USB 1.1 vs Firewire 400 (vs USB 2)

[ant!]

Until now I used for 35mm scans a Minolta Dual Scan III, which uses a straightforward USB 2. Resolution works for me, only the missing hardware dust removal is something I missed sometimes, even though it only works on color films.

I found now a very reasonable priced Minolta Elite Scan II (not the 5400 Elite Scan II!), which is basically the same generation but with the dust removal infrared light, see here. Not sure how else it differs, the claimed dpi is the same as in the Dual Scan III, the density range a bit higher. The connection side is a bit different though, instead of USB 2 I can either use USB 1.1 or Firewire 400. The scanner is still on the way to me, so still have to try it out...

So my question: Using USB 1.1, is it much slower then the Dual Scan III using USB 2? Would using Firewire 400 make sense? How is the speed difference Firewire vs USB 2?
I have a laptop with Expresscard slot (Thinkpad 540p), and there are Firewire 400 Expresscards around for about CAD 40 + shipping. I never used Expresscard or Firewire, is there anything to take care of beside the correct Expresscard size (34mm) and Firewire version (400)? Any drivers needed for those cards? I plan on using it with Kubuntu Linux with Vuescan, but in case this creates problems and Firewire is much faster then USB 1.1, using Windows 10 on the same laptop is an option.

To make it short: Is a Firewire card worth the extra expense? How much faster would it be? Any headaches to expect?

 

[brbo]

So my question: Using USB 1.1, is it much slower then the Dual Scan III using USB 2? Would using Firewire 400 make sense?

FW400 and USB2 will be close in speed. USB1.1 is quite a bit slower, but with the Minolta Elite Scan II 2.820dpi resolution it might not be a (serious) bottleneck. If my math is right, if scanner doesn't have mechanical ability to scan the frame at full resolution in less than 1 minute, USB1.1 (theoretically) should not become saturated so it should not have a big effect on scanning speed.

Theory is one thing, though...

 

[ant!]

Thanks! I'll guess once I receive it I will be comparing the two scanner's speed over USB. If the Elite is significantly slower, I should get the FireWire card to try out if this makes it faster... (I got the Elite for a price similar or cheaper then the best deals on FireWire Expresscards...)

 

[Bushcat]

Does Kubuntu still support Firewire? I thought it was deprecated a couple of years ago. Having said that, a USB to Firewire cable may be cheaper than a Firewire Expresscard.

 

[brbo]

Does Kubuntu still support Firewire? I thought it was deprecated a couple of years ago. Having said that, a USB to Firewire cable may be cheaper than a Firewire Expresscard.

There's no such thing as just a "USB to FW cable" that will work for connecting scanners to PCs. There are some USB-FW solutions, but are very expensive (at least for the scanners that we are talking about here) and won't work in Linux.

 

[ant!]

According to what I find, at least in 2020 FireWire was still supported in all Ubuntu flavors. Some years ago the packages were just replaced by a newer version: https://help.ubuntu.com/community/FireWire and https://askubuntu.com/questions/1209215/is-firewire-still-supported-by-ubuntu-19-10

 

[grat]

USB 1.1 maxes out at 12 mbit/s (1.2 Megabytes/s). USB 2 tops out at 480 mbit/s (53 megabytes), and FW400 stops at just under 400 mbit/s (49 MB/s). Take a look at the sizes of your RAW or TIFF scans, and see how long they'd take to transfer, and whether that's longer than the scan times you're used to.

There are USB <-> firewire adapters, and they appear to be rather cheap, but I have no idea how well they work.

There are PCIe 1394a adapters available, and my 4.19 linux based system recognizes a similar card just fine.

 

[ant!]

According to what I can find, the cheap FireWire to usb adapters, as mentioned already by brbo, do not work, and some might actually damage the hardware. The transfer rates should be indeed useful to guess if FireWire speeds something up. I'll post my tests here...

As said, my laptop has an Expresscard slot, this would be probably the easiest to connect FireWire. There is also an internal m.2 slot and a miniPCIe slot, but both are more difficult to reach and from what I can see prices aren't cheaper for FireWire (for M2 probably even not existing), so the Expresscard which sits on the outside looks more promising...

 

[brbo]

Again, FW-USB adapters do not work. But, as always, you are free to try and be the first to make it work.

If you have USB-C on your computer you could probably make it work with following adapters:

FW400 - FW800
+ Apple FW800 - Thunderbolt
+ Apple Thunderbolt - USB-C

The adapters will cost you almost $100. I'd guess that's more than OP paid for the scanner.

 

[ant!]

If you have USB-C on your computer you could probably make it work with following adapters:

FW400 - FW800
+ Apple FW800 - Thunderbolt
+ Apple Thunderbolt - USB-C

The adapters will cost you almost $100. I'd guess that's more than OP paid for the scanner.

Yes, it is more... And the laptop I mentioned has no usb-c. But good to know this route might work as well in case I need it one day, I found it elsewhere as well (if the usb-c works as a thunderbolt 3 port...).

 

[grat]

PCIe firewire card is about $20-30 USD.

 

[brbo]

OP has a laptop.

 

[grat]

Whoops! I failed literacy today. Been a rough day for me.

yeah-- PC laptops with thunderbolt are kind of rare. I think the OP is stuck with USB 1.1, although since the scanner can theoretically fill a firewire 400 pipe, maybe they'll get the full 1.2 mbyte/s, which wouldn't be that terribly slow.

 

[Les Sarile]

Until now I used for 35mm scans a Minolta Dual Scan III, which uses a straightforward USB 2. Resolution works for me, only the missing hardware dust removal is something I missed sometimes, even though it only works on color films.

I found now a very reasonable priced Minolta Elite Scan II (not the 5400 Elite Scan II!), which is basically the same generation but with the dust removal infrared light, see here. Not sure how else it differs, the claimed dpi is the same as in the Dual Scan III, the density range a bit higher. The connection side is a bit different though, instead of USB 2 I can either use USB 1.1 or Firewire 400. The scanner is still on the way to me, so still have to try it out...

So my question: Using USB 1.1, is it much slower then the Dual Scan III using USB 2? Would using Firewire 400 make sense? How is the speed difference Firewire vs USB 2?
I have a laptop with Expresscard slot (Thinkpad 540p), and there are Firewire 400 Expresscards around for about CAD 40 + shipping. I never used Expresscard or Firewire, is there anything to take care of beside the correct Expresscard size (34mm) and Firewire version (400)? Any drivers needed for those cards? I plan on using it with Kubuntu Linux with Vuescan, but in case this creates problems and Firewire is much faster then USB 1.1, using Windows 10 on the same laptop is an option.

To make it short: Is a Firewire card worth the extra expense? How much faster would it be? Any headaches to expect?

This review of the Minolta Dimage Scan Elite II includes scan times with the different interfaces and shows using the firewire interface to cut scan times in half over the USB.

 

[brbo]

Yes, as I said before, theory is one thing...

One of my scanners (Microtek 4000tf) has both USB1.1 and FW400 connection, so I did some tests...


FW400 (PowerMac G4 PPC 1.25GHz, OSX 10.4, Vuescan (48bit, set RGB exp. to 1, no AF, no ICE)):

4000dpi, 1x multi-sample: 74s
4000dpi, 4x multi-sample: 268s
2000dpi, 1x multi-sample: 15s
2000dpi, 2x multi-sample: 31s
2000dpi, 3x multi-sample: 47s

(times do not include time used by the scanner for pre-scan setup (calibration, holder positioning) - 19s)

USB1.1 (same as above):

4000dpi, 1x, no AF, setup (22s): 156s
4000dpi, 4x, no AF, setup (22s): 622s
2000dpi, 1x, no AF, setup (22s): 39s
2000dpi, 2x, no AF, setup (22s): 80s
2000dpi, 3x, no AF, setup (22s): 117s

(pre-scan setup - 22s)


FW400 (PC i5-3470 3.2GHz, Win 10):

4000dpi, 1x multi-sample: 79s
4000dpi, 4x multi-sample: 289s
2000dpi, 1x multi-sample: 15s
2000dpi, 2x multi-sample: 31s
2000dpi, 3x multi-sample: 47s

(pre-scan setup - 19s)

USB1.1 (same as above):

4000dpi, 1x, no AF, setup (22s): 131s
4000dpi, 4x, no AF, setup (22s): 520s
2000dpi, 1x, no AF, setup (22s): 33s
2000dpi, 2x, no AF, setup (22s): 67s
2000dpi, 3x, no AF, setup (22s): 98s

(pre-scan setup - 22s)


You might want to get that FW card after all. Even when USB1.1's lower bandwidth shouldn't be a problem, there is still a huge performance hit. If you don't use multi-sampling you might still be OK with USB1.1 and scan times around 1min per frame, though...

Computer speed and platform connection standard implementation obviously has an effect too, but it's fairly minor.


(somebody with much more patience than I have might put this number into a thing that Photrio wants to be a table, grrr...)

 

[Les Sarile]

I've tried multi-sampling on my Coolscans but have never been able to see any quantifiable benefit. Anybody seen any benefit to this feature?

 

[ant!]

Thanks for these numbers! So yeah, looks like firewire would make sense. The scanner is still in the post, but I will definitely compare it with the Dualscan III's USB 2.0 first.

Multisampling might be more needed for some slide films? Just a guess, when I tried it out in the past with negatives I also didn't see a difference.

One more question regarding the hardware dust removal by infrared light: I think I had seen in Vuescan a setting to export a raw with 48bit RGB+infrared channel. Does this mean the information is not actually treated for dust removal yet and only in an extra channel, so the removal needs to be done in external software (I am using Darktable with negadoctor)? Or is the information already all treated? Usually I used raw in Vuescan, should I use tiff instead but setting everything to slide and neutral/no corrections?

 

[Les Sarile]

I've tried multi-sampling on slides as well and seen no benefits.

The IR cleaning uses Applied Science which is proprietary and applied at time of scan.

 

[runswithsizzers]

Since I switched to photographing my film with a digital camera, it has been a while since I used my Minolta Dimage Scan Elite F-2900. But as I recall, I did see an improvement in noise reduction by using multisampling via VueScan. Multisampling was especially useful when trying to scan slide film that was somewhat underexposed. In deep shadows, I could see detail on the slide, but the scanner had a hard time telling faint details from black. If I used multiscan, I was able to recover a little bit more shadow detail before noise became too objectionable. As I recall, 3x was only slightly better than 2x, and I could not see any additional benefit from 4x.

Happy to report my present camera-based method is better at capturing faint shadow detail from slides than the old Minolta Scanner was. BTW, my Dimage Scan Elite F-2900 came with a SCSI connection which I was able to use with a SCSI-to-Firewire (400) adapter until I got my most recent iMac which has no Firewire.

 

[jtk]

IncredibVly slow scanner Vs Nikon V. Happily most folks have PCs and don't need firewire, which I think Intel abandoned. Firewire was always an Applescam.

 

[benveniste]

PCIe firewire card is about $20-30 USD.

True, but Microsoft dropped support for Firewire midway through Windows 10. Legacy drivers still work for now, but as is the case with SCSI, support can go away at almost anytime, and I've seen little interest from manufacturers in certifying their own drivers.

My fallback plan (for a Nikon 8000ED) is to use a Minisforum PC with a Thunderbolt 4 port. Once I bought the cables, that worked fine out of the box for Ubuntu, but Windows 11 needed updated drivers which I found on a Lenovo site.

 

[ant!]

I'm a long time Linux user, mostly Kubuntu, and use Windows only on work computers. My private laptop is still dual boot with win10, but seems like I boot into Windows once in two years... And I am quite optimistic I can find a Linux flavor who supports this still quite a while... I'm happy vuescan works great with Linux, and darktable comes anyways from the Linux world, even though it runs now as well in Windows...

And thunderbolt 4 uses the same connector as usb c, and I see many PCs supporting it... Which means the adapter chain described above (FireWire 400 -> FireWire 800 -> thunderbolt 2 -> thunderbolt 3/4)

 

[grat]

True, but Microsoft dropped support for Firewire midway through Windows 10. Legacy drivers still work for now, but as is the case with SCSI, support can go away at almost anytime, and I've seen little interest from manufacturers in certifying their own drivers.

My fallback plan (for a Nikon 8000ED) is to use a Minisforum PC with a Thunderbolt 4 port. Once I bought the cables, that worked fine out of the box for Ubuntu, but Windows 11 needed updated drivers which I found on a Lenovo site.

The OP was suggesting they preferred to run things under Kubuntu-- and I know my Manjaro-based media center can play nice with a PCIe firewire card, because it has one installed at the moment.

Unfortunately, their primary computer is a laptop, which is difficult to wedge a PCIe card into.

 

[brbo]

One more question regarding the hardware dust removal by infrared light: I think I had seen in Vuescan a setting to export a raw with 48bit RGB+infrared channel. Does this mean the information is not actually treated for dust removal yet and only in an extra channel, so the removal needs to be done in external software (I am using Darktable with negadoctor)? Or is the information already all treated?

It's all up to the user. You can have your raw file saved with IR channel exported as an extra (alpha) channel (resulting file is then a 64bit file) that you can use in Photoshop or some other software to remove the dust and scratches or you can have Vuescan to apply corrections for you before saving the "raw" file. For the latter you need to have "Infrared clean" enabled in the 'Filter' tab and 'Raw output with' option set to 'Save' in 'Output' tab.

Usually I used raw in Vuescan, should I use tiff instead but setting everything to slide and neutral/no corrections?

I'm guessing for external software that is not expecting to work with raw files, scanning everything with 'Media' as 'Slide' (or even better with 'Image'), 'Color balance' to 'None' and 'Output color space' to 'Device RGB' will be the best starting point.

I've tried multi-sampling on slides as well and seen no benefits.

Not seeing something (in your use) is not the same as something not having benefits.

For example, I don't see much benefits for scanning with multisampling since my post-processing is usually not really aggressive, but there DEFINITELY is a good side to multisampling - less noise. There are also bad sides - slower scanning times that heat up the negative more and make it curl more.

A dense slide of Velvia 100F (click to get to full resolution files):


(no multisampling)


(4x multisampling)

400% crop:


(no ms)


(4x ms)

No difference, right? But, what if you were to lift the shadows?


(no ms)


(4x ms)

Happy to report my present camera-based method is better at capturing faint shadow detail from slides than the old Minolta Scanner was. BTW, my Dimage Scan Elite F-2900 came with a SCSI connection which I was able to use with a SCSI-to-Firewire (400) adapter until I got my most recent iMac which has no Firewire.

You "only" need to get a couple of Apple adapter already mentioned to get from FW to USB-C or Thunderbolt that your Apple computer has, but probably doesn't make much sense spending money on adapters to get F-2900 working if you already have everything else for camera scanning...

 

[Les Sarile]

Not seeing something (in your use) is not the same as something not having benefits.

For example, I don't see much benefits for scanning with multisampling since my post-processing is usually not really aggressive, but there DEFINITELY is a good side to multisampling - less noise. There are also bad sides - slower scanning times that heat up the negative more and make it curl more.

A dense slide of Velvia 100F (click to get to full resolution files):


(no multisampling)


(4x multisampling)

400% crop:


(no ms)


(4x ms)

No difference, right? But, what if you were to lift the shadows?


(no ms)


(4x ms)



You "only" need to get a couple of Apple adapter already mentioned to get from FW to USB-C or Thunderbolt that your Apple computer has, but probably doesn't make much sense spending money on adapters to get F-2900 working if you already have everything else for camera scanning...

Thank you for taking the time and effort to show any benefits of multisampling.

I too have conducted similar tests from perfectly fine well exposed samples to those that could use major post work and have not see any benefits to it. This includes 16X on very underexposed Kodachrome. Even in Scandig Coolscan 5000 review they state, "An innocent feature with partly groundbreaking effects!" Unfortunately they don't provide any examples showing the benefits in this case which is odd since they are usually very thorough.