CC1101 26MHz CRYSTAL PROBLEM !!!

Ozan AKCAY

Other Parts Discussed in Thread: CC1101, MSP430F2121

Dear Community Members,

I have a big problem about the 26MHz crystal oscillator. I am a electronic hardware design enginner in a water meter factory. On our AMR (868MHz RF) water meter's PCB we are using CC1101 transceiver. In the first 10K pieces of production we used 26MHz crystal with 20pF internal load capacitance (27pF capacitors were on the each pin of the crystal). Our handheld terminals have this crystal too. We are using 26MHz crystal with 16pF load capacitance since the first production. For getting the true matching between crystal's load capacitance and capacitors on each pin of crystal we changed the capacitors from 27pF to 20pF. They are working well too. But the problem is the handheld terminals do not communicate with the new PCBs(with 16pF load capacitance crystal). I think there is a channel shift. The PCB have same specifications, same layout, same schematic. The embedded software(in MSP430F2121) is same. And the register settings of the CC1101 is same too. Everything is same. The only difference is crystal (crystal's load capacitance) and the capacitors connected to the pins of crystal. Does anybody help me? Thank you very much.

over 13 years ago

0 Chris M over 13 years ago

TI__Expert 3595 points

Some questions:

1. What is the accuracy of the crystals (ppm)?

2. What is the data rate?

Design note DN005 (if you put it in the keyword search above, it is the first document you get) tells you how to calculate the needed accuracy and the rest. It could well be that the crystals in the one batch are slightly different, and if you are doing narrow channels the more likely that that small frequency shift can really get you. A quick way to see if this is the case is to widen the RX filter bandwidth and if this allows communications, then frequency shift is the culprit.

Chris

0 H Stewart over 13 years ago

Mastermind 8170 points

It is easy enough to use a frequency counter and measure the crystals output in the radio via a GDOx pin..

We settled on 15pF caps for use with a 26MHz crystal (NX2520SA-26.000000MHZ-B9) used on a CC1110 (same radio) after we completed testing over temperature. We had drop outs and slow starts when using a 22pF cap.

The value of the caps does effect the crystal's frequency, to have 27pF is some units and 16pF in others guarantees there will be an offset between these units. This offset will be more of a problem at low data rates where channel bandwidths are reduced.

0 Sverre over 13 years ago

TI__Mastermind 39860 points

An easy way of determining the correct crystal load capacitor values:

Transmit unmodulated carrier only. Use SmartRF Studio to get the register settings (actually you then use OOK mode and transmit a logic 1).

Measure the carrier. If it is too high, increase one or both of the loading capacitors. If the carrier is too low, decrease one or both of the loading capacitors. The capacitors do not have to be equal in value, both keep them close in value (e.g one 12 pF and one 15 pF is acceptable)

0 Gribo over 13 years ago in reply to Sverre

Intellectual 805 points

You can also measure the crystal frequency with a spectrum analyzer and low capacitance probe, along with 1K resistor (so the crystal will not be loaded).

The two capacitors are in series to each other, in your design, 2 27pF capacitors give 13.5pF total capacitance. Add to this the parasitic capacitance of the pads (about 2pF for 0402 pads) and you got roughly 16pF capacitance which is a bit low for your crystal.

0 Ozan AKCAY over 13 years ago in reply to Chris M

Prodigy 145 points

Thank you everybody,

1. We have two different crystals. First one is the old crystal with 26.000MHz , 10ppm frequency tolerance, 10ppm temperature tolerance, 20pF CL. We are using 27pF capacitors on the pins of this crystal. Second one is the new crystal with 26.000MHz, 10ppm frequency tolerance, 10ppm temperature tolerance, 16pF CL. We are using 20pF capacitors on the pins of this crystal.

2. The data rate is 38.4k. Channel bandwith is 59.9kHz. Our center frequency is 868.03MHz. We are using channel zero for communication.

We measured the new crystal's frequency and saw that it is 25.999MHz. The old crystal has 26.000MHz. Our reader handheld terminals have the old crystal so do not communicate with the modules with new crystal.

I tested one of our AMR module. Our handheld terminal(with old crystal) did not communicate with this module. I changed the 20pF capacitors in the module to 19.5pF and the handheld terminal started to communicate with the module. I measured the crystal frequency of the module. In the first case(the capacitors are 20pF) the frequency is 25.999MHz. After changing the capacitors the frequency increased to 26.000MHz.

I can not change all of the capacitors in the AMR modules. We have thousands of modules.

0 Gribo over 13 years ago in reply to Ozan AKCAY

Intellectual 805 points

You can solve it through software, by changing the base frequency parameter.

Use a terminal with the most accurate crystal (26.000000) you have, make sure its base frequency is indeed 868.030MHz, and then set its base frequency parameter to fit the units with less accurate crystal, whose frequency would be a bit higher (868.09MHz probably for 1KHz error in the crystal) The CC1101 can tolerate up to 40ppm drift, which is 1040Hz difference between 2 units crystals.

0 Chris from Pinterec over 13 years ago

Expert 1560 points

Hello

This is Chris Pinter from Pinter Electronics Consultants. I will help you with your problem today. If you need to get a hold of me or need more help please go to my website. http://www.pinterec.ca/contact-us

The issue you are having is directly related to the load on the crystal. The capacitors you have on your board were tuned for a 20 pf crystal. You will need to re tune the loading capacitors. Please take a look at the datasheet on page 22. http://focus.ti.com/lit/ds/swrs061f/swrs061f.pdf

I see you are using the same value of load capacitors as the crystal load. If this design passed quality assurance during design you might be able to replace your load capacitors with 16 pf and see if that helps. Again I should mention that you need to read the datasheet on the subject and retest for quality for whatever change you make.

I hope this helps,

Chris Pinter

Pinter Electronics Consultants

http;//www.pinterec.ca

0 Ozan AKCAY over 13 years ago in reply to Chris from Pinterec

Prodigy 145 points

Hello Chris,

Thank you for your help. I will try to explain the problem again.

We have an AMR water meter. This meter has an AMR module in it. We used 26MHz(CL=20pF TOL=10ppm) crystal with 27pF load capacitors in the first 10K pieces of production. After that we change the crystal before the second production. All of the design is same. Only the crystal was changed. We bought (from the same crystal manufacturer with first) 26MHz(CL=16pF TOL=10ppm) crystal. After that the second production period started. In this production we used 27pF load capacitors with the new crystal. We did not tune the capacitors according to the new crystal's load capacitance. After production we saw a problem. Our handheld terminal did not communicate with the new modules(second production). We got feedback from the crystal manufacturer and started to tune the load capacitors. After tuning period we found the true capacitor value as 20pF.

After that we test them again. The handheld terminal communicate with the new modules. But still there was a problem. The handheld terminal did not communicate with all of the new modules. After that we compared the frequencies of the readed and not readed modules. The results showed that working module frequency is 26.000MHz but others are 25.999MHz. Does it mean the tuning of the capacitors wrong or critical? When we tune the not working module's capacitor again, we can communicate with it. But in this case we have to tune thousand of modules. This is impossible. What can we do about the tuning?

Now we have 3 different types of modules. First is module with 26MHz(CL=20pF TOL=10ppm) crystal with 27pF load capacitors, second is 26MHz(CL=16pF TOL=10ppm) crystal with 20pF load capacitors working at 26.000MHz frequency and the third is 26MHz(CL=16pF TOL=10ppm) crystal with 20pF load capacitors working at 25.999MHz frequency. Our handheld terminal communicates with first and second types. But does not communicate with third.

I hope I can explain the problem.

Regards

0 Gribo over 13 years ago in reply to Ozan AKCAY

Intellectual 805 points

Is it 25.999000 or 25.999500? yes, the difference is critical. From a pure hardware point of view, there is not much you can do, either change the capacitors or change the crystal.

0 Chris from Pinterec over 13 years ago in reply to Ozan AKCAY

Expert 1560 points

Hello Ozan,

I have seen this problem before in a couple of designs. There is a solution. However, I will need to take a closer look at it. Do you have samples of the different modules? I would like to see them.

Please contact me at the following. http://www.pinterec.ca/contact-us

Thanks,

Chris Pinter

Pinter Electronics Consultants

0 Ricardo Medina over 10 years ago in reply to Ozan AKCAY

Prodigy 185 points

I have the same problem of communication. I have seen that if I increase de RX bandwidth filter modules can communicate correctly

How did you solve?

0 Sverre over 10 years ago in reply to Ricardo Medina

TI__Mastermind 39860 points

There must be a frequency offset between RX and TX in your system. Result is that the signal will be down converted with an offset (i.e offset from ideal IF) and if the offset is large enough the signal (or part of it) will fall outside the RX filter BW.

Do the following:

1) Set the board in TX, unmodulated carrier

2) Measure the carrier frequency

3) If the carrier is at a higher (lower) frequency than the configured frequency you need to increase (decrease) the capacitive load.

0 Ricardo Medina over 10 years ago in reply to Sverre

Prodigy 185 points

We have problems with communications between different productions. My configuration is 38.4 kbps, GFSK, 20 kHz deviation, no DC filter. RX filter: 100Khz. Optimized for sensitivity

I think that could be by Cload (Crystal) and capacitors. We use NX3225SA-26.000000MHZ-B1 (Cload: 10pF) and C1=C2=27pF

In datasheet said (page 36) that CL=1/((1/C1)+(1/C2))+Cp. Then for CL=10pF C1 and C2 should be 15pF. Is it correct?

I solve communications problems increasing RX bandwidth filter from 100KHz to 135Khz. Is this the real solution?

0 Sverre over 10 years ago in reply to Ricardo Medina

TI__Mastermind 39860 points

A crystal specified for 10 pF load cannot use C1 = C2 = 27 pF. The load would then be 13.5 pF + parasitics. The carrier frequency will be lower than what you configure CC1101 to. The solution is to change the capacitive loading capacitors to 15 pF (or possibly one 12 pF and the other 15 pF).

0 Ricardo Medina over 10 years ago in reply to Sverre

Prodigy 185 points

Thanks for your answer.

I am worried because it is supposed that reference crystal and capacitor are the same in all productions.

Details

C14, C15

KEMET C0603C270J5G

27pF 5% 50V NP0 / C0G

DIGIKEY 644-1101-2-ND
NX3225SA-26.000000MHZ-B1
Frequency Stability: ±10ppm
Frequency Tolerance: ±10ppm
Load Capacitance: 10pF
ESR: 50 Ohms
-20°C ~ 70°C

Sample: 1

Sample: 2

Sample: 3

868[Mhz]

868,299866

868,24

868,32

868,24

Δ_f:868 [KHz]

59,866

-20,134

59,866

915[Mhz]

914,999969

914,918

914,954

914,912

Δ_f:915 [KHz]

81,969

45,969

87,969

OK! I will try changing value of capacitors . Thanks for all

0 Ricardo Medina over 10 years ago in reply to Sverre

Prodigy 185 points

Dear Sir,

I have tested devices changing capcitors from 27pF to 15pF and now I have measure more exact frequency (when I transmit in unmodulated and continuous mode

	Sample: 1	Sample: 2	Sample: 3
868[Mhz] 868,299866	868.28	868.28	868.28
Difference f [KHz]	19.86	19.86	19.86
915[Mhz] 914,999969	914.992	914.992	914.992
Difference f [KHz]	7	7	7

But still I have the communications problems between some devices.

CC1110:

Sample	Data	CC1110 Mark
1	February 2012	1CM020G 1149
3	July 2012	1CM020G 1149
2	December 2012	28M2579 1231

I cannot communicate between sample 2 and sample 1, nor sample 2 and sample 3. And to solve it I have to change RX filter bandwidth from 101Khz to 135Khz

Why happe this? Is the solution change the RX filter?

0 Nicolas Normand over 10 years ago in reply to Ricardo Medina

Prodigy 30 points

Hello,

I am working on the C1101 product. I would like to know if there is some bank capacitors inside the die to adjust the crystal load capacitance ? If it does exist, which register control these bank capacitors. If not capacitor value C81 and C101 (corresponding to the typical application on datasheet) are a little bit too high ( 33pF iso 27pF) on my evaluation board.

question is: will the application work anyway ? ( i mean functional point of view)

Nicolas Normand from IPDIA.

0 TER over 10 years ago in reply to Nicolas Normand

TI__Guru**** 317180 points

For CC1101 it is not implemented a on chip possibility to adjust the crystal load cap.

If you are using a unmodified EM you have gotten from TI it should work just fine. How do you know it uses 33pF caps and which crystal is used?

Other wireless

Other wireless technologies forum

CC1101 26MHz CRYSTAL PROBLEM !!!