100MHz SAW filter?

[mrf184]

I'm designing a high IF software defined radio receiver where interference rejection is very important, since the existing direct conversion receivers just don't cut it. The architecture is simple: 900MHz RF signal gets mixed down to a 100MHz IF signal, which is then bandpass filtered and digitized (by undersampling with a 80MSPS ADC). The IF filter needs to be very sharp but loss can be tolerated; it needs to have a bandwidth of about 10MHz.

I did a search on mouser of SAW filters around this frequency range and all are extremely expensive, plus none had the bandwidth required.

I have seen a teardown of a GSM basestation that used this kind of architecture:
https://www.eevblog.com/forum/rf-microwave/huawei-rru3908-base-station-teardown-(part-1-of-2)/
.. and it seems to use a LC filter made from SMD chip inductors and capacitors. However with typical inductor Q of about 20 @ 100MHz for 0603 inductors, I wasn't able to design a filter sharp enough.

Does anyone have experience designing this kind of SDR and know of any other types of IF filters applicable here?

[Gribo]

SAW at 100MHz will be big and fragile -> expensive. You can go to 0402 wirewound or even 0201 to get higher Q.

A quick search yielded the SF2131B by Murata. 96MHz center, 20MHz BW, 10dB insertion loss, 3.7USD at 3K quantity.

[daqq]

Take a look at sawcomtech: https://www.sawcomtech.com/?page_id=17
or http://www.krystaly.cz/en/Products/Filters

[Howardlong]

I use Golledge.co.uk extensively for this kind of thing.


IF SAW filters tend to have larger insertion losses, about 15 to 20dB. You may need LC matching filters too, and don't trust the data sheets if you do! It's not uncommon to need to make up your own test fixture and attach to a VNA to figure out the right LC terminations.

[scatha]

Any particular reason that you've chosen a 100 MHz IF? It's right in the middle of the broadcast FM radio bands and being a bit non-standard it does limit your options. If you use 70 or 140 MHz you can get surplus satcom tubular coaxial filters with pretty decent specs, eg. 1 dB IL, 10 MHz -3dB BW, rolloff to -50 dB at +/- 10 MHz.

[the_janitor]

Not sure how much you can tolerate for IL, but here is a band-pass filter I've made in the past (the center frequency wasn't 100 MHz, so I've tweaked a bit some component values to make it work for 100 MHz). It's pretty cheap to build if it meets your specs, only downside is board space (compared to a SAW filter).

In the simulation I've only used "real" models for the inductors (manufacturer: Murata); you can buy them from digikey. As for the capacitors, they are standard values.

Passband BW is 10 MHz, stopband BW is 20 MHz (50 dB), source/load impedance 50 ohms

[Howardlong]

The problem of using simulations for lumped solutions in relatively sharp or narrowband applications is that in the real world part tolerances have a large part to play.

If it’s for low volume high value solutions, and space isn’t a problem, the costs of using variable caps and manually tuning can of course often be justified, including hand-wound and hand tweaked inductors. For medium to high volume, that usually makes little sense nowadays, and SAW filters are usually the solution, although sometimes commercially produced helical filters can be a solution in low to medium volume.

[David Hess]

Even an over coupled helical resonator is not going to be that wide at 100 MHz.

I would select a different much higher IF and maybe even use low side injection so that a cheap 10 MHz wide filter can be used and then a 2nd IF to get down into a range where you can undersample.