crickets, let's just randomise which sensor we use during boot, that ought to do it!

Here is a render of the enclosure (work in progress): https://i.imgur.com/AEYgLxU.png

And of course we agree on the software side! Soapy + GNURadio are a must. Hopefully my driver-writing skills are up to the challenge!

> What do you mean? Do you have your own DSP stack to deal with that while user receiving/transmitting? If there is, do you provide any control over it?

You can quickly pause whatever is running on the LA9310, push the NXP NLM stack, correct local frequency errors by synching to a cell network tower and then resume normal operation. It's going to cause a glitch, but if you want to maintain frequency accuracy it's a small cost you need to pay. Once you have absolute deviation and drift it should track quite well.

My understanding is that SRS and the like all require beefy desktop-class processors to run in?

Example: https://github.com/nxp-qoriq?q=la93

Feature wishlist: please! Do send them our way. Especially if it's hardware, this is the right time!

The idea is however you get shielding and cooling, sit it on the desk upside down and have all the RFNM boards accessible from the top.

1) Direction finding thanks to the 8x 153 MSPS ADCs and coherent clocks.

2) Mixed domain analyzer: have one daughterboard act as a RF receiver, and at the same time sample an analogue voltage with the other one. This is a capability reserved to the most expensive of test equipment and lets you analyze how a RF switch is behaving (or do side channel attacks?).

3) Sample almost 600 MHz of bandwidth in real time, use the powerful DSP core to run FFTs on it and send the results over to a browser that implements a RTSA display. This lets you have a real-time view of the spectrum around you for just a few watts. Thanks to the double-PPLs on the Granita board, you can also sweep the spectrum very fast.

4) There is enough processing power onboard to enable RFNM as a 5G RedCap node. We are working with NXP to add an eSIM, so with the right software, this can become a fully-functional 5G UE and connect to the normal cell network. Don't care about 5G? You can write your own standard and deploy it on the same hardware (the limitation here is having access to NXP's DSP development tools, which might limit the processing to the beefy i.MX 8M Plus, but some cores will be available as binaries).

5) Technically, anything requiring an insane amount of ADCs and DACs. You can implement your own board, as the heavy lifting (the motherboard) is already done for you. You could prototype something easily with the development board that's on the website and turn it into a real design within weeks.

The Granita is technically unreleased, the manufacturer has been making changes to the packaging of the chip as recently as a couple of months ago. It's based on the same silicon as the IceWings: https://arcticsemiconductor.com/icewings/

But fair feedback, noted. I can't move the optional OCXO below the enclosure, as that would be taller than the aluminium block itself and require a slot breaking the RF shielding, but maybe I can move the connectors. I'll try and get creative.

Edit, surface-mount MMCX to the rescue? https://i.imgur.com/1GJhJa6.png

As a comparison, $550 would be 10x to 20x less expensive than what's on the market today for similar bandwidth specs (Aaronia Spectran V6 and Ettus USRP).

All existing SDR software should be supported via OsmoSDR.

Yes, it should work as a coherent receiver. There are some questions about unmatched I/Q pair trace lengths, but from what I heard from the experts as long as the clock is coherent (which it is), we should be able to correct those in software.

(1) The AFE7903 wouldn't allow for any modularity in the system (look at what we are calling the RFNM interface on the website, I think that's the real reason this platform will work),

(2) Pricing, that single chip would cost in quantity 1k more than our current BOM, and you still need to add FPGAs, frontend, etc. next to it.

(3) Those single chip frontend modules don't have the embedded DSP cores we can use to do things like processing FFTs in real time and feeding them to a browser with no computing on the host, which I think will be very cool (multiple 160 MHz FFTs with a gr-phosphor like visualisation I think has never been done before at this price point).