I occasionally venture off into reading discussions on various camera forums and rumor sites. Lately the rumor mill has been awash in rumors about a new full frame mirrorless system coming from Canon that isn’t going to use the EF-M mount, or as some rumors indicate the EF mount for that matter.
Reading through the discussions on these articles there’s a lot of hand-wringing and anxiety about how this change my happen. One of the biggest concerns that I see is the mount, it’s compatibility with existing EF lenses, and how “non-native” EF lenses will have performance issues.
In my assessment, the whole situation is silly, and this article will try and detail why.
When it comes to talking about “compatibility” there’s two points to consider: physical compatibility and electrical. I’ll start with the physical.
Physical Mount Compatibility and Considerations
I don’t really think this is nearly the issue that some people want to make it out to be.
First, no, I don’t at all think that Canon will reuse the EF mount on a full-frame mirrorless camera. However, I don’t think it really matters that much.
What makes a mirrorless mount better than a SLR mount is that the flange focal distance (FFD) or register can be shorter since a mirror doesn’t need to fit in between. The side effect of that is simple, lenses with longer FFDs can be adapted to mount on the shallower mount with a simple spacer.
Presumably, that spacer will be a stand-alone quick-connect device, like the whole host of mount adapters for current the current mirrorless platforms. However, if Canon is going full in on mirrorless, there’s no reason that they couldn’t make a semi-permanent mount conversion that can be done by Canon service centers. The latter would be stiffer, but it would be an “all in” conversion for EF lens owners.
Looking at Canon’s options for mounts, I strongly suspect that they’ll go for a design that’s thinner than the 44 mm FFD of the EF mount. Even if Canon was to make a camera that’s nearly as thick as the current 5D, a shallower FFD means that cooling, electronics, or larger/thicker articulated display can be put behind the sensor more easily.
However, when it comes to wide angle lenses, there are marked advantages of having a shallower mount.
A Detour in Optics: Shorter Registers are a Big Feature for Mirrorless Cameras
A lens’s focal length describes the distance behind the optical center of the lens where the lens will focus a collimated (parallel rays, or infinitely distance source) rays. Technically it’s more complicated than that because camera lenses are complex lens systems not thin lens approximations, but the abstraction is good enough for the purposes of this discussion.
For a 15 mm lens, an infinitely distant source will focus at a point 15 mm behind it. Obviously this is a problem if the rear most element of the lens is 45 mm in front of the sensor you want to focus the light on.
Lens makers have a trick up their sleeves to deal with this problem, they use what’s called a retrofocus design. In lenses designed this way, the lens acts like a projector and throws the focus of the light behind where the focal length would suggest it should be. It still focuses a 15mm angle of view, but the instead of focusing it 15mm behind the lens, it focuses it on the film/sensor 40-odd mm back.
Interestingly, for digital sensors, retrofocus designs have one main advantage, there is less vignetting with wide-angle. This is because the incident light hits the sensor more straight on in the corners where it can’t be blocked by the micro lenses, color filters, and electronics surrounding the photo-diode. However, modern sensor designs (namely back illuminated sensors), tuned micro-lenses, and post processing, go a long way to mitigating those problems.
On the other hand, retro focus designs make lenses more complicated to design. Moreover, those designs are necessary for all lenses with a focal length shorter or approximately equal than the FFD. With a 44 mm FFD, all EF lenses shorter than about 50 mm (maybe more) have some level of retrofocus design. This is a potentially big consideration in optimizing designs for mid-range, “general purpose” zooms like the 24-70 and 24-105, that have be retrofocus for the 24 mm to ~50mm range, but then need to be more conventional form there to the telephoto end.
Reducing the register of a mirrorless platform to 18-20 mm would mean that most common focal lengths probably wouldn’t need to be retrofocus at all. The only lenses that necessarily would need to remain retro-focus designs would be ultra-wide angle lenses like the 16-35s, 12-24, and 14 mm primes, and even then the magnitude needed is decreased.
Canon’s already demonstrated that they recognize the benefits of shorting the back focus distance; they did this with EF-S lenses and their EF-M mount. Though both cases are crop formats, the crop format demands shorter focal lengths to maintain the same angles of view as a full frame sensor (e.g, 15mm instead of 24mm).
From a physical standpoint, the most sensible solution to me is to make the FFD of a mirrorless camera shorter than that of DSLR, and probably in the 18-20 mm range. The longer FFD DSLR lenses can still be used, by way of an adapter, and still work just as well, the shorter FFD opens up design options for many common wider angle focal lengths.
The second consideration is the electrical interface.
Put bluntly, EF design is maybe mind-blowingly smart. Actually to say it’s smart is probably an understatement. Canon paid, in one fell swoop, in full for the future of their cameras and lenses.
Why was this so brilliant? The EF mount is completely electronic.
If you doubt the brilliance of this move, just look at the rest of the industry. Every new clean-slate mount since the EF mount have been purely electronic: Sony’ E mount, 4/3rds and micro-4/3rds, Nikon’s CX mount, the list goes on. Even more poignantly, Nikon’s F mount has slowly been moving more and more towards being purely electronic; first with in lens focus motors, and now with electromagnetic diaphragms.
By making the EF mount purely electronic, Canon abstracted the lens from the camera completely. To talk to old lenses, the camera only needs the software to talk to old lenses and the right voltages. There’s no complexity to this from the user’s perspective, no tables of will do this on that but not this other thing.
For a purely electronic mount, there’s really only a few potential problems, and none of them are really hard to surmount, let alone insurmountable.
- The overall power demands of the lens (namely the servos)
- The voltage and clock rates for the digital communication
- The logical communication protocol, e.g., what bits mean what to the lens.
Both the EF and EF-M mounts isolate the servo power from the logic power, meaning that adjustments to servo voltage can be made relatively easily if that should even be necessary.
However, it’s almost certainly never going to be an issue. Canon, like most camera makers, uses spring-loaded pin connectors to connect the lens. By my calipers these are 0.05″ (1.3 mm) pins which are pretty standard in the electronics industry. A quick look through Digikey’s catalog at some random parts shows that the vast majority of these connectors are can carry 2 to 3 A at up to 150 Vdc.
Using Canon’s VBAT of 6V, at 2A on the connector that’s 12 W of power. We can be pretty sure that as things currently stand, no lens is using remotely that kind of power. Given a that an LP-E6N battery is only good for 14 Wh of energy, if a lens that drew anywhere near 12W would show massive and noticeable battery drain within minutes of being turned on.
In any event, if more power was needed by the lens servos, 3 things could be done:
- Use a higher power contact, there are sprung pin connectors that can do up to 9 A under continuous load
- Use a higher voltage, since DC power is voltage times amperage (V * A), upping the voltage to say 9 V from 6 V increases the power through a 2A pin by 50%
- Add an additional contact for the higher power lenses, multiple contacts in parallel split the load
None of these pose any real issue, though. Especially since the logic in the lens is powered independently and therefore the lens can negotiate a higher servo voltage if needed.
Signal Voltage and Clock Rates
Canon’s lenses use the SPI bus, which is a single ended serial communication system. Being single ended, as opposed to using a differential signaling solution the bus is limited by it’s own susceptibility to noise and interference as clock speeds rise as well as the ability to switch higher voltages at higher clock speeds.
The SPI bus itself imposes virtually no limits, the data can be encoded anyway desired, and the clock can be generated at any frequency that’s needed. Some SPI devices support clock speeds as high as 20 MHz — though probably not at 5V.
The issue here is that the bus may need to be run at a higher frequency to support more sophisticated communication (e.g., synchronization of in body and in lens IS). To reach that higher frequency the bus must then be run at a lower voltage (i.e., 3.3V or 1.8V not 5V).
At first glance this seems to pose a problem, except it really doesn’t — and in fact, has already been addressed by Canon (see EF-M case study). The solution to this is to use a handshake of some form to tell the camera how to talk to the lens.
The final hurdle is the logical communication protocol. Here the concern is that there’s some new commands that the are needed that the existing EF protocol can’t handle. And here too the answer comes back to negotiating the desired protocol level.
Again, this is not something especially new in the electronics industry. Virtually all interfaces that support multiple speeds and command sets negotiate their capabilities when the link is established. Doing it here, again, is really just a matter of the lens being able to ID itself and the camera knowing which mode to switch the lens too after that.
Case Study: Sony’s A to E Mount Alpha A9 Performance Limitations
I suspect some of the concerns with “non-native” EF lenses comes from the problems Sony’s adapted A mount lenses have on their E-mount mirrorless cameras.
At first glance, it appears Sony was in the same situation Canon is, they had an existing lens mount that they fully understood and they were adapting it to a new mount that they completely controlled and understood.
While Minolta’s Alpha system beat Canon’s EOS to market by a year, like many competitors at the time, Minolta didn’t commit to an entirely electronic lens mount. While the A mount broke from the previous Minolta MF mount, it still used mechanical connections to drive the AF and aperture diaphragm. This was by far the most common approach to AF cameras in the mid-80s.
Electronically lenses appear to use an SPI like bus to communicate with the camera. The most notable difference between the A and EF mounts, is that initially the A mount only communicated from lens to camera; there was no provision to send commands to the lens. Arguably, this was unnecessary at the time since control over the focus and aperture was driven mechanically by the camera.
By 2003, Minolta recognized the benefits of in-lens focus motors, and adapted the A mount to support electronically commanded in lens motors. However, the A mount never adopted an electronically controlled aperture; at least to the best of my research.
Converting A mount lenses to E mount cameras presented several problems. One is the mechanically driven focusing and aperture. The other is an electronics problem of signal conversion. Even if you limit the discussion to lenses with internal AF motors, the aperture diaphragm still needs to be actuated.
Secondly there’s also an apparent need to convert the signals electronically. Sony’s E-Mount cameras use a moderately high signalling rate, reportedly 750 Kbps to 1.5 Mbps and a signalling voltage of ~3 V. I’m not aware of the signaling rate, but the A mount used 5V logic so far as I can tell.
Neither the electrical or mechanical issues are insurmountable, A mount lenses were adapted to E mount cameras. However, the mechanical and electrical differences do mean that the camera and lens aren’t communicating as efficiently. Moreover, the mechanical aperture is likely a huge stumbling block in rapidly actuating the aperture for high FPS shooting.
Canon doesn’t have to deal with the mechanical issue at all, EF lenses have always been purely electronic. Moreover, the logical communication protocol for EF lenses has always been designed to be bidirectional (and full duplex) something that the A mount doesn’t seem to have been.
In short, I don’t see any reason to believe that Canon’s EF lenses would have serious issues on an Canon, but not EF, mount full frame mirrorless camera.
Case Study: EF and EF-M mounts
Of course the biggest reason not to be concerned about a rumored non-EF-M mirrorless mount being a problem is Canon’s own EF-M mount.
While the internet loves to denigrate Canon for being incompetent, digging into the innards of the EF and EF-M mounts shows Canon’s not nearly as stupid as the armchair engineers want to believe.
Moreover, I think Canon’s approach to the EF-M mount should be considered very indicative of how they’ll likely approach any other mirrorless mount.
Electrically, Canon’s EF mount is pretty straight forward. There are 8 contacts at the camera-lens interface, these are (from left to right looking at the pins on the camera):
- VBAT: 6 V power for the focus and IS motors
- PGND: Ground for the motor power
- PGND: A second/redundant ground point
- VDD: 5.5 V digital logic power
- DCL: (MOSI) Serial data from the camera to the lens
- DLC: (MISO) Serial data from the lens to the camera
- DCLK: (SCLK) SPI bus clock
- DGND: Logic/Data Ground
For the sake of this discussion the extra 3 contacts used by teleconverters and teleconverter capable lenses isn’t material. The camera never sees those connections directly.
The only real oddity, of the EF mount (EF-S since it’s electrically identical) is the two ground connections for the servo power.
The EF-M mount changed things slightly, but not all that substantially. The EF-M mount uses 9 pins, instead of 8. Though the position and use of the pins has changed somewhat.
Based on Canon’s patent these are (from left to right when looking at the camera):
- MIF: Indicates that the lens is mounted
- VDD: Logic voltage
- VBAT: Servo drive voltage
- PGND: Servo ground
- DCL: (MOSI) Data, camera to lens
- DLC: (MISO) Data, lens to camera
- LCLK: (SCLK) SPI/logic clock
- DGND: Logic/SPI Bus ground
- DTEF: Lens Type identifier
In summery, the major changes are that PGND is no longer carried on two pins, and two new pins
DTEF have been added. Otherwise, the remining 7 pins do the same things though some have changed positions.
MIF is a new pin, it’s function isn’t. EOS SLRs have a microswitch in the mount that’s closed by the lens’s mount when the lens is fully mounted. Given the space constrains posed by a thinner mount, removing a comparative large mechanical device and replacing it with a smaller electrical connection makes a lot of sense.
The pin is held high by a pull up resistor until the lens is fully connected, then it is pulled to ground by the lens. Moreover, the pin is positioned such that it’s the last pin to physically be connected wen the lens is fully mounted, and it does not pass over any other pins while the lens is being rotated into position.
DTEF appears to be Canon’s implementation of a lens protocol handshake. While I speculated about this earlier, my speculation was that a protocol change might use the logical communication process to switch modes, this does it at the physical level. In some ways, this is an elegant solution akin to Apple’s use of a resistors in their 30-pin (pre Lightning) connectors for fast charging rates.
The signal is an analog voltage set by a voltage divider split between the camera and the lens that’s fed into an ADC.
When mounting a lens to an EF-M camera, the MIF pin is pulled low by the lens indicating to the camera that the lens is properly mounted and ready to be used. The camera then reads voltage of the
DETF line which tells the camera whether it’s talking to an EF or EF-M lens. The camera then configures the integrated level shifters for the proper voltage and the clock generator for the correct clock rate for the SPI bus for that lens.
The solution that Canon came up with for the EOS M is quite elegant. Everything that’s needed to talk to either an EF or EF-M, as a native device, is in the camera. Voltages are adjusted properly, and the camera speaks both the EF and EF-M protocols natively to the lens.
To sum up what’s become a much longer article than I intended, I really don’t see any reason to be concerned about a new Canon mirrorless camera breaking anything with existing lenses. I don’t even expect that there won’t be performance parity. The worst case (and most likely case) I can imagine, is that you will need a physical mount adapter, and even that I don’t see as a real issue. Just leave it on the camera, (or buy one for all your lenses if you want to go that route) while the bulk of your lenses are EF mount.
As for a full frame Canon mirrorless system, looking at the engineering as I’ve done in this article, this is where I think Canon will go.
I expect a new mount will use a shorter flange-focal distance. The advantages that come from future lens designs being able to be less retro focus seems to outweigh the disadvantages. Moreover, even if they retain a thick “SLR-like” body, the shorter FFD makes it that much easier to pack electronics and better yet cooling in behind the sensor.
Further, I expect a new mount will be electrically compatible with the EF mount, in exactly the same way as the EOS M/EF-M mount is.
Beyond that, there’s not a whole lot to really be concerned about, at least in so far as the mount is concerned.
I’m not sure why this was ever necessary though Canon may have been concerned in the mid-80s that lenses would require too much power for a single pin. ↩︎
DETFappears to be connected to a 10-bit ADC according to the table on page 30 of the EF-M patent. Any value between 0x0000 and 0x007F (127) sets the camera to “first interchangeable lens” mode. The range from a value between 0x0080 and 0x027F (511 values) is reserved, then a value between 0x0280 and 0x037F (255 values) sets the camera to “second interchangeable lens” mode. Finally, values from 0x0380 to 0x03ff are “non-complaint” lens.
Given magnitudes, frequencies, and labeling of Figure 13A and 13B, and the presumption that V1 is a 5V signal as it’s output on VDD without any level shifting, and VDD for EF lenses is 5V, and there is no level shifting circuitry to be found in an EF-EOS-M adapter, then it’s reasonably to assume based on figure 13, that first interchangeable lenses are EF lenses using an EF lens mount, and are signaled natively at the EF frequencies. And the second interchangeable lens type are EF-M lenses, which are signaled at a lower voltage (probably ~ 3.3V) and at a higher rate (approximately 2x the clock rate based on the figures.) ↩︎
In writing this it strikes me that there’s almost certainly level shifters in EOS DLSRs too, as a modern Digic chip’s I/O voltages would likely be in the 1.8–3.3V range and not in the 5V range. ↩︎