Minolta 5400 Mark 1 - ZigZag artefacts

[Archiloque]

There was a toolkit for the 5400 calibration process : colors, resolution, etc. It's referenced in the repair manual but probably long gone.

 

[brbo]

Did you do it on the 5400?

I guess it's not that different from other scanner procedure. There is a tutorial by Gleb Shtengel for the coolscan 8000/9000. You "just" need to move the board by very small increments so that the image is centered and not skewed. Feasible I guess.

Exactly.

 

[ScanMan_2]

I've been using a Minolta 5400 scanner for a few years now. Despite having both top of the line Coolscans at home (the 5000 and 9000), I still believe the Minolta 5400 Mk1 is my best scanner when it comes to extracting as much information as possible from a 35mm negative. I have scanned numerous photos with all these scanners and while the Coolscan tend to have better colors right out of the box, the Minolta 5400 diffuse lighting really give a more pleasing and forgiving image, especially when you have some imperfections on your negatives. Also, its resolution and grain rendition is unmatched.

It is slow, it is a brick. But I really believe it was the pinnacle in terms of image quality when it comes to advanced user scanners for 35mm.

Anyway, after these praises for a 20 years old scanner nobody asked for, here is my problem.

I noticed in dark areas the presence of zigzag artefacts. No matter what value of multisampling I choose, they're still here. I know that Coolscans tend to suffer from the same illness. I have an old Coolscan 4000 that suffers from it. But this issue has been well documented and investigated by the community, and the culprit identified. It's a simple 30 cents capacitor involved in the Analog Digital conversion that needs to be replaced.

Unfortunately, the 5400 doesn't attract the same interest from users nowadays. Sadly, their only appeal seems to be their lens, hence the sheer number of units sold without it on Ebay. Anyway, I was wondering if anyone had any idea of what capacitor or IC chip could be falty ?
Here are two examples. One in the shadows of a slide. Another in the densest highlights of a negative. Click to enlarge :

View attachment 336889 View attachment 336890

I plan to have all the chemical capacitors replaced at some point but I wondered if anyone had already encoutered this issue and had some clues. Thanks.

Just signed up to partake in this post cause my 5400 has just started doing this exact same zigzag pattern. How interesting… I have pulled the thing apart twice previously to take care of lines that were caused by lubrication issues but this is new to me. I was going to buy one of the ones on eBay with the lens removed to do a swap but it’s still quite expensive due to shipping for UK to AU so if I can fix the problem by replacing capacitors that would be good.

 

[Archiloque]

Well unfortunately, koraks is sending me the scanner. He tried to intervene on the ccd board and added a capacitor in parralel to the one he was suspecting to be faulty with no visible effects on the image.

I'll try to have all its chemicals capacitors replaced when I get it back and see if that has an effect. I've also ordered a replacement light bulb because a fading backlight might have an effect on the noise generated by the sensor.

 

[koraks]

On the unit I've worked on for @Archiloque, capacitor failure does not seem to be the problem.

I think it's also relevant to point out that the noise problem only occurs with that particular unit under the following combination of conditions:
* Grain dissolve mode is enabled.
* An area of very high density is scanned (i.e. a virtually pitch black slide)
* The scan is made with very high contrast (e.g. to compensate for extreme underexposure of a slide)
In my tests on this unit it produces in my opinion perfectly acceptable, low-noise scans from properly exposed slides. The noise problem is also only very slight on extremely high-contrast scans of extremely dense areas as long as grain dissolve mode is left disabled.

For those running into these zig-zag noise problems, I would recommend to first verify if the problems occur under normal operating conditions, or if they are isolated to fairly extreme/border-line cases like the combination I described above. In case of the latter, I suspect that factors like lamp aging but also production batch variations and even unit-to-unit variations can play a role.

By means of illustration, here are a few example scans I made over the past two days using @Archiloque's unit:

5400dpi 100% crop from a normally exposed (but high-contrast scene) Fuji slide; likely Sensia 100. Grain dissolve was OFF for this scan.

5400dpi 100% crop of the edge of a slide showing the rebate area and a very deeply underexposed area (this looks pitch black when viewed with the naked eye). Grain dissolve OFF. Fuji Sensia 100.

5400dpi 100% crop of part of the same area of the same slide as above, but now scanned with grain dissolve ON. This replicates the zig-zag problem that @Archiloque reported.

Insofar as there is a problem with this unit (and in my humble opinion, there isn't - I think it works just fine under normal conditions), a possible culprit would be inadequate filtering of the analog power supply on the analog/AD board. This is a relevant section of that board:

The white bar top left in the image is in fact part of the CCD assembly itself. In Nikon scanners with capacitor problems, there is a conceptually similar section, with an electrolytic capacitor providing bulk filtering capacity, and smaller ceramic capacitors around the A/D chip to filter its analog and digital power supplies. Of these, the electrolytic capacitor would be the prime suspect, since these capacitors are somewhat prone to aging. Indeed, it's common to find faulty electrolytics (with effectively zero capacitance) in old equipment. Sometimes this is accompanied with physical signs of damage (bulging, leaking). Sometimes there are no physical signs of damage. Note that capacitance cannot be measured in-circuit, so there's no non-invasive way to determine if there is a problem here.

In this particular case, what I did was establish that the contacts for the non-populated C22 are parallel to C32. This is likely an alternative or additional capacitor that was simply never included in assembly because it's redundant. It did mean however that we could use this site to add a 100uF/16V (power supply in this section is 12V) ceramic capacitor. These generally have far lower ESR and better durability than electrolytic caps. If C32 were faulty and present virtually no capacitance, mounting another (new) capacitor parallel to it would alleviate the problem. Here's what that could look like:

Note C22 is populated now with a modern ceramic capacitor; this one is rated for 100uF/16V and effectively offers double the capacitance of the original 47uF/25V C32.

After this fix, I made some new scans, but the noise performance was unchanged. Note that I personally found the performance very well to begin with (the scanner performs in no way any worse than e.g. my own Scan Dual IV - if anything, it's the opposite).

In my personal opinion, it makes no sense to try and change additional ceramic decoupling capacitors around the A/D chip, for two reasons:
1: The newly fitted C22 would have made a noticeable difference since it offers low impedance filtering on the analog supply. Had the decoupling caps on the A/D chip failed, this would at least have made an improvement, even if it would not have constituted a 100% solution.
2: Ceramic capacitors are known to be quite durable. Given the overall production quality of these devices, combined with the favorable conditions under which these decoupling caps operate (no excessive temperatures, very low current loads) I find it very unlikely that they have degraded significantly and that replacing them would have any merit.

I do think that there's a good possibility of lamp degradation exacerbating noise in these scanners. The scanner and the scanning software adjust signal gain to compensate for overall density differences between scanned samples, but this will also (whether intentional or not) compensate for degrading light output of the light source. Since this is a fluorescent lamp with limited lifetime (8000 hour according to to Konica Minolta), it's very conceivable that the lamps in these units, especially heavily-used ones, are starting to fade. I therefore agree with @Archiloque that it's worth a shot to try and replace the lamp and see if it makes an improvement.

In private communication with @Archiloque, I have expressed my view that noise improvement of these units might be possible, but would involve fairly far-reaching interventions:
* Total separation of analog and digital power supplies, and particularly separating the motor drive power supply entirely from the rest of the unit. Note that all power supplies trace back to the mainboard and are fairly highly integrated with each other and the digital electronics, so this would constitute major interventions on the mainboard at least (cutting traces, adding new PCB's and external connectors etc.)
* Additional shielding of parts; it might be possible to shield e.g. the main stepper motor more effectively (surrounding it with copper foil tied to ground, comes to mind) and/or doing the same with the analog and A/D electronics insofar mechanically possible.
* Galvanic isolation of connections between the analog/AD board and the main board. This again would be a major intervention requiring replacing the connector cable with two new cables as well as a new (to be designed & manufactured) PCB with suitable isolation electronics for the various signals.
All this is a pretty labor-intensive project. There is a slight possibility that it will somewhat improve the performance of the scanner, but it would effectively be an attempt to take it beyond the performance levels Konica Minolta envisioned for this device. Nothing wrong with that, but the phrase 'putting the cart before the horse' comes to mind. I would also be hesitant to perform major and irreversible surgery on a unit that to the best of my estimation seems to perform quite well as it is.

Personally, my advice remains to leave the grain dissolver disabled when scanning extremely dense samples. That seems to solve the problem at least with this particular unit just as well.

 

[Archiloque]

I am very grateful for @koraks attempt at changing the capacitor on the A/D board. I am still convinced the 5400 wasn't that noisy when it was released, and, somehow, something is causing this weird shaped noise. Yet, the scanner still scans remarkably well negatives and normal slides. Even with grain dissolver. But in difficult conditions, its shortcomings seem more apparent than before. Also, the grain dissolver was a huge selling point back in the days. So something is definitely wrong.

I'll keep updating this thread with hypothetical discoveries. First will be the swapping of the CCFL tube. I'll report when done.

 

[brbo]

Does every software (Vuescan, Minolta DiMage Scan, Silverfast) produce same pattern?

 

[Archiloque]

I did not try Silverfast - overpriced - but Vuescan and Dimage produce sensibly the same results.

The only difference between vuefast and Dimage seem to be related to CCD calibration. Hot pixels (that produce streaks along the sensor travel) are more visible on Dimage , they're effectively removed with Vuescan but it weirdley duplicates adjacent pixels and create 4 or 5 pixels uniform wide streaks.

That's the only difference I've been able to find in the rendering of both softwares.

 

[koraks]

But in difficult conditions, its shortcomings seem more apparent than before.

Yes, you demonstrated that quite convincingly with a scan of a few years ago and a recent one. Although I think these scans were made on different machines of the same type, right? Either way, I also found the difference remarkable. I have some hopes for improvement if you manage to replace the light source.

Does every software (Vuescan, Minolta DiMage Scan, Silverfast) produce same pattern?

I think @Archiloque scans with the Minolta software and I tried it with Vuescan, and we get the same pattern, so it seems hardware-related. My hypothesis is still that the noise is caused by either noise injected from the stepper motor into the power supply rails, and/or EMI emanating from the stepper picked up by the CCD and accompanying circuitry. It will show up mostly when dealing with very weak signals, resulting in a poor s/n ratio.

Vuescan but it weirdley duplicates adjacent pixels and create weird 4 or 5 pixels wide streaks.

Yes, I noticed these too, and we corresponded about them.

I am very grateful for @koraks attempt at changing the capacitor on the A/D board.

You're very welcome; I enjoyed the excellent exchange on the technical troubleshooting part and your responsiveness in this! I'm sorry I couldn't realize a satisfactory solution at this point.

 

[ScanMan_2]

Yes, you demonstrated that quite convincingly with a scan of a few years ago and a recent one. Although I think these scans were made on different machines of the same type, right? Either way, I also found the difference remarkable. I have some hopes for improvement if you manage to replace the light source.



I think @Archiloque scans with the Minolta software and I tried it with Vuescan, and we get the same pattern, so it seems hardware-related. My hypothesis is still that the noise is caused by either noise injected from the stepper motor into the power supply rails, and/or EMI emanating from the stepper picked up by the CCD and accompanying circuitry. It will show up mostly when dealing with very weak signals, resulting in a poor s/n ratio.



Yes, I noticed these too, and we corresponded about them.



You're very welcome; I enjoyed the excellent exchange on the technical troubleshooting part and your responsiveness in this! I'm sorry I couldn't realize a satisfactory solution at this point.

I don’t use grain disolver but I do use the ICE setting. I will run some scans of various settings over the weekend and see if there’s any examples worth posting.

 

[shijan]

Is it happens with 2x or 4x multisample setting? From my experience it was a visibble improvement with 2x multisample in my scanner when i used it. So i always scanned with 2x on 5400II.

 

[shijan]

I also can confirm that at pixel level Vuescan software gives striped digital texture pattern compare to original Minolta scanner software. At least it was in version somewhere in 2013 when i tested it. It is not related to exposure. It is just some sort of less accurate "demosaic" variation.

 

[Archiloque]

Is it happens with 2x or 4x multisample setting? From my experience it was a visibble improvement with 2x multisample in my scanner when i used it. So i always scanned with 2x on 5400II.

In my case, multisaMpling will get rid of color noise but the pattern will be there nonetheless. Only a bit "cleaner"

I also can confirm that at pixel level Vuescan software gives striped digital texture pattern compare to original Minolta scanner software. At least it was in version somewhere in 2013 when i tested it. It is not related to exposure. It is just some sort of less accurate "demosaic" variation.

I am pretty sure it's related to the way ccd calibration works in vuscan. There is no demosaicing in line CCD sensors. Some pixels are probably considered out of specs (hot, dead) and discarded, then replaced by their nearest neighbors. I don't have hot pixels streaks in Vuescan as in Dimage. But I get these weird blocky lines instead.

 

[scarbantia]

Thanks for sharing your experience. I have a 5400(mk I) with the same problem, but it's not caused by lamp degradation.
The problem appeared overnight, but it has never happened before with dark slide film.
I'm using Silverfast Ai 8 with calibration.

 

[koraks]

Welcome aboard @scarbantia !
Could you post an example of the noise you get? It's interesting that it has appeared overnight. That suggests some kind of rather sudden failure.

 

[scarbantia]

Thanks the quick response in this old thread! Here is a before-after comparison.
The error appeared a month ago, I've never seen it before.

 

[Archiloque]

Yeah. Same problem, really.

I've tested a few of these scanners and they all show the same pattern. Something is badly aging and we don't know what it is. Once again, the Coolscans also show a similar problem that can be cured by the replacement of a particular cap. Unfortunately I do not have the knowledge to determine which one it is on the Minolta.

 

[scarbantia]

Some of my previous scans, all looks good.
Unfortunately, I have no idea at the moment, but this scanner deserves to be kept alive.

 

[koraks]

I did some limited work on one of @Archiloque's scanners, which didn't solve the problem. I replaced the filter caps on the A/D chip, which should have helped to give the chip a cleaner power supply. Apparently, this noise is injected through a different route. It's possible that replacing some electrolytics in the power supply itself might help. However, for proper diagnosis, you'd have to ideally rig up a scope to the scanner and then try and locate sources of noise as the scanner operates. This would work best using two scanners, one with a noise problem and one without. That way it's a little easier to spot the noise that's responsible for the image artefacts (there'll always be noise, everywhere - the question is how much of it in which spot is the problem here).

 

[brbo]

@scarbantia, it would probably help if you scanned the exact same slides now and compare the scans to your previous scans from those slides. Not saying that your scanner didn't change behaviour, but that it would be easier to compare the "before" and "after".

I have 5400 mk1 and will try to dig up some dense slides to see if my scanner have the same problem.

 

[scarbantia]

I tried re-scanning with the same settings, but the new scan is slightly brighter. (.icc profile, analog gain, ME, etc.)

200% crop:

 

[brbo]

So the left side (brighter) is the new scan? It actually shows less noise (though I would atribute that to it being slightly out of focus).

I made some scans myself yesterday and can observe similar pattern. It's much more prominent with grain dissolver enabled and depends on the hardware exposure, the lower the exposure the higher the noise in shadows, which is perfectly understandable. If you use software that doesn't give you (or you don't use it) control over hardware exposure then you will definitely get slightly different results even from the same slide.

 

[scarbantia]

So the left side (brighter) is the new scan? It actually shows less noise (though I would atribute that to it being slightly out of focus).

I made some scans myself yesterday and can observe similar pattern. It's much more prominent with grain dissolver enabled and depends on the hardware exposure, the lower the exposure the higher the noise in shadows, which is perfectly understandable. If you use software that doesn't give you (or you don't use it) control over hardware exposure then you will definitely get slightly different results even from the same slide.

No, the right side is the new scan. The problem is the zigzag noise pattern. The left side has less shadow detail, but the overall scan is slightly darker, or perhaps less contrasty.

 

[koraks]

The left side has less shadow detail, but the overall scan is slightly darker, or perhaps less contrasty.

This implies that it's difficult or even impossible to verify the suspicion that the noise problem appeared suddenly. The scanner may have performed the same, always, but you only noticed it at a certain point.

 

[brbo]

@scarbantia, that's with grain dissolver enabled, right?

Do you feel that you need to enable it on slide film? Unless it is Fuji Provia 100F and a scene with a lot of areas where infamous "pepper grain" is visible I would never enable grain dissolver. You loose a tiny bit of sharpness and scan time basically doubles (depending on the film this can cause the film to curl during scanning).