Medium Format Forum
Register a free account now!
If you are registered, you get access to the members only section, can participate in the buy & sell second hand forum and last but not least you can reserve your preferred username before someone else takes it.
H3D
- Thread starter brucec
- Start date
>Erin was somewhat negative on the M8 in his first reiteration of >his review. I believe some of the text has and opinions brought a >fair bit of controversy when first released. However, this type of >comparison is non-sensible to me. I do not see how he comes up with >a 1.2x difference. Where did that number come from? Moreover, >blowing up an image to 200% in PS is not the way to do the >comparison. One needs to interpolate and compare prints. Moreover, >it was not clear how the RAW images were processed. The work flow is >critical as well. Both cameras serve different purposes and this >type of comparison is meaningless to me.
>[Erin was somewhat negative on the M8 in his first reiteration of >his review. I believe some of the text has and opinions brought a >fair bit of controversy when first released. However, this type of >comparison is non-sensible to me. I do not see how he comes up with >a 1.2x difference. Where did that number come from? Moreover, >blowing up an image to 200% in PS is not the way to do the >comparison. One needs to interpolate and compare prints. Moreover, >it was not clear how the RAW images were processed. The work flow is >critical as well. Both cameras serve different purposes and this >type of comparison is meaningless to me.]
polypal
Active Member
Mark,
Thank you for your contribution.
In fact it comes down to the same thing.
It is useless to compare two different formats.
In this case not only the format but the complete post production
need to be compared and for what?
The Leica M8 can count on a keen group of users.
This is a great camera in its own right.
MF cameras like Hasselblad serve another purpose and are meant for those who need the extra quality.
Thank you for your contribution.
In fact it comes down to the same thing.
It is useless to compare two different formats.
In this case not only the format but the complete post production
need to be compared and for what?
The Leica M8 can count on a keen group of users.
This is a great camera in its own right.
MF cameras like Hasselblad serve another purpose and are meant for those who need the extra quality.
blowupster
Member
Meaning less. M8 is a very good deal and I probably will buy one this week. For quality, even on the test, it shows that 39 Mpix is far better.
Hi Paul,
> It is useless to compare two different formats.
I'm not sure I agree. There is a lot of use out of comparing different results, no matter what the format.
If what you say were true, you really shouldn't be comparing much anything...how could you compare any two digital cameras if they didn't have the exact same size sensor, with the exact same number of sensor cells?
I think you compare based on a set of input criteria (size of camera, dislikes/likes of functions etc.) and the resultant output, without regard to format...especially when comparing digital. The only "format" thing that figures in, is related to how large do you want your output to be and depending on what your output medium is, that determines what the minimum acceptable "pixel" count would be.
Regards,
Austin
> It is useless to compare two different formats.
I'm not sure I agree. There is a lot of use out of comparing different results, no matter what the format.
If what you say were true, you really shouldn't be comparing much anything...how could you compare any two digital cameras if they didn't have the exact same size sensor, with the exact same number of sensor cells?
I think you compare based on a set of input criteria (size of camera, dislikes/likes of functions etc.) and the resultant output, without regard to format...especially when comparing digital. The only "format" thing that figures in, is related to how large do you want your output to be and depending on what your output medium is, that determines what the minimum acceptable "pixel" count would be.
Regards,
Austin
Hi Marc,
> but for ultimate > image quality ... 39 meg, true 16 bit 1.1X 645 sized Kodak sensor...
I'm not sure I believe you get 16 bits of actual data. It may return the data using 16 bit words, but that does not mean there is 16 bits of usable data.
What makes you believe you really get 16 bits of usable data? If that were the case, the exposure would have to be spot on, with absolutely not one bit (literally) of error. Also, the raw data would have to be more than 16 bits, to allow for tonal curve adjustments.
Regards,
Austin
> but for ultimate > image quality ... 39 meg, true 16 bit 1.1X 645 sized Kodak sensor...
I'm not sure I believe you get 16 bits of actual data. It may return the data using 16 bit words, but that does not mean there is 16 bits of usable data.
What makes you believe you really get 16 bits of usable data? If that were the case, the exposure would have to be spot on, with absolutely not one bit (literally) of error. Also, the raw data would have to be more than 16 bits, to allow for tonal curve adjustments.
Regards,
Austin
>[BTW i wanted to say I am new to this forum and wanted to first say >hi. In regards to the comparison, If you take the opinion that the >comparison in principle is ok, I would state the way he did it is >not. I made a previous comment on this point. However, as much as I >hate to use this comparison, it is like comparing a Ferrari to lets >say a Honda Accord (or Accura Sedan), they both can get you from >point A to B, but if you try to achieve maximal performance >(e.g.Speed), the Ferrari is going to win. However, if you are only >using the Ferrari to drive a few miles to the market or store each >day, a Ferrari is not the most efficient or cost effective way to >go. I would also make the point that I personally do not see how >the M8 is a good deal. It is very expensive and for the tradeoff in >image quality for cost compared to other options in the 35mm >spectrum, I personally do not consider it a good value. Albeit many >do...]
The value is in the lenses Mark.
Austin, yes, the term 16 bit data is used a lot without really understanding what it means, but as far as I know, and according to hype from the MF back makers, these are 16 bit processors. The older Kodak ProBack was 12 bit and the Canon 1DsMKII is 12 bit. Canon is making a lot of the fact that the new 1DMKIII is now 14 bit. The Leica DMR was 16 bit, which was engineered by Imacon using a Kodak sensor. Those files show as good or better dynamic range and detail as my Canon 1DsMKII.
Personally, I could care less what bit they are IF the photos are good... and in the case of the H3D/39 ... they are as good as I have been able to get out of any digital camera regardless of bit depth.
As far as your math is concerned I'll leave that to the math minded ... but if those theories are true for 16 bit, then it would be true for 12 bit, 14 bit, or 8 bit wouldn't they?
Austin, yes, the term 16 bit data is used a lot without really understanding what it means, but as far as I know, and according to hype from the MF back makers, these are 16 bit processors. The older Kodak ProBack was 12 bit and the Canon 1DsMKII is 12 bit. Canon is making a lot of the fact that the new 1DMKIII is now 14 bit. The Leica DMR was 16 bit, which was engineered by Imacon using a Kodak sensor. Those files show as good or better dynamic range and detail as my Canon 1DsMKII.
Personally, I could care less what bit they are IF the photos are good... and in the case of the H3D/39 ... they are as good as I have been able to get out of any digital camera regardless of bit depth.
As far as your math is concerned I'll leave that to the math minded ... but if those theories are true for 16 bit, then it would be true for 12 bit, 14 bit, or 8 bit wouldn't they?
jotloob
Well-Known Member
Austin
As far as I can remember : The first generation of digital backs had 8 bit per colour . The second generation had 12 bit per colour and the current backs have 16 bit colour depth RGB .
That's what the documentation says . It says it more clear in the CFV specs than for the CF22 and CF39 .
A byte holds 8 usable bits , but you never talk about the parity bit or any error correction bits .
That means , when the specs say 16 bit colour depth , it is 16 bit colour depth , because the parity and error correction bits , as well as control bits , which shurely exist , are of no interest for the user .
As far as I can remember : The first generation of digital backs had 8 bit per colour . The second generation had 12 bit per colour and the current backs have 16 bit colour depth RGB .
That's what the documentation says . It says it more clear in the CFV specs than for the CF22 and CF39 .
A byte holds 8 usable bits , but you never talk about the parity bit or any error correction bits .
That means , when the specs say 16 bit colour depth , it is 16 bit colour depth , because the parity and error correction bits , as well as control bits , which shurely exist , are of no interest for the user .
Hi Marc,
> Austin, yes, the term 16 bit data is used a lot without really > understanding what it means, but as far as I know, and according to > hype from the MF back makers, these are 16 bit processors.
Sure, but the process data word size has nothing to do with the actual usable number of bits from the A/D.
> ... but if those theories are true for 16 bit, then it would be true > for 12 bit, 14 bit, or 8 bit wouldn't they?
Not sure what you mean. What theories are you talking about?
Typically, the datapath in a scanner and digital back is 16 bits, no matter what the actual number of usable bits are from the sensor. Manufacturers/marketing has decided to ignore the actual capabilities of the unit, and instead just claim they are 16 bits, since they have 16 bit datapaths. There are now some standards for testing to determine the actual number of bits a digital imaging device can provide, but few manufacturers have adopted this...since knowing the truth puts them on the lower end of the specsmanship scale with regards to the generally unknowing public.
Regards,
Austin
> Austin, yes, the term 16 bit data is used a lot without really > understanding what it means, but as far as I know, and according to > hype from the MF back makers, these are 16 bit processors.
Sure, but the process data word size has nothing to do with the actual usable number of bits from the A/D.
> ... but if those theories are true for 16 bit, then it would be true > for 12 bit, 14 bit, or 8 bit wouldn't they?
Not sure what you mean. What theories are you talking about?
Typically, the datapath in a scanner and digital back is 16 bits, no matter what the actual number of usable bits are from the sensor. Manufacturers/marketing has decided to ignore the actual capabilities of the unit, and instead just claim they are 16 bits, since they have 16 bit datapaths. There are now some standards for testing to determine the actual number of bits a digital imaging device can provide, but few manufacturers have adopted this...since knowing the truth puts them on the lower end of the specsmanship scale with regards to the generally unknowing public.
Regards,
Austin
polypal
Active Member
Funny I was thinking of cars too when I wrote my comment about comparing M8 and MF quality.
Maybe further explanation helps here.
The new Leica digital, this is not the first digital camera from Leica, has as much chance against Hasselblad 39Mp as 35 mm Leicas against MF cameras with film.
So there is no point in comparing those. The only thing they have in common is they are both digital.
That is were it all ends.
There is a good market of well motivated users for the M8.
It would have been more usefull to compare the M8 with the digital products from Canon and Nikon.
These cameras are being used for more or less the same fields of photography.
I certainly do not agree that products should have exactly the same specs if they are compared.
I think products that are meant for the same group of users are suitable to be compared.
In the end I am not interested how big the medium is what I use to get what results I want.
The result ie print is what I am interested in.
Maybe further explanation helps here.
The new Leica digital, this is not the first digital camera from Leica, has as much chance against Hasselblad 39Mp as 35 mm Leicas against MF cameras with film.
So there is no point in comparing those. The only thing they have in common is they are both digital.
That is were it all ends.
There is a good market of well motivated users for the M8.
It would have been more usefull to compare the M8 with the digital products from Canon and Nikon.
These cameras are being used for more or less the same fields of photography.
I certainly do not agree that products should have exactly the same specs if they are compared.
I think products that are meant for the same group of users are suitable to be compared.
In the end I am not interested how big the medium is what I use to get what results I want.
The result ie print is what I am interested in.
Hi Jurgen,
> As far as I can remember : The first generation of digital backs had 8 > bit per colour . The second generation had 12 bit per colour and the > current backs have 16 bit colour depth RGB . That's what the > documentation says .
Again, because a manufacturer claims something, or because the device returns the data in 16 bit words does not mean that all of the 16 bits are actually good data. The only way to really tell is to have the sensor's specs for dynamic range (which is the best the unit can hope to achieve), OR, run tests on the entire system.
> A byte holds 8 usable bits , > but you never talk about the parity bit > or any error correction bits .
These, if available, have nothing to do with the width of the acutal usable data, so this doesn't even enter into the discussion.
> That means , when the specs say 16 > bit colour depth , it is 16 bit > colour depth , because the parity and error correction bits , as well > as control bits , which shurely exist , are of no interest for the > user .
No, this is entirely incorrect. Any other bits, such as parity/ECC/control (if at all available) are not included in the color depth/number of bits.
Regards,
Austin
> As far as I can remember : The first generation of digital backs had 8 > bit per colour . The second generation had 12 bit per colour and the > current backs have 16 bit colour depth RGB . That's what the > documentation says .
Again, because a manufacturer claims something, or because the device returns the data in 16 bit words does not mean that all of the 16 bits are actually good data. The only way to really tell is to have the sensor's specs for dynamic range (which is the best the unit can hope to achieve), OR, run tests on the entire system.
> A byte holds 8 usable bits , > but you never talk about the parity bit > or any error correction bits .
These, if available, have nothing to do with the width of the acutal usable data, so this doesn't even enter into the discussion.
> That means , when the specs say 16 > bit colour depth , it is 16 bit > colour depth , because the parity and error correction bits , as well > as control bits , which shurely exist , are of no interest for the > user .
No, this is entirely incorrect. Any other bits, such as parity/ECC/control (if at all available) are not included in the color depth/number of bits.
Regards,
Austin
jotloob
Well-Known Member
Austin
Might be that I did not express good enough what I mean , or I was misunderstood by you .
16 bits of colour depth (per colour) makes 2 bytes of true colour information .(per colour).Parity/ecc/control bits might fill up an other byte or two . I don't know . But I also know , that these control bits have got nothing to do with the true colour bits .
And therefore I understand 16 bits as 16 bits of colour information . Was my previous posting so much to misunderstand ? ? ?
Might be that I did not express good enough what I mean , or I was misunderstood by you .
16 bits of colour depth (per colour) makes 2 bytes of true colour information .(per colour).Parity/ecc/control bits might fill up an other byte or two . I don't know . But I also know , that these control bits have got nothing to do with the true colour bits .
And therefore I understand 16 bits as 16 bits of colour information . Was my previous posting so much to misunderstand ? ? ?
Hi Jurgen,
> 16 bits of colour depth (per colour) makes > 2 bytes of true colour > information . (per colour).
It allows "room" for two bytes (16 bits) of tonal information, that is true. But, it does not mean all 16 bits are actually good/usable bits. You could store 10 bit data in 16 bits, or 12 bit data in 16 bits. It's simply the next most convenient step up from 8 bits, but it doesn't mean all the bits are used.
If you have a 250G hard drive, it does not mean you have 250G of data stored on it.
> Parity/ecc/control bits might fill up an > other byte or two . I don't know . But I also know , that these > control bits have got nothing to do with the true colour bits . > And therefore I understand 16 bits as 16 bits of colour > information . Was my previous posting so much to misunderstand ? ? ?
Here is what you said:
"That means , when the specs say 16 bit colour depth , it is 16 bit colour depth , because the parity and error correction bits , as well as control bits , which shurely exist ,"
There are two confusing things, at least to me, about your statement. One is that it appeared you believed the parity/ecc etc. was part of the number of color bits, and second is why you would bring up parity/ecc etc. in the first place when it is simply not relevant of this discussion.
If I misunderstood you, and you understand that parity/ecc/control is NOT part of the image data/bit depth, then we should eliminate that from the discussion.
Regards,
Austin
> 16 bits of colour depth (per colour) makes > 2 bytes of true colour > information . (per colour).
It allows "room" for two bytes (16 bits) of tonal information, that is true. But, it does not mean all 16 bits are actually good/usable bits. You could store 10 bit data in 16 bits, or 12 bit data in 16 bits. It's simply the next most convenient step up from 8 bits, but it doesn't mean all the bits are used.
If you have a 250G hard drive, it does not mean you have 250G of data stored on it.
> Parity/ecc/control bits might fill up an > other byte or two . I don't know . But I also know , that these > control bits have got nothing to do with the true colour bits . > And therefore I understand 16 bits as 16 bits of colour > information . Was my previous posting so much to misunderstand ? ? ?
Here is what you said:
"That means , when the specs say 16 bit colour depth , it is 16 bit colour depth , because the parity and error correction bits , as well as control bits , which shurely exist ,"
There are two confusing things, at least to me, about your statement. One is that it appeared you believed the parity/ecc etc. was part of the number of color bits, and second is why you would bring up parity/ecc etc. in the first place when it is simply not relevant of this discussion.
If I misunderstood you, and you understand that parity/ecc/control is NOT part of the image data/bit depth, then we should eliminate that from the discussion.
Regards,
Austin
G'Day All:
re : the Brand/Format/8/12/16 Bit Debate in the latter posts above.
The hard-boiled gumshoe author, Mickey Spillane wrote almost all his books on an electric Smith-Corona typewriter. (Never would change over to a word processor.)
I'm looking for a Smith-Corona myself. I think it's time for me to write some best selling crime fighter books.
The Smith-Corona sure worked for Mickey.
Cheers, Colin
re : the Brand/Format/8/12/16 Bit Debate in the latter posts above.
The hard-boiled gumshoe author, Mickey Spillane wrote almost all his books on an electric Smith-Corona typewriter. (Never would change over to a word processor.)
I'm looking for a Smith-Corona myself. I think it's time for me to write some best selling crime fighter books.
The Smith-Corona sure worked for Mickey.
Cheers, Colin
jotloob
Well-Known Member
Austin
My english is very basic , so it could very well be , that I was misunderstood (by you) .
Just as a basic info for you : I have worked 30 years in hardware and software service buisness in the biggest computer company and therefore can very well decide between information bytes and control bytes (parity ecc etc. )
So there is definately a misunderstanding .
Conclusion : we eliminate this from the discussion as mentioned in your posting .
16 bit of true colour information . Yes , thats what the specs for CFV and CF backs say.
If it is really true , I can not tell . I am a consumer of the digital backs , and can believe the data in the specs or not .
But why the hell should HASSELBLAD or other brands cheat ? ? ?
Please can you give me one good reason why they should cheat ? ? ?
I think , we could always doubt , whats right or wrong , or we could just use the gear we have and enjoy it .
I don't see any reason why I should doubt , but its also not my buisness to check and test what's in the specs .
So , Austin , please don't make this forum a mental fighting arena again .
My english is very basic , so it could very well be , that I was misunderstood (by you) .
Just as a basic info for you : I have worked 30 years in hardware and software service buisness in the biggest computer company and therefore can very well decide between information bytes and control bytes (parity ecc etc. )
So there is definately a misunderstanding .
Conclusion : we eliminate this from the discussion as mentioned in your posting .
16 bit of true colour information . Yes , thats what the specs for CFV and CF backs say.
If it is really true , I can not tell . I am a consumer of the digital backs , and can believe the data in the specs or not .
But why the hell should HASSELBLAD or other brands cheat ? ? ?
Please can you give me one good reason why they should cheat ? ? ?
I think , we could always doubt , whats right or wrong , or we could just use the gear we have and enjoy it .
I don't see any reason why I should doubt , but its also not my buisness to check and test what's in the specs .
So , Austin , please don't make this forum a mental fighting arena again .
Hi Jurgen,
Your English is just fine.
Since you mention background, I have been designing digital imaging devices (not exclusively, but for a large part) since 1976, so I have a pretty decent understanding of the principles involved.
Someday you'll have to say what computer company do you consider the biggest. It depends on when one is talking about, and by what metric "biggest" is measured. I've worked as an engineer/scientist for IBM, DEC, Compaq, HP and Sun.
> But why > the hell should HASSELBLAD or other brands cheat ? ? ?
It is a fact in the scanner market, the manufacturers "cheat" as you call it (that's not what they call it ;-). It's not disputable. That is why tests have been written to help bring some honesty to this. Whether the manufacturers do so in the digi-back market, I don't know for a fact as I do in the scanner market...I just haven't taken the time to examine it closely enough to provide conclusive evidence one way or the other. But, given what I know, again, I doubt that any back actually provides 16 bits of valid data.
In digital imaging, you don't want to have 16 bits of valid data and then apply tonal curves to that data in a 16 bit realm. You will end up with gaps in the codes (combing is what it is typically called), you need to "process" data using a larger number of bits. You need "headroom" (additional bits) for this type of processing. If you understand the algorithm behind applying tonal curves (and/or setpoints) you will understand this. If you want a further explanation on this, I can provide it...but you can probably look it up on the web just as well...though there is a lot of misinformation and misunderstanding out there, unfortunately. And many self proclaimed experts as well.
> Please can you give me one good reason why they should cheat ? ? ?
The reasons are simple. One is marketing people write the documentation, typically, and they don't understand the difference between what the A/D can deliver for number of "good" bits, and how many bits of data are returned. Second is simply specsmanship...if another company "exaggerates" in this way, so whould we. And third, there was no standard, so anyone can say anything they want. And, no one holds them to task for this misinformation.
Regards,
Austin
Your English is just fine.
Since you mention background, I have been designing digital imaging devices (not exclusively, but for a large part) since 1976, so I have a pretty decent understanding of the principles involved.
Someday you'll have to say what computer company do you consider the biggest. It depends on when one is talking about, and by what metric "biggest" is measured. I've worked as an engineer/scientist for IBM, DEC, Compaq, HP and Sun.
> But why > the hell should HASSELBLAD or other brands cheat ? ? ?
It is a fact in the scanner market, the manufacturers "cheat" as you call it (that's not what they call it ;-). It's not disputable. That is why tests have been written to help bring some honesty to this. Whether the manufacturers do so in the digi-back market, I don't know for a fact as I do in the scanner market...I just haven't taken the time to examine it closely enough to provide conclusive evidence one way or the other. But, given what I know, again, I doubt that any back actually provides 16 bits of valid data.
In digital imaging, you don't want to have 16 bits of valid data and then apply tonal curves to that data in a 16 bit realm. You will end up with gaps in the codes (combing is what it is typically called), you need to "process" data using a larger number of bits. You need "headroom" (additional bits) for this type of processing. If you understand the algorithm behind applying tonal curves (and/or setpoints) you will understand this. If you want a further explanation on this, I can provide it...but you can probably look it up on the web just as well...though there is a lot of misinformation and misunderstanding out there, unfortunately. And many self proclaimed experts as well.
> Please can you give me one good reason why they should cheat ? ? ?
The reasons are simple. One is marketing people write the documentation, typically, and they don't understand the difference between what the A/D can deliver for number of "good" bits, and how many bits of data are returned. Second is simply specsmanship...if another company "exaggerates" in this way, so whould we. And third, there was no standard, so anyone can say anything they want. And, no one holds them to task for this misinformation.
Regards,
Austin
blowupster
Member
> M8 is a very good deal...
> Would you please explain what makes you draw that conclusion?
It's my conclusion in my case. Usualy digital version of the body is at least twice the prize of film body. The M8 is not so more expensive than the M7. The print usued from a M8 or M7 is about the same level. If you have no tripod posibilities and bad light and must use 800 asa, the M8 will shoe beter result than M7.
The M8 shows that the lens limites will not allow big step in 6 months. 10Mpix could be enought for most of (my) realisations.
For beter quality my Hasselblad V with normal film is there.
Of course a 40Mpix sensor is far better, but it's realy expensive and devaluate too fast. The CFV is perhaps not so far of what I could economicaly justify.
> Would you please explain what makes you draw that conclusion?
It's my conclusion in my case. Usualy digital version of the body is at least twice the prize of film body. The M8 is not so more expensive than the M7. The print usued from a M8 or M7 is about the same level. If you have no tripod posibilities and bad light and must use 800 asa, the M8 will shoe beter result than M7.
The M8 shows that the lens limites will not allow big step in 6 months. 10Mpix could be enought for most of (my) realisations.
For beter quality my Hasselblad V with normal film is there.
Of course a 40Mpix sensor is far better, but it's realy expensive and devaluate too fast. The CFV is perhaps not so far of what I could economicaly justify.