LowPowerLab Forum

Hardware support => RF - Range - Antennas - RFM69 library => Topic started by: joelucid on January 15, 2017, 01:59:18 AM

Title: Datasheet of XTAL currently used in RFM69HW
Post by: joelucid on January 15, 2017, 01:59:18 AM
See attached. This is according to HopeRF. I'm having trouble extracting manufacturer and part number from this doc - but it does have some useful data.
Title: Re: Datasheet of XTAL currently used in RFM69HW
Post by: perky on January 15, 2017, 10:35:09 AM
Nice! As we thought, they're +/-10ppm. The temperature tolerance though (+/-20ppm, which to me appears excessive, possibly an error) appears to conflict with their reliability data which shows it to be +/-5ppm over the entire temperature range. We know the rough shape of those S curves so working say from 0-40 deg C is likely to be half that.

Mark.
Title: Re: Datasheet of XTAL currently used in RFM69HW
Post by: joelucid on January 16, 2017, 05:37:02 AM
Quote
The temperature tolerance though (+/-20ppm, which to me appears excessive, possibly an error) appears to conflict with their reliability data which shows it to be +/-5ppm over the entire temperature range.

I think the temp drift at +/- 20ppm is entirely plausible. For example one gw I calibrated has a drift of 104 hz / C at 433 mhz. That over 60 C is around 6khz which is about 15ppm. I think the reliability data refers to the change in frequency at room temperature after exposure to extreme temps.

That's why we need temperature calibration!
Title: Re: Datasheet of XTAL currently used in RFM69HW
Post by: perky on January 16, 2017, 08:43:19 AM
Yes, you're right that reliability data is measured at room temp after exposure to long periods of temperature extremes. I may have to re-evaluate afcRxBw, Fdev and and re-draw my worst case channel spacing diagram to suit. That'll affect the noise floor of course, butl luckily I do have guard bands between channels that will absorb some of this so I probably won't need to change BR as that would mess up all my timings ;). Thanks for the post it was very useful.
Mark.
Title: Re: Datasheet of XTAL currently used in RFM69HW
Post by: joelucid on January 16, 2017, 09:25:42 AM
Quote
I may have to re-evaluate afcRxBw, Fdev and and re-draw my worst case channel spacing diagram to suit.

Or you finally cave and do temperature calibration ;) - but you're absolutely right. I hadn't even thought of the channel spacing implications of going always AFC.

Joe
Title: Re: Datasheet of XTAL currently used in RFM69HW
Post by: perky on January 20, 2017, 06:48:30 PM
OK, I've caved in ;) The effects of a +/-20 ppm temp drift has had significant impact on my channel spacing stuff and the agcRxBw. If I can reduce it to +/- 10ppm it would be much better.

So I'm going to try a very simple compensation scheme. I've got an accurate temperature sensor on board, and it's actually very linear. Also the compensation curve is quite linear across my temperature range, so this could let me use the temperature directly to calculate a frequency offset. If I make that offset +/-10ppm the worst case is if the actual crystal frequency is +/-20ppm it gets corrected to +/-10ppm. If the crystal is really +/- 0ppm (i.e. flat) in that region I still get a worst case +/- 10 ppm after the (in this case un-necessary) correction.

Mark.
Title: Re: Datasheet of XTAL currently used in RFM69HW
Post by: joelucid on January 21, 2017, 03:13:39 AM
Depending on volume you might want to check with hoperf on the tcxo option. They quoted me a very attractive price. Just 10c more than the standard rfm69hw. The catch: some one time development cost and minimum order 3k pieces.

Might make sense for Felix to offer these too. The usefullness of these modules is pretty limited with their standard XTALs.
Title: Re: Datasheet of XTAL currently used in RFM69HW
Post by: WhiteHare on January 21, 2017, 08:18:59 AM
If you wanted to order a small batch (6 or more modules), you might try pcb.ng. 

Also, if I recall correctly, there's an australian company that sells 10ppm modules in quantity 1.  I could possibly dig up the link if you're interested.

Of course, it would be better if Felix were to offer it, but short of that, these might be stopgap measures to support your prototyping efforts.
Title: Re: Datasheet of XTAL currently used in RFM69HW
Post by: joelucid on January 21, 2017, 01:48:47 PM
Quote
If you wanted to order a small batch (6 or more modules), you might try pcb.ng. 

WH, I don't want to build rfm69hw modules myself. Too small.

Quote
Also, if I recall correctly, there's an australian company that sells 10ppm modules in quantity 1.  I could possibly dig up the link if you're interested.

Really? There's a company other than hoperf that sells custom rfm69hw modules? I want more like +/-1ppm and in particular < 1ppm over the temp range.
Title: Re: Datasheet of XTAL currently used in RFM69HW
Post by: WhiteHare on January 21, 2017, 02:35:47 PM
I don't think you'll find anyone offering mass produced RFM69 modules at 1ppm.

What's the use case for such tight requirements?  What I concluded a while back is that if you need that level of accuracy, then it's for narrowband.  i.e. range is the driving issue.  But then it's just cheaper to get a mass produced Lora than it is to hotrod a custom RFM69 module with a TXCO.  1ppm is going to be expensive for just dumb hardware.  However, maybe you're ongoing tuning gets you there though with the cheaper gear?  Were you able to get it dialed in that precisely?
Title: Re: Datasheet of XTAL currently used in RFM69HW
Post by: WhiteHare on January 21, 2017, 06:00:28 PM
Has anyone tried making their own RFM69 module using 0603 components?  Maybe it's worth a shot.  Then you could hand assemble them yourself while also fixing some of the known issues with the HopeRF modules.  Granted, it would be a bigger module board, but that's just a trade-off.
Title: Re: Datasheet of XTAL currently used in RFM69HW
Post by: joelucid on January 22, 2017, 09:47:11 AM
Quote
But then it's just cheaper to get a mass produced Lora than it is to hotrod a custom RFM69 module with a TXCO.

First I think this is wrong. The LoRa chips are still about twice as expensive than the sx1231h and TCXO's can be had very cheaply. As I said I'm looking at an offer with a 10c premium for a TCXO RFM69HW module.

Secondly LoRa is good for range but not so good for other stuff:

My library supports multiple bitrates on one gateway. It does this by listening at the beginning of each frequency hop in the following order. First at the lowest bitrate until you would see a SyncAddressMatch. Then at the next faster bitrate if SyncAddressMatch hasn't occurred etc. Thus in the time it would take to receive one packet at the lowest bitrate I have covered all bitrates.

This is cool because it lets you use 300kbit for in-house TH motes, but the lake thermometer 1km away can send fine at 2400 baud. Best of all worlds. It's doable because without AGC and AFC I can detect SyncAddressMatch in 5 bytes. I think LoRa needs 10 bytes just preamble to work.

I really think LoRa is overkill for home automation and a TCXO based RFM69HW can handle anything home connected I can imagine fine, with flexibility and very cheaply.

Quote
However, maybe you're ongoing tuning gets you there though with the cheaper gear?  Were you able to get it dialed in that precisely?

Oh sure I can calibrate it over the entire temp range within 120hz or so. The problem is that requires an additional step in manufacturing and a step that's more expensive than the TCXO - or so it seems right now.

Joe
Title: Re: Datasheet of XTAL currently used in RFM69HW
Post by: WhiteHare on January 22, 2017, 12:48:49 PM
I hadn't realized you were undaunted by the 3,000 unit minimum order needed to getting the TXCO's installed for an extra 10 cents per unit.  If you're willing to do that, then, yeah, it's a whole different calculation.  So, if you're actually stretching for the gold ring, then please believe me when I say: I'm rooting for you!
Title: Re: Datasheet of XTAL currently used in RFM69HW
Post by: perky on January 22, 2017, 01:09:06 PM
There'a always the possibility of getting a PCB assembly house to mod the modules with TXCOs. It should be pretty easy for them, that may be a much cheaper option than using LoRa even for relatively low production quantities. Of course HopeRf could decide to sell them with TXCOs and they'd still be cheaper than LoRa.
Mark.