Tim, Glenn, Andrew and Jeromy by phone.
- How many dBm's does the GPS see via the GPS cylindrical patch antenna when the ATV system is on?
We set up Paul's HP 853A + 8558B Option H41 spectrum analyzer at Andrew's cabin. We futzed with the 21 GHz module for a bit, but it was clearly a bit flakey - the LO was very noisy (see picture below). So we used the 1.7 GHz module (the 8558B) even though the frequency was drifting while it warmed up. We connected the spectrum analyzer up to the GPS antenna via a N-to-SMA and SMA-SMA barrel. We calibrated the 1.7 GHz module using the calibration output (which we checked on Andrew's scope first). We measured:
|ATV system on (power amps and all)||-7 dBm|
|ATV 8V PA only||-47 dBm|
|ATV exciter only||-62 dBm|
The bandwidth with the camera running was 30 MHz (center +/- 15 MHz).
With exciter on but no modulation and the power amps on, we got two spurs +/- 600 KHz at -20dBc and two at +/- 6 MHz at -30 dBc (where dBc is relative to the the center peak).
We're not sure about what interference levels the GPS can stand. We hypothesize it's about +30 dBm on top of the actual signal, which is about -130 dBm. So worst case it really might be -130 dBm + 30 dBm = -100 dBm. And with a -7 dBm signal from the ATV, we might need a -93 dB filter (!) in the circuit. (See the following for a more detailed calculation.)
Here's the way we think it works.
The ATV outputs at around 1.2GHz with a bandwidth of 30MHz. The center frequency is therefore about 300MHz removed from the GPS L1 band at 1575.42MHz. We don't think there are significant spurs in the ATV output within the L1 bandwidth of ~+/-1MHz. If this is right, the interference should appear as uncorrelated white noise to the GPS signal tracking process.
The threshold of interest for the interference signal occurs when the interference power just equals the noise floor of the GPS receiver. To find this point we can calculate the noise power in the L1 band per Hz. The power due to thermal noise in Watts per Hz is just equal to Boltzmann's constant. To this we add the estimated receiver noise figure of 3dB for a total noise floor of
- Noise Floor:
- -204 dBW/Hz + 3dB = -201 dBW/Hz
To compare the noise floor to the interference power, we must account for the spreading gain of the system. In the present case with wideband noise, we expect the interference to effectively spread over the whole RF input bandwidth which is determined by the chipping rate or 1.023MHz, so the spreading gain is ~1/1E6 = 60dB. So the ATV peak power measured at the GPS antenna input, which is -37dBW, will be spread to
- Interference Power:
- -37dBW - 60dB/Hz = -97dBW/Hz
Now factor in the ceramic filter on the GPS input, which is estimated to have a stop band around -20dB, so the total required filtering to reduce the interference power to the threshold level is
- Maximum Filtering:
- -97dBW/Hz -20dB -(-201dBW/Hz) = 84dB
This may be a bit pessimistic. Mainly because there is some extra gain in the system. The signal to noise ratio (SNR) at the output of a GPS correlator must reach a certain level to obtain an adequately small Bit Error Rate (BER). For a BER of 10E-5 the required SNR is about 16dB, given the spreading gain of the data stream (1.023MHz / 50Hz, or about 43dB) and assuming a 1 bit correlator (about 2dB loss) the jamming margin is around
- Jamming Margin:
- 43dB - 2dB -16dB = 25dB
So, the bare minimum filtering required is about 59dB. It would probably be much better to try to approach the 84dB level and reserve a healthy jamming margin for unknown contingencies. (Some of this 25dB is required to find the GPS signal in the normal thermal noise, which is typically 18dB over the signal.)
Jeromy called in, and we discussed how to get a filter that would block the ATV signal - a high pass filter, a band pass for the GPS and/or 802.11b, or a notch at the ATV frequencies.
- http://www.digikey.com/ has:
- Panasonic SAW filter EFCH1575TCB1
(-40dB stop band, 1.5dB insertion loss, 4$ from DigiKey in singles)
We talked about adding a SAW filter to the GPS LNA that we're using, but when we opened it up (see picture) we saw there's probably no room to add anything.