• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required as is 20 years of participation in forums (not all true). There are daily reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Snake oil in photography

Offler

Senior Member
Joined
May 27, 2021
Messages
367
Likes
320
I have been dealing with snakeoil claims in gaming computer builds for some time - most commonly how a mouse with higher polling rate will help you win, while people on this forum seem to be particularly knowlegeable about the tech in general I would like to ask their opinion on snakeoil in photography.

Few I recognize is snake oil claim "larger sensor means lower noise because of more light".

Been reading into topic and in general... people do mistake pixel size with total sensor size.

There is a light-sensitive surface and the rest of electronics - usually ADC converter and one or multiple amplifiers. The "non-sensitive" part of sensor is usually not discussed by the photographers at all. Also as demonstrated on solar panels, lens in front of photosensitive surface can be used to concentrate the light and shrink the sensitive part significantly - and to be clear many sensors use microlenses for about 10+ years.

Another which sounds profoundly like HIFI sellers "this lens has creamy bokeh".

Usually true description, but also put on a very basic and heavy manual lens - usually priced in a similar or same way as modern automatic lenses.
 

Keith_W

Active Member
Joined
Jun 26, 2016
Messages
184
Likes
88

BTW, I would be quite interested in a separate thread with examples of snake oil in computing that you have observed. I always wondered whether you really need more than 200fps because people spend stupendous amounts of money to do so. And then there are all those aesthetic things that gamers go for that make no difference to performance and makes maintenance more difficult ... I am certainly guilty of that by jamming a watercooling rig into a tiny early 2000's case that was not designed for watercooling.
 

Timon VDB

Active Member
Joined
Oct 13, 2020
Messages
116
Likes
173
Location
NL -> AT (Vienna)
BTW, I would be quite interested in a separate thread with examples of snake oil in computing that you have observed. I always wondered whether you really need more than 200fps because people spend stupendous amounts of money to do so.
The scientific answer to that is you need at least 500 frames per second to fully defeat human visual acuity. source

Of course for the average Joe it's a bit like 10k euro carbon bikes :)
 

Timon VDB

Active Member
Joined
Oct 13, 2020
Messages
116
Likes
173
Location
NL -> AT (Vienna)
I have been dealing with snakeoil claims in gaming computer builds for some time - most commonly how a mouse with higher polling rate will help you win
Do you mean mouse DPI? The 'default' USB polling rate of 125hz is definitely way too low to beat human perception. source
 

tonycollinet

Major Contributor
Joined
Sep 4, 2021
Messages
3,092
Likes
4,650
Location
UK/Cheshire
I have been dealing with snakeoil claims in gaming computer builds for some time - most commonly how a mouse with higher polling rate will help you win, while people on this forum seem to be particularly knowlegeable about the tech in general I would like to ask their opinion on snakeoil in photography.

Few I recognize is snake oil claim "larger sensor means lower noise because of more light".

Been reading into topic and in general... people do mistake pixel size with total sensor size.

There is a light-sensitive surface and the rest of electronics - usually ADC converter and one or multiple amplifiers. The "non-sensitive" part of sensor is usually not discussed by the photographers at all. Also as demonstrated on solar panels, lens in front of photosensitive surface can be used to concentrate the light and shrink the sensitive part significantly - and to be clear many sensors use microlenses for about 10+ years.

Another which sounds profoundly like HIFI sellers "this lens has creamy bokeh".

Usually true description, but also put on a very basic and heavy manual lens - usually priced in a similar or same way as modern automatic lenses.
It is obvious that a larger sensor receives more light. If we assume that the non sensing electronics per pixel (you mentioned) shouldn't need to get larger as the sensor gets larger, then larger sensors obviously have significantly more area available for light sensing than small sensors.

Of course, they don't have to use it - but then why use a larger sensor?

Plus - we can clearly see that most cameras with large sensors have both better noise levels, and better resolution (even at the same pixel count) than smaller sensors. Lenses also cannot have infinite resolution.

So, no - that is not snake oil)

Lens characteristics also affect smoothness of bokeh. You can see this by looking at the edges of out of focus highlights. Lenses where those are soft will result in smoother bokeh. Hard edges result in less smooth bokeh

Both of these phenomena are measurable.

I'm sure there is snake oil in photography, but the two items you mention are not that - and I can't think of any definite ones off the top of my head. We definitely don't suffer from the likes of magic crystals to put on top of the camera, or $10,000 cables to connect camera to computer.
 

Propheticus

Senior Member
Joined
Nov 18, 2020
Messages
391
Likes
587
Location
Vleuten, Netherlands
Is there a question here? Or is this an invite to pile on our examples of photography snake oil?

Quality of bokeh is very visible and measureable. Its quality is affected by spherical and chromatic aberration (SA, CA) caused by spherical lenses and its roundness by the shape and number of aperture blades.
The quality of the aspherical lens element(s) used to correct SA and CA also has an impact on the uniformity of the bokeh. Less smooth surfaces becomes visible as onion rings.
Lens systems that fully correct dispersion of all 3 base colours to converge on the same focal plane and thereby correcting CA very well are called Apochromatic (APO).
These are (more) expensive to make and are most often found in 'fast' portrait lenses where subject separation with a shallow depth of field is desired... and thus a lot of the background will be out of focus. Having the out of focus areas come out smooth (or creamy so you like) is more pleasing and less distracting to look at.

Interesting study on how to simulate these lens effects in computer rendered graphics.
 
OP
O

Offler

Senior Member
Joined
May 27, 2021
Messages
367
Likes
320
It is obvious that a larger sensor receives more light. If we assume that the non sensing electronics per pixel (you mentioned) shouldn't need to get larger as the sensor gets larger, then larger sensors obviously have significantly more area available for light sensing than small sensors.

Of course, they don't have to use it - but then why use a larger sensor?

Plus - we can clearly see that most cameras with large sensors have both better noise levels, and better resolution (even at the same pixel count) than smaller sensors. Lenses also cannot have infinite resolution.

So, no - that is not snake oil)

Lens characteristics also affect smoothness of bokeh. You can see this by looking at the edges of out of focus highlights. Lenses where those are soft will result in smoother bokeh. Hard edges result in less smooth bokeh

Both of these phenomena are measurable.

I'm sure there is snake oil in photography, but the two items you mention are not that - and I can't think of any definite ones off the top of my head. We definitely don't suffer from the likes of magic crystals to put on top of the camera, or $10,000 cables to connect camera to computer.
I just ran through sensor lineup of Sony:

The important thing is pixel density characterized pixel size in microns.

Now I took 20MP M43 sensor IMX272/269 which is probably in my camera (Olympus EM-5 MKIII), and compared it with IMX410 which is fullframe sensor. Notice the M43 has pixel size 3.3microns while FF sensor has 5.94microns.

So DPReview has database of sample photos in their reviews, where you can select particular cameras:

I compared OM-1 which has Sony IMX472 with EM-5 MKIII which has sony IMX272 or 269 (just different interface, otherwise same characteristics), and compared it to 24,2Mpix cameras by SONY - Alpha 7 III, which should have their IMX410 sensor.

Same manufacturer, same lineup... Alpha 7 III is clearly better as both sensors used in OM/Olympus cameras. But pick different camera by Sony with higher pixel count (I took Sony Alpha 7R III which has 42.2MP sensor). You would notice that the noise is getting worse with higher pixel count, regardless total sensor size - because of smaller pixels. Also for some reason many other 24.2MP Sony cameras are really not on par with A7III in terms of noise.

Unfortunately things are even bit more complicated. I have been noticed by an astophotographer, that different sensors "gear up" on analog signal amplifier at different ISO settings. For IMX272/269 its on ISO 2000, but for IMX472 its on ISO1000. Due this my EM-5MKIII has lower noise on ISO2000 than on ISO1600...

For me... going from IMX272 to IMX472 isnt an upgrade. The noise is practically the same. The sensor use 4 photosensitive pixels instead of 1, but the net gain in terms of quality seems to be 0. It is true though that newer sensor allows higher native ISO.

Also amount of light... Solar panels can be enhanced with lenses to produce same amount of electricity but utilizing much less photosensitive surface. This technology is already used in microprisms or microlenses which focus light on photosensitive part of pixel. But given F2,8 at 1/125s, and pointing at white surface, amount of light exiting the lens should constant and has nothing to do with sensor itself. Yes, different lenses are made for different sensor sizes, but the F value nor exposure time are not characteristics of sensor. Only ISO is.


Also regarding bokeh... My point was bit different.

I can pay lets say 1400 dollars for M.Zuiko 45mm PRO lens which is weather sealed, and fully automatic. Or I can pay same amount of money for manual lens, without any weather sealing. Bokeh is used for marketing, so people forget what is the rest of the product they pay for.
 
Last edited:

r042wal

Member
Joined
Aug 20, 2022
Messages
15
Likes
10
I watched a good interview with Andreas Koch, inventor of the SACD. Andreas has been working with DSD through Sony and Philips for decades.

In his interview, he was comparing camera sensors with DSD sampling telling us more (or bigger) is not better. He said more pixels require more processing and with the additional processing, noise is introduced which then has to be filtered out.

He said DSD is the very same saying the sweet spot in DSD is double DSD or DSD128, not DSD256 which most 'golden ear' audiophiles claim is superior . He said the faster the scan rate, like in quad DSD or DSD256, the closer to the noise floor you come and the poorer the signal to noise ratio. He said to correct this requires processing and filtering which detract from the source.

Where I think snake oil factors into this is where recording studios are taking existing albums and upsampling them to DSD256 so they can sell them for more profit. New recording studios are popping up that record in DSD256 and sell that for a premium while offering DSD128 for less. They lead us to believe that because of the higher sample rate, DSD256 has to be better than DSD128.

I don't know how relevant this was to the conversation but I thought I would throw it out there. Anyone interested in the interview can find it here:

https://w ww.youtube.com/watch?v=eH-WW5s3NBM&list=PLhLUXQPaF2U7vRKL_H8KW1Tm8WNQ0qUl4&index=2
 

tonycollinet

Major Contributor
Joined
Sep 4, 2021
Messages
3,092
Likes
4,650
Location
UK/Cheshire
I watched a good interview with Andreas Koch, inventor of the SACD. Andreas has been working with DSD through Sony and Philips for decades.

In his interview, he was comparing camera sensors with DSD sampling telling us more (or bigger) is not better. He said more pixels require more processing and with the additional processing, noise is introduced which then has to be filtered out.

He said DSD is the very same saying the sweet spot in DSD is double DSD or DSD128, not DSD256 which most 'golden ear' audiophiles claim is superior . He said the faster the scan rate, like in quad DSD or DSD256, the closer to the noise floor you come and the poorer the signal to noise ratio. He said to correct this requires processing and filtering which detract from the source.

Where I think snake oil factors into this is where recording studios are taking existing albums and upsampling them to DSD256 so they can sell them for more profit. New recording studios are popping up that record in DSD256 and sell that for a premium while offering DSD128 for less. They lead us to believe that because of the higher sample rate, DSD256 has to be better than DSD128.

I don't know how relevant this was to the conversation but I thought I would throw it out there. Anyone interested in the interview can find it here:

https://w ww.youtube.com/watch?v=eH-WW5s3NBM&list=PLhLUXQPaF2U7vRKL_H8KW1Tm8WNQ0qUl4&index=2
His comparison with camera sensors is bunkum. More pixels require more processing because there is more data. But there is not more processing per pixel. So noise is not increased in that way. Furthermore, the higher pixel count means each noise "grain" (if you like) is smaller, and so the noise becomes less visible, even when it is the same amount.
 

Keith_W

Active Member
Joined
Jun 26, 2016
Messages
184
Likes
88
His comparison with camera sensors is bunkum. More pixels require more processing because there is more data. But there is not more processing per pixel. So noise is not increased in that way. Furthermore, the higher pixel count means each noise "grain" (if you like) is smaller, and so the noise becomes less visible, even when it is the same amount.

Yes. Furthermore the image can be downsampled which is very effective at getting rid of noise.
 

tonycollinet

Major Contributor
Joined
Sep 4, 2021
Messages
3,092
Likes
4,650
Location
UK/Cheshire
Leica.

Not the camera or the lenses per se but the mystique built around the brand.
You can say the same about any well known or respected brand.

Traditionally Leica have made exceptionally high quality cameras - arguably the best ever, so in that respect, not snake oil.

Sticking the brand on Panasonic hardware though - not so much. For me the brand has been devalued in the digital age.
 

Propheticus

Senior Member
Joined
Nov 18, 2020
Messages
391
Likes
587
Location
Vleuten, Netherlands
Yes. Furthermore the image can be downsampled which is very effective at getting rid of noise.
Indeed, using high pixel counts combined with pixel binning in 2x2 or 3x3 grids can reduce noise because you average values. Individual (sub)pixels that were picking up noise no longer stand out.
The probability that all 4 pixel readouts in the grid were noise instead of signal is of course smaller.
Many recent (very) high pixel count smartphone cameras use quad bayer layout sensors that have 2x2 grids of RGB subpixels. In low light the 4 subpixels per colour are combined for better sensitivity/lower noise. In good light you can apply array conversion to get a high resolution photo instead.

Alternatively you could cool your sensor to around -100°C to reduce the noise, but that's not really practical in a hand held device :)
 
Last edited:

fcracer

Active Member
Forum Donor
Joined
Nov 10, 2020
Messages
114
Likes
194
Leica.

Not the camera or the lenses per se but the mystique built around the brand.
Leica has a few characteristics that make their M range unique that live up to the hype:
1. Super compact full frame camera
2. Very high quality materials and build quality (albeit their software can be a bit funky)
3. Super compact wide aperture lenses that have high resolving capability
4. Rangefinder focusing system
5. Excellent resale value (my M10 setup cost me $50/month to own)
 
OP
O

Offler

Senior Member
Joined
May 27, 2021
Messages
367
Likes
320
Yes. Furthermore the image can be downsampled which is very effective at getting rid of noise.
It is, I used it from time to time, but it goes right against the idea of high pixel-count sensor.

Smaller pixels mean higher noise albeit at finer grain (relative to whole image).

From what I have seen on Sony cameras - A7 III - 24Mpix , A7R III - 42Mp and A7S III - 12Mp, there is definitely a sweet spot on 24Mp.
 
Last edited:
OP
O

Offler

Senior Member
Joined
May 27, 2021
Messages
367
Likes
320
Indeed, using high pixel counts combined with pixel binning in 2x2 or 3x3 grids can reduce noise because you average values. Individual (sub)pixels that were picking up noise no longer stand out.
The probability that all 4 pixel readouts in the grid were noise instead of signal is of course smaller.
Many recent (very) high pixel count smartphone cameras use quad bayer layout sensors that have 2x2 grids of RGB subpixels. In low light the 4 subpixels per colour are combined for better sensitivity/lower noise. In good light you can apply array conversion to get a high resolution photo instead.

Alternatively you could cool your sensor to around -100°C to reduce the noise, but that's not really practical in a hand held device :)
Its called Stacked CMOS sensor.

IMX472 and IMX272 are both 20 megapixel Micro 4/3 sensors.

First one is stacked CMOS, so it in fact has 80 megapixels, which are read in groups of 4 and what is presented is the average. On top of its its back-illuminated sensor.
IMX 272 has "native" 20 megapixels and its front illuminated.

I was about to buy newer camera with this sensor, however I must say I am quite disappointed at its noise levels.

Some of cooling is achieved with Infrared-cut filters which should prevent the "hot" part of the light to heat up the sensor.
 
OP
O

Offler

Senior Member
Joined
May 27, 2021
Messages
367
Likes
320
Leica.

Not the camera or the lenses per se but the mystique built around the brand.
They designed rangefinders and that became their staple.

More recently they designed Micro4/3 system but they never made any products of their own. There are few M43 lenses branded as Lumix Leica DG, and from my experience they are exceptional.

I would not say that the brand is special, when it comes to resulting photography, but their cameras are really well built.
 
Last edited:

Keith_W

Active Member
Joined
Jun 26, 2016
Messages
184
Likes
88
Leica has a few characteristics that make their M range unique that live up to the hype:
1. Super compact full frame camera
2. Very high quality materials and build quality (albeit their software can be a bit funky)
3. Super compact wide aperture lenses that have high resolving capability
4. Rangefinder focusing system
5. Excellent resale value (my M10 setup cost me $50/month to own)

I am a Leica M10 owner too. Some comments:

1. Super compact FF. Yes it was when I bought my M9. It was the smallest FF camera on the market. However the Sony A7IV is about the same size if you discount the hand grip, offers IBIS, and is objectively a superior camera in every way except for build quality and horrible Sony ergonomics. If you want even more compact there's the Sigma FP.
2. Software funky? How so? I find the menu on my M10 to be better designed than ANY camera on the market.
3. No argument from me there. The problem with all other FF systems is that their bodies might be small, but the lenses are obscenely huge. This goes for Leica's own SL range.
4. I don't like rangefinders. I put up with it. I use the EVF close to 100% of the time. I wish Leica would make an M without the rangefinder and with a built-in EVF.
5. No argument there. I have not lost money on any of the lenses i've owned. I've even made money.
 

Propheticus

Senior Member
Joined
Nov 18, 2020
Messages
391
Likes
587
Location
Vleuten, Netherlands
Its called Stacked CMOS sensor
No, it's called a quad bayer colour filter array. Independently of this you can stack the die layers of a CMOS chip in different orders, regular, back-side illuminated (BSI) or stacked which is taking the BSI concept one step further.

p.s. the largest camera in the world in the article I linked to uses 'old skool' CCD instead of CMOS to be able to read out the whole sensor at once (global shutter) to prevent rolling shutter issues during the long exposure shots it makes.
 
Top Bottom