E-mount electronic protocol reverse engineering

Sorry about the tone of that post I've been having quite the day. Glad that code is helping you.

I opted for something way too powerful because I wanted to eliminate that as a concern while in the POC phase. For building a product it would make sense to optimize and use something more appropriate. Best to prove it out first though IMO.

You should pull my Teensy code. It does all the negotiation stages and enters the main message loop. From there it should be pretty easy to do what you are talking about. Also I have docs for the message sequence. Just please don't start from scratch again 3 of us have gotten to roughly this point before giving up. https://github.com/LexOptical/E-Mount

Athiril wrote: bostwickenator wrote: I could fool a camera into detecting an emulated lens and then change the aperture values the camera was receiving. The focus commands are harder to interpret I never got to the point of deciphering the protocol to a fwd/stop/back. It's more complicated than that because the camera can command the lens to an absolute position not just a direction. In fact there are quite a few focusing modes. That said I don't think I was far off that I just haven't figured out that detail. I actually have a stepper motor attached to an old lens to do exactly what you were suggesting lol. If you wanted to make a chip to talk to the camera and return the focal length of a lens for setting the image stabilizer that is possible with what I currently have. I don’t know if it helps but I still have an A7s I, which is CDAF only, I remember gen III coming out and Sony making adapters work way better then they do on previous gen. So I guess the adapters are buggy and gen III is more flexible and compensating for known problems, where as older gen is more ‘strict’ in the protocol

Sorry for the slow reply - I thought I had replied to this before.

The adapters aren't buggy - Sony explicitly limited the adapters to speaking a subset of the protocol that older lenses were capable of "playing nice" with. Among other things, the "native" protocol has some additional AF commands that allow a body to command a lens to not only move to a specific position/in a specific direction, but to do it at a specific speed. The "legacy adapter" subset only allows relative/absolute movements at maximum speed, because A-mount lenses appear to only be commandable in this way.

Moving at a controlled speed is critical to CDAF, which is why legacy-adapter CDAF is so horrible. It still is on even the latest bodies. I don't know what Sony did body-side to add legacy-adapter PDAF, but that's really what changed everything for legacy adapters. Any functionality that requires CDAF has generally taken much longer for Sony to figure out how to do with legacy adapters, or simply has not/likely will never figure out. They eventually figured out eye-AF, likely by using PDAF to get things really close to a CDAF solution and/or using PDAF to get close enough to find an eye and then use PDAF on the eye itself, but video AF is still horrible.

The A7S/S2/original R have no PDAF capability, so they need native lens CDAF to be acceptable. Emulating a native lens is a LOT harder.

Entropy512 wrote: Athiril wrote: Is this far enough along to make a “dumb” adapter? IE rather than translating between two protocols, just take output from an E mount body into an Arduino, and drive a stepper motor directly with a motor driver circuit? All that’s really needed would be “move back” “move forward” and “stop” or “off”. bostwickenator (Lex Optical) got to the point of emulating a "fake" lens and having it accepted by a body... I think? So I think he was close. I haven't touched this in over a year (I'll get back to it eventually...) and I'm not sure what he's done. Doing something like that with a manual focus legacy lens and a focus gear was something I always wanted to experiment with, never got around to it though.

That’s exactly what I was thinking, a gear around the focus ring.

Then the techart adapter came out and I figured I could mod it and take the motor output directly into a tiny duino, and send it to a motor on the focus ring instead snd when it reaches full throw, send it to the motor for the auto extension tube for closer focus, maybe with a switch for close AF like some lenses have.

Though given the price of it I didn’t want to gamble.

Maybe it’s easier to do this via a cheap EF AF adapter since the commands are published.

Edited by Athiril - 03 July 2020 at 12:25

Athiril wrote: Is this far enough along to make a “dumb” adapter? IE rather than translating between two protocols, just take output from an E mount body into an Arduino, and drive a stepper motor directly with a motor driver circuit? All that’s really needed would be “move back” “move forward” and “stop” or “off”.

bostwickenator (Lex Optical) got to the point of emulating a "fake" lens and having it accepted by a body... I think?

So I think he was close. I haven't touched this in over a year (I'll get back to it eventually...) and I'm not sure what he's done.

Doing something like that with a manual focus legacy lens and a focus gear was something I always wanted to experiment with, never got around to it though.

I am curious about the Techart TZE-01. It seems like they got it working quite quickly. Is Nikon using similar protocols to Sony? Why do Samyang AF FE lenses not work on it?

owenn01 wrote: Okay - I'm not scared of asking this question: one learns though asking the dumb question others fear to ask: What , in simple terms, is this all about; and where does this take you in terms of either a finished activity/product/action etc? It all looks very interesting (and technical!) - but what will it deliver for you? There - asked it and not even a tiny bit embarrassed    Thanks and best regards, Neil PS. Don't forget - keep it simple for me!!

I think it varies between the various people involved, but it happens that there's a common interest of understanding E-mount protocols and even having some reference implementations of it.

In my case - I've been unhappy with the state of EF adapters for ages. It's my opinion that we could be doing MUCH better than we are now. Look at what the MC-11 has achieved for SGV lenses. Things that could be done "better" though:
1) Officially support AF-C. Funny thing is, this is one of the only cases where the MC-11 works well and it's officially unsupported!
2) Make AF-S better - MC-11 was driving many lens motors at full speed which breaks native AF-S (hunting and overshoot). It got better with a firmware update at the long end of the 17-70, but the wide end is still way too fast. User tuning would be a huge benefit, or at least clear and concise documentation of "lens X has been attempted to be tuned but we hit the limits"
3) Officially support non-SGV lenses with native functionality. Metabones sort-of does this, but has the position for certain important features of "this doesn't work because we don't know the optical formula of the lens". Thing is, such aspects of the lens CAN be profiled in the field. Look at how the lensfun distortion/CA database is populated for example. There are documented methods for determining exit pupil distance (which is likely the critical item for off-center PDAF). The MC-11 proves that storing a table of "additional lens data" is feasible for an adapter. The problem for an adapter where this information is "baked in" with each firmware release is that the table would be huge - so why not have it populated with a tuning tool and only containing the lenses a user owns/uses?

As to whether I'll ever get there - who knows. I've got a lot of other projects I work on, and those pesky EF spring contacts are Bad News for someone as inept as I am on the mechanical side of things.

bostwickenator (Alex)'s "E-mount film body" project makes the rounds every once in a while. There are some fun challenges there, it remains to be seen whether E mount lens motors are repeatable enough for this to actually work, or if they're designed to fundamentally be part of a closed-loop AF control system and won't work well open-loop. I know he has a Voigtlander manual lens, for that lens just having aperture control would be a big deal and would probably work well on his body.

I think leegong has in the past been involved with projects that amount to trying to have the equivalent of CHDK on non-Canon cameras. It's hard to tell, pretty much if it's a camera product and has firmware, he's shown interest in picking it apart. :)