FM transmitter/receiver - how to avoid directional antenna issues

[aandrew]

Hey everyone,

I'm working on a little project using Silicon Labs si4712 and si4705 chips (fm tx and rx). I have things working fine, but one thing that is driving me absolutely BONKERS is the directionality of the antennas. I get *great* audio in one orientation but if I rotate the tx or rx a little, or if I'm physically near one of them, the audio starts turning to crap (usually an increase in hiss or static). These are both meant to be battery operated and one thing I've noticed is that if I do have the USB cable (charging) hooked up, things improve, presumably due to grounding?

I'm using 10cm-ish lengths of wire for the antennas and have 120nH of shunt inductance between the antenna and ground, as recommended by silabs. I've read over their documentation and to the extent that I understand it, I think I've done everything possible to correctly match the antenna to the chip. The documentation suggests that this should work fine, and my layout and grounding and supply bypassing should all be up to snuff. Their tech support has been pretty good, but tight-lipped; they say as long as I've followed their recommendations then things should be good.

Their tx output/rx input is high impedance and have an internal varactor which is automatically adjusted to match the amp/receiver to the antenna for the tuned frequency. This is different than what I'm used to, where you want to hit a specific impedance (usually 50 ohms) and match the antenna that way. You can read back the tuning capacitance and as long as you're not slammed against one of the rails, it should be matched up properly. I'm never at one end or the other so as far as I'm able to determine, things should be great, yet this stupid directionality persists! 

Does anyone have any hints or tips or perhaps areas to investigate that I mightn't have thought of? The fact that I do get great audio sometimes tells me I must be close.

[hendorog]

Perhaps the usb cable is acting as a counterpoise and balancing the antenna up. What happens when you add a second wire attached to ground on the board for the antenna to 'work against'?

[BrianHG]

Is the problem just with the receiver or transmitter?

Try a different radio and check to see if the transmitter has the directionality problem.

Also, when testing, try 90Mhz, then 107Mhz.  See if your antenna has a weird tuning aspect due to thickness & length VS pcb layout and Q factor of the 120nH inductor.  (I know the data sheet says 120nH, but for FM radio, I would have thought it would be a larger value...)

What's your set transmitting power?

I once made an FM transmitter awhile back and the best result was a 14 gauge solid core, unshielded copper cable cut to the right length.  (it was a 1 watt transmitter...)

Does the receiver have the directionality issue when receiving normal radio broadcasts?

[aandrew]

That is an excellent test -- the receiver does not have the same directionality problem when receiving real broadcast FM, which means the problem's on the tx side.

The inductor I'm using is a Taiyo Yuden HK1608R10J-T, which is a 100nH, not 120nH; I do have some Bourns CE201210-R12J 120nH inductors I can sub in.

The Taiyo inductor has a typical Q of 18, where the Bourns one is 22. I have to admit this is where my analog electronics knowledge is weak. I understand that Q is how its DC resistance and inductance are related, and the "peakier" the curve (higher the Q) the more efficient the inductor is, but I'm lost on how to apply this. I'm also not sure if these are appropriate inductors for this application. They're certainly much higher current than I need.

The transmitting power is very weak; I'm playing in the unlicensed FM transmitter area as defined by the FCC, which IIRC limits my transmit range to a couple of hundred feet (interestingly it's defined as a distance range and not as EIRP like 2.4GHz).

Perhaps the usb cable is acting as a counterpoise and balancing the antenna up. What happens when you add a second wire attached to ground on the board for the antenna to 'work against'?

This is something I will check tomorrow as well. Thank you for the idea!

[BrianHG]

Ok, now that we narrowed the problem down to your transmitter:

The 120nH inductor should have a Q factor of at least 30, if not more.  A Q of 40 or more would be great.  It sounds like your output oscillator is muting at odd angles and it is just a fluke that it appears your working orientation happens to be pointed toward your receiver.
Also, stop your power setting loop in your control software and set the transmit power to maximum and leave it there.

I don't know the size of the 120nH inductor you used, but, if it is large enough or through-hole, you can make your own with some magnet wire.  Though, for testing, you will need a 100mhz scope and you'll test at the 90 Mhz range with your probe clipped over an insulated point portion at the base of the output antenna to isolate the probe's capacitance from the antenna.  Basically you will be probing EMF off the antenna and not the antenna's voltage itself.

[BrianHG]

These are the only 2 inductors I would use for your application:

Murata Electronics North America
P#  LQW2BHNR12J03L
120nH Unshielded Wirewound Inductor 320mA 400 mOhm Max Nonstandard
    
Wurth Electronics Inc.
P#  744762212A
120nH Unshielded Wirewound Inductor 800mA 200 mOhm Max 1008 (2520 Metric)


Or, like I said, get some fine magnet wire from an old motor or transformer, make 3 coils,  2, 3, &, 4 wraps tightly around a jeweler's screwdriver, trim and tin the legs and mount it on you PCB.  One of the 3 should equate the Murata 120nH Unshielded Wirewound Inductor with a Q>40.

[BrianHG]

The transmitting power is very weak; I'm playing in the unlicensed FM transmitter area as defined by the FCC, which IIRC limits my transmit range to a couple of hundred feet (interestingly it's defined as a distance range and not as EIRP like 2.4GHz).

Another test here, if you have an old analog FM radio, older the better.  For your home made si4705 radio, choose an empty static radio station above 99.3MHz and listen for the hiss.  Next, turn on your old FM radio and tune the dial to your receiver's set frequency -10.7MHz.  So, if you are on 99.5, tune that analog radio, volume off, down to around 88.8MHz and you will hear a dead point on your portable radio as the old analog radio transmits a blank carrier 10.7MHz higher than the station it is actually tuned into.  Check the distance which your receiver works cleanly without static.  I have some old radios which would work well over 1000 feet and they have passed FCC as they are certified.  When I was 12, in summer camp, I modified one so it would transmit any audio cassette played in it's tape deck and modified the MIC for voice and had a small radio station which went over 10k feet, but, there were no other strong radio stations to interfere there that far outside of the city.

Note this doesn't work with car radios as they are double conversion receivers.

[aandrew]

The 120nH inductor should have a Q factor of at least 30, if not more.  A Q of 40 or more would be great.  It sounds like your output oscillator is muting at odd angles and it is just a fluke that it appears your working orientation happens to be pointed toward your receiver.

Interesting. May I ask how you came to the minimum Q the shunt inductor needs, and why you think the output oscillator is muting? I'm certainly not questioning your conclusions but rather hoping to understand how you came to them.

Also, stop your power setting loop in your control software and set the transmit power to maximum and leave it there.

No power loop, simply setting the output voltage to 100dBuV. The 4712 ranges from 88 to 115dBuV accurately.

I don't know the size of the 120nH inductor you used, but, if it is large enough or through-hole, you can make your own with some magnet wire.  Though, for testing, you will need a 100mhz scope and you'll test at the 90 Mhz range with your probe clipped over an insulated point portion at the base of the output antenna to isolate the probe's capacitance from the antenna.  Basically you will be probing EMF off the antenna and not the antenna's voltage itself.

Interesting. The footprint is 0805 which is plenty large for that.  I have some of both of the LQW2BHNR12J03L and 744762212A inductors on order, they should be in tomorrow (gotta love Digi-Key's incredible Canadian shipping!). I would like to know how you came to select those, as I also found the AISC-0805-R12J-T from Abracon, which has a higher Q than either of the others. The DCR is similar as well.

Or, like I said, get some fine magnet wire from an old motor or transformer, make 3 coils,  2, 3, &, 4 wraps tightly around a jeweler's screwdriver, trim and tin the legs and mount it on you PCB.  One of the 3 should equate the Murata 120nH Unshielded Wirewound Inductor with a Q>40.

I'm going to attempt this as well tonight. Thank you again for your insights and your time.

Another test here, if you have an old analog FM radio, older the better.  For your home made si4705 radio, choose an empty static radio station above 99.3MHz and listen for the hiss.  Next, turn on your old FM radio and tune the dial to your receiver's set frequency -10.7MHz.  So, if you are on 99.5, tune that analog radio, volume off, down to around 88.8MHz and you will hear a dead point on your portable radio as the old analog radio transmits a blank carrier 10.7MHz higher than the station it is actually tuned into.  Check the distance which your receiver works cleanly without static.  I have some old radios which would work well over 1000 feet and they have passed FCC as they are certified.  When I was 12, in summer camp, I modified one so it would transmit any audio cassette played in it's tape deck and modified the MIC for voice and had a small radio station which went over 10k feet, but, there were no other strong radio stations to interfere there that far outside of the city.

This is interesting -- I don't have a radio like this (I have a cheap analog clock radio I'm testing with along with my own receiver).

[BrianHG]

Interesting. The footprint is 0805 which is plenty large for that.  I have some of both of the LQW2BHNR12J03L and 744762212A inductors on order, they should be in tomorrow (gotta love Digi-Key's incredible Canadian shipping!). I would like to know how you came to select those, as I also found the AISC-0805-R12J-T from Abracon, which has a higher Q than either of the others. The DCR is similar as well.

     That higher Q of 50 is specified for the AISC-0805-R12J-T is at 250Mhz, way outside of FM radio.  The Q of 40/45 for the digikey coils I chose were for 100Mhz and 150Mhz, the 744762212A being the best having a Q of 45 at 100MHz.  (Note that at 250Mhz, this inductor may have a Q in the 70s or higher, Q gets higher as you approach the self resonating frequency).  What's going on is that the output of the transmitter is like an open-collector of a PNP transistor, creating a class C amplifier.  There is no push/pull like a CMOS or Bipolar digital logic output device.  The inductor is pulled up & expected to swing back down to make the transmitted sinewave.  (my best guess since we aren't told the internal wiring of the IC.)  If this doesn't work, then we have to take a look at the PCB layout.
 

This is interesting -- I don't have a radio like this (I have a cheap analog clock radio I'm testing with along with my own receiver).

That radio is perfect.  Cheaper, the better.  High quality modern IC based radios have a really weak internal tuning oscillator.
Just turn it on and tune it's dial while listening on your receiver at an open frequency above 99.5Mhz.  You should hear the static mute out when your clock radio is tuned to around your homemade radio's frequency -10.7Mhz.

Curious, you probably didn't know that most FM radios transmit/broadcast a signal at 10.7MHz above the frequency they are tuned to, did you?  This means just turning on your FM radio at 88.8Mhz means it is transmitting a radio wave at 88.8Mhz+10.7Mhz = 99.5Mhz...  Tuning that radio to 97.2Mhz means it is broadcasting a signal at 107.9MHz.

(Note: the broadcast is dead silence.....)

[LaserSteve]

Don't forget transmit antenna polarization effects.

Steve

[hagster]

As Lasersteve suggests it sounds like a polorization mismatch. You dont state what axis you rotate your antenna by, but I assume you are doing it such that the antenna are either cross polorised or you are in the null on the end axis of the antenna.

To make it less susceptiable to polorisation mismatch you can use a cicrular polorised antenna on one end.

[aandrew]

Curious, you probably didn't know that most FM radios transmit/broadcast a signal at 10.7MHz above the frequency they are tuned to, did you?  This means just turning on your FM radio at 88.8Mhz means it is transmitting a radio wave at 88.8Mhz+10.7Mhz = 99.5Mhz...  Tuning that radio to 97.2Mhz means it is broadcasting a signal at 107.9MHz.

(Note: the broadcast is dead silence.....)

No, I didn't know this, but I assume it has something to do with heterodyning down to some common IF (10.7MHz in this case?) no matter what you're tuned to?

As Lasersteve suggests it sounds like a polorization mismatch. You dont state what axis you rotate your antenna by, but I assume you are doing it such that the antenna are either cross polorised or you are in the null on the end axis of the antenna.

In my case I'm just using two ~10cm pieces of wire for the antenna (one on tx, one on rx). I took a brief look at circularly polarized FM antennas but they're all considerably larger than the approximate saucer plate size that the enclosure will be. My fading in and out seems to occur no mater if I rotate about any axis. The fading is however much worse if the antennas aren't pretty much parallel with each other, which is one of the things I'm hoping to avoid needing to do.

The new inductors should arrive today and I'll be able to try some of these great suggestions out. Thank you everyone for your thoughts and comments, I'm learning a lot and truly appreciate it.

[BrianHG]

No, I didn't know this, but I assume it has something to do with heterodyning down to some common IF (10.7MHz in this case?) no matter what you're tuned to?

Yes, 100% correct.
Also, since this signal isn't coming out of any antenna from your radio tuner, you can get an idea of the kind of distance and omnidirectional you should be getting with the si4712 with nothing more that a tiny home made inductor alone, no antenna.

[aandrew]

Just an update to help keep this thread useful to future people...

Swapping out my original 120nH inductor with the Murata LQW2BHNR12J03L made a *world* of difference. As in I can't believe how much better it's gotten! I changed nothing else about the system but that inductor and the signal quality is amazing compared to before. Thank you!

Now the remaining issue is most certainly on my receiver side. The crappy night-stand clock radio is happy receiving the signal in pretty much any orientation but the two receivers I've built up are far less happy.

One thing I found interesting is that silabs recommended NOT using the LPI (short antenna) input and instead using the FMI (long antenna) input. They've recommended a 70mm long wire antenna on the receiver side which is a little problematic for me, but also interesting to me is that their application notes suggest a 1nF capacitor in series with the antenna, but their applications engineer suggests 240nH of shunt capacitance since the FMI input is characterized as 4k in parallel with 5pF. I am pretty sure the 1nF cap recommended in AN383 is because typically the long antenna is the ground lead for headphones, and the capacitor blocks any audio return currents. These currents don't exist for wire antennas.

Using your earlier post, I found very few inductors on digikey at this inductance which state ANYTHING about Q, but one in particular that drew my attention was Murata's LQW2BASR24J00L.  It's 240nH with a Q of 44 at 250MHz. Like the Abracon inductor I mentioned earlier, this is way out of broadcast FM's band but a look at their gigantic databook the graph for this family seems to indicate that the Q is in the mid-20s at 100MHz for the part that they graphed that is closest to 240nH.

Not ideal, but better than the only other published Q of 4 for the Samsung CIGT201210UHR24MNE.

[BrianHG]

Swapping out my original 120nH inductor with the Murata LQW2BHNR12J03L made a *world* of difference. As in I can't believe how much better it's gotten! I changed nothing else about the system but that inductor and the signal quality is amazing compared to before. Thank you!

Heheheheh.   These inductors need to have a high Q in a detuned state way outside their self resonating frequency so that your transmitter chip can sweep tune that 1 fixed inductor value from 88Mhz through 108Mhz with nothing more than an internal varicap at the antenna input inside the IC.

The Wurth Electronics Inc. #744762212A would be almost twice as good as it had the highest Q of 45 at the lowest frequency of 100Mhz with the a series resistance impedance less than half that of the Murata coil.  Only your home made inductor could have done even better, though, you would have to experiment with building 3, then, in circuit, spreading the coils apart with a plastic screwdriver until you scope read peak amplitude.

[BrianHG]

If you want to bend the law, get a small MPSH10 npn transistor + a nice 5-10watt RF mosfet, flat copper board to layout and some shielding, and with a little play, trimpot, caps , resistors, 12v supply, a thick 14-16 gauge home made round inductor & same material for a 10 foot long antenna, you might squeeze 1-2 watts out of that little thing and this homemade RF amp.  Well placed high off the ground, you'll transmit a few blocks in all directions.

[gregariz]

I'm using 10cm-ish lengths of wire for the antennas and have 120nH of shunt inductance between the antenna and ground, as recommended by silabs. I've read over their documentation and to the extent that I understand it, I think I've done everything possible to correctly match the antenna to the chip. The documentation suggests that this should work fine, and my layout and grounding and supply bypassing should all be up to snuff. Their tech support has been pretty good, but tight-lipped; they say as long as I've followed their recommendations then things should be good.

Semiconductor companies can be notoriously bad at the app notes - or even knowing the performance details of what they actually make. I have been in a number of interviews in recent years surrounding these radio chips and modules and hate to break it to you but apps groups like that which support these products are some of the worst places to be in a company. The people can be the dregs of the engineering world - couple that with marketing mischief and you get what you get.

First thing I usually do when using anything like this is to make up a test board to said device with the bare minimum and no antenna - and actually measure its parameters. I would hazard a guess that silabs have never had that reference design in a chamber so its likely they don't really know the radiation pattern of the thing - other than its matching. The fact that TX is different from RX is your clue right there. It's not the Antenna.