The new Muller HDS scanner, what do you think about that?

[prtprt]

The Company Filmfabriek in the Netherlands is coming with a new scanner for 8mm, 9,5 mm and 16 mm (all in one scanner) called Muller HDS. It can also handle sound. Estimated shipping june 2015.

You can read about it here:

http://www.filmfabriek.nl

What do you think about that?

Charlie

[Tscan]

It looks a lot like a scanner i remember seeing a few years ago that never materialized (MWA). Maybe this is it? But for that price I would spend a little extra money and go for the new Blackmagic Cintel scanner.
https://www.blackmagicdesign.com/products/cintel

[RCBasher]

Maybe this is it? But for that price I would spend a little extra money and go for the new Blackmagic Cintel scanner.
https://www.blackmagicdesign.com/products/cintel

They are different machines, the Muller HDS is 8/9.6/16mm and the Cintel is 16/35mm.

[prtprt]

The German company MWA-Nova have 4 versions of film scanners on their web. You can find them here:

http://www.mwa-nova.com/

But I think they are very expensive, more expensive than the Muller HDS.

Charlie

[vintagefilm]

The earlier Muller scanner was a nice piece of work. I saw the one they had at the Internet Archive a few years ago. It has a nice mechanical design, and uses scientific cameras for the imaging, so you can select your own level of cost versus resolution. Solid European engineering. I think the cost is high but probably worth it in an transfer house. (Actually, when I first looked at these, the cost was much higher, somewhere around 40K, so it has come down, perhaps with the cost of cameras coming down) The most interesting feature for me is not the hardware, but their restoration software http://www.filmfabriek.nl/restoration.html It seems to be based on Virtual Dub, as seen in the screenshots. So this puts it in the same league as the Film9 software. I am still working to get my head around that package. My limited tests look pretty good, but I am not getting reliable output yet, at least not to use it as a production tool. Has anyone done a direct comparison between these two software packages? Also, I should point out that Clive Tobin has been looking at scientific cameras as well. A friend just took delivery of a Point Grey based machine with 2048x2048 resolution. I have not yet seen it. the quest for more pixels never ends.

[JeremyC]

the quest for more pixels never ends.

If I have a 3k camera with a 14 or 16 bit ADC wouldn't that 'image' more detail than a 4k camera with a 10 bit ADC, depending on the image.

[RCBasher]

A friend just took delivery of a Point Grey based machine with 2048x2048 resolution. I have not yet seen it. the quest for more pixels never ends.

If I have a 3k camera with a 14 or 16 bit ADC wouldn't that 'image' more detail than a camera with a 10 bit ADC, depending on the image.

Assuming the PGR 2048 x 2048 is the one with the CMOSIS sensor then you would most likely be correct Jeremy. I have tested that sensor (not in a PGR camera) and the dynamic range was appalling.
The other problem with high pixel count sensors is their requirement for an exceptional lens in order to image individual detail to each pixel. Often all you get is a bigger file with no real gain other than pixel-count bragging points!

[JeremyC]

Frank,

Thanks for the affirmation and very interesting to see your designs being directly adopted by manufacturers of small format scanners.

In your experience with machine vision cameras which are the ones that do provide good dynamic range and are CCD sensors still superior compared with CMOS technology sensors?

[RCBasher]

I have not had the luxury Jeremy of a wide variance cameras coming my way, so my hands-on experiences are limited.

The “iconic” ICX285 CCD sensor seems to have set a good bench mark for many years and still is at the top of many people’s list, especially amongst astronomers who need low noise/high dynamic range. Although only 1384x1036 its 6.45um pixel size has ensured its reputation.

However, the newer Sony HAD II process used in the ICX674 has managed to increase the pixel count within the same size chip but retain the same dynamic range. In fact in my tests it actually performed slightly better than my ICX285 camera. It is also about the same price, so it was a bit of a surprise when I saw a short while back that PGR had introduced a new USB 3 model with the ICX285 alongside the ICX674 model! As I understand it, the reputation of this sensor seems to be prolonging its natural life!

For a higher pixel count with the same quality, the ICX694 should perform the same as the ICX674 as it is the same technology, just a larger chip to accommodate the higher pixel count. If purchased in quad tap readout form then nearly as fast as the ICX674 in dual tap form.

On the CMOS front, as I already mentioned, the CMOSIS CMV2000/4000 sensors are well down on dynamic range. The one I tested showed zero detail in the dark area of a high contrast image with high gamma applied whereas the CCD sensors and now the new Sony IMX174 CMOS sensor show quite a bit of detail. PGR’s own tests (see their GS3-U3_ImagingPerformance.pdf) show the CMOSIS to be around 7dB (more than a stop’s worth) of dynamic range down on the IXC674.

I’ve mentioned the IMX174 before and although I haven’t yet seen the claimed 10dB improvement over the ICX674 it does at least match it, but at a significantly lower price.

That’s pretty much it for the main current generation sensors I’ve had my hands on. Must take this opportunity to thank Christoph Manz for his help in analysing the various sensor performances, especially with FFTs of the dark frames.

[VideoFred]

The most interesting feature for me is not the hardware, but their restoration software http://www.filmfabriek.nl/restoration.html It seems to be based on Virtual Dub, as seen in the screenshots. So this puts it in the same league as the Film9 software. I am still working to get my head around that package. My limited tests look pretty good, but I am not getting reliable output yet, at least not to use it as a production tool. Has anyone done a direct comparison between these two software packages?

Both software packages are based on my Avisynth film restoring script. ;)

Fred.

[Andreas Wideroe]

USB3 is too slow for scanning of images in realtime (24fps) with a resolution of 4K + higher bitdepth. Would love to see support for cameralink, coaxpress or similare high speed protocols in this or other modulare scanners.

Also I wish I would see support for up to 16 bits (bitdepth).

Teledyne Dalsa has some awesome cameras btw, but they cost lots of $$$.

About the restoration software, last time I checked VirtualDub and AVI synth didn't have support for more than 8bit video. Hopefully this has changed, or?

About resolution, these cameras use bayer pattern which means the images in reality are 1/3 in resolution of what is advertised. However, I would choose areascan (bayer) cameras over linescan cameras because of image stability and less possible artifacts.

Cheers,
Andreas

[RCBasher]

USB3 is too slow for scanning of images in realtime (24fps) with a resolution of 4K + higher bitdepth. Would love to see support for cameralink, coaxpress or similare high speed protocols in this or other modulare scanners.

Not true. 4000x3000 sensor is 12Mpixels. Even at 16 bit transfer at 24fps that comes in under the USB 5Gbit limit. Cameralink is hideously expensive at these transfer rates. Anyway, what global shutter 4K sensor can deliver RAW data at 24fps?

Also I wish I would see support for up to 16 bits (bitdepth).

Do you mean 16 bits of useable image data or 16bit packet depth? The later is easily handled, I use it all the time. Image data is only 12 or 14bit at best though. 12bit A/D is good enough for over 70dB of dynamic range and it is a rare sensor that can achieve this. 12bit data can be packed into 16bit data transfer, hence increasing the throughput significantly.

About the restoration software, last time I checked VirtualDub and AVI synth didn't have support for more than 8bit video. Hopefully this has changed, or?

Still only 8bit, but if one can afford the costs of 4K scanning (it takes quite a lot of computing power and storage speed to handle at 24fps, should such a camera exist) then one can afford a professional NLE.

About resolution, these cameras use bayer pattern which means the images in reality are 1/3 in resolution of what is advertised. However, I would choose areascan (bayer) cameras over linescan cameras because of image stability and less possible artifacts.

The green channel is 50% which is where the eye is sensitive to image detail. A good de-Bayer algorithm will produce an image approximately 80% of that for a full RGB sensor (or separate mono scans) of the same nominal pixel count.

[Gelinox]

Both software packages are based on my Avisynth film restoring script. ;)

Fred.

For not to perturb this thread we answer here ->>
viewtopic.php?f=1&t=24564&p=215531#p215531 ;)

[JeremyC]

About resolution, these cameras use bayer pattern which means the images in reality are 1/3 in resolution of what is advertised. However, I would choose areascan (bayer) cameras over linescan cameras because of image stability and less possible artifacts.

Or use a mono camera and take RGB separate frames and merge them.

[Andreas Wideroe]

About resolution, these cameras use bayer pattern which means the images in reality are 1/3 in resolution of what is advertised. However, I would choose areascan (bayer) cameras over linescan cameras because of image stability and less possible artifacts.

Or use a mono camera and take RGB separate frames and merge them.

Yes, that would be an awesome option (modulare upgrade/module).

/Andreas