00:00.36 | lennert | i guess that doesn't answer your question.. ? |
00:00.46 | beewoolie | Wait. Let me read that again. |
00:00.58 | beewoolie | Ah, so it does invert. |
00:01.03 | lennert | yes |
00:01.09 | beewoolie | OK. That helps some. |
00:01.15 | beewoolie | I'll need to edit my schematic. |
00:01.21 | lennert | oh |
00:01.25 | lennert | you made this schematic? |
00:01.27 | beewoolie | The problem is that when it floats. |
00:01.33 | beewoolie | Yeah. I drew it. |
00:01.38 | beewoolie | I didn't design it. |
00:01.41 | lennert | oh, okay |
00:01.48 | beewoolie | The notation on the inputs is wrong. |
00:01.54 | beewoolie | It should read TRST and not nTRST. |
00:02.26 | lennert | in jtag, does TRST have a pullup like TDI/TMS/TDO do? |
00:02.27 | beewoolie | The other problem is that when the base pulled low and the collector floats, the nTRST output takes a long time to float. |
00:02.36 | beewoolie | I think it is supposed to. |
00:02.49 | lennert | well |
00:02.52 | beewoolie | What I've read is that the standard requires pull-ups. |
00:03.16 | beewoolie | But, the signal is fast to fall from 3.3V. It takes a long time to rise back again. |
00:03.22 | lennert | the transistor probably floats right away but the trst still has some capacitance |
00:03.35 | lennert | the transistor itself probably has some c-e capacitance as well |
00:03.41 | lennert | -- you'd need a pullup |
00:04.07 | lennert | i.e. the electrons aren't leaking away very quickly when the transistor floats |
00:04.09 | beewoolie | OK. That's what I thought. |
00:04.18 | beewoolie | I'll see about modifying the circuit. |
00:04.30 | beewoolie | I want to look at the other lines, too. it could be that all of them need pull-ups. |
00:04.39 | lennert | i think they should be pulled up on the board.. ? |
00:04.52 | beewoolie | I thought that nTRST was pulled up, too. |
00:04.57 | lennert | i think my s3 board has pullups for tdi/tms/tck on the board... |
00:05.00 | beewoolie | Once I look at the signals, I should be able to tell. |
00:05.17 | lennert | when i looked at the diagram i assumed that trst would have a pullup too, cause i didn't see one |
00:05.29 | beewoolie | What diagram/ |
00:05.29 | lennert | fluke to the rescue :) |
00:05.30 | beewoolie | ? |
00:05.35 | lennert | the pdf i ftp'ed from you |
00:05.44 | lennert | i thought, hey no pullup |
00:05.57 | lennert | so i assumed that trst would have a pullup on the board |
00:05.59 | beewoolie | I don't understand. |
00:06.03 | lennert | okay |
00:06.05 | beewoolie | Oh, right. So do I. |
00:06.10 | lennert | okay :) |
00:06.12 | beewoolie | I'll check. |
00:06.28 | lennert | i'm going home on monday and first thing i'll do in .nl is get my fluke :) |
00:06.49 | beewoolie | Yeah! |
00:06.55 | beewoolie | OK. the clock signal looks really clean. |
00:07.05 | beewoolie | I'm going to increase the speed and see how it looks. |
00:07.27 | lennert | you have a scope, too.. |
00:07.54 | beewoolie | Yeah. I bought it from an auction site. |
00:07.59 | beewoolie | It was about $150. |
00:08.03 | lennert | wow |
00:08.09 | beewoolie | Maybe $200. |
00:08.16 | beewoolie | It was a reasonable price considering. |
00:08.22 | lennert | taksa should get me one of those |
00:08.46 | beewoolie | :-) |
00:09.01 | beewoolie | Looks like the interface works fine, even at the highest speed. |
00:09.10 | lennert | what's the highest speed? |
00:09.45 | lennert | with this transistor, i think trst can toggle at 10mhz or so if you're pushing it :) |
00:10.40 | beewoolie | The parallel port can only do 270KHz. |
00:10.59 | beewoolie | Should a 10K pull-up be OK? |
00:11.28 | beewoolie | .3ma drain. |
00:12.16 | lennert | i think so.. |
00:13.01 | lennert | the transistor has about 4 pF output capacitance |
00:13.27 | lennert | (according to the datasheet) |
00:14.01 | lennert | i.e. 10k will pull it up in 200ns or so |
00:17.37 | beewoolie | lennert: how do you figure that out |
00:17.38 | beewoolie | ? |
00:18.10 | beewoolie | Given a charged capacitor, how long will it take to discharge through a known resistance. Or in this case, charge up. |
00:23.36 | ka6sox | RxC |
00:25.03 | lennert | 4 times the RC time |
00:25.16 | lennert | (as a guideline) |
00:25.21 | lennert | what ka6sox said |
00:25.28 | beewoolie | ka6sox: Hey. |
00:25.45 | beewoolie | it looks like my UBE jtag interface is fine for the task. |
00:26.04 | beewoolie | In fact, it even has nTRST. |
00:27.38 | beewoolie | lennert: so, RC in this case would be 40ns, 4pf*10KOhms. |
00:27.53 | beewoolie | But, it's really an exponential, right? |
00:28.25 | beewoolie | e^-kt or some such nonsense. |
00:28.27 | lennert | yes, it's exponential |
00:28.27 | lennert | yes |
00:28.35 | lennert | R * C is the "RC time" |
00:28.53 | lennert | i.e the time it needs to go to x % of its value |
00:28.58 | lennert | (when discharging) |
00:29.11 | *** join/#openjtag AchiestDragon_ (n=dave@whipy.demon.co.uk) |
00:29.29 | ka6sox | beewoolie, w00t |
00:29.33 | ka6sox | glad to hear that |
00:29.40 | lennert | (63%, which is 1 - 1/e) |
00:29.52 | lennert | see http://en.wikipedia.org/wiki/RC_circuit |
00:29.59 | ka6sox | ya..what he said |
00:30.04 | lennert | heading "Time domain considerations" |
00:30.07 | beewoolie | ka6sox: thx. I built it so long ago... the openwince jtag software doesn't work with it, but I'm not sure why. |
00:30.20 | ka6sox | ah |
00:30.38 | lennert | usually you take 4 or 5 times the RC time to consider it 'done' |
00:31.11 | lennert | (at 5 times the RC time, (1/e)^5 = 0.7% of the charge will be left) |
00:31.22 | beewoolie | I thought that this board was broken. I finally got around to looking at it and the only problem I cna see is this nTRST rise time. |
00:31.36 | beewoolie | lennert: Ah, I see. |
00:45.15 | lennert | any math geeks here? |
00:50.22 | beewoolie | lennert: what's the Q |
00:50.45 | lennert | i want to convert a 50mhz clock to a 1.8432mhz clock |
00:50.56 | lennert | that doesn't really divide nicely |
00:51.02 | lennert | so |
00:51.36 | lennert | what you can do is that you generate a clock every 28 cycles for 73 times in a row, and then generate one every 27 cycles for 576-73 times in a row |
00:51.43 | lennert | that will give you exactly 1.8432mhz |
00:51.45 | lennert | but a lot of jitter |
00:52.07 | lennert | (50 mhz / 1.8432 mhz = 27 + 73/576) |
00:52.29 | lennert | what is the optimal way of dividing the 27 and 28 over the 576 slots? :) |
00:52.41 | lennert | not sure how to phrase it more clearly |
00:52.44 | lennert | does it make any sense? |
00:53.02 | beewoolie | Let me think about that for a moment. |
00:55.15 | beewoolie | How accurate do you want it? Do you want 5 digits? |
00:55.39 | lennert | the optimal way |
00:55.50 | lennert | 28 28 28 28 28 28 28 27 27 27 27 isn't very optimal |
00:56.01 | lennert | 27 27 27 28 27 27 27 27 27 27 27 28 27 27 27 is much better already |
00:57.33 | beewoolie | Are you planning to do this with some flip-flops? |
00:57.38 | lennert | there are 6169588227439311236793765122563713685468080277641945401304843362102684539482558630358706188800 possible ways according to 'bc' |
00:57.42 | lennert | yeah |
00:57.49 | beewoolie | Seems cheaper to get the right crystal. |
00:57.49 | lennert | i have it using the non-optimal way now |
00:57.57 | lennert | hehe |
00:58.01 | lennert | it's for the spartan3 board |
00:58.10 | lennert | i do have a 1.8432mhz extra crystal on it |
00:58.15 | lennert | but that way my projects aren't 'portable' |
00:58.22 | beewoolie | Oh. So the flip-flops are cheap...er...free? |
00:58.33 | lennert | the flipflops come with the spartan3 board ;) |
00:58.37 | beewoolie | :-) |
00:59.12 | lennert | i _think_ the 'greedy' algorithm is optimal |
00:59.18 | lennert | but i can't prove it :-/ |
00:59.50 | lennert | i.e. your input clock is 50mhz, 20ns period |
00:59.53 | beewoolie | That's where I'm going. |
01:00.09 | ka6sox | they are "free" essentially |
01:00.11 | lennert | you want the output to have a 542.53ns period (1843200 hz) |
01:00.12 | beewoolie | You accumulate error until you overflow. |
01:00.39 | lennert | 27 cycles gives you a 540ns period but 28 gives you a 560ns period |
01:00.48 | lennert | so, the first cycle it's better to wait 27 cycles |
01:00.56 | lennert | cause 542.53 is closer to 540 than 560 |
01:01.05 | lennert | but then you have 2.53 error |
01:01.11 | lennert | etc. |
01:01.15 | ka6sox | it sounds like you need a "real 1.8432MHZ" |
01:01.39 | lennert | ka6sox: sure, and i have one, but it doesn't come with the sp3 board by default |
01:01.54 | lennert | ok, i admit, it's a bit of an academical exercise |
01:02.04 | ka6sox | heh |
01:02.21 | beewoolie | After about 8 27's you put in a 28. |
01:02.27 | lennert | yeah |
01:02.45 | lennert | the greedy method gives ---+-------+-------+- |
01:02.49 | lennert | where - is 27 and + is 28 |
01:03.16 | beewoolie | That gives 542.22 |
01:03.24 | lennert | (the distance between +s is sometimes 7, sometimes 8) |
01:03.41 | beewoolie | 1.8442 |
01:03.46 | beewoolie | 1.8443 |
01:04.23 | lennert | as long as you do 73 '28's per 576, you won't drift |
01:04.26 | beewoolie | You'd have to alternate between 8 and 9. |
01:04.39 | lennert | it's just for the jitter |
01:04.47 | lennert | hmmm |
01:05.01 | beewoolie | If that's what you care about, you could just do 73 28's and then the rest as 27's. |
01:05.19 | lennert | yeah, but that gives awful jitter |
01:06.44 | beewoolie | it's not a simple ration. |
01:06.47 | beewoolie | ratio. |
01:07.11 | beewoolie | Is this audio? |
01:07.26 | lennert | 1.8432mhz = 16 * 115200 |
01:07.32 | lennert | it's for serial |
01:07.40 | lennert | (where the jitter also doesn't matter much..) |
01:07.46 | lennert | (as i said, academic exercise..) |
01:07.58 | lennert | (just one of those problems that grab you and won't let go) |
01:12.34 | beewoolie | It's a good one. |
01:13.24 | beewoolie | Seems like you can model it with an error calculation, There is an ideal period and there is the current period. It's a DDA. |
01:13.37 | beewoolie | The trouble is that the target is irrational. |
01:14.04 | beewoolie | You may need 576 data points to get it right. In fact, you could precalc the steps and just code them directly. |
01:16.00 | lennert | well, in hardware you can do it with a 576 bit array if you do it the ube way |
01:16.23 | lennert | i'm sure you can use some tricks to get the logic down |
01:16.23 | beewoolie | It isn't really that bad once you calculate the ideal pattern. |
01:16.26 | lennert | yeah |
01:16.36 | lennert | so, is the greedy algorithm ideal? |
01:16.42 | beewoolie | It's really a short c program. |
01:16.49 | *** join/#openjtag vmaster_ (i=vmaster@p549B6E39.dip.t-dialin.net) |
01:16.52 | beewoolie | I don't know that by name, I suspect yes. |
01:17.31 | vmaster_ | okay... i got disconnect 5 minutes ago - gotta love 24h disconnect |
01:17.32 | beewoolie | You can calculate the frequency at every step. When it goes below the target, add a 28. |
01:17.40 | lennert | i have a c program that calculates the root mean square of the 'jitter' |
01:17.41 | beewoolie | :-) |
01:18.00 | beewoolie | lennert: I don't think it's that hard. |
01:18.05 | lennert | the rms is remarkably better if you do it the greedy way |
01:18.56 | lennert | http://www.wantstofly.org/~buytenh/clocks.c |
01:23.53 | vmaster_ | oh, beewoolie is marc singer? |
01:25.36 | lennert | vmaster_: yes.. |
01:25.51 | lennert | vmaster_: how long did that take you? :-) |
01:26.09 | vmaster_ | mhh... idling this channel for 3 months |
01:26.27 | vmaster_ | i just had a look at his jtag schematic |
01:26.51 | lennert | people should use their real names on irc anyway ;) |
01:28.20 | vmaster_ | hehehe |
01:29.02 | vmaster_ | [02:29] [freenode] -!- Nick dominic is already in use |
01:29.10 | vmaster_ | bad luck, i guess |
01:29.31 | lennert | i don't know vmaster_'s real name either.. :) |
01:29.46 | lennert | well, that's what the realname field in /whois is for :) |
01:31.47 | beewoolie | vmaster_: dood step off! |
01:32.33 | lennert | 'step off'? |
01:32.46 | beewoolie | lennert: Sure, you can use the error and optimize the transitions to get closest. |
01:33.07 | beewoolie | It isn't polite to bandy folks' names about on a public channel. |
01:33.15 | beewoolie | It's kinda like passing people's email addresses. |
01:33.30 | lennert | ah, right. just wasn't sure what 'step off' means. |
01:34.02 | beewoolie | it's still a DDA algorithm. |
01:34.22 | beewoolie | I've done it countless times when drawing pixel lines |
01:34.45 | lennert | oh, Bresenham? |
01:35.00 | lennert | i know that as the bresenham algorithm |
01:35.14 | lennert | wonder if wikipedia has a proof.. |
01:38.46 | lennert | it doesn't |
01:49.58 | beewoolie | lennert: interesting. 10K doesn't work. |
01:50.24 | lennert | as in.. doesn't pull it up fast enough? |
01:50.28 | beewoolie | Nope. |
01:50.28 | lennert | how long is it taking to pull up? |
01:50.46 | beewoolie | the resistor make the curve a little taller and squarer, but not significantly. |
01:51.02 | beewoolie | Presently, the rise takes about 2ms |
01:51.05 | lennert | is this while it's connected to the target? |
01:51.10 | beewoolie | Yeah. |
01:51.13 | lennert | okay |
01:51.26 | lennert | maybe there's a lot of capacitance in the target |
01:51.48 | lennert | does it go faster when it's disconnected from the target? |
01:52.27 | beewoolie | I haven't tried... |
01:53.01 | lennert | can you give that a try? |
01:53.07 | beewoolie | Yeah. :-) |
01:53.14 | lennert | :) |
01:53.17 | beewoolie | Without a pullup, the output signal isn't nice. |
01:53.25 | beewoolie | Hang on. I need to move the ground. |
01:54.00 | beewoolie | It's flat without a pull-up. |
01:54.10 | lennert | how do you mean? |
01:54.10 | beewoolie | Let me try with the pull-up, too. |
01:54.16 | beewoolie | There is no output signal. |
01:54.19 | beewoolie | It's 0v |
01:54.26 | lennert | okay, hmmm |
01:55.55 | lennert | well, without the target connected there's no vcc :) |
01:55.58 | lennert | i think |
01:56.03 | lennert | if i'm reading this thing right |
01:56.30 | beewoolie | Yeah. Just noticed that. I have no pullup without the target. |
01:56.58 | beewoolie | Let me look at the schematic. |
01:57.03 | beewoolie | For the target... |
01:59.02 | beewoolie | 10K pullups on the board for TDI, TMS and TCK. |
01:59.09 | beewoolie | and tdo. |
02:00.00 | lennert | 10k should be fine |
02:00.36 | beewoolie | Wierd. They've got a 10K pull-down on tck as well. |
02:01.04 | lennert | to make it symmetric, probably |
02:01.21 | lennert | so the rise time and fall time will be about equal? |
02:01.53 | lennert | you're probably supposed to drive TCK both ways |
02:02.05 | lennert | (instead of driving it one way and floating it the other way, as with tdi/tms) |
02:02.33 | lennert | you don't know the pullup value on $RANDOM_BOARD |
02:02.38 | beewoolie | Seems like there should only be one. |
02:02.48 | beewoolie | Remember, this is on the board itself. |
02:02.51 | beewoolie | It's got both. |
02:02.52 | lennert | why? it makes sense to me. |
02:02.53 | lennert | yes |
02:03.16 | lennert | let's see what the sp3 board has |
02:03.24 | beewoolie | All it's going to do is create a current drain. |
02:03.37 | lennert | and make the clock more symmetric |
02:03.44 | beewoolie | Oddly, trstn isn't beside the other lines. |
02:04.51 | lennert | on the xupv2p tck only has a pull-up |
02:05.22 | beewoolie | I admit I'm no pro at this, but I don't see how a pull-up and pull-down will make the signal more square. |
02:06.01 | lennert | if you only have a pull-up, the fall time is determined by your drive strength (in mA) and the rise time by the size of the pull-up |
02:06.33 | lennert | drive strength is determined by what's on the jtag adapter, while the pull-up is on the target board |
02:07.00 | lennert | instead of floating it when you want to make TCK rise, you can actively drive it to vcc |
02:07.16 | lennert | but then the rise time and fall time will still be different because you have the pull-up |
02:07.26 | lennert | does that make any sense? |
02:08.16 | lennert | if you use a current source to pull the bus down, the fall will be linear, while the rise will be exponential (due to the pull-up) |
02:10.04 | beewoolie | I'll accept it without understanding it. for now. |
02:10.14 | beewoolie | It looks like the board doesn't have a nTRST signal. |
02:10.24 | beewoolie | No, the *chip* doesn't have nTRST. |
02:10.30 | beewoolie | They wire nTRST to reset. |
02:10.33 | beewoolie | POR. |
02:11.08 | lennert | okay |
02:11.09 | lennert | 'POR'? |
02:11.37 | beewoolie | Power On Reset. |
02:11.41 | lennert | right |
02:11.49 | lennert | hm |
02:11.52 | beewoolie | it's a full reset instead of a warm reset. |
02:12.01 | beewoolie | So, that means that there could be a lot of things using that line. |
02:12.06 | lennert | yeah |
02:12.14 | beewoolie | I think that I have to accept that the JTAG schematic is really OK. |
02:12.27 | beewoolie | I just need to let the reset be slow. |
02:12.28 | lennert | 2ms sounds like it has not 4pF but 40nF capacitance.. well, that's still not unrealistic |
02:12.43 | beewoolie | That's what I thought. |
02:13.03 | beewoolie | Let's see if the data sheet has a value for it. |
02:13.21 | lennert | i wonder why they don't use a simple buffer for TRST |
02:13.35 | lennert | but a transistor instead |
02:13.45 | lennert | probably because they expect the transistor to be off most of the time, hm |
02:15.18 | beewoolie | lennert: do you mean on my schematic? |
02:15.25 | lennert | beewoolie: yes |
02:15.32 | beewoolie | I think because it inverts. |
02:15.40 | beewoolie | And, there are no more lines. |
02:15.44 | lennert | you can get inverting buffers too |
02:15.52 | beewoolie | The 244 has only four outputs. |
02:15.59 | lennert | right |
02:16.07 | beewoolie | Another chip? or a transistor? |
02:16.10 | lennert | but an inverting buffer isn't much bigger than a transistor.. |
02:16.16 | lennert | another chip |
02:16.20 | beewoolie | Lots more pins. |
02:16.23 | lennert | one that can swing trst up as well as down |
02:16.27 | beewoolie | Remember that I made this by hand. |
02:16.32 | beewoolie | I agree. |
02:16.40 | beewoolie | I don't recall where I got the schematic. |
02:16.46 | lennert | if you can drive trst both ways you won't have this problem |
02:16.46 | beewoolie | I must credit it somewhere. |
02:16.55 | lennert | right, you said it wasn't yours |
02:16.55 | beewoolie | I'm not sure it is a problem. |
02:17.05 | lennert | well, no, it's only the reset line |
02:17.07 | beewoolie | I'll have to see how it performs on the slug. |
02:17.19 | beewoolie | Also, the design makes it possible to ignore the voltage of the target. |
02:17.22 | beewoolie | for ntrst. |
02:17.48 | beewoolie | The xlinx design is good this way too, but it has *no* reset capability. |
02:18.51 | beewoolie | ep1220 uses a transistor ...let me check. |
02:20.29 | beewoolie | He uses a FET for the RESET pin and a 245 (buffer) for nTRST |
02:20.38 | lennert | right |
02:22.17 | beewoolie | He's got four outputs and four inputs plus a reset. |
02:22.45 | beewoolie | No, that's wrong. |
02:23.10 | beewoolie | He's got four outputs. two inputs, and two selectable as either. |
02:23.16 | beewoolie | Plus reset. |
02:23.34 | beewoolie | Darn, I wish I had that last resistor. |
02:24.02 | lennert | how do you mean? |
02:25.35 | lennert | http://www.standardics.philips.com/products/lvc/pdf/74lvch322245a.pdf is the datasheet for a 32bit bidi buffer.. the test circuit on page 11 has both a pullup and pulldown on the output, too |
02:29.47 | [g2] | hey lennert ! |
02:29.55 | lennert | [g2]! |
02:30.12 | lennert | how's you? |
02:30.36 | *** join/#openjtag toi (n=pleemans@d5152D12D.access.telenet.be) |
02:32.48 | [g2] | Great lennert thx |
02:34.11 | lennert | what's funny, over here they always say 'normal' instead of 'good' or 'great' |
02:34.18 | lennert | "i'm normal" |
02:34.53 | lennert | well, neither can i :) |
02:35.00 | lennert | they just mean that things are as usual |
02:35.08 | [g2] | that's _why_ we IRC so much :) |
02:35.19 | [g2] | I fully comprehend |
02:35.31 | lennert | to me, 'good' sounds like 'normal'. but they think that 'good' is for when you win the lottery or something. |
02:35.43 | lennert | subtle difference in language that i keep forgetting and making mistakes with |
02:35.49 | [g2] | life is all relative :) |
02:36.13 | lennert | yup :) |
02:36.45 | [g2] | I haven't seen scarface but I heard a line from there "Any day above ground is a good day" |
02:37.02 | lennert | hehe |
02:37.16 | lennert | my profession is more boring than that |
02:38.08 | [g2] | beewoolie what was the sticking boint on the JTAG ? |
02:56.28 | beewoolie | [g2]: I don't really know what was wrong. I think that the openwince jtag software has some sort of problem. |
02:56.54 | beewoolie | [g2]: I've got jtag code that works fine with it. |
02:56.59 | beewoolie | Search me. |
02:57.13 | beewoolie | I'll be finishing my code so that I don't have to worry about openwince. |
02:57.14 | [g2] | Ok your code works but openwince doesn't |
02:57.20 | beewoolie | right. |
02:57.35 | beewoolie | I can read the target ID as well as some of the coprocessor registers. |
02:57.39 | beewoolie | I'm confident that the HW is fine. |
02:57.46 | [g2] | awesome |
02:57.56 | beewoolie | So, it's just a matter of hacking at JTAG to get the rest of the functionality. |
02:58.09 | beewoolie | It's only been...a year since I built the thing. :-) |
02:58.20 | [g2] | are you going to move on to flashing or debugging fiirst ? |
03:00.40 | beewoolie | I'm not sure. |
03:00.47 | beewoolie | I need to get access to the processor register. |
03:00.56 | beewoolie | Once I do that, it should be easy to do flashing |
03:01.07 | beewoolie | Debug is going to take more work, but the basics should be easy. |
03:01.23 | beewoolie | [g2]: I'm working on it. ;-) |
03:01.28 | [g2] | heh |
03:09.35 | lennert | ok, 4am, bed time for me |
03:09.37 | lennert | nice chatting |
03:09.39 | lennert | g'nite all |
03:09.43 | [g2] | sweet dreams |
03:09.44 | beewoolie | nitey nite |
03:09.49 | beewoolie | thanks |
03:09.50 | [g2] | heading home tomorrow ? |
03:22.03 | *** join/#openjtag _AchiestDragon (n=dave@whipy.demon.co.uk) |
03:42.47 | wookey__ | bee-dinner: there is a very similar jtag device schematic here: |
03:43.00 | wookey__ | http://www.lart.tudelft.nl/projects/jtag/ |
03:43.23 | wookey__ | essentially equivalent to the wiggler device but just wired up a differently due to different buffer chip |
03:45.47 | wookey__ | sadly we use one of each for cpld and cpu jtag ports on B2 ('cos that's what the software is set up for and it was easier to get two dongles than fix the software :-/ |
03:46.17 | wookey__ | funny - that pdf comes up on a blakc background for me with gpdf. It didn't used to. |
03:46.55 | wookey__ | bee-dinner: Do you really find it impolite to use real names on IRC? I didn't realise this convention existed. |
03:47.15 | wookey__ | I'm with lennert - much easier if people use real names so you can work out who's who. |
03:47.49 | wookey__ | <PROTECTED> |
03:47.58 | wookey__ | anyway - late here - night. |
03:48.56 | wookey__ | (I mention the lart JTAG dongle just as something to compare with as there seems to be a lot of chat above about the one you have maybe not working? - I've only skimmed though - ignore me if it's all sorted) |
03:51.06 | [g2] | hi wookey__ |
03:53.04 | [g2] | wookey__ how close is the B3 hw ? |
04:09.04 | wookey__ | layout is done, boards are being made now |
04:09.23 | wookey__ | prototype production run is soon (jan/Feb - not exactly sure) |
04:09.59 | [g2] | do you know how much the boards will cost ? |
04:10.10 | wookey__ | nope, not yet |
04:10.44 | [g2] | I'm looking for a guessitimate range in $100 increments |
04:10.49 | wookey__ | I do know there will be a respin as soon as we show that it basically works, as a couple of things got missed off v1 in order to 'get it out'. |
04:11.08 | wookey__ | we were aiming for GBP 100, but have no doubt missed. |
04:11.28 | wookey__ | should be in the GBP 100-200 range, so $150-300 |
04:11.57 | wookey__ | Also depends if you have CPLD or FPGA, how much ram, flash etc fitted |
04:12.24 | [g2] | the design is open right ? |
04:12.30 | wookey__ | yes |
04:12.36 | [g2] | I've got 2 ideas in mind |
04:12.47 | wookey__ | I think the exact licence is still to be decreed, but should be like B2 |
04:13.17 | [g2] | that's something like if you make big changes you need to send them back thingy |
04:13.24 | wookey__ | (which let you build your own, but not produce lots of small variation on the board |
04:13.51 | wookey__ | (otherwise there is no reduction of costs by sharing larger runs) |
04:14.47 | wookey__ | if you want to make changes - design an expansion board, was the idea |
04:14.56 | wookey__ | what are you ideas? |
04:15.21 | [g2] | 1) send you money and have extra board tacked on to the initial run to send out to devs |
04:15.21 | wookey__ | (I really ought to go to bed - it's 4:15 am here and I have woodchopping to do tomorrow morning) |
04:15.41 | wookey__ | OK - that should be no problem |
04:15.56 | [g2] | 2) have the design reviewed and commented on by a professional design house |
04:16.13 | wookey__ | We have been using itechnic.co.uk for that |
04:16.50 | wookey__ | Dave Bisset there is ex-dyson (vacuum cleaners) and thus familiar with production engineering/cost reduction etc |
04:16.53 | [g2] | ok. Well thx for your time. sleep well and get great rest for the woodchopping |
04:17.03 | [g2] | I'll be around tomorrow and the next day etc.... |
04:17.15 | wookey__ | But if you have someone else who wants to get involved that would be worth talking about |
04:17.31 | wookey__ | night |
04:17.31 | [g2] | well I've got a hardware vendor building my boards |
04:17.41 | [g2] | cheers sweet dreams |
05:32.47 | *** join/#openjtag ByronT (n=byron-po@adsl-068-209-197-246.sip.bhm.bellsouth.net) |
06:36.28 | *** join/#openjtag toi (n=pleemans@d5152D12D.access.telenet.be) [NETSPLIT VICTIM] |
06:36.28 | *** join/#openjtag cpage (n=cpage@netblock-68-183-40-90.dslextreme.com) [NETSPLIT VICTIM] |
06:36.28 | *** join/#openjtag [g2-lap] (n=g2@cpe-066-057-008-035.nc.res.rr.com) [NETSPLIT VICTIM] |
06:36.28 | *** join/#openjtag dwery-away (n=dwery@nslu2-linux/dwery) [NETSPLIT VICTIM] |
06:43.46 | *** join/#openjtag cpage (n=cpage@netblock-68-183-40-90.dslextreme.com) [NETSPLIT VICTIM] |
06:43.46 | *** join/#openjtag [g2-lap] (n=g2@cpe-066-057-008-035.nc.res.rr.com) |
06:43.46 | *** join/#openjtag dwery-away (n=dwery@nslu2-linux/dwery) [NETSPLIT VICTIM] |
06:46.01 | *** join/#openjtag toi (n=pleemans@d5152D12D.access.telenet.be) |
06:48.51 | *** join/#openjtag [g2-lap] (n=g2@cpe-066-057-008-035.nc.res.rr.com) |
06:50.33 | ka6sox | Lennert: ping? |
06:57.33 | key2 | g2-lap: still there ? |
06:57.57 | [g2] | key2 yes |
06:58.41 | key2 | g2: in the digilent spec, they say 2 PPC but I wonder if it's possible to run a linux with multi proc |
06:58.55 | ka6sox | yes it is. |
07:00.17 | key2 | and then let's say I wanna add a PCI device (or cardbus), what do I have to do ? compile the driver for the chip and map the PPC's PCI base address ? or is it more complicated than that ? |
07:00.39 | key2 | let's say hostap for a wifi card |
07:00.45 | ka6sox | get a chip that supports the timing requirements of PCI |
07:00.59 | ka6sox | (something that has a PCI interface already) |
07:02.11 | key2 | g2: with your blackdog, can't you use the MMC connector to transform it into a JTAG programmer ? |
07:03.11 | [g2] | key2 that was my whole point |
07:03.46 | [g2] | with a little FPGA magic and possibly off connector hw (cpld or components) it'd be a plug in |
07:03.53 | key2 | yeah but you have to be capable of recompiling a linux on it plus remake all the cores it needs for the USB and sheet like that |
07:04.03 | key2 | wonder if the PHY is inside the fpga or a chip outside |
07:04.18 | [g2] | key2 no it runs Sarge on the PPC core |
07:04.43 | key2 | oh so you won't touch that ? |
07:04.54 | [g2] | you'd just change the section that remaps the MMC connector |
07:05.05 | key2 | yeah but do you have the source ? |
07:05.15 | [g2] | I think it's open |
07:05.33 | key2 | you haven't already tryed to fedle with it ? |
07:05.52 | [g2] | I haven't looked deeply into it, but you can put serial on the thing |
07:06.00 | [g2] | for the PPC core |
07:06.13 | key2 | where do you stick the max232 ? |
07:06.27 | [g2] | the TX/RX pins :) |
07:06.32 | [g2] | and VCC/GND |
07:06.39 | [g2] | they're labeled |
07:06.47 | key2 | oh so you have to open the shit |
07:07.09 | [g2] | well that'd be for development |
07:07.19 | key2 | gotcha |
07:07.33 | [g2] | but I think this may be OBE |
07:07.42 | key2 | fuck it's 8 AM and i haven't slept yet |
07:07.43 | key2 | .- |
07:07.43 | [g2] | Overtaken By Events |
07:08.00 | key2 | gni? |
07:08.19 | [g2] | dunno gni |
07:08.45 | key2 | you have an URL |
07:08.45 | key2 | ? |
07:09.03 | [g2] | URL for blackdog ? |
07:09.07 | key2 | yeah |
07:09.08 | key2 | dev |
07:10.13 | [g2-lap] | http://www.projectblackdog.com/ |
07:11.49 | [g2] | key2 are you a fpga programmer ? |
07:11.57 | key2 | verilog |
07:11.59 | key2 | not fpga |
07:11.59 | key2 | .- |
07:12.01 | key2 | :) |
07:12.19 | [g2] | verilog |
07:12.24 | [g2] | versus vhdl |
07:12.42 | key2 | not vhdl |
07:12.43 | key2 | fuck |
07:12.48 | key2 | i'm so tired .- |
07:12.50 | key2 | :) |
07:12.59 | key2 | that's what I meant |
07:13.23 | key2 | by the way |
07:13.42 | [g2] | you are a vhdl programmer ? |
07:13.58 | key2 | what's the best between cadence or mentor graphic for designing pcb ? |
07:14.11 | key2 | no, i programm in verilog, and not in vhdl |
07:15.01 | [g2] | and do layout right ? :) |
07:15.09 | key2 | :) |
07:15.21 | key2 | i use orcad |
07:15.33 | key2 | but heard mentor graphic was good for that kind of thing |
07:15.57 | ka6sox | I've used PCB and it works pretty well. |
07:16.03 | [g2] | I'm sure it is but last I heard it was quite expensive |
07:16.15 | ka6sox | orcad is. |
07:16.16 | [g2] | kinda like abatron for JTAG or BDI |
07:16.23 | key2 | even P2P ? :) |
07:16.29 | ka6sox | PCB is free and does up to 8 layers. |
07:17.24 | key2 | yeah but not sure it has the same features |
07:18.58 | ka6sox | its pretty good including using netlists. |
07:19.13 | ka6sox | eagle PCB is good too. |
07:19.28 | key2 | yeah for home use :) |
07:20.00 | ka6sox | [g2] what are you wanting to use it for? |
07:20.33 | [g2] | ka6sox the blackdog ? |
07:20.34 | key2 | the blackdog is great but I don't know a single thing usefull that runs on it |
07:21.04 | ka6sox | its a training tool..not a "real" linux system. |
07:21.15 | [g2] | oh I couldn't resist a hardcore running debian on a at least semi-open platform |
07:21.17 | [g2] | for $199 |
07:21.48 | [g2] | I've only waited like 3-4 years to get a softcore/hardcore on a fgpa |
07:22.43 | [g2] | its really a real linux system ~400 MHz PPC 64MB memory USB 2.0 it's more powerful than a slug |
07:22.43 | key2 | g2: what do you do in life ? |
07:23.07 | [g2] | I've got an embedded linux hw/sw company |
07:23.26 | key2 | url ? |
07:23.39 | [g2] | http://www.giantshoulderinc.com |
07:23.46 | [g2] | http://www.giantshoulderinc.com/hw-4533 |
07:23.53 | [g2] | http://www.giantshoulderinc.com/ab3/case.jpg |
07:24.12 | [g2] | There will be a web update real soon |
07:27.13 | key2 | g2: who developed the hardware ? |
07:28.47 | [g2] | A California company |
07:29.02 | key2 | why not yourself ? |
07:29.36 | [g2] | because you weren't around to give me a layout :) |
07:29.45 | key2 | :) |
07:29.54 | [g2] | time-to-market, cost, risk |
07:30.05 | key2 | riskS |
07:30.15 | [g2] | costS |
07:30.31 | [g2] | I've been in the industry for nearly 25 years |
07:30.37 | key2 | really ? |
07:30.41 | [g2] | I'm pretty familiar with stuff |
07:30.46 | key2 | how many people works in your company ? |
07:30.51 | [g2] | me |
07:31.02 | key2 | on your own ? |
07:31.12 | [g2] | and the misses does some bookwork |
07:31.19 | [g2] | yup all on my own |
07:31.37 | [g2] | and a bunch of ppl on the internet helping me informally |
07:31.42 | key2 | why someone would buy this thing and not just a simple PC ? |
07:31.56 | [g2] | <5W power |
07:31.59 | [g2] | POE |
07:32.13 | [g2] | 14-15K Debian apps |
07:32.27 | [g2] | Appliance builds |
07:32.30 | key2 | people don't really care aboput that |
07:32.35 | key2 | about |
07:32.43 | [g2] | ok then it'll fail |
07:33.07 | [g2] | doesn't really matter that much |
07:33.59 | [g2] | the point is I've got a full-on embedded linux company shipping hw and embedded linux distros |
07:34.17 | [g2] | I can easily consult to more than pay for all the hw |
07:34.26 | key2 | yeah |
07:34.40 | [g2] | It's not the "goal" per se |
07:34.51 | [g2] | it's who you become trying to reach the goal |
07:35.55 | [g2] | so now I've got many of the skills I've wanted to work on and the proof is in the product |
07:35.57 | key2 | plagiarism.com ?? |
07:36.01 | key2 | :) |
07:37.12 | [g2] | it's late I'm missing the poirt |
07:37.18 | [g2] | s/poirt/point |
07:37.24 | [g2] | s/poirt/point/ |
07:37.46 | [g2] | it's an open source company |
07:37.55 | key2 | no |
07:38.13 | key2 | talkin about that: it's who you become trying to reach the goal |
07:38.28 | key2 | anyway |
07:38.40 | [g2] | yes, i heard it on a tony robbins tape |
07:38.42 | key2 | at this time of the night/morning i can only talk shit |
07:38.50 | key2 | it's 8:45 and am DEAD :) |
07:38.57 | [g2] | sweet dreams |
07:39.03 | key2 | u |
07:39.05 | key2 | 2 |
07:39.07 | [g2] | soon |
07:43.49 | *** join/#openjtag velinp (n=velinp@c88-163.cable.netissat.bg) |
07:44.06 | velinp | ka6sox: ping |
07:45.02 | *** part/#openjtag [g2-lap] (n=g2@cpe-066-057-008-035.nc.res.rr.com) |
08:40.30 | ka6sox | yo |
08:42.04 | ka6sox | velinp, pong |
09:28.24 | velinp | ka6sox: I sent a summary (including ka6sox) of our discussion to nslu2-linux, (#10838) for comments; |
09:33.23 | velinp | maybe it's late night for you and I was too slow; sorry |
10:01.29 | velinp | hey, just noticed: openjtag turns 1 yr today! happy anniversary and openjtagg'ing to all! |
12:07.19 | lennert | ka6sox: pong |
12:08.11 | lennert | velinp: thanks :) |
12:38.12 | velinp | lennert: hi, still kicking the poor arm platform of yesterday:)? know whenabouts wouter comes back ? |
12:49.21 | lennert | velinp: i took a slight break from kicking that platform :) |
12:49.27 | lennert | no idea when wouter will be back |
12:49.34 | lennert | i saw him post to debian-devel today though |
12:51.41 | velinp | lennert: thanks; will try to find him; you a DD now? |
12:54.00 | lennert | i don't even have an AM assigned yet! |
12:54.09 | lennert | i don't expect to be a DD until 2007/2008, if at all |
13:00.48 | velinp | lennert: but you will be; good to hear that; I saw wouter's post; hope he comes back to my slug by his own will :); time for me to do some work. |
15:01.50 | *** join/#openjtag _AchiestDragon (n=dave@whipy.demon.co.uk) |
15:41.41 | [g2] | morning lennert |
16:31.07 | *** join/#openjtag kas11 (n=kas11@244-120.8-67.tampabay.res.rr.com) |
16:48.27 | *** join/#openjtag ByronT (n=byron-po@nslu2-linux/ByronT) |
16:50.10 | lennert | morning [g2] |
20:06.29 | *** join/#openjtag bullet (n=bullet@66.59.62.81.cust.bluewin.ch) |
20:26.58 | key2 | morning g2 |
20:27.17 | key2 | g2: what do you think about the gumstix |
20:42.31 | [g2] | key2 the gumstix is nice depends on what you want to use it for |
20:45.20 | [g2] | to me the gumstix is more of an embeddable product and an appliance |
21:38.36 | key2 | i see |
23:41.13 | *** join/#openjtag jamie (n=jamie@pdpc/supporter/active/jamie) |
23:44.00 | [g2] | hi jamie |
23:47.36 | jamie | hello g2 |
23:48.12 | [g2] | welcome to openjtag |
23:48.38 | jamie | thanks. I've been here before, but never gotten much involved. |