« Peregrinations on the East Side with Art | Main | Random Excellence: Stephen Gillette »

Tuesday, 03 November 2009

Comments

Worth reading, thanks, even if the writing was hard to follow at times. But the only issue addressed in the article that seems especially complicated by the big-sensor-small-body brief is autofocus speed. It's an important issue and the reps do a good job of explaining the challenge, but it made the absence of a voice from Panasonic doubly disappointing.

It seems a lot of armchair engineers start in on these cameras with 'I don't see why they just didn't....' I've certainly have been guilty of that rant. I guess if it was easy everyone would be making them and they would cost $300.

Interesting article. However, isn't it erroneous to say the absence of an AA filter is the cause of higher noise at high ISO in the M9? It would seem the AA filter would have exactly zero impact on the noise characteristics of a sensor.

Sounds like they are a generation or two from making it work. That being said, I really wouldn't mind a Canon G12 or G13 (whatever number they're at these days) that only had 3-4 Megapixels and better low light performance - plenty for a 5x7.

From the last page of the linked article:
"Though the M9 was the best of the lot, with such a huge image sensor you'd think it would not even be close. However, like many medium-format digital camera companies, Leica chose not to use an anti-aliasing (or "blur") filter over the sensor. The benefit is more detail and increased sharpness in bright light. The downside is you increase the incidence of noise in high ISO images."

Hogwash.
Moiré and/or stairstepping in some cases, yes. High ISO noise? Ridiculous.

Some statements from the pdngearguide.com article make me think the author is not really an expert on technology:

"In the case of the EP-1, the path the focusing signal must travel from the image sensor so it can communicate with the circuitry of the interchangeable lenses, adds to the focusing time."

The time an electrical impulse takes to travel the distance between sensor and lens is less than a nanosecond. It's the processing of the sensor's signal and the movement of the lens that take time.

"...Leica chose not to use an anti-aliasing (or "blur") filter over the sensor. The benefit is more detail and increased sharpness in bright light. The downside is you increase the incidence of noise in high ISO images."

High-ISO noise (which is an effect taking place in the sensor's, amplifier's and A/D-converter's electronics) cannot be decreased by putting an anti-aliasing filter (which is an optical device) over the sensor. In fact using a AA filter requires more sharpening of the resulting picture which would in turn increase the visibility of noise.

I´m sorry, but I still don´t get it.
There had been gazillions of compact point and shoot cameras in the film era that were autofocus AND had 35mm sensors [aka film] behind.

The article focuses on autofocus difficulties, more than on size difficulties.

It looks more like an excuse all this stuff of autofocus engines and gears and sensors.

All in all it seems that price is the main driving factor for not having big sensors on a point´n´shoot camera.

I guess I must be missing something obvious, but it doesn't appear to be possible to print that article; which is a shame as I would really like to read it properly and share it with someone at home.

It was very professional of PDN to swipe the sensor-size comparison graphic from the wikipedia without attribution.

http://en.wikipedia.org/wiki/File:SensorSizes.png

James,
That sensor size graph is public domain and doesn't need to be attributed.

Mike

the higher grade lenses are actually a hindrance to autofocusing speed. In the case of the EP-1, the path the focusing signal must travel from the image sensor so it can communicate with the circuitry of the interchangeable lenses, adds to the focusing time.

I got to this same point and was completely baffled. How are better lenses a hindrance? Because the better IQ that they allow makes focus errors easier to see? If they were saying that it takes a while to process all the data coming off the sensor to make focus decisions, I could understand that. If they were saying that the smaller DOF compared to point and shoots makes focusing a harder problem, I would totally buy that.

My guess is that what we are seeing here is non-english engineer-speak poorly translated and filtered through a tired journalist's consciousness before being cut for length by an editor.

Really, it's a shame, since I think the autofocus challenges for mirrorless sensor cameras are very interesting. I am a little puzzled that there aren't more design tricks used to prefocus lenses, so cameras don't have to hunt across the whole focus range.

Ask soft questions, get soft answers. Worse still, he didn't actually answer the question he sought to, as he let the marketing messages from the companies completely distract him.

1. There is nothing in particular that would stop a large sensor compact from focusing as fast as the fastest small sensor compact.
2. The excuse that "adding a dedicated AF part" would make the compacts more expensive is belied by the fact that Olympus sells traditional DSLRs for less than the E-P1. I call bull****. So would any good reporter.
3. The speed at which a lens' mechanical focusing mechanism moves is determined by motor size versus mass. Use too small a motor, the mass will move slowly. The real issue here is that the companies in question are using older sample/test/move/sample/test/move algorithms.
4. The contention that noise wasn't as well handled as sensor size would project is totally unsupported (and even contradicted in the article). Start with a noisy sensor, end with a noisy camera would be the real problem. But I suspect his evidence is 100% anecdotal, not actually tested.

Funny thing is, he didn't actually uncover the real reasons why large sensor compact cameras are difficult to make right.

Right, yet the Panasonics blow the AF speed argument out of the water. A few posters have already picked on the weird statement about the AA filter.

I was hoping to read about things like the difficulties the lens-sensor interactions in a small package, compacting all the electronics, power issues... but nada?

"the M9 is a bit of an anomaly because of its high price and stubborn rangefinder traditionalism. (Only manual focus and lens-based aperture adjustment.)"

how horrid.

A point that was glossed over in the PDN article is that Leica avoided the whole autofocus speed issue by using a manual rangefinder. I don't hear any M9 owners complaining about focusing speed and accuracy.

This option is equally available to the makers of smaller sensor cameras, which because of their shorter focal length lenses and greater depth-of-field require less focusing precision. I would settle for lenses that had a decent focusing scale and depth-of-field scale so I could set the focus directly and manually--but of course that would be "a step backward."

I don't see how you can have an article about large sensors in small cameras and miss the "angle of the impinging light on the sensor" - umm - angle.

From the article:
---
"With present mechanical engineering, it would've been pretty hard to put an AF sensor in there and still keep the body as slim as we did," said Richard Pelkowski, also an Olympus Product Manager. "And if you look at the construction of the camera and everything it does have, an HD movie mode, a 3-inch LCD etc; if you add in an AF sensor, it would've been impossible to deliver it at a cost that's going to allow people to actually purchase it."
---

I, for one, would prefer they include the AF chip and forego the HD video. But maybe that's just me. It does seem that speedy AF would be more in-line with the ideal DMD than video, but maybe the DMD is less salable than the current trend towards multiple shiny features.

Even if it weren't laughable that the signal path from the AF circuitry to the lens is responsible for slow AF, how would a "high grade lens" make it worse ? (A "high grade lens" may be more difficult to focus quickly from a mechanical point of view, but that's not what the author is suggesting). How does the GF1 focus noticably faster ? And how does a DSLR with a longer registration distance not suffer from this increased circuit path ?

"Noticeably absent from PhotoPlus this year was Panasonic". I really wish Panasonic were there. I had to go to B&H to see a GF1 and couldn't see the 20/2.7. But they weren't there in 2007 or 2008 either so it's not surprising.

"All of these cameras were disappointingly noisy at ISO 800 and above. " Anyone reading that would think that DSLRs beat these cameras hands down. I found ISO 1600 samples from the EP1 I used there to be pretty darned close to those from my Sony A700 and certainly usable. Noise at pixel peeping 100% view, sure, but nothing that can't be cleaned up and nothing that's even visible in, say, an 8x10.

@ Iñaki "There had been gazillions of compact point and shoot cameras in the film era that were autofocus AND had 35mm sensors [aka film] behind."

I agree. But most, if not all of these had the autofocus looking out through a window near the lens. They were not very compact cameras. My Ricoh FF9, for example, is 120 X 68 X 42 mm and has a fixed 35 mm lens. The new Pen is 121 X 70 X 35 mm, plus lenses.

To put an AF sensor window where it would not get shrouded by any lens you put on it would increase the size of the camera. Also, with this set up, the lens and camera have to talk to each other. That would (as far as I know) make putting a lens on via an adaptor manual focus only, even if the lens did not obscure the AF window.

If there could be some sort of semi silvered mirror between the lens and sensor that let an internal AF sensor see to focus, we might see faster AF. The mirror would be a darker, fuzzy spot if using live view, and would just swing out of the way, maybe sideways, during exposure.

Just an idea.

Mike, I didn't say PDN was legally required to provide attribution on the graphic, I just said it was unprofessional and a little slimy (thus "swiped") not to attribute it.

One cannot help but be impressed with the sophistication of the readers of TOP. When an article does not convey a sense that the author really grasps a situation, the readers of TOP notice it.

A few responses to the responses :

-- When the article mentions high grade lenses, I assume it really means interchangable lenses. In that case, needing to communicate with the circuitry in the lens does add some delay, since built-in lenses are typically smaller, and don't have circuitry located away from the rest of the camera. And it's not necessarily a direct path with no handshake or processing going on, so you can't assume the only delay is the signal propagation time.

-- The lack of an AA filter does not change the sensor noise characteristics. However, it does allow noise to arrive at the sensor and show up in pictures more than it would with an AA filter. The AA filter serves to smooth out noise, similar to how some noise reduction software works. So, yes, images may appear to be noisier without an AA filter, but you're capturing at least as much info and can deal with it in software - probably better than you could were an AA filter in place.

-- I believe it's the combo of making the body small enough and adding the AF sensor that makes it so expensive. The fact that other SLRs WITH an AF sensor are cheaper doesn't invalidate that, since they all have significantly larger bodies.

-- Panasonic may have much better AF in their cameras than the others, but my understanding is that it still doesn't match an entry level SLR in many situations.

I don't think it was a particularly well written article either, but bad writing doesn't necessarily mean all the info was wrong. Perhaps just poorly stated. And I really thought there'd be something about correcting lenses or algorithms for the larger angles of the light as well.

Not that great of an article. Just pick up any of the panasonic micro 4/3's cameras and you will see some blazingly fast AF. My G1 focuses faster then my dslr.

I think the writer needed to do a little more research, instead of just pumping this article out overnight. At least that is how it reads.

The issue he mentioned with "high grade" lenses was their greater mass of glass to be moved. I believe that's real, if not all that dominant.

Some people seem to be missing the complexity of the dedicated AF sensor. The mechanism used by DSLRs, the through-the-lens phase-detection AF system, requires the AF sensor to be located in the image plane. Right where the sensor already is. Which is physically impossible. The DSLRs work around this by having semi-transparent spots in the mirror AND a secondary mirror on the back of the main mirror, reflecting the image down to the AF sensors in the base of the body. This is complex, noisy (and shaky), space-consuming, and hard to build precisely enough (and hard to keep aligned precisely enough).

There are, of course, other approaches. One could use a SMALL moving mirror that just relayed the center AF point to a dedicated AF sensor (and moved out of the way for exposure). That only works for a single AF point, though. One could use the kind of infrared AF that 35mm P&S cameras used in the 1980s (though most of them didn't have the zoom range of something like the G11; not sure the IR-AF works at the long end).

Or one could license whatever patents Panasonic has on whatever makes the GF1 so much better :-).

'I would settle for lenses that had a decent focusing scale and depth-of-field scale so I could set the focus directly and manually--but of course that would be "a step backward."'

To be fair, the author did cite the M9's esoteric "traditionalism". But I hear you. In some minds this would make something like the DP1 more usable. No doubt a risk today for a mass market item, even at the "luxury" level, especially given that 4/3 cameras are shrinking and can take most manual lenses.

But what if we step further "backward"? What about zone focusing? A slowish high-quality wide lens; a zone focus lever; simple finder. Maybe a couple of AE presets for DOF or shutter speed bias. Without the costs of mechanical rings, linkages, lens markings and sturdier barrel, it might be affordable enough--as a daylight street shooter, a professional's pocket snapper, a hipster thing... in many body color options, of course.

But this bring up another point that the article glossed over: because of massive DOF, compact digicams can get away with sloppy AF (in effect zone AF). That similar AF strategies do not translate easily to larger formats was made painfully obvious when live view came along. And, again, frustrating that the party most successful at tackling this problem was omitted.

btw, did anyone understand the relevance of that analogy about the hose and the bucket? I suspect it was left over from a discussion that got cut.

"There had been gazillions of compact point and shoot cameras in the film era that were autofocus AND had 35mm sensors [aka film] behind."

Iñaki, most of those cameras were not very fast or reliable at autofocus. Seems to me that the current generation of large sensor compacts is at least at or past that level (excepting maybe Sigma). The problem isn't autofocus per se, but autofocus at or near DSLR speeds, while keeping costs down. So, yes, price is the issue, but cost contributes to price, and apparently speedy autofocus costs money or space.

Phat Photographer's interest in future Canon G-series camera with fewer pixels prompts me to ask an oblique question. To what extent can you get the effect of fewer better, cleaner pixels by downsampling a larger image. Take a Canon G10 for example and cut its 14.7 MP down by a factor of 4 in postprocessing to get the desired 3-4 MP result.

If downsampling works acceptably, then both the majority interest (lotsa pixels) and minority interest (fewer/better pixels) would be satisifed by the same camera.

@David Bostedo "-- The lack of an AA filter does not change the sensor noise characteristics. However, it does allow noise to arrive at the sensor and show up in pictures more than it would with an AA filter. The AA filter serves to smooth out noise, similar to how some noise reduction software works. So, yes, images may appear to be noisier without an AA filter, but you're capturing at least as much info and can deal with it in software - probably better than you could were an AA filter in place."

Sorry, but I don't understand what you're saying. What noise is arriving at the sensor?

The lack of an AA filter means that all the light that the lens transmits is allowed to strike the sensor. Noise is not a part of that light, it's an electronic anomaly that's added to the image by the sensor itself. An optical filter in front of the sensor can't have any effect on high-ISO noise because that noise only exists after the filter, and it's not an optical phenomenon. The presence or absence of high-ISO noise is in no way connected to the presence or absence of an AA filter.

@Inaki: "There had been gazillions of compact point and shoot cameras in the film era that were autofocus AND had 35mm sensors [aka film] behind."

Yes, but they also had wonderful f/5.6-13 zoom lenses and Kodak marketed "Max" film for their use. Would you accept a digital camera that needed to be set to ISO 800 to get acceptable exposure times even on a sunny day?

You could probably make something quite small and fast using a prime, but that market is extremly limited. Most people will want a zoom, but you can't make those fast AND small. Using a small zoom lens thus takes away from the biggest advantage of larger sensors (lower noise) because you'd have to crank it up to 800 on most days where a smaller sensor with f/2 lens would be fine at ISO 100.

robert e:
"Iñaki, most of those cameras were not very fast or reliable at autofocus."

I had a point and shoot that had three auto focus settings: close, not so close and far.

Dear David and Dave,

I believe the issue with noise and AA filters is spatial frequencies. You can do things to limit high frequency noise in an AA-filtered camera because there's no subject detail at the single-pixel level. In an unfiltered camera, one has to tread much more carefully.

RAW or no, the cameras do a LOT of signal processing before you get to see the image.

pax / Ctein

@bas et all:
I know that all this autofocus stuff will be sluguish, or was slugish.

I just don´t care. What I mean is that, partially, it seems we are overall stagnant at some point in the techinical evolution.

There had been previous cameras with side moving mirrors [Oly did it], as well as side viewfinders [Oly again] or 110 reflex systems with, guess what, MINIATURE LENSES THAT HAD SPECTACULAR APERTURES.

Thing is all this had been done previously, and for instance, Panasonic still does some impressive zoom lenses with propper apperture range throughout the zoom. And lens aperture IS a physical characteristic of it. So it does not change because of the size of the sensor.

There had been vertically developed zooms [Sony], and spiraling retractile mechanisms [Pentax Espio and Optio].

For god´s sake, companies DO HAVE I+R+D departments to SOLVE the problems, not to hide behind them, don´t they?

And sorry about the tone, but cameras have become expensive enough and obsolescent enough [lack of driver development] to forgo all those sins.

The comments to this entry are closed.