamwa | wpwrak, finally i knew what you meant that ..."if anyone on the Milkymist team has an oscilloscope" sigh ... ;-) | 03:08 |
| cladamwa | wolfspraul, another real example did successfully on TMDS directly connection to xilinx. confidence goes up. ;-) | 03:11 |
| wolfspraul | good | 03:11 |
| wolfspraul | he did *a lot* of hdmi/dvi work | 03:11 |
| wolfspraul | and is waiting for an R4 to help us even more, which is wonderful :-) | 03:11 |
| cladamwa | he could publicly more activities though, but thanks him surely. ;-) | 03:12 |
| kristianpaul | hmmm, just wondering if lattice is/already targeret the portable devices market | 03:13 |
| kristianpaul | it tought iglo was small, but thise ecp3 from lattice looks quite interesting | 03:13 |
| wolfspraul | what is iglo and ecp3? | 03:16 |
| kristianpaul | iglo is a fpga familly from actel | 03:17 |
| kristianpaul | small packaging, jsut imagine a fpga could fit on the nanonote board .) | 03:18 |
| kristianpaul | and ecp3 seems is a lattice fpga that seems to have also small footprint | 03:18 |
| kristianpaul | http://www.latticesemi.com/products/fpga/ecp3/ecp3minidevices.cfm | 03:18 |
| kristianpaul | just thinking... if we really want to have a fpga based mobile device. well... | 03:20 |
| wolfspraul | s-6 fits on a nanonote-size board easily I think | 03:20 |
| kristianpaul | perhaps | 03:20 |
| kristianpaul | so the only un explored field will be power consuption | 03:22 |
| wpwrak | cladamwa: here's a BJT variant: http://downloads.qi-hardware.com/people/werner/m1/tmp/r142.pdf | 03:24 |
| cladamwa | wpwrak, great. you simulated. ;-D | 03:28 |
| wpwrak | cladamwa: of course, what's not so nice there is that it burns ~300 mW in case of a short | 03:28 |
| cladamwa | wpwrak, btw, i also have seen a nice article: http://www.bonavolta.ch/hobby/files/MotorolaAN1542.pdf | 03:28 |
| cladamwa | see figure. 4 in page 8. ;-) | 03:29 |
| cladamwa | wpwrak, yeah. i also need to calculate later. :) | 03:31 |
| wpwrak | qucs is your friend :) | 03:32 |
| kristianpaul | argh, no again my system get mess after i plug a second jtag-pod | 03:36 |
| kristianpaul | not same time ! | 03:36 |
| kristianpaul | just another one.. | 03:36 |
| Action: kristianpaul brings the laptop | 03:37 | |
| kristianpaul | with mess i mean flterm/gtkterm just show garbage, uart is unhappy.. | 03:37 |
| wpwrak | strange | 03:38 |
| wpwrak | that shouldn't happe | 03:38 |
| wpwrak | n | 03:38 |
| wpwrak | you should just get a /dev/ttyUSB1, i guess | 03:38 |
| kristianpaul | i did: | 03:38 |
| kristianpaul | 1. plug jtag-pod-a check flterm all ok | 03:39 |
| kristianpaul | 2. un-plug jtag-pod-a | 03:39 |
| kristianpaul | 3. plug jtag-pod-b check flterm, got garbage | 03:39 |
| wpwrak | ah, unplugging jtag-pod-a may have done that. perhaps insufficient error checking in flterm | 03:40 |
| kristianpaul | hmm | 03:40 |
| kristianpaul | but also gtkterm shows garbage | 03:40 |
| xiangfu | kristianpaul, you doing that with m1 power-on? | 03:41 |
| kristianpaul | xiangfu: yes | 03:41 |
| kristianpaul | as always | 03:41 |
| xiangfu | kristianpaul, that is dangerous | 03:41 |
| kristianpaul | but i do powerdown and poweron just in case | 03:41 |
| kristianpaul | xiangfu: why? :-) | 03:41 |
| Action: kristianpaul like danger | 03:41 | |
| xiangfu | kristianpaul, I remember in the IRC. have talk about this once time. | 03:42 |
| kristianpaul | ha i jsut did some rmmod modprobe and problem went away :) | 03:42 |
| xiangfu | the jtag-pod is not design for hotplug | 03:42 |
| wolfspra1l | xiangfu fried a jtag-serial board recently | 03:42 |
| wpwrak | ah .. "un-plug jtag-pod-a" .. that's the USB or the M1 side ? | 03:42 |
| kristianpaul | xiangfu: i dint hot pluged it | 03:42 |
| wolfspra1l | and was more than happy/lucky that the m1 was not fried as well :-) | 03:42 |
| kristianpaul | fried, ouch.. | 03:42 |
| wpwrak | barbecue M1, yumm .... | 03:42 |
| xiangfu | yes. m1 still fine. but I broken the jtag board | 03:43 |
| kristianpaul | ah, well you asume jtag-pod-b belongs to another M1 board too ;) | 03:43 |
| kristianpaul | sorry i was not clear | 03:43 |
| xiangfu | kristianpaul, . "un-plug jtag-pod-a" .. that's the USB or the M1 side ? | 03:43 |
| kristianpaul | xiangfu: yes i just touch usb cable for unplug/plug | 03:44 |
| kristianpaul | usbserial and ftdi_sio was all i rmmod/modprobe problem solved ! | 03:44 |
| kristianpaul | http://devbisme.webfactional.com/blogs/devbisme/2011/12/08/big-ugly-pig-bring-phase-1 | 04:06 |
| lekernel | wpwrak: (55mA) so what about connecting the DDC pin directly to 5V? after all, there's already the input PTC | 08:01 |
| lekernel | this xula board is cute. though I'd like to see DDR2/DDR3 (which is cheaper than SDR anyway) | 08:02 |
| xiangfu | lekernel, I am test the William's patch on libbase/uart.c | 08:12 |
| xiangfu | I have a question: why we needs such code: | 08:13 |
| xiangfu | unsigned int stat = CSR_UART_STAT; | 08:13 |
| xiangfu | CSR_UART_STAT = stat; | 08:13 |
| lekernel | because if the UART sends a TX interrupt too fast after RXTX is written, the CPU would reach CSR_UART_STAT = stat; after the UART has sent that TX interrupt and lose it | 08:23 |
| lekernel | this won't happen with the original MM UART because it only sends the TX interrupt after the character has been transmitted on the line, which is plenty of time for the CPU | 08:23 |
| xiangfu | lekernel, thanks. I will commit the code from email "Improvement for libbase - uart.c' by William | 08:35 |
| GitHub182 | [milkymist] xiangfu pushed 1 new commit to master: https://github.com/milkymist/milkymist/commit/832e989a21fd86ba16f1e238b280294bee392564 | 08:38 |
| GitHub182 | [milkymist/master] improvement for libbase uart.c - William | 08:38 |
| Action: xiangfu switch back to rc3 to test these midi controller. | 09:24 | |
| xiangfu | ok. now I can sure we have regression on MIDI stuff. | 09:30 |
| lekernel | xiangfu_: what sorts of regressions? | 09:32 |
| lekernel | because of your USB HID patch? | 09:32 |
| xiangfu_ | <xiangfu> the 'Latest active controller' give nothing. whatever press on LV3. | 09:32 |
| xiangfu_ | <xiangfu> but RC3 do give some values. | 09:32 |
| xiangfu_ | <xiangfu> like "7 (0)". when I switch to channel 8. and move the FX1 under lv3. it also give values uner RC3 images. | 09:32 |
| xiangfu_ | <xiangfu> but nothing when I use the latest daily build. | 09:32 |
| xiangfu_ | <xiangfu> oh. another thing is . my connection is like : LV3 ----> laptop ----> usb-midi-convert ----> m1 | 09:32 |
| xiangfu_ | lekernel, no. I use the MIDI port. | 09:33 |
| xiangfu_ | my connection is like : LV3 ----> laptop ----> usb-midi-convert ----> m1's MIDI port | 09:33 |
| xiangfu_ | under RC3 image : works fine. the 'Latest active controller' do give values | 09:34 |
| xiangfu_ | under recently build: the 'Latest active controller' give nothing. | 09:34 |
| xiangfu_ | great. I have midi2osc that can dump all midi message. :) | 09:38 |
| xiangfu_ | http://downloads.qi-hardware.com/people/xiangfu/tmp/qjackctrl-midi2osc-connection.png | 09:44 |
| xiangfu_ | wpwrak_, is the DX2 work out of box? I try to connect the DX2 with the usb midi converter to my laptop. then connect it to midi2osc. | 10:22 |
| xiangfu_ | what even I press. nothing output under midi2osc. :( | 10:22 |
| xiangfu_ | I connect the LV3 output to DX2 MIDI in. when press on LV3. the DX2 led(sys/mon) blinks | 10:24 |
| qi-bot | The Firmware build was successfull, see images here: http://fidelio.qi-hardware.com/~xiangfu/build-milkymist/milkymist-firmware-20120210-0952/ | 10:40 |
| xiangfu_ | hmm... whateven I do the DX2 just not working. :( | 10:57 |
| xiangfu_ | s/whateven/whatever | 10:57 |
| peterkirn | Hi Milkymisters finally writing up a proper look at the milkymist for createdigitalmotion.com. I wondered, is there any sort of canonical summary of what the MM hardware is built on, that is, the FPGA engineering aspect? | 12:06 |
| wolfspraul | hi Peter, nice to see you here | 12:08 |
| wolfspraul | canonical summary, hmm | 12:09 |
| wpwrak_ | lekernel: (55 mA) hmm, that may still burn the connector. that DVI specicication adam found specifies a contact rating of 1.5 A. | 12:09 |
| wolfspraul | you mean what an fpga is, why it's well suited to deliver high audio and video performance? | 12:10 |
| wolfspraul | but then there's a lot of high-end chips on m1, and what holds it together is the Milkymist software | 12:11 |
| lekernel | hi peterkirn, nice to see you here! | 12:11 |
| wpwrak_ | xiangfu's qjackctl setup looks scary ;-) | 12:11 |
| wolfspraul | (with 'software' I mean including the SoC soft-design) | 12:12 |
| lekernel | what's inside of the FPGA is described http://milkymist.org/thesis/thesis.pdf p. 21 - there have been some slight changes since then, but that's basically it | 12:12 |
| wolfspraul | computers built around fpgas are sometimes called 'reconfigurable computers' | 12:13 |
| lekernel | and the device specs are on http://milkymist.org/wp/for-developers/ | 12:13 |
| lekernel | (device = board etc.) | 12:13 |
| peterkirn | @wolfspraul: exactly. why use an FPGA for milkymist, in terms of performance, why it might be especially applicable to open source hardware, and how the mm pulls it together. But maybe that's even an entirely different article, it occurs to me. I'll certainly link to that thesis and specs! | 12:13 |
| wolfspraul | to express that there is a deeper layer of 'configurability' (=performance) that an fpga can reach and expose, that a non-fpga microprocessor cannot | 12:13 |
| peterkirn | heh, oops, temporarily forgot I was on irc& start speaking twitter/facebook/google+ like all these crazy kids... | 12:14 |
| peterkirn | Yes, and I do point that out. I think deep configurability is the soul of the project. | 12:14 |
| wolfspraul | yes | 12:15 |
| wolfspraul | with software we take it for granted that things are 'configurable', because software can be programmed at will | 12:15 |
| wolfspraul | but when you try to do something innovative with audio or video, you will very quickly run into limitations of even your latest quad-core Intel computer | 12:15 |
| wolfspraul | say something as 'simple' as separating and processing audio signals for a multi-speaker configuration | 12:16 |
| wolfspraul | if you want to filter a bit, or send some signals a few milliseconds earlier or later | 12:16 |
| wolfspraul | (to some speakers) | 12:16 |
| wolfspraul | don't even try with your notebook :-) | 12:16 |
| wolfspraul | but for an fpga that is possible because the reconfigurability (programmability) reaches deeper into signal processing, audio and video signals serving as a good example | 12:17 |
| wolfspraul | lekernel: is this accurate? | 12:17 |
| wolfspraul | I am trying to put these things in words so hopefully it helps peter, or myself, or others :-) | 12:18 |
| wolfspraul | sometimes it's important to spend some time explaining what it is we do... | 12:18 |
| wpwrak_ | wolfspraul: speaking of reconfigurability, in the hypothetical monthly news, the LED array could serve as a simple example for the benefits of having an fpga. in our case, it was a day's work to add a multi-channel PWM with matrix controller that takes care of all the work "in hardware". | 12:18 |
| wolfspraul | wpwrak_: ah, you can also comment on whether you think my attempts at a more high-level explanation are still accurate or not... | 12:19 |
| lekernel | wolfspraul: well, actually I think you can do this delay using a notebook soundcard, since they should support sample-accurate synchronized channels *g* | 12:19 |
| wpwrak_ | wolfspraul: that's functionality nobody foresaw in the original design. in a more traditional setting, you'd have to solve it in software. which means hammering the CPU with interrupts at > 10 kHz. not pretty, if it works at all. | 12:20 |
| wolfspraul | ah yes sure :) I just try to give examples... | 12:20 |
| wpwrak_ | wolfspraul: (interrupts) or build more complex external hardware, e.g., with a dedicated controller, its firmware, etc. | 12:20 |
| lekernel | but generally, yes, that's the idea - have software tightly integrated with the hardware, so that the system has very low latency | 12:21 |
| wolfspraul | if a 1ghz cpu processes an audio signal, let's say at a simple 44khz, that's 'only' about 20,000 machine level instructions per sample, and the entire CPU is busy | 12:22 |
| lekernel | another advantage of the FPGA (which we aren't exploiting a lot at the moment, since GPU's can also do texture mapping etc.) is you can have fully custom compute-intensive algorithms | 12:22 |
| wolfspraul | so if you want to process multiple signals, mix, merge or manipulate them in real-time, and you have a bunch of flexible software running to accomplish that, you quickly run out of power | 12:22 |
| peterkirn | No, that seems fair. I think people will make the comparison to the GPU, but the ability to switch a device on and have access to this immediately is a big deal from an end user / coder-scripter perspective. | 12:22 |
| peterkirn | Yes, I'll leave going into the precise details on why this can have a performance benefit for another time, but should at least be able to provide a big picture. | 12:23 |
| wolfspraul | comparison to gpu is fair | 12:23 |
| lekernel | and of course... all integrated, small, low power consumption, reconfigurable, open source at the HDL level | 12:23 |
| wolfspraul | yes, the entire Milkymist One runs at ca. 5W | 12:23 |
| wolfspraul | I think it's fair that people compare with notebooks and say we can do all this and 100 times more with a macbook pro and quartz composer/resolume, etc. | 12:24 |
| wpwrak_ | perhaps we could say it's like a GPU (i.e., dedicated and optimized hardware) but with the flexibility to try new things. so if we can think of, say, some new video effect, we can add it with little effort. plus, we can upgrade all units in the field. | 12:25 |
| wolfspraul | the Milkymist One today can only do a few things, but they are implemented in an *entirely* different way than that notebook+vj software setup, creating the promise of more different things to come in the future | 12:25 |
| wolfspraul | people always like 'different' (in reality few do ;-)), and the Milkymist One really is about as different from a notebook+vj software as can be :-) | 12:26 |
| peterkirn | Well, the biggest problem with the computer is that we don't have a simple mixer on the market that has, say, an analog input (connected to the MM) and a couple of HDMI or DVI ins (for your MBP) ;) But that's another conversation. | 12:26 |
| peterkirn | I imagine a lot of us would love to use both. We just have to make that affordable mixer happpen... | 12:26 |
| wolfspraul | so the few that actually do appreciate innovation, should be interested in diving into Milkymist land and trying to understand what's going on | 12:26 |
| wolfspraul | that's a fair angle imo | 12:26 |
| lekernel | wolfspraul: I think you're underestimating it. especially after we'll have added support for small video loops with FFMPEG. | 12:26 |
| peterkirn | Oh, I can see it being practical even without having to have it be different for the sake of it, but you'll see my angle, and then I'm sure the discussion will continue after that. | 12:26 |
| wpwrak_ | there are two effects: a) if lowers our barrier of entry. we can do things companies like nvidia do, but without needing their budget to roll out new chips, etc. b) it also makes the units in the field more long-lived, since they can be upgraded | 12:26 |
| wolfspraul | lekernel: one by one. I am trying to break this down to a 'fair' level where a reporter/journalist can feel comfortable. | 12:26 |
| wolfspraul | not for the sake of it | 12:27 |
| peterkirn | I'm going to lunch, but continue to discuss. ;) | 12:27 |
| wolfspraul | there is a point, but it's not as immediately obvious yet as say the swoosh effect on an iPad | 12:27 |
| wolfspraul | definitely not different for the sake of it | 12:27 |
| wpwrak_ | (like nvidia) of course, we don't have the same raw processing power. there are limits :) | 12:27 |
| wolfspraul | none of the people driving Milkymist forward are that superficial | 12:27 |
| peterkirn | right, but we can at least say there would be a significant investment of money and time before you matched the milkymist on latency. there's not an instant-on solution. | 12:28 |
| peterkirn | I wouldn't underestimate the interest of folks like me who would play with both, again, which is why I'm still jonesing for that mixer& ;) | 12:28 |
| wolfspraul | yes, both - agree | 12:28 |
| wolfspraul | instant-on, definitely | 12:28 |
| wolfspraul | and that's another aspect we hopefully keep improving | 12:28 |
| wolfspraul | the m1 should make vj'ing or playing with visuals as easy as nowhere else | 12:29 |
| wolfspraul | but I think it's 'should' today, not there yet | 12:29 |
| peterkirn | I didn't expect to be successful, but I did go on a failed mission to find a VGA - composite converter yesterday, since now I'm stuck with a MBP that, at best, I can only get to DB15. | 12:29 |
| wolfspraul | (I'm conservative again, lekernel probably thinks it is already :-)) | 12:29 |
| peterkirn | and yes, you can plug a milkymist into a V4& ;) | 12:29 |
| wolfspraul | ah too bad. we have some tentative plans on including such a converter cable in the box. | 12:30 |
| peterkirn | (if anyone knows where to find such a thing in berlin between today and & uh, tomorrow & I'm all ears!) | 12:30 |
| peterkirn | ha good idea. | 12:30 |
| lekernel | peterkirn: I have one (which I actually used with a V4 ;), can lend it to you if you want | 12:30 |
| wolfspraul | I sourced a few and have one myself, and Sebastien has one. but the fpga doesn't support it yet :-) :-) | 12:30 |
| peterkirn | lekernel: really? okay. ;) I'm eager to try it out; I haven't used one yet. | 12:30 |
| wolfspraul | lekernel: is this working now? vga->composite adapter cable? | 12:30 |
| peterkirn | okay, late for a lunch meeting but back soon to finish this up! | 12:30 |
| lekernel | wolfspraul: we're talking about a different thing here. the thing that the FPGA doesn't support is the passive adapter, but scanconverters work just fine. | 12:31 |
| peterkirn | (no reason a db15 > composite shouldn't work. when I have more time, I should try to make one&) | 12:31 |
| wolfspraul | ah ok | 12:31 |
| wolfspraul | that's a good small example of reconfigurability :-) | 12:31 |
| wolfspraul | even though m1 currently does not support it - yet | 12:31 |
| lekernel | peterkirn: we're also thinking about supporting "dumb" cables, eg connect the red channel of VGA to the composite signal, and the FPGA does some DSP to generate a proper signal | 12:31 |
| lekernel | peterkirn: at the moment there's a proof of concept that displays a checkerboard pattern, but it's not integrated in the SoC (no framebuffer etc.) | 12:32 |
| wolfspraul | btw, advance notice: I found a projector that consistently refuses to display the m1's vga signal saying "unknown signal" | 12:32 |
| wolfspraul | more about that soon | 12:32 |
| wolfspraul | (for peterkirn - that's a bug we have been hunting for a while...) | 12:33 |
| wolfspraul | the projector displays from a notebook just fine | 12:33 |
| wolfspraul | but from m1 - all black | 12:33 |
| wolfspraul | xiangfu will have that projector on his desk in a few days... | 12:34 |
| wpwrak_ | that'll be interesting to hunt down | 12:52 |
| wolfspraul | peterkirn: in whatever article or blost post you write, if you want to mention where Milkymist was used before, one event where I have all data is from 8static in Philadelphia | 13:06 |
| wpwrak_ | GRRR. the patch editor chokes in double quotes in the file !@*%#$! | 13:06 |
| wpwrak_ | s/in/on/ | 13:06 |
| wolfspraul | that was the group Cheap Dinosaurs [cheapdinosaurs.com] with visuals by no carrier [no-carrier.com] at 8static on November 11, 2011 [8static.com]. video is here http://www.youtube.com/watch?v=ZcO3qslAmTo | 13:07 |
| wpwrak_ | and i was wondering how the heck i would have managed to cause a memory corruption that made the editor come up empty when trying to load my patch ... | 13:08 |
| wolfspraul | and photos here http://www.flickr.com/photos/su1droot/sets/72157628115535073/with/6390867571/ | 13:08 |
| wpwrak_ | nice search engine optimization :-) | 13:09 |
| wolfspraul | Milkymist was actually used all around the world on selected occasions, but that's one case where I have all data - event/date, group, vj, video, photos... | 13:09 |
| wpwrak_ | video quality is pretty good. more of this ! :-) | 13:10 |
| wpwrak_ | a world map with little links to gig would be nice :) | 13:10 |
| wolfspraul | here's another video of the same group/event/performance (but I think the 1st vid is better) http://www.youtube.com/watch?v=u_oqaPflyTw | 13:13 |
| wolfspraul | wpwrak_: well, it's a terrible amount of work to get all details and pieces together | 13:13 |
| wolfspraul | and it's not like we have a rolling stones biography project where multiple people sit down for how long to piece every memorable day together... | 13:13 |
| wpwrak_ | would be nice to have a description on the videos. one with strategic keywords, like "milkymist" :) | 13:13 |
| wolfspraul | so I rather have a few cases like that cheap dinosaurs one, and that serves as proof that all this fancy Milkymist stuff actually works and actually makes sense for serious performing artists | 13:14 |
| wpwrak_ | maybe we need to get the message out more that we want footage | 13:15 |
| wolfspraul | if I find another such well documented case, of course I will pay attention, but I'm not running after people asking them to tell me where and when and how they used mm - infinite time sink... | 13:15 |
| wolfspraul | I'm still waiting for the infamous 25gb agata video for example :-) | 13:15 |
| wpwrak_ | the dinosaurs are a good reference point. set the bar reasonably high, so it won't be all blurry smartphone cam pictures | 13:15 |
| wolfspraul | video art historians will still regret the loss of that file in 500 years | 13:15 |
| wpwrak_ | ;-) | 13:16 |
| wpwrak_ | the effects are stronger on the 2nd video | 13:16 |
| wolfspraul | I just wanted to give peterkirn one well documented case if he wants to mention that in a side sentence somewhere | 13:16 |
| wpwrak_ | and nicely matched to the tone of the music | 13:16 |
| peterkirn | Hi, back :) | 13:40 |
| peterkirn | wolfspraul: just looking did you have that case? not everything needs to fit into this article, either. | 13:40 |
| peterkirn | oh, 8static& that's the same as this NO CARRIER video? | 13:41 |
| lekernel | peterkirn: we have two NO CARRIER videos, one with Nullsleep and another with 8static | 13:41 |
| lekernel | s/with/at | 13:42 |
| GitHub166 | [flickernoise] wpwrak pushed 31 new commits to direct-midi: http://git.io/IHWgkw | 13:43 |
| GitHub166 | [flickernoise/direct-midi] stimuli: make MIDI channel numbers one-based - Werner Almesberger | 13:43 |
| GitHub166 | [flickernoise/direct-midi] stimuli: added MIDI channel wildcard (WIP) - Werner Almesberger | 13:43 |
| GitHub166 | [flickernoise/direct-midi] test/stim: update for one-based channel numbers - Werner Almesberger | 13:43 |
| peterkirn | lekernel: yeah, I know, got both of them in the story already ;) | 13:48 |
| peterkirn | hopefully next time, better documentation! | 13:48 |
| wolfspraul | peterkirn: which case? sorry don't understand | 13:52 |
| wolfspraul | Thihi_: thanks a lot for your feedback, super tired now, so I will reply tomorrow | 13:53 |
| wolfspraul | of course I agree with you, just thinking about the quickest path to improvement now... | 13:53 |
| wolfspraul | more tomorrow, n8 | 13:53 |
| wolfspraul | (Thihi is a Finnish journalist who is reviewing the m1 for some potential articles/posts...) | 13:54 |
| wpwrak_ | perhaps we should explain the FPGA more in terms of it being the technology that allows us (as opposed to big industry) to build such a device, without trying too much to praise FPGA benefits on a general basis | 13:54 |
| wpwrak_ | the FPGA has of course more benefits (as well as weaknesses), but they're more of a secondary nature | 13:55 |
| GitHub194 | [migen] sbourdeauducq pushed 1 new commit to master: https://github.com/milkymist/migen/commit/945d655d451f513095a1ea3cb603b7ebe97b3b89 | 14:26 |
| GitHub194 | [migen/master] bus: ASMI hub (untested) - Sebastien Bourdeauducq | 14:26 |
| larsc | what's ASMI? | 14:27 |
| lekernel | the next memory system | 14:27 |
| lekernel | with out-of-order execution and parallel execution of commands to different DRAM banks | 14:28 |
| larsc | ah, ok | 14:34 |
| lekernel | http://createdigitalmotion.com/2012/02/milkymist-is-digital-visual-synthesizer-and-processor-built-on-sophisticated-open-source-hardware/ | 14:55 |
| lekernel | peterkirn: very good article ! | 14:56 |
| Scopeuk_ | the first picture makes me want to buy a fader fox | 15:08 |
| wpwrak_ | Scopeuk_: it conveys a sense of power, doesn't it ? :-) | 15:17 |
| peterkirn | Thanks! | 15:19 |
| peterkirn | heh, had that impact on me& well, have at least one faderfox. | 15:19 |
| wpwrak_ | nice borg reference ;-) | 15:24 |
| peterkirn | heh& right? Surely the Borg would have liked FPGAs. Organic-hybrid FPGAs, anyway... | 15:25 |
| wpwrak_ | very good article, thanks a lot ! | 15:27 |
| wpwrak_ | playing with FPGAs actually does convey a bit of a not-from-this-world feeling :) well, at least for me, who's new to this. but i was pleasantly surprised that it wasn't all that hard to get something done with the FPGA. | 15:30 |
| Action: Scopeuk_ has always cheeped out on midi controllers, I built my current one. one of these daysi'll buy a proper one | 15:31 | |
| wpwrak_ | for midi controllers, i think a combination of LV3 and something with pad would be ideal | 15:34 |
| wpwrak_ | a pad is great for interactive use. allows you to move. | 15:35 |
| wpwrak_ | the LV3 is more an engineering instrument. something where you quietly fine-tune some parameter. | 15:35 |
| Scopeuk_ | touch pad or drum/trigger ? | 15:35 |
| wpwrak_ | touch pad | 15:36 |
| Scopeuk_ | fair enough | 15:36 |
| Scopeuk_ | I had a lot of fun with an M-audio trigger finger | 15:36 |
| Scopeuk_ | although there is an old maudio keyboard with an optical 3 controller sensor on one end x25 photon iirc, which looks interesting to me, but i havetn been able ot find one at the right price | 15:37 |
| wpwrak_ | i have a korg kaossilator pro, the faderfox lv3, an icon icreativ, and a korg nanoKONTROL2. i love the feel of the kaossilator for interactively manipulating effects | 15:37 |
| wpwrak_ | but the kaossilator doesn't have nearly enough controls | 15:37 |
| Scopeuk_ | very nice | 15:37 |
| Scopeuk_ | how do you find the nanoknotrol? | 15:37 |
| wpwrak_ | i don't like it. feels too cheap and fragile. | 15:38 |
| Scopeuk_ | fair enough, I suppose they are fairly cheep | 15:38 |
| wpwrak_ | i like the solid feel of the icreativ. and also that it packs a lot of different controls though some clever multiplexing | 15:39 |
| Scopeuk_ | ever used any of the behringer b control series? | 15:39 |
| wpwrak_ | however ... they screwed up on the pad resolution. it's only 16x8, whatever you do | 15:39 |
| wpwrak_ | (nanokontrol) yeah, it feels like disposable hardware ;-) if i went to a place where i'd be certain it'll get stolen, the nanokontrol would be my choice for bringing along ;-) | 15:40 |
| Scopeuk_ | :) | 15:40 |
| wpwrak_ | i looked at the behringer specs but didn't get any. they look nice enough, though. | 15:40 |
| Scopeuk_ | one of my friends has a bcr2000 have never had a chance to play with it, it's much much bigger than I was expecting | 15:41 |
| Scopeuk_ | feels supprisingly solid | 15:41 |
| peterkirn | I have utterly abused the KORG nano series. | 15:42 |
| peterkirn | They may be light, but they're not fragile. ;) | 15:42 |
| wpwrak_ | novation have some interesting combi devices. with basically all types of controls, rotary, fader, drum pads, keyboard, joystick, x/y pad | 15:42 |
| peterkirn | people like the BCRs, one of the better Behringer pieces and there are a lot of them around. | 15:42 |
| peterkirn | I'll probably be using an Akai micro keyboard; those are handy and easy to drop in a bag. | 15:42 |
| wpwrak_ | might be interesting to have at least from a developer's point of view. alas, a bit pricy. and also a little big. | 15:43 |
| Scopeuk_ | they hold there value though, I'm looking to pick one up when I can grab a second hand one for £50 | 15:43 |
| Scopeuk_ | did anyone get to play with the modular system that showed up a few years back? was just a back pane covered in jack sockets and then various control types you could just plug in in any configuration | 15:43 |
| wpwrak_ | haven't heard of that | 15:45 |
| Scopeuk_ | let me see if i can find it | 15:45 |
| wpwrak_ | modular usually means expensive, though | 15:46 |
| Scopeuk_ | http://www.gearjunkies.com/product_info.php?cPath=6_76&products_id=13703 | 15:47 |
| Scopeuk_ | I believe the company may have gone bust | 15:47 |
| wpwrak_ | yeah, a domain grabber has taken over mawzer.com. that's death in the internet age. you don't rest in peace - they drag out your corpse and publicly defile it. | 15:51 |
| wpwrak_ | could be that things like lemur and then i/droid pads killed it | 15:52 |
| wpwrak_ | in the future, i want a pad on milkymist, too. then we don't need to chase low-end midi controllers :) | 15:53 |
| Scopeuk_ | midi is daisy chain, it should be a doddle to build a set of simple single control/bank devices that chain, and it should be possible to add a sysex signal to auto conf the channels in a predictable way | 16:08 |
| wpwrak_ | new | 16:12 |
| wpwrak_ | oops | 16:12 |
| Scopeuk_ | damn it I'm now trying to design the above in my head | 16:38 |
| wpwrak_ | ;-)) | 16:38 |
| Scopeuk_ | i'll get as far as a design, then i'll look at how much encoders cost again and decide against it | 16:41 |
| Scopeuk_ | too many test patterns of late, ff ft tt tf looks more normal than ff ft tf tt | 16:48 |
| wpwrak_ | are encoders that expensive ? the ones i see at digi-key are. but i wonder if that's a good place to look for such things. | 16:49 |
| Scopeuk_ | i've never seen much below £1:50 per unit | 16:51 |
| wpwrak_ | that doesn't sound too bad | 16:52 |
| Scopeuk_ | which is fine until you want 10 or 20 | 16:52 |
| wpwrak_ | with the UKP nice and cheap ... ;-) | 16:52 |
| Scopeuk_ | think i'm still stuck in a student mind set | 16:52 |
| Scopeuk_ | :P | 16:52 |
| Scopeuk_ | some of us have to live here | 16:52 |
| GitHub78 | [migen] sbourdeauducq pushed 1 new commit to master: https://github.com/milkymist/migen/commit/ef436a1ec9dbc37b5cfa0ec64627238f9a5f59fd | 16:54 |
| GitHub78 | [migen/master] bus/asmibus: add get_slots, fix get_fragment - Sebastien Bourdeauducq | 16:54 |
| Scopeuk | wpwrak_, http://pastebin.com/dZLMBxG4 | 17:27 |
| Scopeuk | think I might have a little too much spare time | 17:28 |
| nrlnd | : D | 17:28 |
| wpwrak_ | lots of hardware. seems that you have some MIDI equipment that isn't easily configured and you need to adapt to it ? | 17:35 |
| Scopeuk | wpwrak_, well mine isn’t that easily configure I have to re-program the eeprom, | 17:41 |
| Scopeuk | but that also only outputs usb | 17:41 |
| Scopeuk | idea was just a minimal auto conf, I'm sure I would run into lots of problems with implementation | 17:42 |
| Scopeuk | I'll put it down for a day or so and look it over | 17:44 |
| lekernel | peterkirn1: welcome on board ;) | 18:00 |
| GitHub92 | [flickernoise] wpwrak pushed 4 new commits to direct-midi: http://git.io/Qvn32w | 18:44 |
| GitHub92 | [flickernoise/direct-midi] ptest: new options -v and -m to exercise MIDI message processing - Werner Almesberger | 18:44 |
| GitHub92 | [flickernoise/direct-midi] test/stimin: added MIDI injection to test cases - Werner Almesberger | 18:44 |
| GitHub92 | [flickernoise/direct-midi] stimuli: ignore value 0 in midi_proc_diff_button - Werner Almesberger | 18:44 |
| kristianpaul | fpgaminer: Hi | 23:30 |
| kristianpaul | Your Verilog_Xilinx_Port works with a 50Mhz clock right? | 23:30 |
| kristianpaul | I just got in sinthezised on my M1, but i'n sending data (64bits) and got no reply | 23:31 |
| kristianpaul | i just noticed you use hash_clk for serial_transmit clock | 23:31 |
| kristianpaul | but this is set to 50Mhz freq | 23:31 |
| kristianpaul | but main pll is dividing clock freq by 2 it seems | 23:33 |
| kristianpaul | ah this could ne, that i should enable the commented line with CLK2X.. | 23:34 |
| kristianpaul | btw your code is very nice well documented ;-)!! | 23:34 |
| kristianpaul | ah indeed thats the problem, ClkFrequency should be 25000000 | 23:43 |
| --- Sat Feb 11 2012 | 00:00 | |
Generated by irclog2html.py 2.9.2 by Marius Gedminas - find it at mg.pov.lt!
