Good Wifi chip?

[richardman]

What is the "user-favorite" Wifi chip these days? Is the ESP8266 still "the best" or has something better comes up?

Robustness of the Wifi protocol implementation, good QA, i.e. quality over price, plus availability, are the criterions.

Thanks

[ralphd]

Espressif has a new chip coming down the pipe, but for now nothing can beat the esp8266 at the $2 price point, or even double or triple that price.
http://nerdralph.blogspot.ca/2015/10/2-esp8266-4mb-esp-12-e-module.html

[coppice]

The only really good thing about the ESP8266 is its price, and that has certainly been low enough to get people's attention. If you want good, and are less sensitive to price, there are a number of SoC and radio chip + MCU options. Some, like the TI CC3200, can achieve extremely good power consumption for applications which send data intermittently, and make long term battery operation practical. What is still rather weak across the industry is 5GHz support in wifi chips for small embedded systems.

[nowlan]

Dont those esp chips die after 4 connections? I didnt they were very stable from what ive read.

[ralphd]

Dont those esp chips die after 4 connections? I didnt they were very stable from what ive read.

I did read about memory leaks in older firmware, but nothing so bad that it failed after 4 connections.
My experience is the modules are stable, with perhaps a stronger pullup and a cap on RST to avoid possible spurios resets.
The modules can have power busts from below 80mA to near 200mA, so poor regulator circuit design can lead to problems.

[richardman]

I should add that my interest is in making an M0+Wifi and an M4+Wifi board. Throw our compilers and JumpStart API at it, and Bob's your Uncle. So in that sense, a "dumb" Wifi chip that talks to a micro easily is preferred.

[Rasz]

Dont those esp chips die after 4 connections? I didnt they were very stable from what ive read.

https://news.ycombinator.com/item?id=10375154

[Mr.B]

ESP8266 gets my vote.
I have been using them for about 4 months now... R&D only at this point, nothing serious developed yet.
Prior to that I was using Roving Networks RN-171, very good, but very pricy - about USD 30 each.
The ESP8266 is peanuts for the punch it packs.

[westfw]

There are a wide range of "wifi chips."  The ESP8266 is "smart" with a complete TCP/IP implementation and (at least initially) a rather clunky uart-based interface (has that changed?  I would have thought that adding an SPI interface, or improving the uart interface, would have been a natural thing to do given the "programabilty" of the boards.  But I haven't heard any noise to that effect.  (OTGH, I haven't been paying too much attention.))  Uarts tend to be in short supply on small microcontrollers, and less easily bit-banged than SPI/I2C.  Because of the "A" part.
TI CC3xxx have TCP/IP, but with an SPI interface.  Other modules are (I think) interfaced below the IP layer...

The ESP8266 is the only one I've heard of that is actually user-programmable in any normal sense.

[coppice]

There are a wide range of "wifi chips."  The ESP8266 is "smart" with a complete TCP/IP implementation and (at least initially) a rather clunky uart-based interface (has that changed?  I would have thought that adding an SPI interface, or improving the uart interface, would have been a natural thing to do given the "programabilty" of the boards.  But I haven't heard any noise to that effect.  (OTGH, I haven't been paying too much attention.))  Uarts tend to be in short supply on small microcontrollers, and less easily bit-banged than SPI/I2C.  Because of the "A" part.
TI CC3xxx have TCP/IP, but with an SPI interface.  Other modules are (I think) interfaced below the IP layer...

The ESP8266 is the only one I've heard of that is actually user-programmable in any normal sense.

You just listed something that is very much user programmable - the TI CC3200. MTK, Atmel and others have devices with the TCP/IP stack on board and user programmability.

[richardman]

List prices:
TI C3100 - $20
ESP8266 - $2-$3

hmmm...

[nctnico]

ESP8266 gets my vote.
I have been using them for about 4 months now... R&D only at this point, nothing serious developed yet.
Prior to that I was using Roving Networks RN-171, very good, but very pricy - about USD 30 each.
The ESP8266 is peanuts for the punch it packs.

That may be but you have to kick a device in the nuts to see if it really works in all situations and really meets FCC regulations. $3 each may seem nice but I would be very wary to use the ESP8266 in a serious product. There may be problems sourcing it and how well is the TCP/IP stack implemented? Security? Stability? I'd rather use something I can take control over (fix bugs) if necessary.

[trevwhite]

The problem I have found with Atmel and Microchip offerings is that you are tied into using their development software and the documentation on the actual wifi devices seems very deliberately restricted.

I would really like a wifi device myself that I can just setup over uart and have full documentation for and actually does work properly.

Are there actually some examples of people using the ESP8266 and it actually performing under any form of duress? Every time I read something about them it sounds good but then you dig deeper and find it never really remains reliable/stable and locks up, crashes, freezes you out of the communication uart. I was very much put off using them because I could see them taking up a lot of my time and ending up with something unreliable. Maybe the firmware for them will mature over time..

[Mr.B]

ESP8266 gets my vote.
I have been using them for about 4 months now... R&D only at this point, nothing serious developed yet.
Prior to that I was using Roving Networks RN-171, very good, but very pricy - about USD 30 each.
The ESP8266 is peanuts for the punch it packs.

That may be but you have to kick a device in the nuts to see if it really works in all situations and really meets FCC regulations. $3 each may seem nice but I would be very wary to use the ESP8266 in a serious product. There may be problems sourcing it and how well is the TCP/IP stack implemented? Security? Stability? I'd rather use something I can take control over (fix bugs) if necessary.

I agree with you 100%.
I have not designed it into any serious product yet and will not be doing so until I have proven its stability and reliability.
My current serious product is still using the RN-171.

[kripton2035]

from what I've read (and experimented) the esp8266 can be programmed : with a lua interpreter, or like an arduino, or directly with the expressif SDK
the lua system leads to not stable esp8266 (I saw that, often the esp reboots or hangs with simple programs)
with the arduino ide programming, I have no problem so far
I did not try the expressif sdk but I've read that it seems more stable than the arduino ide system

you can also use an arduino with a wifi shield, you can get one with a fcc certification. but you will pay some $50 for it and for me it's quite the same as the $3 esp8266 !

[richardman]

I talked to the "hardware guy", and we are definitely leaning toward the C3100. It is more robust and proven. More expensive, but at 1K+ QTY, the difference is only $4 or so.

Our target here is the OEM market, where designers can just drop the module in.

[blueskull]

The cost saving is more than $4, IMHO.

Espressif has a public available forum and constantly updating SDK for their chips, that means you can run your own C code in it, and get the most performance out of it. Also, the 80MHz CPU can be abused up to 160MHz without any problems at commercial temperature, so you can further push it to higher performance.

Unlike a CC, it can be expanded with only a GPIO expander to make a true SoC solution, no external MCU or CPU needed, so this is another $3 down.

Finally, this chip has available hardware design that is not FCC certified, but FCC verified, so you can simply copy the design and you know it will pass the certification.

[richardman]

Finally, this chip has available hardware design that is not FCC certified, but FCC verified, so you can simply copy the design and you know it will pass the certification.

Please educate me on the difference between the two? Also, is FCC <xxx'ed> carry weight in non-US markets?

Thanks.

[blueskull]

Finally, this chip has available hardware design that is not FCC certified, but FCC verified, so you can simply copy the design and you know it will pass the certification.

Please educate me on the difference between the two? Also, is FCC <xxx'ed> carry weight in non-US markets?

Thanks.

According to FCC regulations, if you passed radiation test from a test house, then it is verified. If you actually get a cert, then it is certified. The certification takes very little money, but you can not manufacture others' design despite their design was certified. You need to re-certify it. Also, radio modules without on board regulators can not pass certification, so their modules can not be certified from the very beginning. But since their design passes radiation tests, you can copy the layout and know you will pass it as a final product.

I think there is a thread here discussing this problem. They claimed it to be certified, which is not. But later they said it is verified, not certified, and it passed some of pre-compliance tests.

[nctnico]

Finally, this chip has available hardware design that is not FCC certified, but FCC verified, so you can simply copy the design and you know it will pass the certification.

Also, is FCC <xxx'ed> carry weight in non-US markets?

Nope. Certification tests are slightly different so you'd need to certify for CE.

[blueskull]

Finally, this chip has available hardware design that is not FCC certified, but FCC verified, so you can simply copy the design and you know it will pass the certification.

Also, is FCC <xxx'ed> carry weight in non-US markets?

Nope. Certification tests are slightly different so you'd need to certify for CE.

I'm not anywhere close to an expert in intended radiators regulation, but I do know some of non-intended radiators. CISPR 22 is quite similar to FCC 15. They require different test apparatus, bot using inverse square law, they can be equivalated, and they are quite similar, with a few dB difference in different frequency band limits.

[nctnico]

Still any sane importer will want to see proper certification papers. The importer is responsible for bringing products onto the EU market so any non-compliancy will be the importer's problem.

Finally, this chip has available hardware design that is not FCC certified, but FCC verified, so you can simply copy the design and you know it will pass the certification.

Unfortunately getting your board / product certified is not as simple as copying a design an expect it to pass verification. It is highly likely it passes certification but you can still mess things up and fail certification or there can be different rules applying to your particular product. For example: I'll be spending the next couple of days at a test house to re-certifiy the modifications to an existing design. Because it can be used in both industrial and domestic situations the most strict limits for each situation (domestic/industrial) applies. If Espressif only tested their design in an industrial or commercial environment it is likely not to pass in the other.

[blueskull]

Still any sane importer will want to see proper certification papers. The importer is responsible for bringing products onto the EU market so any non-compliancy will be the importer's problem.

Importers don't care about whether the chip is certified or not, they care about the final product. If your product passed CISPR 22, then it is just fine.

Well, the fake certified module don't have a change at large quantity, though, unless the importer is totally nut.

[richardman]

OK, lets imagine this scenario:

Company X (i.e. us :-) ) wants to provide a drop-in OEM wifi-micro module for designers to use. Let's say we have a choice of using a Verified chipset or a Certified chipset.

Our end user, Company Y, takes our module and makes a Widget Z. Regardless our board's status, they will need to get their product Z FCC and other agency's certifications.

If we want to give them the "best chance" to succeed, is it better for us to use a chip that is Verified, or Certified, or does it not really matter?

It looks like the ESP8266 is Verified and the TI C3100 is certified.

Thanks.

[blueskull]

OK, lets imagine this scenario:

Company X (i.e. us :-) ) wants to provide a drop-in OEM wifi-micro module for designers to use. Let's say we have a choice of using a Verified chipset or a Certified chipset.

Our end user, Company Y, takes our module and makes a Widget Z. Regardless our board's status, they will need to get their product Z FCC and other agency's certifications.

If we want to give them the "best chance" to succeed, is it better for us to use a chip that is Verified, or Certified, or does it not really matter?

It looks like the ESP8266 is Verified and the TI C3100 is certified.

Thanks.

There is no "certified" chips, but there are certified modules. Whatever you use ESP8166 or CC3100, you NEED to certify your module before selling them. If you USE modules from third parties, then CC has an advantage that you can buy verified module, while ESP doesn't. If you are building modules yourself, then it doesn't matter.

ESP has verified, that is to say, can be certified with little modifying (add a regulator and some caps), ref designs, and I believe CC also has verified designs.