Raw RF interface, possible?

[RobinTheunis]

Let's talk about using the wifi PHY for other things.

With the current SDK I think we have zero control of the physical layer because there is a wifi stack running on top of it. For an application I want to play with different types of packets and signals. My goal is to make my own protocol without the latency and software overhead of wifi. Also I want to experiment with doing ranging/ distance measurements on raw ofdm signals.

I guess there almost no information about the PHY/modem available? Changing the software would violate any FCC/CE approval.

Is there a possibility that the modem would be open so we can define our own protocols on top of it? In the modem basically an synchronizer and a FFT/IFFT are implemented, maybe some channel estimation hardware. Maybe also some error correcting blocks. Which of those can we repurpose to do other stuff?

[WiFive]

Like the idea, but it's probably not going to happen without RE.

[ESP_Sprite]

FWIW, we do have a sniffer interface, and I think we will be getting an interface to send arbitrary (non-management, however) WiFi frames. If you want, you could build your own protocol over that when.

[RobinTheunis]

Something more low leveller wouldn't be possible? I don't know how the WiFi baseband is implemented on the ESP32 but I would figure that some more low level interface would be possible.

For example is it possible to get the raw subcarriers after the FFT on the receiver? Before mapping it into a QAM or QPSK constellation? Is it possible to send a specific symbol with the subcarriers, a specific modulation scheme?

[RobinTheunis]

Mhh, there is no low leveller interface yet? With the sniffer interface there is still some wifi stuff hidden away. The goal is to get rid of it. Ideally something like, I want this OFDM scheme with these sub carriers with each running a BPSK constellation. That not every OFDM scheme is possible I understand, it depends on the FFT hardware. A view into the wifi baseband would be nice so I can check what is possible with it. That a fully software defined radio isn't possible, I understand.
What would be nice is to have access for example to the raw subcarrier data of an OFDM symbol before any mapping to bits from a QAM constellation.

It would be nice to open up the wifi baseband and rf path. I think that there are definitely more applications possible than doing WiFi alone with this chip,

[flashpackets]

I can only second the request for additional physical layer interface info.

For instance, I have one application where I need to adjust the output power level down to it's minimum power level.

Scanning through the available code base, it wasn't clear where one should go to study up on the magic at the physical layer interface. Would anyone care to point to a good spot to start?

Thanks,

[ESP_Sprite]

Let me put it like this: We are kind-of hesitant in opening the lower layers more. Not only because of IP reasons, but also because we've been bitten by these things before: when there was a method to send manager frames available, the first (and only, for a while) project that used them was a mass-deauthenticator throwie. Yes, sure, there are other bits of hardware that you can use to do this, but when it hits the news, it's still an Espressif product that's placed in a bad light. So I'm willing to ask around to see what we can do, but I think knowing what, specifically, you would want to use this for would help a lot in in talking the engineers responsible into adding functionality like this.

Also, we have a merge request for a method to adjust the Tx power of WiFi and BT, so that should arrive in the master branch shortly.

[WiFive]

what, specifically, you would want to use this for

802.15.4, backscatter, TMD, ...

...definitely nothing rhyming with hammer.

[RobinTheunis]

specifically, you would want to use this for

I have done some lot of stuff about indoor ranging and localization. My method is based on doing some computational stuff on the channel estimation when I send a known sequence. The method uses a lot of properties of OFDM signals, like time domain correlation is a very computational intensive operation, but moving to the frequency domain it become a lot easier.

My idea is to send a few OFDM symbols where the sub carriers are modulated with BPSK. An running my algorithm on the received signal after the FFT processing.

To be able to be sure something like this is possible I need to investigate some of the lower layers. In what different modes the baseband can be configured, what are the limits there, is there an accurate time-stamper?

[Abdo_mg]

These are very interesting ideas that I thought about them today (2022 ^^). I want to make sure that I send the same "exact" sequence again and again, which requires a full control over the sent bytes.

Did anyone succeed in getting control over the Phy layer?