Fools rush in where angels fear to tread. Only mad dogs and Englishmen go out in the noon-day sun. I’m going to ask a question in the spirit of these two quotations.

The April 30 IJM newsletter discusses the higher resolving power of piezography. There’s the statement: “Piezography K7 curves can resolve over 1000 pixels per inch”. For followers of Piezography, this statement is familiar. It’s been made before and is included in the new Piezography Manual on p68.

The discussion surrounding this statement is usually about using 7 inks and QTR’s different dithering patterns which mean that there’s no white space between the dots, unlike the Epson K3 driver. I’m no expert, but I don’t have any trouble understanding and believing the resolving benefits of K7 over K3 and QTR dithering over OEM. There was that small font test image which was used as proof once upon a time. What I don’t understand is the benefit of sending more than 720dpi to QTR.

Anyone who reads Lula will know that there has been a debate raging for years about whether or not to resample when printing using the Epson driver. I think there’s now a consensus that you’re better to send a file to the driver in the driver’s native resolution (for an Epson - 360dpi or 720dpi depending on printer and settings) than to let the combination of the OS and driver do some low-grade resampling for you.

But what about QTR? Roy made a post on Yahoo QTR a while back:
Yahoo! Groups
Now at first blush this seems to agree with Jon - he seems to say to set the print size to whatever you need without any resampling, so that the dpi adjusts to suit, and let QTR do the rest. But he also says “QTR works on 720ppi input data so on Mac OSX that is provided by the system, on PC there is an extra step to resample to 720ppi for QTR”.

Given this, what’s the point of sending more that 720dpi to QTR? As I read this, QTR is going to resample down to 720dpi. We understand that K7 and better dithering is going to provide more detail, but in what sense is Piezography resolving 1000dpi? Are you really saying that you can see a difference between a 1000dpi image sent to QTR and one resampled to 720dpi and then sent to QTR? Or have I misunderstood what Roy seems to be saying?

Roy holds the view that it’s hard to spot the difference in real world images if you resample before sending to QTR, but if it’s true that QTR does want 720dpi as the input resolution then surely it is open to the user to use high quality resampling tools, .e.g. Qimage on Windows, to deliver a 720dpi image to QTR if they feel that they can see the difference.

You are not going to be able to take advantage of this without having a lot of high standard optical resolution. You can’t invent that by sampling. You will actually end up with the opposite result either softening the image or having the K6 / K7 curves reveal the patterning of the sampling algorithms when you use special software.

When Piezography K7 was first released many of its users were not yet shooting digital. Drum scanning of large format film was actually very popular. PMT drum scanners are capable of awesome optical resolution. The CCD scanners do not produce the same type of acuity and they often step or interpolate.

So, its like the old Welsh saying, “you either have it or you don’t” and I’m not taking about the Boom-Boom. Optical resolution is the native resolution that an image is captured with or scanned at (without interpolation).

Sony is rumored to release an A7r II with 59 megapixels. But, 9000 pixels over 8x10 print is difficult to perceive. Too small to look at all that detail. It’s there. What you want is 24,000 pixels across 24" which is something you can perceive. You won’t get it without large film and a drum scanner. You can’t invent the pixels by sampling. That would be like putting a cheap tele-converter on a Leica Summilux ƒ1.4 to judge whether the Summi is as sharp as they say it is.

Scanning film in itself is also subject to different standards. I can spot a FlexTight scan when I see one. It has a pronounced patterning. Yet, its better than using an Epson V desktop. The PMT drum scanners are where you want to be. The finer the grain film, the better.

So where does that leave everyone else shooting DSLR? In a really great place. At Cone Editions Press we routinely print iPhone images at 20" x 24" and even 24" x 24". Piezography does an amazing job at 180 dpi even 120dpi.

So, it’s relative to what you have and what you expect.

In a nutshell - you either have the boom-boom or you don’t. If you don’t - then don’t dwell on resolution. If you want it - then go buy yourself a 4x5 and some Dektol! Our Hell 3400 drum scanner cost way over $100,000 when we bought it back in 1994. They do not even sell on eBay when listed at $1,000. They also weigh 3,000 lbs. But this Howtek with 8,000 dpi optical resolution sold for $1,500 and fits on a desktop. Put a 4x5 on it and you have 32,000 x 40,000 pixels. Photoshop can now handle that. You can make 40" x 50" tac sharp prints.


Jon, thanks for the reply, but something remains unclear. Roy says that files are resampled to 720dpi for QTR. In Windows, QTR does it, but in MacOS apparently the printing pipeline does it, which is supposedly worse. So what use is all that extra optical resolution - from a drum scanner or hi-resolution digital - when it’s going to be resampled to 720dpi?

a good question and I do not have an answer for you. certainly the Mac OS has changed since 10.4 - especially towards printing.

But, do you have a way to have some fine grain film drum scanned so you can determine for yourself?

Are you able to do any high optical resolution printing or is this just a hypothetical question?

Is this just a hypothetical question? Yes and no and no, in that order.

Yes, because at the moment I don’t have a digital camera that produces enough megapixels to print at more than 720dpi at my normal print sizes. It might be an issue if I chose to print small enough. Nor do I intend to purchase a drum scanner for my medium format negs.

No, it’s not hypothetical, in the sense that high megapixel cameras are starting to become available, and I could purchase one if there was a clear benefit. Your 30 April newsletter seemed to be an invitation to do so, in order to use all those extra megapixels. But given Roy’s description of how QTR resamples, I just can’t see how that makes any sense, beyond a certain (720dpi) point. Perhaps for mega-sized prints.

And no, because there is the general issue of workflow and resampling or not for QTR. This is the real reason for my question. For some time I resampled my images (up) to 720dpi for K7 printing with QTR. I don’t recall why I did this, but I must have read somewhere that QTR expects 720dpi. Then somewhere along the line I read your comment to just give QTR whatever optical resolution you have, be it high or low, and let it deal with it. The implication to me was that QTR wasn’t resampling. So I did this. It was a more convenient workflow, as it meant that I didn’t have to save temporary images for printing, I could print the master copy.

But then came Roy’s recent post indicating that my initial understanding was correct - QTR expects 720dpi, and either resamples the image itself to get it as an input resolution (on Windows), or relies on the OS printing pipeline to do it (MacOS). As I mentioned in my initial post, this is precisely the issue that has caused so much debate, esp on Lula, about how to print using the Epson driver. Do you resample or not, if so what to resample to and how best to do it, and if you don’t is there any resampling occurring anyway, and if so what software is doing it and how is it doing it?

This statement of Roy’s on Yahoo-QTR has caused a fair bit of off-forum consternation and argument about what the best K7 workflow is. In reading what you and Roy have said, there is the appearance of an inconsistency. True, Roy like you says to just throw whatever optical resolution that you have at QTR. But for different reasons. He thinks that any resampling effects are not likely to be seen. I am not a Mac user, but based on what I’ve read on Lula I’m really surprised that he says this, given the reports of noticeable resampling artefacts that the MacOS printing pipeline produces. At least on Windows any resampling is done by QTR and therefore was programmed by Roy and therefore hopefully is better quality. There are some, including those who occasionally post on this forum, who think you should never resample, and only print at the physical size that corresponds to the optical resolution at 720dpi. Personally I think that’s one step too far.

You may well say, why not make comparative prints yourself and see? Because I’m more interested in making images rather than multiple test prints and staring at them. How would you ever know whether your tests cover all types of images? It ought to be possible to specify the optimal workflow and just stick it. So that’s why I raised the question - not because I necessarily want to print 1000dpi images, but because I want to understand the optimal QTR K7 workflow.

Thank you for your responses.

I’m going to be that Englishman that ventures out in the noon-day sun. Given that I’m a self-proclaimed amateur, not even printing piezography yet… I really must be a fool! Part of my reasons for venturing out on a limb is I am trying to figure out if piezography is worth it for me to try. Heck, all I need is a $250 R1430, some ink and cartridges and I’m ready to start. But would it be worth it for my amateur perspective and 13" wide prints? I’ve been investigating various prints and what potential they have. I also had an interesting experience with resizing and resampling. Here goes.

I hear what Jon says about resizing, especially upsizing. Three years ago I took a whole set (600 photos) in VGA mode by accident. I was devastated because that was a big family trip and the prints were just no good for printing on anything larger than 2x3 inches. Then I tried upsizing and resampling. True, it won’t invent detail, but it can smooth edges. The eyes in the pictures were heavily pixelated in the VGA sized images, but when upsized from 640 to 2560 (a huge amount), I was surprised how good the detail in the eyes was. I was not expecting such a result. There were a lot of artifacts remaining, and I wouldn’t want to print a large 8x10, but upsizing was not as bad an idea as I was expecting. The reason is that it’s interpolating the pixels.

Because of this, I would not expect moderate upsizing to “damage” a good image. I think it comes down to how much upsizing can it take before artifacts become noticeable. In fact, because upsizing managed to smooth many curved lines (around eyes especially), I would expect it to improve transitions from dark to lighter tones. However, I’d expect that with Jon’s equipment, the images he’s working with begin with plenty of pixels to begin with.

Brian’s post has raised a new question for me which may give me the high-noon sunburn. If QTR expects a 720 dpi image, or QTR will resize to 720 dip, why can you choose 1440 or 2880 dpi in resolution? Are these two different things? If they are the same thing, why would QTR downsize to 720 only to print at 2880 dpi?

Anyway, I really hope some one more knowledgeable responds for Brian here. I don’t want to be the only Englishman to be hung out to dry. If it helps, I’m only half English anyway. And I live in Canada. That may be worse though…