CC1101, optimized design

Hi Stefano,

No problems with many questions :) We shall answer your questions asap so you can move forward with the new design spin. The range that you are reporting is very low and there is something wrong here. If you can send me your existing schematic and layout (gerbers), then I can review these as well. Please address the request to Richard Wallace.

Regards,

   Richard.

1. I see on the reference board the PCB is a 2 layer FR4, 0.8 mm thick; moreover, two connectors are present on it, which carry all signals of the MCU front end. I need to use a 4 layer PCB, as the board is complex, and it could be better if the RF section could not be on a separate board. Is it possible to use 4 layers and remove those connectors? If it is, what margins have to be considered in order to minimize noise towards the RF section?

***  A single 4 layer board with copper fill connected to ground on Layer 1 and 4 and a continuous ground plane on layer 2 is a better design that having two boards. The only critical area is the RF matching network. When transferring the reference design you must maintain the impedance of the output transmission line by maintaining its width/height (dielectric thickness) ratio. Height is currently .8mm. On a 4 layer board the height of the dielectric is from the transmission line to the ground plane under it on layer 2 and is on the order of .3mm thick. In this case the width of the line will be reduced to .3/.8 of the reference design width.

2. I read this post http://e2e.ti.com/forums/t/1726.aspx in which the problem of the mechanical resistance of the SMA connector is raised. I'm afraid I will have the same problem; thus I would like to know if there is any news about that (Richard Wallace states that a through-hole SMA connector will be used in the next versions of EM boards).

*** I use a micro UMC connector on the board with a short piece of coax to a SMA (RP) connector at the antenna. The board never sees the antenna mechanical load.  I have broken a SMA off a CC1100Em module in test due to ridged couplers and other heavy connections, it is not easy.

3. On the reference board a 0402 ferrite bead is included in the power line of CC1101, but there's no reference to any specific part. Have you got any suggestion?

***  The BLM15BB121SN1D from Murata is commonly used and is a good start. The best part value will come down to where your spurs pop up.

4. What kind of other precautions do I have to take in order to improve the stability and reliability of the RF section? I read this thread: http://e2e.ti.com/forums/t/136.aspx and am a little confused... should I have to put an ESD diode, a 10k+ resistor, a coil, an ESD protection chip... one of them or all of them? And how can I insert them in the reference design without losing performance?

*** Follow the Ti Reference design closely for the placement of matching and bypass components, use a continuous ground plane and use vias to ground close to component grounds and to break up areas of copper fill. If you do not have a metal case, attach points for board level metal shielding is a good idea to mitigate potential problems getting certified.  There are standard sizes on the market and quick turn custom suppliers.  We have fielded numerous units and have not seen an ESD problem using the Ti reference design matching network without extra components.

5. In some documents a filter to cut noise at 699 MHz was added near the antenna; should I use it even if I work at 433 MHz?

6. In a place with high electro-magnetic fields (i.e. a railway station) I noticed that the boards absorbed more current, thus reducing the battery life. Could the absence of ferrite beads or other protections lead to similar results?

*** Any interference or noise ranging from broad band noise entering on battery leads to other user in the 433MHz band all contribute to lower sensitivity. Increased current is more likely caused by large (near saturation) signals coming in via the antenna affecting the Rx front end or actual packetized signals close enough in waveform to cause the radio to try and sync. In one situation we added a SAW filter due to interfering signal near our band. A tight 4 layer single board design should help.

7. In CC1101 data sheet it is stated that the use of wire wound inductors at 868 and 915 MHz can lead to a small improvement in the output power. Is it true at 433 MHz as well? If it is, what wire wound inductors could be used?

*** I like the Murata line similar to the LQW15AN12NJ00D. Coil Craft also has excellent parts. They are copper wire windings on a 0402 core as opposed to etched patterns on a ferrite material. You will have a little less improvement in loss at 433 than high frequencies but the big pay off is sharper band edge roll off due to the higher Q.

8. Until now, I've been using a 27 MHz crystal. Is it a good choice? What are, basically, the differences between a 26 and a 27 MHz crystal? The only differences (of the 26 MHz with respect to the 27 MHz) I can see are a smaller channel filter bandwidth and longer typical times (calibration and others). Is there something more?

9. I read that in a relatively short time CC1190 will be available, which could lead to have a simple way to extend CC1101 range. But will it work at 433 MHz? And, if not, how can I add a LNA in order to improve CC1101 sensitivity (I prefer not to use a PA, as, from what I can understand, 10 dBm is the law limit in Europe).

***  Adding a LNA is easy since it is done on the output of the matching network so everything is in a 50 ohms system. With modest gain you can go from about a 14 dB noise figure to <3dB which is a 11dB gain. You will give back < 2 dB due to two RF switches and a SAW filter which becomes necessary when the input saturation level goes from -20dBm to -40dBm due to the added 20dB LNA gain. It is still a 9dB increase.

10. I use a switching DC-DC converter to power CC1101 at 2.5V. Is it good, or should I take something more into account?

*** You need to scope out the output of the converter to see what is there to decide its goodness.  Two good methods are to use the switcher down to a LDO linear regulator input which provides active ripple and noise cancellation. The other approach is to use the switcher to drive a Li-Ion battery management IC with a small single cell Li-ION battery. This really cleans up the supply and gives you short term support during power glitches and outages.  I believe I've seen a paper showing a bit more range using 3.2V

H Stewart said:

***  A single 4 layer board with copper fill connected to ground on Layer 1 and 4 and a continuous ground plane on layer 2 is a better design that having two boards. The only critical area is the RF matching network. When transferring the reference design you must maintain the impedance of the output transmission line by maintaining its width/height (dielectric thickness) ratio. Height is currently .8mm. On a 4 layer board the height of the dielectric is from the transmission line to the ground plane under it on layer 2 and is on the order of .3mm thick. In this case the width of the line will be reduced to .3/.8 of the reference design width.

I will certainly discuss this with the PCB designer. Do I have to reduce the width of all lines? Or can I leave some lines as the reference design? 

H Stewart said:

*** I use a micro UMC connector on the board with a short piece of coax to a SMA (RP) connector at the antenna. The board never sees the antenna mechanical load.  I have broken a SMA off a CC1100Em module in test due to ridged couplers and other heavy connections, it is not easy.

That's a good idea, I guess. 

H Stewart said:

***  The BLM15BB121SN1D from Murata is commonly used and is a good start. The best part value will come down to where your spurs pop up.

I will use it. Surely I will eventually try others, but... first of all I have to get experienced with RF design :)

H Stewart said:

*** Follow the Ti Reference design closely for the placement of matching and bypass components, use a continuous ground plane and use vias to ground close to component grounds and to break up areas of copper fill. If you do not have a metal case, attach points for board level metal shielding is a good idea to mitigate potential problems getting certified.  There are standard sizes on the market and quick turn custom suppliers.  We have fielded numerous units and have not seen an ESD problem using the Ti reference design matching network without extra components.

Ok. Again, I will talk to the PCB designer about the metal shielding attach points. 

H Stewart said:

*** Any interference or noise ranging from broad band noise entering on battery leads to other user in the 433MHz band all contribute to lower sensitivity. Increased current is more likely caused by large (near saturation) signals coming in via the antenna affecting the Rx front end or actual packetized signals close enough in waveform to cause the radio to try and sync. In one situation we added a SAW filter due to interfering signal near our band. A tight 4 layer single board design should help.

So you advise to insert a SAW filter, something similar to what is explained in SWRA296 (AN072). Where can I find a suitable SAW filter? And what about the 699 MHz notch filter explained in SWRA168A (DN017)?

H Stewart said:

*** I like the Murata line similar to the LQW15AN12NJ00D. Coil Craft also has excellent parts. They are copper wire windings on a 0402 core as opposed to etched patterns on a ferrite material. You will have a little less improvement in loss at 433 than high frequencies but the big pay off is sharper band edge roll off due to the higher Q.

Sorry, I am not sure to understand what you are saying. Do I have to use Murata parts as stated in the datasheet or is there anything better?

stinf said:

8. Until now, I've been using a 27 MHz crystal. Is it a good choice? What are, basically, the differences between a 26 and a 27 MHz crystal? The only differences (of the 26 MHz with respect to the 27 MHz) I can see are a smaller channel filter bandwidth and longer typical times (calibration and others). Is there something more?

I guess there's nothing more to say about this point... 

H Stewart said:

***  Adding a LNA is easy since it is done on the output of the matching network so everything is in a 50 ohms system. With modest gain you can go from about a 14 dB noise figure to <3dB which is a 11dB gain. You will give back < 2 dB due to two RF switches and a SAW filter which becomes necessary when the input saturation level goes from -20dBm to -40dBm due to the added 20dB LNA gain. It is still a 9dB increase.

So I could simply add a LNA between the balun and the antenna, am I right? Have you got any part to suggest?

H Stewart said:

*** You need to scope out the output of the converter to see what is there to decide its goodness.  Two good methods are to use the switcher down to a LDO linear regulator input which provides active ripple and noise cancellation. The other approach is to use the switcher to drive a Li-Ion battery management IC with a small single cell Li-ION battery. This really cleans up the supply and gives you short term support during power glitches and outages.  I believe I've seen a paper showing a bit more range using 3.2V

The converter is TI TPS62205, with a +/- 3% 2.5V output. If I use its output to a LDO input, I think I have to power CC1101 with, say, 2.0V. But what happens if the MCU I/Os are at 2.5V? I don't think CC1101 can afford 2.5V inputs, if it is powered at 2.0V, and I am not sure the MCU can afford input voltages as low as 2.0V.

About the other solution, I think it would be too complex to implement it in my design, as I should add the battery management IC, the Li-ION battery and some glue logic to interconnect the CC1101 I/Os.

Many thanks for your help, which I greatly appreciated (also in your answers in other threads, to tell the truth...).

Stefano

I will certainly discuss this with the PCB designer. Do I have to reduce the width of all lines? Or can I leave some lines as the reference design? 

*** The majority of the lines are at the xtal frequency or much less so nothing special is required when changing the W/H ratio.  It is important to maintain the layout of the matching network while paying extra attention to avoiding extra pad size or stubs beyond what is needed for the components. 433MHz is considerably more forgiving the 2400MHz.

So you advise to insert a SAW filter, something similar to what is explained in SWRA296 (AN072). Where can I find a suitable SAW filter? And what about the 699 MHz notch filter explained in SWRA168A (DN017)?

*** A filter of some type is cheap insurance against an out of band emitter disrupting service. The possibility of a two way 400 MHz band radio or similar source being close to your units is real. A single key fob for turning on a light is one thing;  a high node count network with high availability expected in a public setting is another. You want a filter that is affordable, covers your band, is low loss (<.5dB), small and will not distort at your Tx power.  There are a lot of options on the market.  I’m not familiar enought with the 433MHz band to understand the 699 MHz issue. If you decide to use a filter it is very likely to mitigate any issue at this frequency and should be part of your selection criteria.

Sorry, I am not sure to understand what you are saying. Do I have to use Murata parts as stated in the datasheet or is there anything better?

*** The part number I gave is an example of a high performance part due to its construction and what I use in my designs.

So I could simply add a LNA between the balun and the antenna am I right? Have you got any part to suggest?

*** It is place after the balun but the LNA must have a RF switch such as a Ma/Com  (now Kobham) SW-438 V2 or similar before and after it to permit it to be bypassed during Tx.  The LNA should be turned off during Tx as well.

What sort of antenna do you use? does it require a ground plane, if so, what is the size of your ground plane?

What is the exact crystal frequency? You can tune it via a register in the CC1101.

AFAIK, the 433 band has only 1.7MHz available in most countries. how do you manage to get 256 channels of 100kBps each, without transmitting outside of the ISM band?

The 433.92MHz frequency is used by car keys and alarm systems, you might want to stay away from it, if your system allows it.

Gribo said:

What sort of antenna do you use? does it require a ground plane, if so, what is the size of your ground plane?

At the moment I am using an M4, compressed helical, 1/4 wavelength antenna; but I think I will move to an SMA one, which I still haven't chosen.

Gribo said:

What is the exact crystal frequency? You can tune it via a register in the CC1101.

The crystal frequency is 27.000 MHz. I know I can tune it, but my question was about the choice between 26 and 27 MHz, as per CC1101 datasheet.

Gribo said:

 

AFAIK, the 433 band has only 1.7MHz available in most countries. how do you manage to get 256 channels of 100kBps each, without transmitting outside of the ISM band?

The 433.92MHz frequency is used by car keys and alarm systems, you might want to stay away from it, if your system allows it.

You are definitely right.

I have to admit I overlooked this aspect (as I did with many others...). I am actually using the system well above the ISM band, and I will have to find a way to reduce the number of available channels.

One way I am thinking of is to move from 433 MHz to 868 MHz, but to be honest I don't know if I am jumping out of the frying pan into the fire :)

Thanks for your help.

Stefano

Did you measure the 27.000 frequency? the CC1101 requires 40ppm total deviation (Both temperature and aging) from the crystal. if your crystals deviate too much from each other, you will have reception and sync problems.

At 27MHz your crystal can deviate 1KHz at most from each other.

Gribo said:

Did you measure the 27.000 frequency? the CC1101 requires 40ppm total deviation (Both temperature and aging) from the crystal. if your crystals deviate too much from each other, you will have reception and sync problems.

At 27MHz your crystal can deviate 1KHz at most from each other.

I didn't. According to its datasheet, the frequency tolerance at 25°C is +/- 20 ppm, the frequency stability over temperature is again +/- 20 ppm and aging for the first year is +/- 2 ppm. The total is slight more than 40 ppm; could it be possible it has something to do with my problems?

Those are the same crystal parameters we use in our design, so it should be fine, as long as the actual frequency is indeed 27.000MHz.

I have been tripped up with incorrect loading caps, very important too. Check the crystal datasheet to verify actual loading caps at the 27Mhz. Handy way to check if your transmitting where you expect is to output an unmodulated signal and verify actual output frequency on spectrum analyzer.