guidomo Posted October 15, 2009 Share #101 Posted October 15, 2009 Advertisement (gone after registration) The more one trends towards being an "early adopter", the more it costs. I think you have this wrong. Those that really "lost" money are people who bought the M8.2 shortly before the M9 introduction. It cost more than the original M8 and the second hand prices are rock bottom now. I have had almost >2.5 years of use with my M8 and did not spend money on upgrading. So my balance sheet is 4200(cost)-1500(resale)=2700 Euro in 32 months, i.e. less than 100 Euros/month. Not too bad for a serious hobby. Link to post Share on other sites More sharing options...
Advertisement Posted October 15, 2009 Posted October 15, 2009 Hi guidomo, Take a look here Erwin Puts: The Leica M9: part 5: M8/9 noise and dynamic range. I'm sure you'll find what you were looking for!
rosuna Posted October 15, 2009 Share #102 Posted October 15, 2009 It doesn’t make any difference. There are no really “bad” sensors anyway. For example, the Panasonic Live-MOS sensors in all the (Micro) FourThirds cameras from Olympus and Panasonic (and Leica) were much maligned for their noise, just like the Kodak CCDs used before; still if you compare raw images of, say, an Olympus E-30 against those from a reportedly low-noise Nikon D3X, the difference amounts to just one f-stop. With regard to noise, ISO 400 with the E-30 roughly corresponds to ISO 800 with the D3X. That is just what one would expect from pixels roughly half the size, everything else being equal. If the pixel size was the same, so would be the signal-to-noise ratio. I’ve done lab tests of about 90 percent of all the DSLRs introduced since 2002. Some vendors have switched to CMOS sensors for some of their models during those years, but I’ve never come across a case where an improvement of image quality in general or signal-to-noise ratio in particular was attributable unequivocally to the sensor technology being CMOS or CCD. What I have found is that there is one major factor determining the signal-to-noise ratio in raw files, namely pixel size, and one major factor determining the signal-to-noise ratio in JPEG files, namely the noise reduction algorithm employed. Pixel size determines the signal but noise have many sources and CMOS handle noise much better when you increase ISO. You cannot replicate the RAW from a top class CMOS design by manipulating a RAW from a CCD. You are defending something that is counterfactual. There is a relation between pixel size and SNR, that is true, and obvious. All CMOS sensors are the same, that is not (compare the sensors of the Sony A900 and Nikon D3X). CCD (passive devices) are the same regarding image quality than CMOS, no matter the ISO level? That is not true. It isn't software "cleaning", it is noise control in the sensor itself. Signal may be lower due to circuitry, but last generation CMOS sensors handle noise in an incredible manner. Reading some comments I get the feeling of an insistent negation of the evidence. Link to post Share on other sites More sharing options...
rosuna Posted October 15, 2009 Share #103 Posted October 15, 2009 Stunning image quality is never obsolete. Someone who wants to shoot at ISO 100000 should get the Nikon D3s. Be realist. We are losing contact with reality here. I don't know how good is the IQ of the M9, but regarding the M8, it is mediocre at ISO 640 and a disaster over it. Even the X1 will have much better image quality at high ISOs than the M8 does (and probably, the M9). The "fill factor" of the X1 sensor will me much smaller, and the SNR lower, but... The Nikon D3s has impressive performance (specifications at least are impressive), but also is the 5D Mark II, Nikon D3X, Nikon D700, 1Ds Mark III... and the soon to be announced 1D Mark IV (with a high density sensor)... the evolution of the past 3 years (from the 5D and 1Ds Mark II days) has been substantial... If the only improvement of the M9 regarding noise is explained by the larger image size... well... Link to post Share on other sites More sharing options...
carstenw Posted October 15, 2009 Share #104 Posted October 15, 2009 I am being realistic. I find myself going beyond ISO 320 perhaps 5% of the time, and above ISO 640 1% of the time. Moving to a Nikon or Canon for so few shots is not worth the loss of sharpness, colour and look for me. The AA filters on those sensors yield results which are very unattractive to me, and the built-in noise reduction on those CMOS sensors, while yielding low noise, also smear the results in a way that I find unacceptable. I will take the M8 combined with NoiseWare over a high-ISO Canon or Nikon any day. Link to post Share on other sites More sharing options...
mjh Posted October 15, 2009 Share #105 Posted October 15, 2009 Pixel size determines the signal but noise have many sources and CMOS handle noise much better when you increase ISO. You cannot replicate the RAW from a top class CMOS design by manipulating a RAW from a CCD. You are defending something that is counterfactual. Looks like all my efforts understanding this stuff theoretically and all the time spent analyzing its practical effects was wasted then … Link to post Share on other sites More sharing options...
Tony Wright Posted October 15, 2009 Share #106 Posted October 15, 2009 Looks like all my efforts understanding this stuff theoretically and all the time spent analyzing its practical effects was wasted then … On the contrary - I found your posting extremely valuable - thank you. I made a fairly curmudgeonly post in the first few hours of this thread because I was distressed by Erwin's results but I wonder if you can speculate on the reasons why the M9 sensor (substantially the same as the M8 but for size) was shown as performing significantly worse than the M8. After several years of potential development it seems counterintuitive that Leica would install a sensor that gives an inferior performance. This led me to question the validity of the exercise. cheers Tony Link to post Share on other sites More sharing options...
rosuna Posted October 15, 2009 Share #107 Posted October 15, 2009 Advertisement (gone after registration) Looks like all my efforts understanding this stuff theoretically and all the time spent analyzing its practical effects was wasted then … If your conclusion is that the image quality of the M9 is the same or better than the image quality of the 5D Mark II or the Nikon D3x (smaller pixel pitch, lower fill ratio), from RAW files, then a revision is necessary. Link to post Share on other sites More sharing options...
thrid Posted October 15, 2009 Share #108 Posted October 15, 2009 Originally Posted by roey I might be missing something, but doesn't that mean that he compares the quality of the in-camera JPG conversion rather than the dynamic range and noise characteristics of the sensors? That appears to be a succinct description of what he did, yes. In that case isn't the test pointless and doesn't say anything about the true performance of the sensor? RAW is what really counts. Who shoots jpeg, except to plink around a little? Link to post Share on other sites More sharing options...
luigi bertolotti Posted October 15, 2009 Share #109 Posted October 15, 2009 If your conclusion is that the image quality of the M9 is the same or better than the image quality of the 5D Mark II or the Nikon D3x (smaller pixel pitch, lower fill ratio), from RAW files, then a revision is necessary. Making rigorous tests (as rigorous as one can do) on measurable values like s/n ratio is one thing, affirming that "Image Quality" is better from a certain camera is another thing: IQ is not a measurable quantity, and is highly personal : the previous post from Carsten is a typical example : one can like the IQ of a pic taken with an excellent high ISO camera, another can prefer a someway "dirtier" high ISO look, for instance for it has a feeling more resembling the classic high ISO films... Link to post Share on other sites More sharing options...
delander † Posted October 15, 2009 Share #110 Posted October 15, 2009 If your conclusion is that the image quality of the M9 is the same or better than the image quality of the 5D Mark II or the Nikon D3x (smaller pixel pitch, lower fill ratio), from RAW files, then a revision is necessary. What defines image quality is a big question and it is not confined to ISO and DR performance. I have a 1ds3, an M8 and now an M9. To my mind the 1ds3 produces relatively soft images and I'm not that impressed by its high ISO or colour performance. But that is just my view from using the cameras. Jeff Link to post Share on other sites More sharing options...
t024484 Posted October 15, 2009 Share #111 Posted October 15, 2009 Hans is read noise the same as shot noise? I'm confused, how many types of noise are there? Jeff Read noise is the analog noise that is induced because of imperfections in the way the charge of a pixel has to be translated in a voltage that is offered to the A/D converter to be digitized. The content of a pixel is transferred to a capacitor that is charged with the electrons collected by the exposed pixel. This is the charge to voltage conversion. An amplifier translates this voltage depending on the ISO setting to a low impedance voltage that has to be digitized by the A/D converter. When this has been completed, the capacitor is discharged to a preset level, determined by the black level, and charged again with the content of the next pixel. The bandwidth of the whole channel has to be extremely high, because conversion in the M8 takes place at a rate of 24 MHz or every 42 nsec. This is done in parallel for two channels each processing 5 Mio pixels, being the left and the right half of the picture. Since the M9 also needs 0.5 seconds per picture just like the M8, either the processing is ramped up to 43Mhz, which is rather unlikely, or the M9 has a sensor with 4 outputs, each processing 4,5 Mio pixels. My guess is that the M9 has a lower clock of 22 MHz, because of the same 0.5 sec conversion time as the M8. Now comes the conflicting situation for the developer. The read amplifier has to be fast with a settling time below 42 nsec , but also have a very low noise, a low input capacitance and a high input impedance. All in one go is very difficult to solve. So there should be looked for the best possible compromise. What I found as an example is the AD8001. This amplifier has a noise of 2nvSqrt(Hz), meaning that for its bandwidth of 700Mhz, 529 microvolt of noise is generated. I haven’t found fast amplifiers with less noise. The specifications of the sensor are 20 electrons read noise, with a conversion factor of 25 microvolt per electron, meaning 500 microvolt read noise. This is very much in line with the AD 8001 amplifier. What is actually measured for the M9 is a read noise of 640 microvolt (or 6.4 ADU). So a quick check shows that probably most of the read noise is generated by the read amplifier. So how can this read noise be reduced? The answer is in using more amplifiers in parallel wit, enabling slower amplifiers with less noise. If the bandwidth could be restricted with the same AD 8001 amplifier to 50 MHz, the noise already drops to 140 microvolt or 1/3 of the initial value. That would be the level of the D3X. Now after this long explanation on Read Noise, I can be quite brief on Shot Noise. Shot noise has to do with light, and cannot be influenced. Shot noise is directly related with the sqrt to the number of electrons captured. The larger the pixel, the more electrons are captured resulting in relatively less noise. Hans Link to post Share on other sites More sharing options...
rosuna Posted October 15, 2009 Share #112 Posted October 15, 2009 I am being realistic. I find myself going beyond ISO 320 perhaps 5% of the time, and above ISO 640 1% of the time. Moving to a Nikon or Canon for so few shots is not worth the loss of sharpness, colour and look for me. The AA filters on those sensors yield results which are very unattractive to me, and the built-in noise reduction on those CMOS sensors, while yielding low noise, also smear the results in a way that I find unacceptable. I will take the M8 combined with NoiseWare over a high-ISO Canon or Nikon any day. MF cameras are studio cameras, and low ISO quality may be a priority, but the M8/M9 is a reportage camera. I also go beyond ISO 320 only when is strictly necessary, but it is due to the poor quality of the files I get. It is a "pain in the ass". I would like to have much better performance until ISO 1600 or so (I really don't need that stunning ISO 102,400 of the Nikon D3s). I suppose the high ISO files from Canon or Nikon are smeared, compared to the basic ISO but I find the quality impressive, and much better than I get from a M8 file after processing... any possible processing I would say. AA filters aren't an intrinsic characteristic of CMOS sensors, and color filters may have different properties. You say "I will take the M8 combined with NoiseWare over a high-ISO Canon or Nikon any day". I would prefer the M8 for ISO 160 o 320 pictures, but I will take the Canon or Nikon for anything else. They are miles ahead. It is not a marginal difference. Link to post Share on other sites More sharing options...
wattsy Posted October 15, 2009 Share #113 Posted October 15, 2009 I remember when the M8 came out, the BJP did an article that analysed the camera's dynamic range (amongst other things) and surprisingly concluded that the M8 JPEGs showed greater dynamic range than C1 converted DNGs (or something like that). I'm not quite sure what my point is other than I wouldn't pay too much attention to analysis of M8 (and M9) JPEG output. Link to post Share on other sites More sharing options...
rosuna Posted October 15, 2009 Share #114 Posted October 15, 2009 So how can this read noise be reduced? The answer is in using more amplifiers in parallel wit, enabling slower amplifiers with less noise. If the bandwidth could be restricted with the same AD 8001 amplifier to 50 MHz, the noise already drops to 140 microvolt or 1/3 of the initial value. That would be the level of the D3X. Now after this long explanation on Read Noise, I can be quite brief on Shot Noise. Shot noise has to do with light, and cannot be influenced. Shot noise is directly related with the sqrt to the number of electrons captured. The larger the pixel, the more electrons are captured resulting in relatively less noise. Hans So, the D3X sensor has more shot noise than the M9 sensor (smaller pixels, lower fill ratio) but less read noise due to many parallel amplifiers, correct? Is it possible to multiply the number of amplifiers in a CCD sensor like Nikon does for a CMOS sensor? Is it as simple as that? Link to post Share on other sites More sharing options...
tgray Posted October 15, 2009 Share #115 Posted October 15, 2009 I'm sure this was a typo, but shot noise is from photons, not electrons. Link to post Share on other sites More sharing options...
carstenw Posted October 15, 2009 Share #116 Posted October 15, 2009 If your conclusion is that the image quality of the M9 is the same or better than the image quality of the 5D Mark II or the Nikon D3x (smaller pixel pitch, lower fill ratio), from RAW files, then a revision is necessary. At low ISO this is transparently the case. I don't know anyone who prefers the results of Canons or Nikons to Leicas at low ISO. At high ISO, it is a question of taste, not superiority. Link to post Share on other sites More sharing options...
Angelos Viskadourakis Posted October 15, 2009 Share #117 Posted October 15, 2009 2 years ago while i was testing cine digital cameras-i'm a cinematographer- in MOTIONFX.GR lab using IMATEST i tested my LEICA M8 against my CANON 5D and i was wrong about 5d having higher dynamic range to M8 -two years have passed. this is kindly shared with E.Achilopoulos who has the copyright of his publication.See atachments below:p Welcome, dear visitor! As registered member you'd see an image here… Simply register for free here – We are always happy to welcome new members! Link to post Share on other sites Simply register for free here – We are always happy to welcome new members! ' data-webShareUrl='https://www.l-camera-forum.com/topic/100096-erwin-puts-the-leica-m9-part-5-m89-noise-and-dynamic-range/?do=findComment&comment=1076069'>More sharing options...
t024484 Posted October 15, 2009 Share #118 Posted October 15, 2009 So, the D3X sensor has more shot noise than the M9 sensor (smaller pixels, lower fill ratio) but less read noise due to many parallel amplifiers, correct?Yes, that is correct. 5.94 versus 6.8 microns gives 14% more shot noise with a possible correction for different fill factors. And yes because of the great amount of parallel amplifiers, the D3X has far less read noise. When ISO is doubled, read noise is also doubled, but shot noise goes up with the SQRT. so at higher ISO, shot noise is becoming more and more important. Is it possible to multiply the number of amplifiers in a CCD sensor like Nikon does for a CMOS sensor? Is it as simple as that? The principles of reading a CCD versus CMOS are completely different, but a CCD can be divided in more channels, or the output per channel can be divided over more capacitors+amplifiers giving the same reduction in processing speed. Hans Link to post Share on other sites More sharing options...
Angelos Viskadourakis Posted October 15, 2009 Share #119 Posted October 15, 2009 as we all can see in my test of my own M8 & 5D the M8 ranks better compared to 5D and i remind myself that 5D is a full frame 24x36 sensor quite advance at that time like M9 but it rank lower than M8.Please someone send that to ERWIN PUT in order to know that other people around the globe realize and measure the same thing.The top plate in my previous posting includes the readings from RED camera -one from E.Achilopoulos in ATHENS GREECE and one other from A.PARRA in SPAIN using IMATEST and their results are pretty close so IMATEST results are not godspell but is a technicall analysis of significant values.I remind that my M8 & CANON 5D are tiffs from raw that i shot one after the other and i convert in C1 on the spot. all below shared kindly with mr.E.Achilopoulos MOTIONFX.GR and are copyrighted by him and published by me after his generous permission. Welcome, dear visitor! As registered member you'd see an image here… Simply register for free here – We are always happy to welcome new members! Link to post Share on other sites Simply register for free here – We are always happy to welcome new members! ' data-webShareUrl='https://www.l-camera-forum.com/topic/100096-erwin-puts-the-leica-m9-part-5-m89-noise-and-dynamic-range/?do=findComment&comment=1076087'>More sharing options...
t024484 Posted October 15, 2009 Share #120 Posted October 15, 2009 2 years ago while i was testing cine digital cameras-i'm a cinematographer- in MOTIONFX.GR lab using IMATEST i tested my LEICA M8 against my CANON 5Dand i was wrong about 5d having higher dynamic range to M8 -two years have passed. this is kindly shared with E.Achilopoulos who has the copyright of his publication.See atachments below:p Your figures are completely in line with my figures. 7.3/7.73, 9.0/9.17, 10.1/10.1 all at ISO 160, but quite different from Erwins results. The point is that we are looking at is at higher ISO values, because ISO 160 is accepted by all as a no problem zone. Hans Link to post Share on other sites More sharing options...
Recommended Posts
Archived
This topic is now archived and is closed to further replies.