Why is Ethernet not a basic interface?

LeafLabs Garden » Support » Maple Hardware Support

Why is Ethernet not a basic interface?

(13 posts) (5 voices)

Started 5 years ago by mlindeblom
Latest reply from mlindeblom

mlindeblom
Member

I have researched various Ethernet options such as W5100, ENC28J60 and native interfaces on ARM chips from TI and NXP. ST is expected to have one next year.

I have run some code on the mbed LPC1768 using the IWIP library.

The question I have is based on current mouser.com pricing why is the LPC1768, the LAN8720a and a magjack not in more common use.

My application is small residential HVAC control systems. Ethernet wiring by an certified electrician is easy to obtain. Home Depot stocks all the wiring parts and supplies. The local electronics store has the router.

If I try this with RS485, CANbus or LINbus, I need specialists and don't get electrical isolation. The price of the micro-controller hardware is about $7 less.

USB doesn't work over great distances and without human intervention reliability is questionable.

I would like to avoid electro-smog and wireless options.

Comments Please.

Posted 5 years ago #
mlindeblom
Member

Could the moderator move this post to Community, Developement please

Posted 5 years ago #
soundcyst
Member

the particular STM32 variant that LeafLabs are using doesn't have any dedicated ethernet hardware. the stm32f107 is part of their connectivity line, and has a MAC controller on chip, but still requires a chip to convert to the proper PHY layer.

Posted 5 years ago #
mlindeblom
Member

The LAN8720a is a PHY chip for $1.74 at Digikey.
The LPC1768 has an RMII interface included for 9.48 at mouser.com
A 25 MHz crystal and a magjak complete the interface.

Posted 5 years ago #
soundcyst
Member

ok... so what else do you need to make it? as long as the pinout of the maple is compatible (i.e. the pins attached to the ethernet aren't used for say, USB, DISC, or RTC), buy a maple, rip off the stm32f103, rework an stm32f107, and design a shield. =)

Posted 5 years ago #
josheeg
Member

well I hope the dr-brog-arm board inspired and mostly similar to the maple r3 design is in kicad so if you wanted to add thowse parts why not wrap your own and give it back to the community.

Posted 5 years ago #
larryang
Member

if you have Ethernet on it, people expect a TCP/IP stack and assorted software and libraries to use it.

Posted 5 years ago #
mlindeblom
Member

See my original post.

http://savannah.nongnu.org/projects/lwip

lwIP is freely available (under a BSD-style license) in C source code format and can be downloaded from the development homepage.

The focus of the lwIP TCP/IP implementation is to reduce the RAM usage while still having a full scale TCP. This makes lwIP suitable for use in embedded systems with tens of kilobytes of free RAM and room for around 40 kilobytes of code ROM.

This stack has already running on Cortex M3 platforms.

Posted 5 years ago #
gbulmer
Moderator

I like Ethernet also. I agree it has a lot of benefits.

My concerns are:
- the cost of parts for each ethernet end point is quite significant (> MCU),
- needs a router (which increases the cost),
- needs structured wiring i.e. one 'spur'/end-point, so many wires,
- smallest MCU is quite big (64K+)

RS485 seems lower cost (transceivers < $2), supports simple two-wire network, and one MCU can support multiple RS485 networks in hardware (if USART supports it). A MCU which can use RS485 can be tiny, i.e. 8-pins, under $2. So entire smart sensor might be under $5.

So maybe individual sensors on RS485, and 'rooms' on Ethernet?

Posted 5 years ago #
mlindeblom
Member

If you check out current MCU pricing at digikey or mouser there isn't much difference in price between a 103 and a 107 part (i.e. without and with Ethernet).

Yes, RS485 is cheaper and has similar data transfer rates.

We are in an internet or web based world. I suggest the $5 to $7 increase in cost may be worth making it a standard.

Yes, you need more memory.

Posted 5 years ago #
larryang
Member

For comparison purposes, there's the $59.95 Netduino Plus (http://www.netduino.com/netduinoplus/specs.htm).

Posted 5 years ago #
gbulmer
Moderator

If you check out current MCU pricing at digikey or mouser there isn't much difference in price between a 103 and a 107 part (i.e. without and with Ethernet).

If all that is required is one node connected to a PC, then that is fine. It makes sense to use Ethernet to connect a single node to a PC.

But, I would expect HVAC to have at least one sensor/room, say 10 rooms, so I don't think the price of a 103 vs a 107 is the basis for Ethernet vs e.g. RS485 comparison.

Isn't the comparison 10 smart sensors wired with Ethernet vs e.g. RS485?

If that is the case, the comparison is between STM32F107+LAN8720a+magjack+crystal, over $18 vs e.g. ATtiny+RS485 transciever, under $3, a cost difference of roughly 6x on digikey part prices.

Clearly an ATtiny based solution should require a smaller PCB, and fewer parts, so the actual production cost difference is bigger. If something is to be sold through retail, I guess the retail price is 3x-5x production cost. Say 3x, so $9 vs $54 per node (using part cost, but manufacturing cost will favour the simpler solution).

For 10 rooms, that would be $90 vs $540, and a 10-port router for Ethernet.

Within the same building, RS485 might be twisted pair, and doesn't need a router to build a multi-node network. It might use RJ11 terminations. Don't electricians install telephone cable?

Posted 5 years ago #
mlindeblom
Member

Yes, your analysis is valid.

An SP485 coupled with a STM32F100 series would work for sensors. Modbus or any Industry standard data interchange would be difficult to fit in 8K flash. I would prefer 32K.

8 pins would be too restrictive. 4 are required for power and communication. Humidity would require 3. Temperature at least one. Control ?

Installation costs would most likely overpower sensor system costs unless excessive. I an not sure which route would be cheaper overall.

Posted 5 years ago #

RSS feed for this topic

Reply

You must log in to post.