====== RF Filters ======

Having installed an 868 MHz band LoRa base station on a roof, I found it had very high noise levels, initially indicated by poor SNR. After noticing a mobile phone sector antenna pointing at us from a few hundred meters away, I checked the signal level on surrounding frequencies using a HackRF. I found strong adjacent signals which dropped 20dB when the receiving antenna was moved behind a wall, shielding it from the base station.

Moving the LoRa station would not be ideal for other reasons, but one possible solution is an RF filter.

===== Bandpass Filters =====

These reduce signal outside a wanted passband. The main types advertised for these frequencies are SAW and cavity. There are also LC filters.

Cavity filters appear to have excellent specs with very little pass band loss, high power handing, etc, but they are expensive - 200 GBP typical.

SAW filters are very cheap - a few pounds on Aliexpress - have good stop band rejection, but have high passband loss and low power handling.

I found, after a bit more searching, some LC filters with fairly good looking specs.

===== Aliexpress 868MHz bandpass =====

[[https://www.aliexpress.com/item/1005007539598090.html]]

It seems to say it is an LC filter.

The listing claims:

  * Center frequency: 868 MH
  * 1dB bandwidth: 10 MHz
  * With insertion loss:<2.0dB
  * Out of band inhibition:> 42dB@700MHz, ＞ 50dB@1.2GHz
  * Impedance: 50 Ohms
  * Maximum load capacity: 20dBm (100mw)
  * Interface: SMA (external screw and internal hole)
  * 915M bandpass filter
  * Center frequency: 915 MHz
  * 1dB bandwidth: 30 MHz
  * With insertion loss:<2.0dB
  * Out of band inhibition:> 58dB@600MHz, ＞ 40dB@1.2GHz
  * Impedance: 50 Ohms
  * Maximum load capacity: 20dBm (100mw)
  * Interface: SMA (external screw and internal hole)
  * 1090M bandpass filter
  * Center frequency: 1090 MHz
  * 1dB bandwidth: 10 MHz
  * With insertion loss:<3.0dB
  * Out of band inhibition:> 65dB@915MHz,＞ 42dB@1.5GHz
  * Impedance: 50 Ohms
  * Maximum load capacity: 20dBm (100mw)
  * Interface: SMA (external screw and internal hole)

I tested it with a [[LiteVNA |LiteVNA ]]. Passband insertion loss was about 1.7dB, it was properly tuned, and performance looked generally good.

LiteVNA plots of two examples of this filter:

{{:868_filter_d1.png?600|Filter 1}}
{{:868_filter_d1.png?600|Filter 2}}

I believe I was testing at +5dBm. Device power was set to "-4dB" on low frequency transmitter, and "+5dB" on high frequency one, which a forum, but not the manual, says changes over at 140 MHz. I would like to try on higher power when I get an amplifier.

Chip marked Gr 2J, or Qr 2J?

Ohmeter shows open circuit on input, output, and between them.

The ports are marked in and out. I tried testing the board backwards, and the results were almost the same, but I think there is a small difference in RL.

Testing one board, at 870 MHz, RL forwards was -17.41dB, backwards -17.04dB. S21 forwards -1.69dB, backwards -1.67dB. I think it probably doesn't matter which way it is connected.

I noticed the return loss numbers are significantly different from earalier measurements, which as seen in picture above are around -30dB, not -17dB. I don't think I changed anything. Maybe because the LiteVNA was left to warm up? That is a huge difference though. I tried recalibrating and then got numbers around -25dB, closer to what we had before.