CC1101 ACG freeze in production test environment

Erwin Steffens said:

When testing the product now we see some fluctuations in the RSSI values we read while nothing is changed in the test environment.

We think this is related to the Automatic Gain Control (AGC) of the CC1101 chip. I saw there is a possibility to manually freeze the AGC in the AGCCTRL0 register and you can set the AGC values manually but I cannot find out where this can be set manually. Is this the AGCTEST register where this can be written ? The datasheet says this register should not be written and is for testing only. When it is allowed to write during production testing, what is the meaning of the values in this register ?

Did you ever get an answer on this?  We are experiencing the same thing.  We're using a CC430'5137 talking to a CC1101.  The '5137 never exhibits the problem.  The CC1101, however, shows problems as signal strength is increased.  In the presence of strong signals (around -40dBm) it will occasionally  show an RSSI value dropping 20dB or more and sometimes will lose signal altogether.  During this time, the '5137 continues to report steady RSSIs.  It appears that something goes wrong with the AGC loop, gain is reduced, and the lower gain is not reflected in the RSSI calculation, resulting in incorrect RSSI reporting.  When no signal is received during this failure mode I suspect it is due to the AGC settings being too far out of whack. 

The gain can be set manually as described here http://e2e.ti.com/support/low_power_rf/f/155/p/180664/730390.aspx#730390 

Sverre: Have you have seen a failure mode in which the AGC appears to change and it's not reflected in the RSSI values?  We're seeing some boards with CC1101 work perfectly at a distance.  As signal strength increases and RSSI approaches the -45 range, suddenly RSSI jumps up by 20 to 30dB.  RSSI on the CC430 communicating with the CC1101 does not change, so we know something happened in the CC1101 receiver.  In this failure mode, if the CC430 is brought even closer to CC1101, it begins working again, but it shows RSSIs 20 to 30 dB lower.  Sometimes it goes vacillates between a reasonable RSSI and one that's 20 to 30 dB low.  Sometimes packets are lost when it's in the low condition.  It appears as if a gain stage is wrongly reduced and not accounted for in RSSI values.  

I attempted to adjust the registers limiting the gain of the various stages, which LNA the loop adjusts first, and the target amplitude at the ADC.  --> None appeared to impact this failure mode either way.

We are now having trouble replicating it.  We are just on the lookout for it if it happens again.  Have you ever observed this?

Thanks!!

CJ

This seems to an ESD issue. Please ref to attached file which shows how the CC1101 behves if there is anESD damage to one of the RF pins

Thank you for sharing this report!

Is it possible that transmitting at full power into certain impedances could develop enough voltage or current to damage the CC1101

ESD failure is due to manufacture / handling. Transmitting at full power ("into certain impedances") will not damage CC1101

Someone else having the same problem contacted me through my website asking for the resolution to this.  We do not believe the problem was ESD damage.

I don't have a definitive answer.  We had a few boards exhibiting the problem.  At first we guessed it might be ESD damage.  We tinkered with the AGC values on those boards, and it did not reduce the problem at all.  I could not cause the problem on good boards by tinkering with AGC values either.

Finally we cleaned up the soldering between the radio and the SMA.  The problem appeared to go away.  We were skeptical that cleaning up the solder joints on the SMA made a difference because they were not that bad to begin with.  Moreover, I have done extensive testing on CC430F5137, connecting its output to all kinds of impedance mismatches, and I never saw the problem on CC430F5137.

Our design with the CC1101 never went to production, but we did do quite a bit more testing with it and never saw this "AGC issue" again.

I suspect that CC1101's AGC has an issue under certain conditions. 

One thing to try is to replace the balun, matching, and antenna or antenna connector, and see if that affects the problem. 

Charles,

Did you measure the RX input return loss on a network analyzer before the solder joints were "cleaned up"?  Maybe on that one board you had intermittent matching network connections that were attributed to an AGC issue.

For other boards, did you evaluate the AGC performance with a CW signal rather than a modulated signal as a test?  

Did you try adjusting AGC settings depending on the received modulated signal?  Some received signals may need different AGC settings.

Did you evaluate AGC over temperature as a test?  

Regards,

Eric Hooker

Eric Hooker said:

Did you measure the RX input return loss on a network analyzer before the solder joints were "cleaned up"?  Maybe on that one board you had intermittent matching network connections that were attributed to an AGC issue.

I did not.  I've done a lot of impedance measurements on a board using the CC430.  I never saw this problem regardless of impedance. 

I say I cleaned them up b/c this was months ago, and I can't recall exactly what I did.  They looked okay to start out, but I fixed something about the solder joint going to the SMA.

Eric Hooker said:

For other boards, did you evaluate the AGC performance with a CW signal rather than a modulated signal as a test?  

I think I only did modulated signals, with different modulations.  I can't recall doing a CW test.

Eric Hooker said:

Did you try adjusting AGC settings depending on the received modulated signal?  Some received signals may need different AGC settings.

Yes.  Boards that exhibited the problem exhibited it without regard to AGC settings.  I couldn't make the problem occur on good boards.

The problem disappeared after that.  The design never went to production b/c client switched to BLE.  If I ever see it again, I will check it with the VNA.  Measuring S11 at the 50 ohm single-ended point would have been a good test.  I will certainly do that next time if I see the problem.