| wpwrak | wolfspra1l: btw, how do you go about properly pricing components, for volume production ? do you ask adam to get quotes ? do you ask around yourself ? | 05:27 |
|---|---|---|
| wolfspra1l | not sure what you mean | 05:28 |
| wolfspra1l | mostly I focus on how I can actually get the components, not the price | 05:28 |
| wolfspra1l | at low volumes (that can go into the thousands), any sort of overhead associated with sourcing is typically more expensive than a component price difference | 05:29 |
| wolfspra1l | other than that I think you know the whole sourcing situation quite well | 05:29 |
| wolfspra1l | digikey | 05:29 |
| wolfspra1l | official distributors in us, taiwan | 05:29 |
| wolfspra1l | and then the special case china where 47% of all global semiconductors are being sold nowadays | 05:30 |
| wolfspra1l | so maybe it's not the special case after all, we can assume it to go >50% this year? :-) | 05:30 |
| wolfspra1l | in china competition is so crazy intense that I can get the 'volume' price down to volume of 1, more or less | 05:31 |
| wpwrak | i'm more thinking of millions. find out what chips really cost in the end. digi-key sometimes only go to tiny quantities, like 100 units. most distributors stop at a few 1000 - a bunch of reels, not more. | 05:31 |
| wolfspra1l | well, that's a different matter | 05:31 |
| wpwrak | that's why i ask ;-) | 05:31 |
| wolfspra1l | in those quantities depending what it is you need to hookup with the manufacturer anyway | 05:31 |
| wolfspra1l | for forecasting etc. | 05:31 |
| wolfspra1l | otherwise you can order 5 million but you won't get it, plain and simple | 05:32 |
| wpwrak | sure :) | 05:32 |
| wolfspra1l | for large quantities, really large - contact manufacturer directly | 05:32 |
| wpwrak | but do you have a means to obtain a rough estimate of the price without actually committing to anything ? | 05:32 |
| wpwrak | could be digi-key price * factor | 05:32 |
| wolfspra1l | if they don't know you they will probably not waste 1 minute talking with you | 05:32 |
| wolfspra1l | no, difficult | 05:33 |
| wolfspra1l | you can calculate from raw material costs upwards | 05:33 |
| wolfspra1l | digikey is a totally different world then | 05:33 |
| wolfspra1l | if I want to order 10 million lcd controllers from ilitek, for example - they assume I know all raw materials and manufacturing costs anyway | 05:34 |
| wolfspra1l | there is not 1 buyer who would not know | 05:34 |
| wolfspra1l | with a little research you can find solid pricing on wafer starts etc from tsmc, umc, etc. | 05:35 |
| wolfspra1l | if you are really buying in those quantities you know all these numbers | 05:35 |
| wolfspra1l | and that's the bottom of your negotiation :-) | 05:35 |
| wolfspra1l | then there are lots of chips also with special manufacturing, special materials, special whatever | 05:35 |
| wpwrak | well, the materials per se are a given. it's most likely tssop-20. the rest depends on how much value the respective company puts on what they do with the silicon. | 05:36 |
| wolfspra1l | but again, if you are a serious potential buyer for such chips in millions of units, you will also know exactly | 05:36 |
| wolfspra1l | in large quantities, you can definitely think raw materials/processing up | 05:36 |
| wolfspra1l | large = millions at least | 05:36 |
| wolfspra1l | there's a lot of competition | 05:36 |
| wolfspra1l | so why would they not take your money, even if their profit is just 10k usd... | 05:37 |
| wolfspra1l | the only exception are chips with really unique and valuable ip | 05:37 |
| wpwrak | so, attiny167 or attiny87. first, it would be interesting to know if there's a significant difference between the two. | 05:37 |
| wolfspra1l | they are priced by whoever holds the ip :-) | 05:37 |
| wolfspra1l | you would need to talk to atmel sales | 05:37 |
| wpwrak | (difference ... in terms of price) | 05:37 |
| wolfspra1l | how about die space etc? | 05:38 |
| wolfspra1l | the problem is most likely those sales guys will not talk to you, or me | 05:38 |
| wolfspra1l | because they know as much as we 'inquire', most likely we will not buy | 05:38 |
| wpwrak | same package. the rest would have the be found out by x-ray if we're expected to bring that information to the table | 05:38 |
| wpwrak | well, or solvents :) | 05:38 |
| wolfspra1l | I cannot talk to atmel sales anyway | 05:39 |
| wpwrak | why ? | 05:39 |
| wolfspra1l | they won't answer | 05:39 |
| wpwrak | you tried ? | 05:39 |
| wolfspra1l | no | 05:39 |
| wolfspra1l | but it's always the same, and makes lots of sense | 05:39 |
| wpwrak | or did you make enemies already ? | 05:39 |
| wolfspra1l | a buyer for large quantities doesn't just appear on the street | 05:39 |
| wolfspra1l | no enemies in business | 05:39 |
| wolfspra1l | if someone can make money from me, they will overcome any difficulty, right? :-) | 05:40 |
| wolfspra1l | you can try | 05:40 |
| wolfspra1l | but like I said, all semiconductor sales I know work with short lists | 05:40 |
| wolfspra1l | short lists of known customers | 05:40 |
| wpwrak | there are a lot of people who can pay a few USD 100k (for, say, 1 M chips). they can't possibly expect to know them all. | 05:40 |
| wolfspra1l | for the type of question you have, I'm afraid only the semiconductor sales people themselves can help | 05:41 |
| wolfspra1l | the distributors won't | 05:41 |
| wpwrak | atmel isn't that niche. i could imagine a very restrictive policy for things like telephiny chips, but an avr ? | 05:42 |
| wolfspra1l | they have a set discount table, and if someone can really outgrow that they will outgrow the distributor anyway | 05:42 |
| wolfspra1l | I have not seen 'restrictive policies' | 05:42 |
| wolfspra1l | it's just about money | 05:42 |
| wolfspra1l | how many sales staff do they have, and how much revenue does each have to generate each day? | 05:42 |
| wpwrak | restrictive policy = they won't talk to you | 05:42 |
| wolfspra1l | and how many chips have to be sold to make that happen | 05:42 |
| wolfspra1l | etc. etc. | 05:42 |
| wolfspra1l | no of course not | 05:43 |
| wpwrak | of course, at 1 M you may still be too small for the manufacturer | 05:43 |
| wolfspra1l | they just want to find the money, the real money | 05:43 |
| wolfspra1l | and as we know well, there must be about 1 million times or more talk in the world than money, right? :-) | 05:43 |
| wolfspra1l | the nature of money maybe | 05:43 |
| wolfspra1l | yes, maybe | 05:43 |
| wolfspra1l | it's all just pennies | 05:43 |
| wolfspra1l | but does it matter to you? | 05:44 |
| wolfspra1l | you can work with raw materials costs anyway | 05:44 |
| wolfspra1l | or do you depend on some specific atmel ip? | 05:44 |
| wolfspra1l | everything can be changed... | 05:44 |
| wolfspra1l | so it's just "how much money does it cost you to switch to ip from vendor x?" vs "how much above cost of raw materials is atmel?" | 05:44 |
| wolfspra1l | it can't be that much | 05:45 |
| wpwrak | i need to know which MCU i can use. changing MCUs is a bit effort in this case. | 05:45 |
| wpwrak | s/bit/big | 05:45 |
| wolfspra1l | there are few semiconductors with enough strength to really stay away from this raw material comparison | 05:45 |
| wolfspra1l | or maybe only for a short period of time, like a few months | 05:45 |
| wolfspra1l | yeah a 'bit' effort, fine | 05:45 |
| wpwrak | it's not the MCU's IP by itself, but what's built on top. and that's very architecture-specific | 05:45 |
| wolfspra1l | but you talk about money | 05:45 |
| wolfspra1l | effort so what | 05:45 |
| wolfspra1l | your effort | 05:45 |
| wpwrak | time is money | 05:45 |
| wolfspra1l | if you save/make money, it may still be worth it | 05:46 |
| wolfspra1l | yes, but it sets a limit as to how much above raw materials atmel can be | 05:46 |
| wolfspra1l | and can be sustainable | 05:46 |
| wolfspra1l | sustainably | 05:46 |
| wolfspra1l | the digikey prices are on a totally different level | 05:46 |
| wpwrak | so you're saying Atmel don't put an unusual margin on top of raw materials, for their AVRs ? | 05:46 |
| wolfspra1l | in large quantities, if there is a dollar difference, that's a lot of engineering potential to change chips :-) | 05:46 |
| wolfspra1l | I doubt they can | 05:47 |
| wolfspra1l | you tell me | 05:47 |
| wolfspra1l | switching costs how much? | 05:47 |
| wpwrak | i.e., going to TI, microchip, or NXP would yield roughly the same result ? | 05:47 |
| wolfspra1l | 100k usd? | 05:47 |
| wolfspra1l | so that's very little | 05:47 |
| wolfspra1l | it depends on the chip and what's special about it and what of that you really need | 05:47 |
| wpwrak | switching costs = 2-4 months | 05:47 |
| wolfspra1l | yeah, so all zero in relation to anything 'large volume' | 05:47 |
| wpwrak | well, de facto betting on the wrong chip would mean the project will not materialize | 05:48 |
| wolfspra1l | understand | 05:48 |
| wolfspra1l | but again, there is nothing special in those chips from your perspective | 05:48 |
| wolfspra1l | I wouldn't know why you couldn't get them < 1 usd | 05:49 |
| wolfspra1l | what's your price now? | 05:49 |
| wolfspra1l | even less, why not. but the issue at some point becomes whether it's worth for anyone to talk about another 10 cents less | 05:49 |
| wolfspra1l | ilitek sales told me they are now negotiating over 0.5 cents :-) | 05:50 |
| wolfspra1l | but think about it, even for a volume of 10 million, that's just what... | 05:50 |
| wpwrak | about 1 USD. but digi-key only go up to 100 units. | 05:50 |
| wpwrak | they don't even have T&R | 05:50 |
| wolfspra1l | 50k usd, right? | 05:50 |
| wolfspra1l | can you imagine at a volume of 10 million (!) how many things can go wrong that will immediately cost more than 50k usd? | 05:51 |
| wolfspra1l | so someone having the patience/guts to sit down and squeeze those 50k usd out, that is quite extreme | 05:51 |
| wpwrak | it's all noise + offset :) | 05:51 |
| wolfspra1l | maybe changing one word in the headline of an advertisement will make a bigger difference | 05:51 |
| wolfspra1l | but anyway they will still sit down and try to squeeze those 50k usd out as well | 05:52 |
| wolfspra1l | if it's 1 usd at digikey, then I am sure in really large volume it can easily be 50 cents or less, 30, 20 | 05:52 |
| wolfspra1l | packaging becomes a big issue then | 05:52 |
| wolfspra1l | and someone needs to think that taking the time to sit down with you and talk about another 5 cents reduction is worth it for anyone (!), including you | 05:53 |
| wpwrak | heh, good point :) | 05:56 |
| wpwrak | i also need some LEDs. they're trickier, because there may be .cn equivalents that are cheaper than the ones digi-key have. | 05:57 |
| wpwrak | and a USB connector. same thing. | 05:57 |
| wolfspra1l | not sure, wait | 05:57 |
| wolfspra1l | usb connector - don't buy one from a mainland chinese company | 05:57 |
| wolfspra1l | the only advantage they have is that it's really really lousy quality | 05:57 |
| wolfspra1l | so pretty much any serious brand company will buy one from a taiwanese or japanese manufacturer | 05:58 |
| wpwrak | it doesn't have to have to be premium quality | 05:58 |
| wolfspra1l | which may very well have factories in china, of course | 05:58 |
| wolfspra1l | and they may sell them primarily in china | 05:58 |
| wolfspra1l | but just trying to steer you away from mainland brands in this case | 05:59 |
| wolfspra1l | their niche is the rusty niche ;-) | 05:59 |
| wolfspra1l | for those connectors... | 05:59 |
| wpwrak | yeah, you mentioned connector issues before :) | 05:59 |
| wolfspra1l | but of course, for those you can get pricing only in china | 05:59 |
| wpwrak | oh, if the thing corrodes after a year in the field, that's perfectly fine | 05:59 |
| wolfspra1l | what if it corrodes after 2 weeks? :-) | 05:59 |
| wolfspra1l | stay with taiwan/japan brand on those | 06:00 |
| wpwrak | that would be bad. should last at least 3 months | 06:00 |
| wpwrak | (of light use) | 06:00 |
| wolfspra1l | japan are the best but also a little more expensive, so then taiwanese | 06:00 |
| wolfspra1l | leds don't know | 06:00 |
| wpwrak | alright, leds and connector aren't so critical at the moment, because they're trivial to swap | 06:01 |
| wolfspra1l | sure so it's again raw materials price | 06:01 |
| wolfspra1l | and processing/manufacturing overhead | 06:01 |
| wpwrak | mcu is a different kind of beast | 06:02 |
| wolfspra1l | that's why the mainland companies cannot take market share from taiwan/japan brands on the connectors | 06:02 |
| wolfspra1l | they could never recoup their investments | 06:02 |
| wolfspra1l | so they focus on the rusty niche instead :-) | 06:02 |
| wolfspra1l | hire a few guys from the street, and start producing something... | 06:02 |
| wpwrak | the leds have one special requirement: they should be in the "knee" for brightness vs. price. the brighter the better - as long as the price doesn't explode | 06:02 |
| wolfspra1l | leds don't know, this is an area of extreme investment and competition right now | 06:02 |
| wpwrak | yeah :) | 06:03 |
| wpwrak | at least the color is simple: red :) and they're in a standard package, 0603. no fancy lenses or anything. | 06:03 |
| wpwrak | actually, even the color doesn't matter that much, as long as the forward voltage is around 2 V and they're bright. but red seems to be the most likely candidate. nothing else is as good as red. | 06:08 |
| wpwrak | one issue with the mcu is that it'll run outside specs. the oscillator is supposed to deliver 8 MHz but can be calibrated up to something like 13 MHz. i need 12 MHz. but it seems one can get away with that fairly reliably. | 06:12 |
| wpwrak | and that ability exists in a whole range of similar chips (by atmel), so it's unlikely to change over night | 06:13 |
| wpwrak | particularly not in such an ultra-conservative product line | 06:14 |
| qi-bot | [commit] Werner Almesberger: bacon/prog/Makefile: remove references to C2, inherited from f32xbase (master) http://qi-hw.com/p/wernermisc/7fe9ae1 | 13:04 |
| gargon | hello. sorry for this newbie question: how can I set syntax highlighting in vim for eg. python language? For the :syntax on command, I get "cant open file /usr/share/vim/vim73/syntax/syntax.vim" | 18:22 |
| pang | did you ask google? | 18:26 |
| viric | in the nanonote you mean? | 18:26 |
| viric | gargon: it should tell why it can't open it. | 18:27 |
| gargon | it simply does not exist | 18:28 |
| viric | you should ask people of the distribution you use. | 18:29 |
| viric | or whoever was in charge in your system of having put those files there. | 18:29 |
| gargon | viric: it is the "official" distribution | 18:33 |
| viric | ah, so in the nanonote? | 18:33 |
| gargon | viric: yes, on the nanonote, I forgot to say it | 18:51 |
| viric | gargon: maybe #openwrt people can help better, I don't know. I don't use openwrt in the nanonote. | 20:37 |
| gargon | viric: just for curiosity: what are you using on it? | 20:38 |
| viric | nanonixos | 20:39 |
| --- Wed Aug 1 2012 | 00:00 | |
Generated by irclog2html.py 2.9.2 by Marius Gedminas - find it at mg.pov.lt!