Resolution is a very vague term because it has so many different meanings. In this article I’ll be focusing on the true detail resolving capability of Epson flatbed scanners. Most people often incorrectly refer to this as “effective resolution” which has lead to endless confusion and sub par results with these scanners. I have an Epson v550 which along with the 8xx series are the newest models, Epson released these back in 2014. The v600 came out in 2009 and the v7xx series came out in 2006/7. My v550 is the low end model, the 8xx and 7xx series are significantly different and SHOULD certainly get far better results than my flatbed.
Detail resolving capability is an actual measurement of the smallest details an imaging system is capable of resolving. This is almost always measured in line PAIRS per millimeter. Film, lenses, and various other imaging equipment follow this standard of measurement. Even when something is listed as “lines/mm” it is almost always line pairs per millimeter (cycles/mm is conceptually the same thing too). There are exceptions like when manufacturers try to make their product look better than it is by listing the actual “lines” per mm, but this is RARE and is an equivalent to dots per mm or pixels per mm so it is easy to spot.
Dots or pixels per [UNIT] is not really a measurement of detail resolving capability it is a measurement of sample rate. You can somewhat extrapolate one from the other but they are two different things. To resolve say, 100 line pairs per millimeter you would need a sample rate of at least 200 dots or pixels per mm. In the real world though it would almost always be more than 2x since details will get lost due to rasterization. When dealing with extremely small analog details of photographic film, cycles per mm is preferred since the image will start to soften the smaller the detail or less contrasting it gets.
Cycles are not static on/off bars but full transitions from white to black. If just two pixels represented that cycle then all of the transnational gradations between that cycle will be lost. Since film is an analog medium, there could technically be an infinite number of pixels representing the single cycle. Of course there will always be a point of diminishing returns. You can extrapolate a minimal dpi requirement to resolve a line pair or cycle measurement but it is a lot harder to determine how many line pairs can be resolved at a given dpi without using a test target. In practice the resolving capability in cycles or line pairs per mm should be something less than half of the dots/pixels per mm measurement.
“Effective resolution” is a made up buzzword that has become widespread with people who don’t know what they are talking about. This measurement was popularized by a single online scanner review blog based out of Germany, ScanDig. This blog claims that every single Epson flatbed scanner has an “effective resolution” of 2300 dpi and this value is then parroted by every know-it-all on-line. ScanDig does not go into details about their actual testing methodology besides explaining how test targets work, which they sell on their shop of course!
ScanDig also claims that when using Silverfast scanning software (which they also sell) the “effective resolution” increases to 2600dpi. This is a major red flag and that alone shows their entire testing methodology is flawed. Every single Epson flatbed they test also gets the same “effective resolution” rating which is quite suspicious. They likely do not test between multiple different sample units and probably do not mount their target at the proper height for their scanner. In fact I don’t know if they even use transparent test targets at all!
On scanners like mine I ASSUME that scanning in transparent mode or reflective mode does not alter the detail resolving capability, though it might. I have tested this and it does seem slightly worse but not enough to give a definitive answer. What I do know is that on the higher end Epson flatbeds it DOES make a difference. The 7xx/8xx series of scanners have what is called the “dual lens system”. This dual lens system is pretty simple, you have one focal plane that should be directly on the surface of the flatbed’s glass bed, and the other will be somewhere above the surface of the glass. Using reflective mode, or not selecting the correct lens in the Epson scan software will give you worse results than the flatbed is capable of.
The flatness of the film and the height above the glass that the film is scanned at is the most important factor in achieving good scans out of flatbed scanners. Due to manufacturing tolerances, age, and a number of other very minute details, each flatbed scanner is UNIQUE! I have found accounts of people who have my exact scanner talking about how they get the best results when they raise the film up higher. Personally I get the best results when I scan my film directly on the glass. You must find the correct focal height of your scanner if you want the best results, otherwise the film will be out of focus and your detail resolving capability will be much much worse. The film must also be perfectly flat, which is usually achieved using a sheet of special glass.
Some film scanners like the Nikon Coolscan series have an auto focus system that automatically adjusts the height of the scanner to get the best results. These Coolscans are significantly older and more primitive in other ways, but they are known for getting results that are better than the Epson perfection series despite using the exact same linear CCD image sensing technology. Epson themselves rates their scanners at 6400dpi “optical resolution”. Most people assume this is the resolution of the optics but this is not the case, it is the sample rate of the linear CCD itself, and it is accurate.
The CCD on an Epson scanner is capable of producing 6400 samples per inch, however due to the focus height issue and a non ideal optical system; the actual detail resolving capability is lower than what the CCD can produce. I can demonstrate this with my scanner. Below you will see three images, all three have approximately the same amount of resolved details but each is scanned at a higher dpi setting. As the Epson scanner increases its sample rate (dpi) you will see more chromatic aberrations appear.
These chromatic aberrations(CA) exist in the optical system, they are not on the film. As the scanning DPI increases the CCD is able to see the CA more clearly. This shows that the actual sensor is capable of resolving details finer than the optics of my scanner can resolve. Higher end Epson scanners have better optics but are still not safe from this problem. Nick Carver’s flatbed scanning techniques video unintentionally demonstrates the CA issue. For his setup the CA is more prevalent when mounting film at a higher or lower height than the exact perfect focal point of the scanner.
Because of this issue, I only scan at 4800dpi on my scanner. 4800 dpi gives me practically no CA and also gives me lots of smooth gradations between the resolved details of the image. You may look at the previous diagram and think that the 3200dpi scan looks “sharper” this is because the image has been upscaled and you are seeing the actual sharp edges of each pixel creating contrasting edges with each other. It is an illusion that tricks your brain into thinking there is more detail where there isn’t. 4800 is the ideal sample rate for my scanner but higher end scanners with better optics may need the full 6400dpi the CCD is capable of.
So what exactly IS the detail resolving capability of my flatbed? Well I can’t give an exact number since I don’t own a proper test target but maybe I’ll buy one in the future. What I can do is completely disprove some of the claims ScanDig makes about “effective resolution” by using my own test target. I took a photo of my computer screen using not the greatest setup but got surprisingly usable results.
I then scanned this 35mm photo using my Epson v550 and my Panasonic Lumix GH4. Both of these were released in 2014 so I would say it is a pretty fair comparison in technologies, but a full frame camera from 2014 would of course be ideal. The GH4 has a 16MP sensor, however it is a 4:3 aspect ratio so when filling the frame up as much as possible with the 3:2 35mm frame I only use about 13MP of the available pixels. If you extrapolate this out into dpi it is the equivalent of approximately 3100dpi. As for the scan from the Epson, well see for yourself:
CLEARLY my Epson v550 has resolved more details. You can notice this the most on the letter Z it is quite easy to make out the square pixels that make up the inner part of the letter. On the GH4 scan ignoring the noise, the image is blurred more and you can’t make out the sharp edges of the pixels. At face value this would mean my Epson is resolving more detail than the 3100dpi scan my GH4 made. However as I mentioned earlier DPI is a measurement of sample rate and not actual detail resolving power. I have no idea how much detail resolving power my GH4 is capable of.
In a perfect world where every two pixels on a GH4’s sensor was able to resolve a line pair then it would come out to about 60lp/mm for these scans which is far less resolution than what most 35mm film is rated at. In reality the GH4 is most certainly resolving slightly less than that, or maybe a little more. Cameras are a little different when it comes to how they resolve things. But whatever it is, the Epson is doing a little better but they are not THAT far apart. The biggest issue I have with ScanDig is their follower’s persistence on “only scanning at the scanner’s effective resolution”. Even if the Epson was only capable of getting X dpi who in their right mind would only want a single pixel to represent the smallest possible detail? What if that smallest detail is a more complex object that doesn't fit perfectly onto the grid of pixels?
Once again this is why I always scan at 4800 dpi on my scanner. Your scanner is likely different since every flatbed scanner is unique. I would assume the v7xx/8xx series is capable of more true detail resolving power than my v550. If that is the case (and you don’t get CA’s) then I would say you should scan at 6400dpi.
The dpi settings higher than 6400 dpi are interpolated values. If the Epson was truly resolving details at or beyond the ability of the CCD then interpolation would come in handy to generate smoother gradations between those smallest single pixel details. Luckily this is certainly not the case so interpolation is usually not ever needed.
There’s something else that almost perfectly lines up with my results and the claims from Epson, and that is the Nyquist rate. The Nyquist rate is the minimal number of samples required to resolve a given frequency, which is always 2X the frequency. In our case, let’s say the Epson flatbeds are optically capable of resolving 6400dpi. If we assume that 6400 is the Nyquist rate then that means the actual smallest frequencies (details) the scanner is capable of resolving are 3200dpi, which lines up PERFECTLY with my results.
One last thing I’ll mention is that sometimes manufacturers DO list their detail resolving power in dpi, this is especially true for drum scanners that do make actual dot scans of an image, and therefor a dpi measurement is somewhat accurate, but once again still does not specifically tell you much about how many line pairs are truly resolvable using that system.
Hopefully in the future Epson or some other company will create a flatbed or really just any kind of film scanning device that is affordable and gives results that will blow any existing technique out of the water. I have some ideas of my own that could be really game changing but I would need crazy $ to test them. I hope you found this article helpful. Go subscribe to the ShyStudios YouTube channel because I will have video companions to these articles up sometime in the future.