LowPowerLab Forum
Hardware support => RF - Range - Antennas - RFM69 library => Topic started by: Felix on May 26, 2016, 09:44:24 PM
-
What is this about?
I see a lot of interest in this forum from folks who experiment with all kinds of RF techniques, antennas and matching etc.
I bragged around that I made a RFM69 clone, which is true but it was just an experiment, which I am sure could be much improved, so this is an effort to reach out to the RF experts and ask them to voice their RF wisdom and share advice. I don't have the expensive RF tools to do tests or measure loads matching etc (VNAs and whatnot).
Here's my version of the RFM69 next to Hope's:
(https://farm1.staticflickr.com/584/21088513486_b5e2ea546f_n.jpg) (https://flic.kr/p/y8w9m7) (https://flic.kr/p/y8w9m7)
Some more background
We've seen threads like this one (https://lowpowerlab.com/forum/index.php/topic,1841.0.html) where discussion starts to go into the details of the RFM69/semtech design of the module.
HopeRF has a sx1231h board that has followed the Semtech design almost to the letter but they cut a few corners to allow the same PCB to assemble into a RFM69W or RFM69HW. They add a few components like a transistor to invert the RF switch control.
They use a specific GaAs RF switch (the G4C marked SC70 package chip) which is discontinued but they seem to have vast amounts of this or source it from asia somehow since all the radios I see many of their radios contain this RF switch (for instance the LoRa radios, HW radios, RFM22b etc). The W radios (13dbm like RFM69W) do not contain this switch or the transistor that drives it and the reason is obvious from the transceiver DS, and has a slightly different output path but uses the same PCB.
One thing that confuses lots of people is the main chip used on the RFM69 which is marked as "RF69". This is a custom QFN28 packaged sx1231h, instead of the original QFN24 "SX1231h" which you can buy from digikey, see photo above for comparison and package pin count difference.
This blue board I made can be used to make Ws or HWs. I used a RF switch that had the logic table inverted so I didn't need the extra transistor+resistor driving it.
SX1231h Datasheet (http://www.semtech.com/images/datasheet/sx1231h.pdf)
RF Switch on the RFM69 Datasheet (http://datasheet.octopart.com/UPG2179TB-E4-A-NEC-datasheet-26379.pdf)
What I'd like to focus in this discussion:
- use the sx1231h chip to design a board that is 100mW capable, similar to the blue one above
- discuss whether it's worth using a common PCB that can morph into a W or HW or whether optimizing the design for a single variant can reduce any losses and make a better module
- discuss the possibility of a 4 layer and what the real gains are - middle layers would have GND and VCC to follow RF design patterns and minimize capacitive coupling etc
- discuss the necessity of a metal can cover
- discuss the possibility of replacing the matching network with a single component - i've seen these on some RF boards but I don't know what/how/if this can replace the many 0402 components. If it's possible this would reduce complexity, perhaps reduce losses, maybe even cost
- RF chips usually are very dense and use 0402 components as industry standard, but we've seen the RFM12b using 0603 which is MUCH less challenging to assemble on a pick and place
- pinout should be unchanged, with the exception that the NC (not connected pin) could be used for something useful like telling the difference between a W and HW, which could then be used in software (a logical HIGH on that pin could mean HW for instance)
- substrate PCB should be no thicker than 1mm
- of course, made in the USA :)
- price will be more for many reasons: a basic BOM is easily a few dollars (sx1231h, good quality crystal, rf switch). The quest here is to see if a better design is possible and if the tradeoff with cost makes it justifiable to produce.
I will add to this list as ideas come up.
Please share your advice, RF experts!
Let's watch this space produce something great, but keeping it focused, with information backed by experience and data.
-
It's not an RF issue per se, but since you're wanting to sell these, one obvious "improvement" would be to move from a 2.0 pitch to a 2.1 pitch. That would obsolete the need for breakout boards, making it an easily obtained win-win differentiator. Also, maybe doing that would give you enough extra space that you could use the slightly larger components that you've said would be easier for you to pick-and-place. :)
-
It's not an RF issue per se, but since you're wanting to sell these, one obvious "improvement" would be to move from a 2.0 pitch to a 2.1 pitch.
Who uses 2.1mm pitch? 2.0 pitch IS a standard, it's used on quite a few castellated pin modules, eg ES-12E, RTX4140, as well as all the 'Bees (Zig, X, etc)...
Tom
Updated to correct phrasing on 'Bee modules.
-
Yes, I obviously meant a 2.54mm pitch, and should have said so instead of 2.1mm. I guess I somehow conflated it with 0.1" to yield the erroneous figure. Anyhow, 2.54mm would translate to a lower cost for some users (those buying and using breakout boards because of the wonderful 2.0mm "standard"), and that's all I was trying to say. If you prefer to hand that market to adafruit and sparkfun rather than harvest an easy target, then by all means stick with 2.0mm. Maybe they'll send you a thank-you card.
-
0.1" pitch would indeed be a welcome improvement for beginners as I still remember soldering little bits of wire to each leg to make my own breakout board. Sure, for further stages many would prefer to design their own pcb and would care less but for those starting with these modules I think it would help them to get on board (pun intended) quicker.
Is there a need to have all pins lined up in two rows? Once you change the footprint you may as well have pins on all four sides.
One other suggestion I can make is to see if a bigger decoupling cap is still required. In some posts I've read that the 100mW RFM6HW is pushing the envelope a bit and would welcome some extra caps close to the pa.
I like the idea of an rf shield over the relevant components. Maybe just have pads for mounting pins at first.
Not sure whether 4-layer is required. If you can get a nice GND/shield layer underneath with a 2-layer board then go for it.
-
Yes, I obviously meant a 2.54mm pitch, and should have said so instead of 2.1mm. I guess I somehow conflated it with 0.1" to yield the erroneous figure. Anyhow, 2.54mm would translate to a lower cost for some users (those buying and using breakout boards because of the wonderful 2.0mm "standard"), and that's all I was trying to say. If you prefer to hand that market to adafruit and sparkfun rather than harvest an easy target, then by all means stick with 2.0mm. Maybe they'll send you a thank-you card.
Well the marked up market might exist in the breakout board domain, but the volume market is clearly on the side of 'smaller is better'. I'd buy one breakout board to 100 PCB mounted radios in the smallest form factor possible. And WRT the RFM69, you don't even need a breakout board to fan out the 2mm pitch pad/holes to a 0.1" breadboard strip. You just need about 8 - 10 22ga wires cut to 1/2". Works like a champ at very low cost. Did this with both RFM69 & XBee footprinted WiFi radios.
Tom
-
What I'd like to focus in this discussion:
- use the sx1231h chip to design a board that is 100mW capable, similar to the blue one above
Wow. These chips are really getting good. The spec sheet shows some extremely good TX/RX capabilities.
- discuss whether it's worth using a common PCB that can morph into a W or HW or whether optimizing the design for a single variant can reduce any losses and make a better module
While it appears that you need only a coupling cap for 11dBm output, It may be tricky to utilize the "PA_Boost" output because of the antenna matching network needed at the higher power. Also, those components will be slightly different for the 433 vs 915-Mhz bands.
- discuss the possibility of a 4 layer and what the real gains are - middle layers would have GND and VCC to follow RF design patterns and minimize capacitive coupling etc
The schematic does not look to me as if more than 4 layers would be needed. The only "critical" part will be the antenna matching components & layout. The others are digital control lines, bypass caps and the crystal. Those items should not be much of an issue.
- discuss the necessity of a metal can cover
Even at 915-Mhz you are unlikely to need a shield over the chip or antenna matching network. Besides adding to the cost and complexity, it will increase the operating temp of the chip. I would avoid RF shielding if possible. However, just as the data-sheet says over and over, you will need lots of bypass caps.
-
Now if only Felix could make the RFM69CW not only better, but also smaller!
Joe
The CW only has passives and a crystal. So if all that passive network could be replaced by a single matching network component this could be much smaller yes, and perhaps better tuned. I just don't know if that's possible, that's why I started that other thread (https://lowpowerlab.com/forum/index.php/topic,1930.0.html) to see if others can steer me in the right direction.
-
I just don't know if that's possible, that's why I started that other thread to see if others can steer me in the right direction.
I had looked at that some time back. What I took away is that these components are custom Integrated Passive Devices (IPD). Usually the radio chip provider partners with an IPD company to create a custom chip for a given radio. See https://www.silabs.com/Support%20Documents/TechnicalDocs/AN904.pdf for example.
I didn't find any for the SX1231(H).
-
BTW the performance of the IPD in the doc I linked is really great compared to discrete matching. It would be more than a size advantage if we could find one that fits for the SX1231H.
-
Regarding the antenna matching network - It is usually very desirable to carefully match all the impedances throughout the system in order to achieve minimum losses.
However, there are also a couple of options that could be looked into.
1 The antenna can be tuned to match the output impedance of the final drivers
(without a matching network).
2 A small balun can be purchased or wound to match the type of antenna connected.
(the balun could be different for loops, 1/4-wave, 5/8-wave, dipole, etc.)
3 Within wide limits, it is not necessary to have a matching network between the antennas / drivers.
(You can still receive "good" signal levels even though there may be significant mismatches.
You have to decide if it is "good" enough to accomplish what you want to do.)
John
AE5HQ
-
You could also use the spare IO to manually over-drive the mux input of the RF switch (series resistor between the TXRX of the si1231h and the RF mux, brought out to the spare IO). You could then have full power range available by driving the IO pin to permanently select RFIO.
-
This is the difference between all passives vs some passives and IPD:
(http://i.imgur.com/qDnIRgO.png)
I didn't see a partnership mention in the case of this example, i know there are vendors that make these types of matching integrated components, like the one in this example.
BTW in the linked example the RF driver chip has a differential output, which is different than the Si1231h.
-
You could also use the spare IO to manually over-drive the mux input of the RF switch (series resistor between the TXRX of the si1231h and the RF mux, brought out to the spare IO). You could then have full power range available by driving the IO pin to permanently select RFIO.
I understand the wiring of that, but I guess I'm not following how that would give access to the full power range? Which range are we talking about here, the TX power range?
Looking at the 20dBm schematic, my best guess is you mean that in TX we can override the RXTX pin to tun off the RFIO part maybe? That could be a very erroneous guess :P
But could you expand on that a little please? This could be an interesting development.
-
I've got one for ya, Felix. How about you find a way to unequivocally encode whether the radio is HW or W!!!! No more "I didn't know that I was supposed to uncomment that line..."
Tom
PS: I know this isn't an RF feature but it sure beats trying to get people to RTFM!
-
I've got one for ya, Felix. How about you find a way to unequivocally encode whether the radio is HW or W!!!! No more "I didn't know that I was supposed to uncomment that line..."
Tom
PS: I know this isn't an RF feature but it sure beats trying to get people to RTFM!
You bet.
Tom, I am sure you see this as often as I do and it's one of the most annoying things, both to the user and to us trying to help. Should have been a feature of the RFM module. I am thinking a new RFM module could sacrifice one of the less significant DIO pins (if the NC is used for something more important) to mark the difference between a W and HW. Boy would that solve a lot of issues.
In all honesty the RTFM in this case is really in the top 10-15 lines of settings and headers of any sketch, and it's commented with a !
How that is so hard eludes me. Anyway this is definitely on the must-do list of this module, whenever and if it will ever happen.
-
Why don't you just burn frequency and module type into flash with the bootloader. Then rfm69 could read these settings and it wouldn't be necessary to configure them. That would solve the problem until someone burns their own bootloader at which point they likely know what they're doing.
I mean why really should I have to tell a preassembled Moteino which components came with it.
-
Why don't you just burn frequency and module type into flash with the bootloader. Then rfm69 could read these settings and it wouldn't be necessary to configure them. That would solve the problem until someone burns their own bootloader at which point they likely know what they're doing.
I mean why really should I have to tell a preassembled Moteino which components came with it.
Hmmm, sounds like putting the cart before the horse. The bootloader should be able to read the interface to determine whether its an HW, W, LoRa, and what frequency. HOW that's done is interesting. Sounds like an SC-70 serial rom on the radio module. If this doesn't exist then I'll take credit for inventing it 8)
Looks like Microchip (and probably others) have beat me to it: 24AA01T-I/LT (1K bit in SC-70-5, 2x2mm, $0.19 each).
And, being I2C interface, I'll bet that the pins could be multiplexed with SPI pins so the ROM disappears after booting and configuration.
Tom
-
The bootloader should be able to read the interface to determine whether its an HW, W, LoRa, and what frequency. HOW that's done is interesting. Sounds like an SC-70 serial rom on the radio module.
In my setup I just flash all of that info plus aes key, uid, network etc into one flash page. Then the bootloader and applications can read the data from there. And since the bootloader is already capable of self-modification the gw can just send a patched bootloader with updated settings to a node if it wants to make modifications.
I'd be hesitant to add a component just for that functionality. But I'll admit being able to flash a different bootloader and still have the settings available in the external component is an advantage.
Joe
-
The bootloader should be able to read the interface to determine whether its an HW, W, LoRa, and what frequency. HOW that's done is interesting. Sounds like an SC-70 serial rom on the radio module.
In my setup I just flash all of that info plus aes key, uid, network etc into one flash page. Then the bootloader and applications can read the data from there. And since the bootloader is already capable of self-modification the gw can just send a patched bootloader with updated settings to a node if it wants to make modifications.
I'd be hesitant to add a component just for that functionality. But I'll admit being able to flash a different bootloader and still have the settings available in the external component is an advantage.
Joe
It's more about having the module's parameters WITH the module. Assuming the module was manufactured (ie, ROM programmed correctly) then ANY driver can read the data and correctly configure the radio. Without this 'someone' will have to put the parameters into the bootloader and we're back to the RTFM problem.
@felix, one comment made me wonder if you're thinking of having all of the SX123x signals fanned out to your modules pins. I do not see the value in that. Your value add would be making this module as small as possible fanning out ONLY the signals needed by a Moteino or Moteino-like design.
Tom
-
It's more about having the module's parameters WITH the module. Assuming the module was manufactured (ie, ROM programmed correctly) then ANY driver can read the data and correctly configure the radio. Without this 'someone' will have to put the parameters into the bootloader and we're back to the RTFM problem.
Absolute nobrainer there.
@felix, one comment made me wonder if you're thinking of having all of the SX123x signals fanned out to your modules pins. I do not see the value in that. Your value add would be making this module as small as possible fanning out ONLY the signals needed by a Moteino or Moteino-like design.
Tom
I started thinking more about that. What value do we have for all the GPIO anyway?
I think 2 PCB designs optimized for a W and/or HW would be an additional improvement.
-
I started thinking more about that. What value do we have for all the GPIO anyway?
There are a set of people who use those, especially the OOK crowd and I know that Joe has played around with them, but, from my perspective, this really is about getting the smallest radio possible - I'd pay extra for that. Otherwise the HopeRF RFM69Cxx give me everything I need at an affordable price.
I think 2 PCB designs optimized for a W and/or HW would be an additional improvement.
That would be great for me, but may be overkill. If you go IPD and just wire SPI & Interrupt (and add the EEPROM for identification) then you've got a winner in my book. I think there is an advantage to having a single footprint - it allows upgrading a design that needs the higher power of an 'HW'.
Tom
-
The PCB has to be well balanced and I still think compatibility is something to strongly consider. But that can be debated later.
Having made a batch of my experimental RFM69s, I learned a few things that I plan to apply in a future run.
One thing in particular is the components are quite different (W vs HW) and spotting any placement issues with 0402 components is very hard when there are blank pads around. I don't have an AOI so my eyes are the tool for that. The hope design is all-in-1 because on the W half the pads are unused.
I think a lot of complication is because of the RFswitch circuit and all the other components to support it, that introduces unnecessary 0ohm resistors in the W variant, to link the output path properly. Having 2 optimized designs would eliminate a lot of that noise BOM. The W would then be a lot easier to make, in the absence of a very good RF switch like the "G4C" which is not source-able in the USA.
RE IPD - doesn't look like there is one such device for the SX1231/h.
-
this really is about getting the smallest radio possible - I'd pay extra for that
Yeah the radio is the limiting piece at the moment in terms of size. For these super small applications you probably don't need the RF switch (+13dBm is sufficient for coin cells). Once one fixes the antennas it's amazing what you can get out of a rfm69w.
It might be worth making one that's usable with a cr1632 - soldered on of course, who has space for a battery holder ;). The battery contacts would have to straddle the radio board.
-
For these super small applications you probably don't need the RF switch (+13dBm is sufficient for coin cells).
I agree with Joe on this one.
Tom
-
It might be worth making one that's usable with a cr1632 - soldered on of course, who has space for a battery holder ;).
That would give you at most 15mm for the pcb: http://www.digikey.de/product-detail/en/panasonic-bsg/BR-1632-HFN/P299-ND/447507
-
We have to realize if I try to make this, it's not going to be as easy as other things I assemble. I have to make lots to be able to sell them and eventually make a profit otherwise I'd rather use my very very limited time on other things. So lots of people have to like and want it and be willing to pay a higher price. It will be a hard sell if we niche this too much. In the end the economics plays a big role when making more than just a few units to play around at home. Hence some of the points I raised in my initial post.
A 13dBm design should be much easier to get started with.
IMHO the CW size 16x16mm is not too large.
-
IMHO the CW size 16x16mm is not too large.
IMNSHO a 13dBm radio could be done in a 10 pin 12x12mm package...
Tom
-
IMNSHO a 13dBm radio could be done in a 10 pin 12x12mm package...
Just the right size for a CR1225 :D - 50mAh, but a least a new challenge.
-
You could also use the spare IO to manually over-drive the mux input of the RF switch (series resistor between the TXRX of the si1231h and the RF mux, brought out to the spare IO). You could then have full power range available by driving the IO pin to permanently select RFIO.
I understand the wiring of that, but I guess I'm not following how that would give access to the full power range? Which range are we talking about here, the TX power range?
Looking at the 20dBm schematic, my best guess is you mean that in TX we can override the RXTX pin to tun off the RFIO part maybe? That could be a very erroneous guess :P
But could you expand on that a little please? This could be an interesting development.
Basically yes (sorry for the delay in responding!). If the RXTX pin had a series resistor to the RF mux then this signal could be made available on the spare IO pin. This would allow the MCU to force the mux to always select the RFIO pin, that gives the same operation as a RFM69CW. Essentially the MCU will be able to overdrive it because of the resistor independent of the state of RXTX. If the MCU tri-states its driver (or is driven by a FET) then the RXTX takes over, so then you have RFM69HW operation. In theory the MCU can select the entire TX power range.
Mark.
-
I've got one for ya, Felix. How about you find a way to unequivocally encode whether the radio is HW or W!!!! No more "I didn't know that I was supposed to uncomment that line..."
Tom
PS: I know this isn't an RF feature but it sure beats trying to get people to RTFM!
You bet.
Tom, I am sure you see this as often as I do and it's one of the most annoying things, both to the user and to us trying to help. Should have been a feature of the RFM module. I am thinking a new RFM module could sacrifice one of the less significant DIO pins (if the NC is used for something more important) to mark the difference between a W and HW. Boy would that solve a lot of issues.
In all honesty the RTFM in this case is really in the top 10-15 lines of settings and headers of any sketch, and it's commented with a !
How that is so hard eludes me. Anyway this is definitely on the must-do list of this module, whenever and if it will ever happen.
You could the spare IO pin for this. If you had a RFM69CW only variant you could force this to be the opposite normal state that it would be for a dual version (i.e. one that uses the same IO pin to force the mux to RFIO), just a pull-up or pull-down/GND would do it given that you won't have the RF switch.
Mark.
-
I recently reviewed the FCC certification process a bit and found this: http://www.emcfastpass.com/rf-modules/
Turns out you can use a pre-approved module and then save big on the device certification.
So that could be another differentiator for a custom module: FCC module pre-certification. BTW, how do others approach this? Perky, you sound like your building commercial products at times. Do you certify each product separately or do you use the module certification path?
Joe
-
Module certification simply gives you a warm feeling that the final product will pass, but it all depends on the final system. Self certification is possible as long as you can prove due diligence in court, and choosing a certified module (which BTW will have been tested in a typical representative setup) goes some way towards it. Unfortunately you can't just slap in a certified module into a product and guarantee compliance.
Mark.
-
Module certification simply gives you a warm feeling that the final product will pass, but it all depends on the final system.
I think it goes beyond in the US (see https://apps.fcc.gov/eas/comments/GetPublishedDocument.html?id=50&tn=916170):
A modular grant for a device eliminates the need for a host product to obtain its own separate certification for the specific transmitter component; however, a host product is still required to comply with all other applicable equipment authorizations for other FCC regulations, requirements and equipment functions not associated with the transmitter module portion.
Once you have the module certified you don't need to do RF certification of the end product.
Joe
-
Once you have the module certified you don't need to do RF certification of the end product.
You have to be careful with terminology here. Technically you will need to state that your product meets the relevant specifications, this is a form of self certification. So what this is really saying is that if you have chosen a pre-certified module, *and* you have been diligent in your layout, antenna choice and careful about all the other circuitry, then you could get away with self certifying and a court would see that as enough due diligence. But the onus is on you to ensure the final product complies, and if it doesn't and it hasn't been formally tested you will need to show you have done everything 'reasonable' to ensure it does.
Mark.
-
Technically you will need to state that your product meets the relevant specifications, this is a form of self certification.
Ok, just read a little deeper and indeed I've been inaccurate. For ISM devices you need a declaration of conformance (DoC) instead of certification. A DoC can only be provided by an FCC accredited lab which will do the required RF testing. The FCC is not involved as you say although the lab has to provide the test results at request of the FCC.
It does seem that if you use a pre-certified a module you don't need the declaration of conformance as intentional radiator. So practically speaking using such a module still would greatly reduce conformance costs.
Joe
-
In Europe we have to affix a CE mark. You can self-certify, but you have to provide a CE 'technical file' with everything you have done to ensure conformance so you can prove in court if necessary you've been diligent. I don't believe there is a specific mandatory requirement to have it formally tested using Directive 1999/5/EC (R&TTE), but it's very risky IMO not to at least do some testing:
http://www.nwemc.com/reference/white-papers/us-and-eu-emc-compliance-of-wireless-devices
Mark.
-
In Europe we have to affix a CE mark. You can self-certify, but you have to provide a CE 'technical file' with everything you have done to ensure conformance so you can prove in court if necessary you've been diligent.
Right. I was aware that it's different in Europe. And of course it's always possible that the overall device design causes directional radiation so it definitely seems appropriate to measure the whole product prior to shipping in Europe - even if the module is pre-certified.
-
I looked at a RFM69HW with SX1231H yesterday from Semtech , If the device is $3 wholesale from the manufacturer , could you design/build cheaper or equal that is ETSI/FCC compliant ?
Edit:
That price is restricted to employees :'(
The chip is $1 & not for export unless from a registered reseller >:(