<shill mode> you can run SRSRan on a small self contained SDR such as the Seeve board. Throughput is not going to be good as a custom stack as it's unoptimised, though.
Strange. I haven't changed a thing. Just to confirm, the problem yesterday was that the website would be stuck, not that you couldn't move down further by scrolling in an unnatural way, right?
PPS input/output is off to the side, look for the "User Interface Connector" paragraph on the website.
> 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?
Is this with an older Firefox version maybe? I've tried with a relatively recent one and couldn't reproduce, but I'll look at the logs and try to fix this now that the HN storm has passed :-)
NXP are open sourcing a lot of the support stack for the LA9310 to go along with this product release, but I'm afraid they consider the VSPA DSP to be highly proprietary, yes.
The schematics will be released publicly when the product is shipping, but it won't be OSHW. I don't want to end up like the HackRF, where one small company is burdened with supporting an ecosystem of what is mostly clones. The HackRF designer himself said he wouldn't do it again.
You'd probably want something way more sensitive than that, right? Something that goes well into the micro/nanovolts rather than millivolts. But it should be reasonably simple to do. Also the RTL board when designed with a good frontend should be a viable option.
8x ADCs and 2x ADCs @ 153 MSPS -> 4x RX I/Q pairs and 1x TX I/Q pair. The way the math works, you can sample a 153 MHz signal at 153 MSPS using I/Q, or you can use each ADC line to sample at nyquist, and in that case you get 8x RF channels, each sampling 80 MHz at most. All untested, of course.
We need volumes, so the best I can hope for is people join the waitlist and pre-order when we show a functioning prototype. We already placed sizeable orders with the manufacturers and have the first batch planned for September, but further batches are probably going to be delayed into 2024 (we can't start asking for preorders before the product is tested, which is going to take time, so we can't preorder those chips early like we did for the first batch).
Feature wishlist: please! Do send them our way. Especially if it's hardware, this is the right time!
A few practical use cases that are not easily covered by similar platforms:
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.
No, I think your feedback is fair, MMCX connectors are certainly more durable. Maybe the tradeoff there is that they are not as common, so I might need to include adapters as well. I would like to have them on the opposite side of the board, so that we can maybe enclose them with the CNC enclosure which would make it truly pull resistant.
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/
I don't like them either, but I couldn't find a better way to provide the clock in/out feature and still fit in that area of the board. The optional OCXO on the opposite side of the U.FL connectors doesn't help (can't do edge-launch mmcx, for example), and if you move down you would ruin the legend for the user interface connector.
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.
I can't commit to pricing before I have decided on a partner for the distribution (DIY vs Kickstarter vs Crowdsupply, they all come with different costs and tradeoffs), but $550 is my target retail price for the Motherboard + Granita + Dev Board + CNC enclosure combo. $450 if you are happy with the Granita Lite board. Everything will be sold separately.
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).
The Granita board should already work down to 10 MHz. If you need to go lower, I (or someone else) would need to design a custom RFNM daughterboard for low frequencies. You could do direct conversion up to 150 MHz, so it should be a simple task and come with good performances.
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).
We should have a fair number of boards going around in a couple of months (10-20), when the first batch is done and basic software is complete. Mostly as loaners at first as there will be more requests than boards, but everyone that contributes should definitely get one from the first mass-production batch :-)
https://rfnm.com/blog/introducing-seeve