• State Resolved
  • Locked Locked
  • Replies 5 replies
  • Subscribers 16 subscribers
  • Views 1092 views
  • Users 0 members are here
Support feedback
Options
Options
Related

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

CC1101/CC1190 combo - CS detection happening during out-of-band interference

Hayden Brown
Prodigy 70 points
Other Parts Discussed in Thread: CC1101, CC1190

Hi,

Our product (based on the C1101-CC1190EM 869MHz Reference Design Rev2.1.0), is currently undergoing immunity testing as part of its CE certification.

The product, a datalogger transmitting GFSK data in the 868.1MHz to 869.0MHz band, complies with all tests under ETSI EN300 220. However, during immunity testing we have found that the GDO 'CS' interrupt (which is used for Listen Before Talk), is asserted when an interferer in the range of 80 MHz through to 300MHz is present.

The presence of the interferer (well away from the band of operation), causes the CS to assert -- and therefore our subsequent transmits are blocked because the product then [falsely] believes another device is using the channel it is intending to transmit on.

In terms of our register setup, we have configured the CS threshold detection as follows:

AGCCTRL1 = 0x40

AGCCTRL2 = 0x43

With an over-the-air baudrate of 40.5 kbps, 100kHz receive-BW and 100kHz channel-spacing, we believe this gives us an absolute CS-threshold set at -97.5 dBm.

Indeed, this is approximately what was measured by the test-lab during the LBT-limit testing for EN 300 220.

In contrast to the odd behaviour with an interferer at 80MHz, if we add an interferer in an adjacent channel of operation (say an interferer at 868.2MHz whilst operating on 868.1MHz), no CS is asserted unless the interferer is moved to the operating channel. This implies that the CS is working OK (and is quite selective in terms of frequency) -- at least around the 868MHz band.

We were wondering if you would be able to comment on what sorts of things we should look out for that could cause a very low frequency interferer to influence operation in a much-higher operating band?

Many thanks in advance for your help. If you need any further information on the application in order to comment, I would be happy to provide this.

Best regards,

Hayden

  • 0
    TI__Mastermind 39860 points

    There will be spurs that act as LO's and spurious signals can thus be down converted to IF.

    As an example, there will spurs at N x 26 MHz away from RF and when operating at 868 MHz there will be degraded blocking at 842 MHz compared to 841 and 843 MHz. If you have followed the CC1101-CC1190 reference design there is a SAW filter included in your design. I would thus assume the spurious signal is not coming through the antenna. The way to improve blocking (reduce likelihood of "false" CS) is then to add a shield. 

  • 0 in reply to Sverre
    Prodigy 70 points

    Thanks Sverre.

    I'm currently waiting on our new PCB spin to be tested for the same issue. I will advise once I have a bit more information as to whether this is a one-off problem with a single unit, or more widespread.

    Thanks for your help so far.

  • 0 in reply to Hayden Brown
    Prodigy 70 points

    Hi again Sverre,

    Apologies for the delay in getting back to you with more info on this. We have finally had some interesting developments after resuming today.

    We received samples of our new board spin and immediately sent them up for immunity testing. The problem was confirmed to also be present with the new PCBs. We then got the boards back and have set up an HP8640B signal generator and near-field probe in our lab to try to reproduce the issue.

    Fortunately it looks like we can reproduce the issue with our limited setup (we can trigger the false carrier sense) and we've observed the following:

    Aside: the symptoms below are for our 868MHz board using the latest ref design.

    • false-carrier detection with unit subjected to a ~3V/m field at 433.7MHz effective up to about 434.2MHz
    • false-carrier detection with unit subjected to a ~3V/m field at 216.9MHz effective up to about 217.1MHz

    It is interesting that these look to be sub-harmonics of our 868MHz operating frequency -- also that they are quite narrow frequenecy-bands of influence.

    More-over,

    • The false-carrier detection seems to be most strongly felt at the 2nd sub-harmonic (434MHz), and less so at the 4th sub-harmonic (217MHz). We observe this through the probe-proximity required to trigger the problem. The probe needs to be much closer when applying a 217MHz field.
    • The problem also seems to be most evident when presenting the near-field probe to the vicinity of our internal antenna compared to other areas

    The other interesting thing is that

    • we haven't been able to identify any other sensitive frequencies so far; nor any that are at multiples of 26MHz.

    Hypothetically, if spurs were to cause the false-carrier detection problem, would we expect to see problem frequencies that are separated at 26MHz intervals?

    We are also considering re-testing with:

    1. an rf-shield attached
    2. with the CC1190/CC1101 868 EMK boards
    3. and finally, with our 916MHz variant of the same design

    We will continue our investigation tomorrow -- but thought you might be interested in the developments so far.

    Based on this new information, are there any tests you might be able to suggest we can try to help point out where the susceptibility may lie? Thanks in advance.

    Best regards,

    Hayden

  • 0 in reply to Hayden Brown
    Prodigy 70 points

    After further testing today we've discovered:

    • The false carrier-detection actually seems to occur only at the sub-harmonic frequencies (868MHz / n), and we have noted problems at n = 2,3,4 ....8.
    • We have verified this behaviour on multiple units (including two of our own CC1101 designs as well as the CC1101-CC1190EM 869MHz ref 2.1 board).

    We used RF studio in continuous-RX mode with default settings to test with the CC1101-C1190EM module. The susceptibility was present across all default configurations for the CC1190 range extender mode.

    We also verified the same susceptibility using the CC1101EM (ref1.0) in the same way.

    Our lab test setup uses an HP8640B signal generator and passive magnetic probe, so it is possible we are creating more than the required 3V/m RF field, however the change in RSSI shown in RF Studio is substantial at these frequencies, even at relatively low HP8640B power levels.  It is hard to see how Listen-before-talk (LBT) can work if we have to perform carrier sense correctly at no more than -90dBm, but the RSSI is higher than this when the unit is subjected to an RF field at these other sub-harmonic frequencies.

    Any suggestions would be greatly appreciated.

    Best regards

    Hayden

  • 0 in reply to Hayden Brown
    Prodigy 70 points

    Have continued investigation in this area over the last few weeks and finally have a resolution of immunity issues. Thought I'd add the latest findings in the hope it aids others in future.

    • Carrier detection (and therefore LBT) does indeed occur when an RF immunity interferer is located at any of the sub-harmonics of the operating frequency (exhibited problems down to the 10th subharmonic).
    • It was determined that this was due to harmonics of the test signal triggering the Carrier Detect interrupt.
    • There doesn't really seem to any way around this (as the harmonics of the test gear are -30 to -40 dBc at some frequencies)
    • However, using the "Narrow band responses of receivers" clause In section EN301 489-1, Section 4.4, we were able to designate these sensitive frequencies as narrowband responses. These frequencies can therefore be excluded from immunity tests.

    We identified a further issue whereby a RF interferer with sufficient field strength (3 V/m) was able to also trigger the CS (and therefore LBT) at other frequencies. This was found to occur only when a 'strong' transmission from another device was received prior [~2ms] to a the EUT wanting to transmit (and starting the LBT listen-interval).

    i.e.

    • With the RF field and no prior transmission, no LBT detection would happen (unit operated normally).
    • With the RF field and a prior transmission, a false LBT detection would happen (unit would abort the pending transmit).

    To resolve this issue we were able to add a further 10ms of delay prior to starting the LBT operation, seeming to give the radio circuit more time to 'settle' after the strong previous transmit. The false-LBT in the presence of the RF field then ceased to occur.

    • We attributed this issue down to some "settling" phenomenon with a time constant of several milliseconds.
    • We're not really sure what part of the receiver takes this long to recover (as AGC times should be in the order of a few hundred microseconds at our 40.5kbps over-the-air baudrate). Perhaps some sort of ringing effect. Would be interested in any explanations anyone may have.
    • Have not confirmed whether this last issue was also present on the CC1101+CC1190EM (but would expect it would, as other evidence indicates our hardware functions similarly in performance to the reference design EM performance)