IQ5 rumors

[Paul Spinnler]

Exciting stuff.

2025 is gonna be a big year for P1 bespoke ... upcoming opportunities are significant:

+ XT XL
+ 90 SB Tilt
+ New P1 XT glass (72 MK II FL, 120 ASPH re-hash)
+ XC universal (smaller XT) or XC35 HR / 40 HR
+ IQ5-250/275 (higher res, DR, new CPU, new battery system, concurrent WiFi and Bluetooth, higher res screen –> BOOM)

Dealers will have a fantastic year with all the upgrades and re-invigoration of the market.

If they get upgrade pricing right – its still a difficult market – we are in for an en of the drought next year!

And if Leica hits back with F2 APO MF glass with AF and .... MANUAL FOCUS AND MANUAL APERTURE it would be a complete re-boot of high res MF optics.

Unfortunately manual aperture has been cut in the last years due to weather sealing ... but one can hope!

X-shutter should be make sharp imagery with IQ5 fine given its a high-end vibration free leaf shutter system.

 

[jng]

Within the contex of "none of this really matters in a print," I also wonder if the smaller sensor is good or bad for some of our other lenses that are not 138/120asph/90hr shiftable. On the one hand, pixels are smaller so airy disks will creep into the equation earlier. On the other hand, we could be shifting with fewer mm, so the outskirts of the image may improve. It will depend on the lenses we have and how we use them.

Dave

Good point about staying closer to the lens’s center and therefore sweet spot. However this also means a bigger enlargement factor when printing which is the primary reason why I tend to favor the larger overall sensor or film format. I’d be interested to know the crossover point where for any given lens, diminishing resolution is offset by smaller magnification (enlargement).

John

 

[cunim]

There are all sorts of discussions we could have about when sensor density reaches impracticality. Frankly, I was pretty happy with my IQ180 and only moved to the IQ4 150 because of other features, not resolution. Same with an IQ5. They won't sell me a new back because the pixel density increases to the point that I need an optical bench and the focusing eyes of a teenager. Teenagers don't buy $50K cameras and no one carries massive camera supports any more. No, something else has got to be there.

Pity the poor MF manufacturers. They are trapped in a silicon cage. We do see imaginative new products built around downsized sensors and I love my GFX for what it is, but it certainly is not an IQ4. There is still enough substance in the IQ4 to keep it above mini-MF, but how long can that continue?. Many of us would agree that serious pushing of the SOTA envelope lies with a new larger MF sensor. Sure, it's easy for me to say damn the costs and go big - it's worked in cinematography - but I guess that project is just too big to be justified in today's market.

So the real progress in the IQ5 will be features. Those features better be pretty damn charming if they are to get my money. For example, a superb EVF with a flexible connection to the back would do it while a 300 MP sensor would not. Seems to me that super-dense sensors are just guilding the pig.

 

[Paul Spinnler]

I mean higher DR and better battery life, snappier start-up, etc. would go a long way ... one of the main things I do love about the Leica S3 is that the battery life is truly astounding. I easily go on long extended weekend trips with one charged battery.

The IQ4 eats multiple batteries on a day out easily. Agree there are a lot of brownie points to be fetched with an update of the whole SoC, including latest gen WiFi and BLUETOOTH. One should be able to remote trigger the back with a low energy bluetooth connection which opens up new grip options.

And they should add another hardware port to trigger the back while having an x shutter connected so that Arca, Alpa users can use the X shutter lenses while at the same time having a trigger cable attached.

The CPU should be beefy enough so that it can do parallel processing, meaning wireless file transfer in the backgorund while you shoot into a buffer for example.

 

[BigBoy]

Are there any rumors on the sensor hassy/fuji will be using

 

[Paul Spinnler]

By historic extrapolation it would be 44x33 at 2.8 pitch which equates to 180 MPX. In the poll thread I posted a sensor map for the 811 platform and cut down sensors.

It means the Leica M12 will be 100 megapixels. Which is perfect for the APO 50 and 35M lenses.

Leica will be the only brand with APO F2 optics for MF designed in the mid 2020s with all the latest advancements incorporated including new AF and perfect CA control (industry rumours) which should provide extremely crisp 180 MP w/o diffraction.

The S4 will be quite a shake up for the MF scene as Leica is performing better than ever financially and they are out to kill with the new system meaning full on newly developed optics without compromises as their clientele also is willing to put up with a higher price point. Combine that with a 811 class sensor, compatibility with Leica classics Leica the Summilux range in M, the SL APO and Leica S reuse ability and you all of a sudden have a new heavyweight contender in the building. It’s gonna be wild frankly.

Also, this time around, they had three gens of SL cameras to finally learn AF … the S was designed as their first digital AF camera back in the day with obviously awful focusing performance.

2025 will be showdown time between IQ5 Rodie vs S4 Leica on 811 chips.

Or put differently: F4 APO MF optics from the end 2000s vs F2 APO optics from mid 2020s

 

[BigBoy]

Do you think M lens can handle 100mp?

 

[Paul Spinnler]

The APO 35 and 50 can; the Summiluxes too are very sharp stopped down, but as said even with a tiny bit of blur it will look great. It will give you a good base to post process and sharpen if needed. Leica glass throughout is excellent sharpness wise.

My Summilux 24 images look super sharp on the 60 megapixel chip. I think it’ll look nicely organic on 100. The M12M is going to be wild for sure with the APO35. And what’s even wilder is that you will be able to mount all Leica glass on the S4, curious whether they will open up some creative crop modes. Eg 36x36 or even 40x40.

Some Leica lenses fill out 44x33 with vignetting so to have an S4 which accepts all prior Leica glass will be truly cool.

The best combo will be an M for Mono and the S4 for the rest. Done.

 

[vjbelle]

Good point about staying closer to the lens’s center and therefore sweet spot. However this also means a bigger enlargement factor when printing which is the primary reason why I tend to favor the larger overall sensor or film format. I’d be interested to know the crossover point where for any given lens, diminishing resolution is offset by smaller magnification (enlargement).

John

I guess only you can determine the crossover point for your needs. I remember shooting with my Fuji 50S in Japan and was amazed how well the images held together at 48 inches. That's the largest I usually go on the long side. GAI is marvelous at upsampling so even a 'measly' 50mp file can be made into a whopping 48 inch print that you can put your nose into for detail peeping. At 100MP the task for GAI is even less and 48 inch prints are a breeze with all of the detail I could ever want. I never saw a difference in 'Print' between a 100mp file and a 150mp file. So now with a smaller pixel size diffraction becomes a concern - at least for me. I don't ever want to shoot any more stopped down than f11 and I would like to avoid that if possible. The problem for even a 33X44 sensor is DOF. F8 means fairly limited DOF which can be compensated for by focus stacking or if the scene warrants tilting ( if the camera has movements ). There is no sharpening in post that can get back the detail lost to diffraction. Mitigation - yes but the real detail will be lost. How much of that is going to be an impact for a 48 inch print? Don't know yet but that is my crossover point.

Victor B.

 

[buildbot]

but of course efficient for Sony to cut it down to 35mm,

I don't think they can really take a die masked out for an IMX811 and cut it at all. There is non-pixel stuff all around the edge of a normal sensor. (They could in theory I guess, design a sensor with readout electronics only on one side to allow them to redefine one edge, but I've never heard of anyone doing that). I tangentially work in the field so I would have expected to catch wind of it. For example, Waferscale does something similar to how large image sensor are made by stepping the mask over and precisely aligning the next exposure, across an entire wafer (Thu, waferscale!). They can do this because all 4 edges of the design are made to connect to 4 more tiles. They can't cut a tile in the middle and still have this work! I really think they just decided to switch to 16:9 as it was a bit odd that before everything was 16:9 up until the 44x33 and 54x40mm sensors.

Agree there are a lot of brownie points to be fetched with an update of the whole SoC, including latest gen WiFi and BLUETOOTH. One should be able to remote trigger the back with a low energy bluetooth connection which opens up new grip options.

There are no newer AMD/Xilinx SoCs yet that I am aware of. Everyone is still using the same Zync MPSoC from 2018 - 16nm Finfet. The team that I work for helped with parts of a design that taped out at 5nm recently, for an example of where the industry is at. Xilinx probably won't update this line for a few years yet would be my bet. These parts have 10 year life spans or more.

Wifi and bluetooth will always (okay, 95% of the time) be a seperate chip from this FPGAs anyway. It's actually literally a USB device in the IQ3 as I recall. You might be able to swap in a different Wifi PCB if the chip used the same drivers!

I have a wifi grip that'll trigger my IQ3! Bluetooth LE control would be nice though. Very Leaf Valeo!

 

[Paul Spinnler]

Do you think they can do a custom 54x40? Or is it a goner and IQ5 is 250?

On the SoC – what is your bet of what we'll see in terms of generational improvements between IQ4 and IQ5?

 

[Paul Spinnler]

I guess only you can determine the crossover point for your needs. I remember shooting with my Fuji 50S in Japan and was amazed how well the images held together at 48 inches. That's the largest I usually go on the long side. GAI is marvelous at upsampling so even a 'measly' 50mp file can be made into a whopping 48 inch print that you can put your nose into for detail peeping. At 100MP the task for GAI is even less and 48 inch prints are a breeze with all of the detail I could ever want. I never saw a difference in 'Print' between a 100mp file and a 150mp file. So now with a smaller pixel size diffraction becomes a concern - at least for me. I don't ever want to shoot any more stopped down than f11 and I would like to avoid that if possible. The problem for even a 33X44 sensor is DOF. F8 means fairly limited DOF which can be compensated for by focus stacking or if the scene warrants tilting ( if the camera has movements ). There is no sharpening in post that can get back the detail lost to diffraction. Mitigation - yes but the real detail will be lost. How much of that is going to be an impact for a 48 inch print? Don't know yet but that is my crossover point.

Victor B.

On the diffraction topic. I really can't see that this is such a big factor.

Thats the Rodie 35 HR at F4 centre at 200% percent, slightly sharpened. The loss in detail in the next pic is mainly lack of shadow detail because my shutter time compensation was not perfect and I had to lift the image quite a bit more in the C1 via the exposure slide (so don't mistake the edge of the tree as diffeent in detail – its literally underexposed at F16)


Here is the same cutout at F16, a bit more sharpened, also 200%: At 100% I can see minimal differences and in print it'll go under anyhow, especially if you use a bit of grain to take away the digital bite.



Its gonna be fine with 250 megapixels on all realistic print scenarios ... and that's a rough sharpen in C1, not advanced stuff with high-pass overlays in PS or AI sharpen layers which are blended in, for example.

I think it is so overblown. If we go to F22, F32 yes, it starts to get a bit blurry, but also not that much at 100% post-sharpen. And if you just add a bit of grain anyhow it looks actually great and more natural.

In practice for the next chip you'd just aim for F8-F11, use tilt for diffraction free sharpness front to back or combine in Halycon. On top you have many sharpening tools or you keep it a bit blurry and work with grain for a nice texture.

 

[vjbelle]

Paul..... you and I just come from slightly different points of view. I don't disagree with much of what you are saying but a 48 inch print doesn't lie. It's amazing how much detail can be seen from a print made with Imageprint. If I can see a difference at 100% pixels then I can see it in the print. That's my benchmark at least up until now. I simply won't stop down beyond f11. I would rather take the extra time and shoot maybe 3 images with a little front and back focus for a little more DOF. But I don't even like f11 and try to usually stay at around f9 or more wide open. This is how its been for me with my decades of printing large prints. With Fuji things really get a little better because their lenses can easily be shot at f5.6 or 7.1 and the detail is really staggering. Still maybe three shots are needed for a little more DOF but it's easily worth it.

Best.....

Victor B.

 

[Jon Warwick]

The best combo will be an M for Mono and the S4 for the rest. Done.

This is what I’m thinking too if all Leica lenses were to be compatible on a S4 via Leica adapters. At the moment, for color images, I have a Leica M11 simply for use with M lenses (I really like the options of the different renderings that the M lenses provide, from very modern to more classic to outright vintage) and i also have a GFX100S. It’s possible the S4 could bridge both of those in a one-camera system for colour images. + for B&W, the M Monochrom sensor really “organic” (ie, digitally unprocessed in its rendering), especially with the right lenses.

 

[rdeloe]

Do you think M lens can handle 100mp?

Roger Cicala from the LensRentals blog has a good post that explains why this question doesn't actually make sense in terms of how things work. https://www.lensrentals.com/blog/2019/10/more-ultra-high-resolution-mtf-experiments/ Scroll down to "Appendix: Why Perceptual Megapixels are Stupid"

 

[Paul Spinnler]

Paul..... you and I just come from slightly different points of view. I don't disagree with much of what you are saying but a 48 inch print doesn't lie. It's amazing how much detail can be seen from a print made with Imageprint. If I can see a difference at 100% pixels then I can see it in the print. That's my benchmark at least up until now. I simply won't stop down beyond f11. I would rather take the extra time and shoot maybe 3 images with a little front and back focus for a little more DOF. But I don't even like f11 and try to usually stay at around f9 or more wide open. This is how its been for me with my decades of printing large prints. With Fuji things really get a little better because their lenses can easily be shot at f5.6 or 7.1 and the detail is really staggering. Still maybe three shots are needed for a little more DOF but it's easily worth it.

Best.....

Victor B.

I agree F11 will be the max going forward ... and F5.6-8 optimum; that's why its great XT now features tilt across the range. It'll be even more important with the IQ5 ... On the print. I guess it is an art – personally I do like to actually work with grain which really equalizes the result a bit.

 

[Paul Spinnler]

This is what I’m thinking too if all Leica lenses were to be compatible on a S4 via Leica adapters. At the moment, for color images, I have a Leica M11 simply for use with M lenses (I really like the options of the different renderings that the M lenses provide, from very modern to more classic to outright vintage) and i also have a GFX100S. It’s possible the S4 could bridge both of those in a one-camera system for colour images. + for B&W, the M Monochrom sensor really “organic” (ie, digitally unprocessed in its rendering), especially with the right lenses.

I found actually that the SL was the better camera for manual focusing M glass; given monochrom is only available on M I figured that' the dream combo. EVF based Leica for colour, M for mono. The EVF on the M11 is ok, but the SL2/3 is a lot better. The S4 hopefully will up the ante with 9 MPX and higher mangification and OLED for true blacks ...

 

[jng]

On the diffraction topic. I really can't see that this is such a big factor.

Thats the Rodie 35 HR at F4 centre at 200% percent, slightly sharpened. The loss in detail in the next pic is mainly lack of shadow detail because my shutter time compensation was not perfect and I had to lift the image quite a bit more in the C1 via the exposure slide (so don't mistake the edge of the tree as diffeent in detail – its literally underexposed at F16)
View attachment 211862

Here is the same cutout at F16, a bit more sharpened, also 200%: At 100% I can see minimal differences and in print it'll go under anyhow, especially if you use a bit of grain to take away the digital bite.

View attachment 211863

Its gonna be fine with 250 megapixels on all realistic print scenarios ... and that's a rough sharpen in C1, not advanced stuff with high-pass overlays in PS or AI sharpen layers which are blended in, for example.

I think it is so overblown. If we go to F22, F32 yes, it starts to get a bit blurry, but also not that much at 100% post-sharpen. And if you just add a bit of grain anyhow it looks actually great and more natural.

In practice for the next chip you'd just aim for F8-F11, use tilt for diffraction free sharpness front to back or combine in Halycon. On top you have many sharpening tools or you keep it a bit blurry and work with grain for a nice texture.

We might be saying the same thing here. My point is that when shooting at apertures (physically) smaller than f/4, the smaller pixel pitch of this theoretical sensor won’t buy you any additional resolution over the current generation sensor. Of course, what you intend to do with those pixels afterwards is a different matter….

John

 

[buildbot]

Do you think they can do a custom 54x40? Or is it a goner and IQ5 is 250?

On the SoC – what is your bet of what we'll see in terms of generational improvements between IQ4 and IQ5?

They can custom make anything, and they already have the pixel arch/readout arch done, just need to tweak the number of rows/cols. The trick is, someone needs to pay them for it or they need to see the business case to do so. New masks for new designs are around a million at 90nm and tens of millions at 5nm. Get it wrong? In one of your 35 or so masks (you need different masks for each layer!)? Same cost again.

The IMX411 is in a ton of non-phase things. I don’t think (I don’t know though) Phase had anything to do with the IMX411, just they got the best chip they could. The IMX811 is the same - it’ll be in astro and inspection cams far more than any Phase back that might use it.

Unrelated slightly but good info - https://semiengineering.com/scaling-cmos-image-sensors/

 

[Paul Spinnler]

They can custom make anything, and they already have the pixel arch/readout arch done, just need to tweak the number of rows/cols. The trick is, someone needs to pay them for it or they need to see the business case to do so. New masks for new designs are around a million at 90nm and tens of millions at 5nm. Get it wrong? In one of your 35 or so masks (you need different masks for each layer!)? Same cost again.

The IMX411 is in a ton of non-phase things. I don’t think (I don’t know though) Phase had anything to do with the IMX411, just they got the best chip they could. The IMX811 is the same - it’ll be in astro and inspection cams far more than any Phase back that might use it.

Unrelated slightly but good info - https://semiengineering.com/scaling-cmos-image-sensors/

So which Nanometer mask would apply for a modded 811?

So CPU wise you are saying we are stuck with the same old system?

Super frustrating that we are facing the prospect of 3:2. It’s horrible. I have gotten so accustomed to 54x40.