Created Sat, 08 Oct 2011 10:25:14 +0000 by Ddall
Sat, 08 Oct 2011 10:25:14 +0000
Hello! Here's a crazy idea: A simple Ethernet shield for MAX32: Something cheap, WITHOUT all the stuffs that makes the current network shield so expensive. I looked at the schematics of the current Network shield, and the parts you would need to build such a shield seem pretty cheap. A complete product around 15 - 20$ would make the MAX32 platform even more attractive.
Sat, 08 Oct 2011 19:28:17 +0000
Please, don't forget the UNO32!
The :arrow: MRF24WB0MA from [color=#FF0000]Microchip[/color] at US$25, for example!
Sat, 08 Oct 2011 22:51:28 +0000
The pic32 on the Max has an ethernet MAC built-in, so in theory it should be cheaper to add a shield to than the Uno32. ("In theory" because the price of a Wiznet chip, or a ENC28J60, which have MAC+Phy, is not so much larger than the price of many just-Phy chips. Sigh.) (Max32+phy should have higher performance, though.)
Sun, 09 Oct 2011 17:43:53 +0000
Yes we need a simple network shield for the uno32, Up to now if a ethernet link is needed it's cheaper to use Arduino solutions than chipKIT
Sun, 09 Oct 2011 23:44:30 +0000
if a ethernet link is needed it's cheaper to use Arduino solutions than chipKIT
Why is that a problem? That's the POINT of being Arduino-compatible; you get to use the COTS arduino shields when appropriate and when they work. Since ChipKit Uno32 has more code space than an AVR Arduino, you can probably use the cheaper ENC28J60 based ethernet shields rather than the WizNet-based shields. Other than that, the Uno32 doesn't have any inherent ethernet capabilities that would make it cheaper to add ethernet to than an AVR Arduino.
Mon, 10 Oct 2011 00:55:39 +0000
I made a small plug-in module that goes to the back connectors only. Home made PCB uses LAN8720, resonator, few resistors and capacitors Back row does not have 3.3V (5V only) so run wire or add on-board regulator Having used only back row allows standard shields to be used at the same time on the same level.
But it's kind of difficult to etch and solder 0.5mm pitch package without exposed pins if you don't have hot air tools.
Total cost components is less than $10 I can order blank factory PCBs if any intrest.
Mon, 10 Oct 2011 10:45:55 +0000
Looks like I'm not the only one who doesn't want a 50$ network shield :roll:
Mon, 10 Oct 2011 16:03:20 +0000
doesn't want a 50$ network shield
Of course not. Especially when you can go out and buy an ethernet print server, complete with pretty case and power supply, for far less than that. But actually implementing something is easier said than done.
Total cost components is less than $10
heh. you realize that there's a "rule of thumb" somewhere that says a "reasonable" MSRP for low-volume electronics is "about 5x component cost."
But yeah; if the current cost of the ethernet shield is dominated by (for example) those stackable connectors (and that doesn't really seem unlikely), it would be nice to see a cheaper version without them. SOME shield always has to be "at the top."
Mon, 10 Oct 2011 16:34:20 +0000
it all depends on the actual number behind "low volume" For 10 pieces - yes it is 5x-10x at least For 100 pieces I would say it is closer to 3x or may be 2.5x
In this case you can just use pin headers and RJ45 transformer will become the most expensive component with PHY chip close to it.
So for 100 pieces $20/pc is very realistic.
Thu, 13 Oct 2011 09:24:00 +0000
I made a small plug-in module that goes to the back connectors only.
Would you publish the schematics/PCB?
Like this, right? Needs a regulator, but that's not a bad tradeoff...
Mon, 17 Oct 2011 16:16:35 +0000
This looks nice and cheap :) If only I could buy one... :roll:
Tue, 18 Oct 2011 08:13:00 +0000
This looks nice and cheap. If only I could buy one.
It's looking like there is a fair amount of black magicXXX experience required to lay out those 50MHz PHY chip connections on a 2layer board. I don't think I'm up to the task. :-( (matching chip to magnetics, getting all those RC terminations and power supply decoupling done reasonably, and that 50MHz clock run all the way to the PIC32... grr.)
Tue, 18 Oct 2011 14:29:48 +0000
Like this, right? Needs a regulator, but that's not a bad tradeoff...
Mine is something like this, with connector facing outside Back row has 5V power so adding 5V->3.3V regulator is not a problem You will need few other components: pull up resistors, capacitors etc. Also you can add USB device connector next to it.
Wed, 19 Oct 2011 01:04:40 +0000
I guess that wasn't TOO awful after all, though I don't have lots of confidence in my ability to build this, or whether it will work even if "properly" built. Does anyone want to review? Anyone with fine-pitch SMT experience want to try to build one if I send off the "big bucks" to Seeed to get boards made? The discretes are all 0805, and most are on the bottom so that "skillet reflow" will get most of the components on there. Pretty conservative design rules (9mil, except for the PHY chip itself)...
This uses the PulseJack library from microbuilder.eu; I don't actually know what that is, where to get it, or whether it's actually compatible with the 8270. Scary non-standardization.
Wed, 19 Oct 2011 19:37:31 +0000
Maybe Digilent will make an official shield based on this idea... that was I was trying to get by starting this thread :roll:
Sun, 23 Oct 2011 02:24:48 +0000
I just started looking at PIC32 for a project requiring ethernet. Looking at available PIC32 development boards I am glad I found the chipKit boards. Just ordered a couple UNO32 yesterday to get things going.
I am hoping to design a UNO32 style board with PIC32MX695F512H or PIC32MX795F512H and on board phy and ethernet jack. So far I haven't found anything like it all on a single board. There is the Microchip PIC32 Ethernet Starter Kit II but it does not appear easy to work with due to its fine pitch expansion connector and price.
So I am looking forward to eventually see a single board solution with ethernet.
Sun, 23 Oct 2011 05:40:46 +0000
I suspect if Digilent wants to offer a cheaper network shield, they'd be better off partially populating their current shield with non-stackable connectors.
I don't have enough guts or interest to go much further with my design (yes, I do EAGLE as an idle passtime.) If anyone is interested, I'm willing to pass on what I've got so far, but it's not 100% clear that you wouldn't be better off starting from scratch. The main cost-saving factors are:
Fri, 28 Oct 2011 17:02:02 +0000
You would need crazy mad skill to make such a shield on your own...
Tue, 13 Dec 2011 05:02:51 +0000
mount soapbox
I would like to point out that the Ethernet Shield is a four layer board. If you were to purchase your boards from some place like zoompcb.com with a 2 week turn you would end up paying $67 per blank board. If you did decided to build 100 of these you could get the board cost down to $5 per board, but you would have to outlay $500 just for the blank boards. This doesn't could component cost, assembly cost, time spent trouble shooting the boards when they come in, revisions if there was a mistake (assuming your not perfect). Testing time of each of the 100 boards prior to shipping. Time spent taking the orders, time spent fulfilling the orders. Time spent working on the code to be able make the boards worth using. Time spent setting up message boards and and supporting the product.
Before you know you have a full time job just to build a cheaper Ethernet shield that you end up making less than working at McDonald's yet requires far more skills than a typical Micky D employee.
I have a board that I'm working on that is based on the Max32 + Ethernet Shield and we plan on retailing it for $299. Currently list as the PLAC100 on pontech.com (but the name is subject to change). The first five boards cost me well over $5000 to make.
The UAV100 board is also chipKIT based and is currently listed for $249 but the price will probably go up when everything is said and done.
So, if this if this is a hobby and you want to make some for your friends then you might be able to save a few dollars and learn a whole bunch in the end. I'm sure you will also have lots of fun doing it. But if you want to build a business then I think the price will have to be a little higher.
I think its amazing that Digilent is able to sell the boards for so little and suspect that they are subsidized by Microchip to be able to do so.
umount soapbox
Jacob
Tue, 13 Dec 2011 15:43:25 +0000
Jacob,
Agreed with the point of your post above.
However, my understanding is that Digilent is not being subsidized in any way by Microchip other than the standard volume discounts that are negotiated with any customer that buys enough parts. Obviously, I'm neither Microchip nor Digilent, but I've been on the chipKIT dev team for a while now and have asked this question and gotten the response above.
I can tell you that based on a project going on at my day-job, the discount that a silicon vendor like Microchip can give if you buy enough parts every month is jaw-dropping.
I agree that the price Digilent has set for the boards is incredible, and hats off to Gene and the other guys at Digilent for making that happen. They clearly know what they're doing and do it quite well.
*Brian
Tue, 13 Dec 2011 23:14:21 +0000
Jacob, Brian,
Digilent hasn't been subsidized by Microchip on any of the boards we have done that have Microchip parts on them. They do give us a reasonably good price break on most parts. They gave us a really good price ($0) on the microcontrollers on the first 1000 chipKIT boards, but that is it. Digilent has covered the full cost of all of the engineering and production of the chipKIT boards. We do get tremendous benefit from the marketing value of Microchip promoting the boards.
I was a bit disappointed when our final costs came in and I realized what price we were going to have to charge for the Network Shield. I was hoping it would end up having a lower retail price. The silicon on the board is pretty inexpensive, but the real killer for cost is the connectors. The RJ45 with integral magentics is somewhat expensive, but the dual row header connectors with the long tails are unbelievably expensive. They also have to be hand soldered, so that makes them even more expensive.
We had to do one spin on the board before going to production, so there were two builds of prototypes (at ~$5000 each time). Then we had to make a silk screen change after the first production build, so it was about $300 for new masks for that.
One real killer on the first build was that after we had built 500 boards we were getting about a 40% failure rate in manufacturing test. After several days of scrambling, we tracked it down to the shield connectors. The connectors we had originally procured were cra* and we had to replace them. So it cost us about $9 per board plus the cost of the new (more expensive) connectors to rework the boards.
I don't know the actual numbers, but it wouldn't surprise me to find out that we haven't broken even on that board yet.
Something that you didn't mention in your detailing of the cost of doing business is margin for distributors. Most distributors want at least 30% and some won't touch products unless they can get a 40% margin.
I'd like for us to be able to sell our stuff cheaper, but we have make enough to keep the doors open and pay the salaries.
Gene Digilent
Sat, 17 Dec 2011 13:11:20 +0000
Gene,
perhaps what you guys need to consider is to redesign the board to be more useful as a networking shield, I felt the board was expensive when I purchased one, but I know the costs and work involved to produce and distribute the stuff.
I don't find the CAN interfaces very useful, the 24LC256 is awfully slow and small to have a more elaborated website server running on the board, a 25LC256 or even a 64Mb SST serial FLASH will be much better, even better a microSD slot somewhere.
I was surprised that the Max32 didn't have the 32K crystal and still trying to understand what the use or rationale for D7UG2 comes from, but I can see that on the network shield you added the external 32KHz clock.
Having access to the USB ports is a plus, another thing I'd love to have somewhere are the pads to drop in one of the MRF IEEE802.15.4 modules, but with those connectors walls around three edges kind of hard to find where to put the stuff to make it accessible.
Another option is to have a board that only has the LAN8720 and the 32KHz osc as an option.
Cheers Jorge
Mon, 24 Dec 2012 06:05:26 +0000
I'm a new user, and I'm going to be a dissenting voice, of sorts, here.
While it might be nice to have a lower-cost ethernet-only network board for the chipKit, there are already several inexpensive ethernet Arduino shields that could probably be adopted for this purpose - and one might even be able to find a PHY breakout board or sample board somewhere since the ethernet MAC is embedded in the max32 silicon.
For me, the pair of (quite powerful) CAN controllers in the 32mx795 is a big plus, and a major reason for choosing the max32 in particular. While I am going to be working first with a couple of CAN transceivers on a proto shield, I will almost definitely buy the network shield if/when I want to add ethernet to this project. The CAN transceivers are easy enough to work with by hand, but working with the form factor of the ethernet PHYs I've seen just isn't worth it - and I've done a fair bit of fine-pitch SMD soldering.
While the price is high, there's a ton of functionality on the existing network shield. I plan on integrating an LCD as well, which really needs to be on top, so the pass-through headers aren't something I'd want to lose. In short, the chipKit network shield seems to fill a niche that isn't otherwise occupied, and fill it very well. If all you want is cheap ethernet (maybe with RTCC crystal) there are probably existing cheaper modules you can adapt or hack to get it.
JB
Mon, 24 Dec 2012 06:19:14 +0000
Edit: Delete double-post, probably due to embargo of posting by new user. Sorry!