Minolta 5400 Mark 1 - ZigZag artefacts

[brbo]

Which is in line with what I suggested before; broad peaks, centered around the wavelengths I mentioned earlier. Apparently, the overlap isn't problematic enough even with a white light source (CFL).

Yes, pretty broad peaks, which might explain why Nikon LED scanners have bit higher saturation and a bit better colour accuracy in raw scans.

 

[Archiloque]

Yes, see what the continuinity testing gives. I'm interested whether you can verify that C42 connects to both pins 20 & 21 on the ADC chip. If so, it would be worthwhile replacing it with a new one and see if that improves matters. It's a good lead.

Ok, I think we have our breakthrough here. There is continuity between C42 and pin 20 and 21. I think we can safely assume we have a very likely culprit to this infamous zigzag plague.

I mean, same symptoms as Coolscans, same ADC, same electrolytic connected to the same pins... Okham's razor and all that. I might order a few caps and learn how to solder SMD's. Shouldn't be THAT hard, right ? My soldering skills aren't that good but it may be worth the calculated risk.

 

[scarbantia]

A lot depends also on signal processing; this is pertinent in your digital camera example.
Spectral sensitivity of the sensor is just one aspect and by itself, it doesn't say all that much.

I would need to make some specific measurements with the scanner, but if I set the backlight to match the colour filters of the CCD, I should be able to get a pretty good result. (like Nikon Coolscans or Frontier SP3000) The current CCFL backlight suffer from significant spectral problems, and require too much correction with IT8 calibration and R-G-B analogue gain settings. A backlight that can be adjusted per colour channel would open up a lot of possibilities, especially for negatives.
You are right that a good white light source can be satisfactory, but it is too exciting to experiment with the RGB LEDs.

Digital cameras are clearly weaker in this area than better scanners, because high sensitivity is more important than colour reproduction. (Phase One's trichromatic sensor is a unique exception, with a base sensitivity of ISO 35; but the Bayer filter is of course also a disadvantage).

 

[scarbantia]

Sony ILX739KA.

The author of this article once commented that Minolta 5400 and Canon FS4000US share the same CCD.

Interesting. I can't find a datasheet for that one, but I suspect the ILX734KA is related and that one has the following spectral sensitivities:

Thank you all, these are useful results!

 

[scarbantia]

Perhaps an ILX548K or ILX558K sensor variant might actually be in there. Or specifically, one is the 5400 and the other is the 5400II sensor. (5Mhz vs 10Mhz, different colour filters)

 

[koraks]

Ok, I think we have our breakthrough here. There is continuity between C42 and pin 20 and 21. I think we can safely assume we have a very likely culprit to this infamous zigzag plague.

I mean, same symptoms as Coolscans, same ADC, same electrolytic connected to the same pins... Okham's razor and all that. I might order a few caps and learn how to solder SMD's. Shouldn't be THAT hard, right ? My soldering skills aren't that good but it may be worth the calculated risk.

Yes, I agree, give that a try.

Replacing those electrolytics is in principle straightforward. I get the best success removing the old one by heating one pad with a regular soldering iron and a not too big, chiseled tip, and pushing the cap away from it so as to lift it slight on that side. Then do the same on the other side; it should pop off. Be careful not to damage the sensor as you'll be working on the same side of the board.

For replacement I'd suggest to try initially with a ceramic cap (the little brown box type); something like 1206 size 10uF will be easy to solder even with a regular iron. Only if that somehow doesn't work (I don't see how) or if you feel you absolutely must replace the cap with a similar electrolytic one, I'd try that. In that case, I'd consider not using an SMD part, but a regular leaded part, and solder the leads to the pads. Soldering an SMD cap of this type to an existing PCB is tricky and it often results in poor connections. Plus, if you use an electrolytic cap in this position, the same issue of degradation plays a role - although that will likely take another few decades to become a problem.

Be sure to get regular lead-containing solder, not the lead-free stuff. The latter is far more difficult to work with.

 

[scarbantia]

I bought a backlight unit of a Coolscan 5000ED scanner. Hopefully it will fit the Minolta 5400 with a little modification.
The optics are good, I hope to get a homogeneous light source. I removed the PCB and it is incomprehensible to me how such old microscopic LEDs can give enough light. Unfortunately I cannot identify them, maybe they could be replaced with Cree XE-Gs.

 

[koraks]

I removed the PCB and it is incomprehensible to me how such old microscopic LEDs can give enough light.

Firstly, LEDs are really, really small. What makes them 'big' is the substrate they're on and the package they're in, which is there for optical, mechanical and thermal reasons. The actual LED die is tiny, as shown in your photos. Secondly, the reason that unit is so efficient is because it's a condensor design that makes very efficient use of the radiant power of the LEDs - nothing is wasted. Conceptually it's similar to a condenser enlarger.

The LEDs in this case are a custom assembly made specifically for this application. You could try to kludge CREEs in there, but thermal considerations may make things a little complicated. Note how the thermal arrangement is in the module you posted: the actual LEDs are on a gold-plated copper structure that also acts as the anode or cathode. Note how there's only green solder masking applied in places where it's strictly necessary so that the traces can radiate heat away from the module. Then, most importantly, the LED assembly is mounted on a relatively beefy chunk of aluminum which acts as a heat sink as well as a mechanical base for the prism and the diffuser.

Instead of trying to fit other LEDs into this (which will never be a snap-in replacement anyway), I'd just use it as-is. However, if your 5400 came with original CFL light, I don't think there'll be LED drivers in there, so you'd have to add a LED power supply and constant current driver for these LEDs as well. Keep in mind you have two pairs of two LEDs in parallel there and one string of four LEDs. I suspect the 4-LED array is the green array and the others are red (easily identifiable) and blue. Drive current will have to be guessed unless you can find specifications or an annotated schematic or LED driver used in the Coolscan this module is intended for.

There's no guarantee that the light source will physically fit and project the correct beam angle in the Minolta scanner. You'll have to try and luck out with that.

Overall, it's not a straightforward project. If I had to personally do this, I'd start with the CFL module of the original scanner and modify that for LED illumination instead. Or maybe just try to find the appropriate CFL tube as a snap-in replacement in the hopes of getting a few more years of good use out of the scanner. That would be by far the easiest and most dependable solution.

 

[Archiloque]

Yeah. Seems like a complicated way to achieve your goal.

I just received a very dense led strip with reasonable CRI. I am simpy going to try to swap the CCFL tube for it and see where that takes me.

I thought about the transparent condenser but I am going to trust the pseudo scanhencer already present in the 5400 to tame the light a bit and make it even. It's not very ambitious I am effraid.

On the recapping front. I actually learned soldering, bought for 200 euros worth of magnifier, soldering stations, test PCB to train, the whole nine yards. I recapped in a very ugly way but hey, it works, the C42 capacitor. aaaaaaaand.... That made absolutely no difference

But I did it with a clean iteration of the scanner. Meaning the zigzag effect was very feint. So I took another ccd unit laying around that I knew was terrible. I made some test scans. Recapped it, and swapped the whole ccd unit into the working scanner. It still looked like utter shit. No difference at all.

My take is that it's a lot more complicated than I thought , and out of my paygrade. But hey at least I can solder now.

 

[koraks]

I just received a very dense led strip with reasonable CRI. I am simpy going to try to swap the CCFL tube for it and see where that takes me.

Just to be sure, yo are aware that the power requirements are fundamentally different, OK? So you need to feed the LED strip with whatever DC it's made for; the CFL itself is fed from something like 110VAC at a relatively high frequency. So don't try to connect the LED strip directly to the wires that go to the CFL tube as that will create fireworks.

That made absolutely no difference

Oh...bummer.

 

[Archiloque]

Don't worry. I know.

I bought a 12v led strip and hope there is some kind of 12v rail somewhere to make it clean, concealed in the scanner. Otherwise I'll use an additional power supply.

 

[koraks]

OK, gotcha. There's a good chance you'll be able to find a 12V source internally, but...if there is, it's likely to be shared with the motor power supplies. This can induce flicker in the LEDs, causing output signal noise. The easiest and most dependable solution would be to add a dedicated 12V supply for the strip.

 

[Anon Ymous]

I bought a backlight unit of a Coolscan 5000ED scanner. Hopefully it will fit the Minolta 5400 with a little modification.
The optics are good, I hope to get a homogeneous light source. I removed the PCB and it is incomprehensible to me how such old microscopic LEDs can give enough light. Unfortunately I cannot identify them, maybe they could be replaced with Cree XE-Gs.

Or maybe just try to find the appropriate CFL tube as a snap-in replacement in the hopes of getting a few more years of good use out of the scanner. That would be by far the easiest and most dependable solution.

Yes, finding a replacement lamp would be the most obvious and straightforward solution. This site might have what you need, have a look:

- CCFL lamps 桅3.0 mm

CCFL lamps 桅3.0 mm

www.lcd-chip.com

 

[Archiloque]

I've tested one of these off the shelf CCFL replacement tube. But from CCFL warehouse. It was a lot dimmer than the original tube, a lot colder and had a strong magenta cast.

 

[scarbantia]

Firstly, LEDs are really, really small. What makes them 'big' is the substrate they're on and the package they're in, which is there for optical, mechanical and thermal reasons. The actual LED die is tiny, as shown in your photos. Secondly, the reason that unit is so efficient is because it's a condensor design that makes very efficient use of the radiant power of the LEDs - nothing is wasted. Conceptually it's similar to a condenser enlarger.

Yes, but the chips are only ~0.3x0.3mm, and the light emitting area is even smaller for the green and blue LEDs. These are from 2001. I looked at it under a microscope at 40x magnification. Perhaps the spectral distribution also helps efficiency.

The LEDs in this case are a custom assembly made specifically for this application. You could try to kludge CREEs in there, but thermal considerations may make things a little complicated. Note how the thermal arrangement is in the module you posted: the actual LEDs are on a gold-plated copper structure that also acts as the anode or cathode. Note how there's only green solder masking applied in places where it's strictly necessary so that the traces can radiate heat away from the module. Then, most importantly, the LED assembly is mounted on a relatively beefy chunk of aluminum which acts as a heat sink as well as a mechanical base for the prism and the diffuser.

The original heat dissipation is brilliant, but fortunately I can fit a larger heatsink in the other side. (see pic) I need a copper based PCB and a heat sink or maybe a heat pipe.

Instead of trying to fit other LEDs into this (which will never be a snap-in replacement anyway), I'd just use it as-is. However, if your 5400 came with original CFL light, I don't think there'll be LED drivers in there, so you'd have to add a LED power supply and constant current driver for these LEDs as well.

Yes, I will need an external power supply because the ccfl use an inverter.

Keep in mind you have two pairs of two LEDs in parallel there and one string of four LEDs. I suspect the 4-LED array is the green array and the others are red (easily identifiable) and blue. Drive current will have to be guessed unless you can find specifications or an annotated schematic or LED driver used in the Coolscan this module is intended for.

There are 8 LEDs in total, green, blue, red and IR. The red and IR are clearly distinguishable. Unfortunately I don't have any documentation for it. I'm trying to find the opening voltage with a regulated DC power supply, but I don't know how far I can go. Another problem is that in the original circuit, each LED had a different resistance, so the colours were balanced.

There's no guarantee that the light source will physically fit and project the correct beam angle in the Minolta scanner. You'll have to try and luck out with that.

If it's not optically perfect, I can always try with the grain dissolver, and increase the illumination. But I'm really curious about the collimated light of the condenser lens. Yes, it's a harder light with additional problems.

 

[scarbantia]

Yeah. Seems like a complicated way to achieve your goal.

I just received a very dense led strip with reasonable CRI. I am simpy going to try to swap the CCFL tube for it and see where that takes me.

I thought about the transparent condenser but I am going to trust the pseudo scanhencer already present in the 5400 to tame the light a bit and make it even. It's not very ambitious I am effraid.

On the recapping front. I actually learned soldering, bought for 200 euros worth of magnifier, soldering stations, test PCB to train, the whole nine yards. I recapped in a very ugly way but hey, it works, the C42 capacitor. aaaaaaaand.... That made absolutely no difference

But I did it with a clean iteration of the scanner. Meaning the zigzag effect was very feint. So I took another ccd unit laying around that I knew was terrible. I made some test scans. Recapped it, and swapped the whole ccd unit into the working scanner. It still looked like utter shit. No difference at all.

My take is that it's a lot more complicated than I thought , and out of my paygrade. But hey at least I can solder now.

I'm sorry it didn't work out. Thank you for trying.

 

[koraks]

Yes, but the chips are only ~0.3x0.3mm

Yes, that's what I explained. The actual LED dies are tiny. That's normal, also on modern power LEDs.

There are 8 LEDs in total, green, blue, red and IR

OK, makes sense; so you've just got to work out which are blue, green and IR, but that's easy. Just power them up.

I'm trying to find the opening voltage with a regulated DC power supply, but I don't know how far I can go. Another problem is that in the original circuit, each LED had a different resistance, so the colours were balanced.

Forward voltages are quite predictable. Anyway, shouldn't be too hard to make some current limiters for something like 50mA or so. Control current, let voltage fall where it may. If you know the value of the current limiter resistors on the Nikon PCB and the supply voltage, you also know the current to aim for.

I doubt the balancing of the channels is very critical as I expect that the scanner's firmware will do an automatic adjustment of sensor sensitivity during startup or even at the start of a scan. And you can always scan as a positive and then do the color balancing yourself.

It's an interesting approach; if you insist on the LEDs, I'd definitely try the ones you've got first before trying to hack other LEDs into this assembly. It would save you a lot of headache.

 

[Archiloque]

Alright. I tried a crude but effective mod this afternoon after getting a very dense COB Led strip with an integrated diffuser on top of it. It is supposed to be 528 leds/meter. So 5 leds/cm (I am a mathematician), I've put around 6/7cm of it into the scanner. Pretty even light, at least with my naked eye.

It's sold as "White color" (6000-6500K) with a CRI>90 and indeed looks like a pretty neutral white when lit. I've dismantled the CCFL tube and replace it without any modification by the strip that holds in place without any trouble. It fits pretty well in the plastic support that maintained the 2.6mm CCFL tube. Effortless.

I just screwed back in place the light casing and let the wires out through the removed wiring of the CCFL.

Ok, I know, it's nothing spectacular. But I've tested some already scanned pictures and here are my takes :

The backlight is absolutely even. Even without the grain dissolver. The scanner is 2 to 3 times faster when the grain dissolver is engaged which slowed the scanning almost to a halt when scanning a dark slide. It used to take up to 12 minutes. Now a dark slide will not take more than 3:30/4 minutes.

As for the IQ results ? Well. It seems to be at least equivalent color wise. I even have noticed some slightly different (better) results with negatives. Below you can find two scans of the same slide.

The first is from my CCFL minolta :

This one if from the Led converted one :

The difference is minimal.
As for the noise, it's slighlty reduced. Here is a pretty dense Provia 100F slide :


And a 400% zoom of its 5 stops lifted shadows :




I guess I should make the wiring a little longer, drill a hole in the back of the scanner for the LED power supply and call it quit. OR I am also thinking of getting a small 110VAC to 12VDC to convert the CCFL power supply wiring to 12V DC so that I don't have another power plug. I don't think it will heat up the scanner that much. I'll test that. I know it's a pretty light mod but for a first, I find it really encouraging.

 

[Anon Ymous]

Impressive

 

[brbo]

EXCELLENT!!!

 

[koraks]

a small 110VAC to 12VDC to convert the CCFL power supply wiring to 12V DC

I wouldn't recommend that. You could try, but don't freak out if it doesn't work. The risks include damage to either of the power supplies.

Your results so far are absolutely convincing and in the noise shot, I don't see any of the zigzag-artefacts the whole thing began with!

 

[Archiloque]

It's still there in the absolute black of the sides but yeah, the noise level has been lowered quite a bit with the amount of additional light. Pretty interesting result.

Note that this particular unit was pretty low on the zigzag effect. I realise I had 4 of these scanners (yeah, I am insane). One is a spare unit. But I'll try to convert the other two and see if that makes a more significant difference

 

[koraks]

It would be very interesting to see whether a badly affected unit could be brought back to like-new performance this way.

 

[scarbantia]

Wow, that looks good!

I tried Silverfast Multi-Exposure, but increased the brightness significantly on the second exposure. Even the darkest details were well exposed with no noise. However, if the brightness is too high, the merging of the two images will be incorrect. Scanned as separate images, they can be perfectly aligned using Photoshop's blending option. My external light source was quite cumbersome, but it would be worth experimenting with.

Brightness control would be a great upgrade for the scanner.

 

[Archiloque]

I wouldn't recommend that. You could try, but don't freak out if it doesn't work. The risks include damage to either of the power supplies.

I've thought about it and you're right, instead I'll order the 24V version of this led strip (they exist and for whatever reason, I orderered the 12V version) and make a smal wiring to the pins of the 24V power supply at the back of the PCB, should be easy and risk free, what do you think ?

Edit : even simpler, I just noticed there's a 24V wiring (measured it) going into a small transformer at the botton of the scanner that - I guess - turns the 24V into high voltage for the CCFL. I'll just get the 24V there and unconnect this transformer totally. Useless now.