I don't believe 3-CCDs are needed anymore.

Ignacio Rodriguez · January 10th, 2005, 01:14 PM

> I think it is as follows... the 3CCD is noticeably more sensitive -
> somewhere in the neighbourhood of 3 times because the
> sensors are much smaller in the 1CCD camera - less light gets
> to them (a bit better than 1 f-stop sensitivity improvement)

Correct. Perhaps somewhat less than three times, because some light is lost to the color dividing optics, but three sensor cameras are usually more sensitive than same-size 1 sensor cams.

Actually I think this is why Sony had to make one of it's new palmcorders 3-CMOS. One high-res CMOS at such a small size probably doesn't have enough sensitivity for your average birthday party.

David Kennett · January 10th, 2005, 03:28 PM

For years we have underestimated the progress of inovation. When I first saw the JVC HD-10, I was impressed with the balance of priorities. The single CCD has a separate white and green pixel areas. Since green is more than half luminance info, this already gives a pretty high percentage of the CCD area to luminance. The yellow and cyan (or is it magenta?) elements also contribute a high percentage of the luma (especially the yellow). Such a combination allows a better luma S/N at the expense of the chroma. This is exactly what happened in the HD-10, as the chroma S/N suffers more.

I thought this was an excellent compromise, as there is seldom much chroma present in very low light. And there is more to come! It's risky to make predictions that a particular technology will not be superceded. I specifically remember that I thought that CCDs would never replace Plumbicons.

Toke Lahti · January 10th, 2005, 05:04 PM

<<<-- Originally posted by Toke Lahti: Example: white light coming in.
3ccd: 1/3 of the light is green and that gets into green ccd.
1ccd: 2/3 of the light is filtered from green pixel, so 1/3 of the light gets into green pixel.
Where's the difference just considering sensivity? -->>>

So I need to continue a little bit:
Let's assume that 1ccd and 3ccd has same size pixels.
Then they will get same exposure.
If you want same resolution from both then 1ccd should of course have more pixels than one chip from 3ccd. And that means bigger chip for 1ccd.

If I had a choise I would take a camera with 2/3" 1ccd than 1/3" 3ccd.

Ignacio Rodriguez · January 10th, 2005, 06:39 PM

> I would take a camera with 2/3" 1ccd than 1/3" 3ccd.

But you would get a better image with a camera with three 1/2" CCDs, and that would probably be less expensive than a single 2/3" CCD. Sensor size is critical to the cost of a camera system because it directly affects the size of the lenses and because larger sensors are much more expensive to make because yields are much lower, this means more CCDs have to be made to get a good one, thus more money is spent. This is how things have been for the SD world for some time now with CCD technology.

With higher resolutions, aligning three sensors accurately is probably more difficult and costs more money, hence the single sensor concept makes more sense than in the SD world, but large CCDs are expensive. However, large CMOS sensors are common now in digicams. And CMOS technology has gone a long way in the last few years. I think we will end up using giant (i.e. 35mm) single CMOS sensors for HD video, but not just yet. It will take some time, those thing are still expensive at large sizes, even if not as expensive as CCDs.

Juan M. M. Fiebelkorn · January 12th, 2005, 11:00 PM

In fact the sensor itself isn't the biggest cost of a camera these days.
Think that a normal sensor costs around 200 dollars for a 1920x1080 one.
The cost itself goes on housing mechanics, IC and the like.
The sensors used on a consumer camera indeed cost just tens of dollars....
ICs cost more than them.Not to mention all the motors and mechanicals involved in tape mechanisms.

Toke Lahti · January 13th, 2005, 07:53 AM

not to mention the lens...

Ignacio Rodriguez · January 13th, 2005, 08:21 AM

> 200 dollars for a 1920x1080 one

Really? But at what size? 1/6", surely not 1/2" and definately not 35mm, right?

At 1/6", even with three sensors, sensitivity is low, such is the case with my PDX10. Even though it doesn't do HDV, it has a high-res sensor which accounts for it's great downsampled SD image, better than the PD170 under certain conditions. It's sensitivity seems similar to that of the FX1/Z1, although Sony has done a better job with gain and noise and the sensor is larger. I doubt that the sensor array in this new cam costs anything below US$800.

And as Toke hints, the larger the sensor, the larger the optics and they are also a very expensive component, especially if you want the 800 lines or more of res for HDV.

Toke Lahti · January 13th, 2005, 08:39 AM

<<<-- Originally posted by Ignacio Rodriguez: But you would get a better image with a camera with three 1/2" CCDs, and that would probably be less expensive than a single 2/3" CCD. Sensor size is critical to the cost of a camera system because it directly affects the size of the lenses and because larger sensors are much more expensive to make because yields are much lower, this means more CCDs have to be made to get a good one, thus more money is spent. -->>>

3x1/2" use more silicon than 1x2/3" so if quality of production line is ok, then 3x1/2" should be more expensive. These hd chips have been made many years now, so the yield issues should be in the past.

But like Juan, I don't think that ccd/cmos costs are so critical anymore or at least in the future.

And then the critical part will be the lens.

You can already see that in consumer still cameras. Because megapixels are advertised, you can buy a cheap 8Mp camera with lens that has resolving power less than half of the chip.

A good lens for consumer dslr costs many times more than the body.

Zeiss digiprimes are the sharpest lenses in the world and their resolving power is just the needed 100lp/mm for 2/3" hd. But they would be very expensive to make even for mass market.

For 1/3" lens to have resolving power of 2Mp, it needs to have 200lp/mm and that would be even more expensive.

So to design a successful new hd camera to pros/prosumers with optimal price/quality ratio, one needs to find the sweet spot in lenses.

So what would be the optimal choise for imager's size if one considers the resolution, aperture, size, weight, price, etc.?

I believe that 4/3" would be ideal format for overall use.
Biggers are too bulky for document use and smallers lack of sensivity and resolution.
If it would become mass market product maybe lenses for that size might have even 4k resolving power for reasonable price.

Ignacio Rodriguez · January 13th, 2005, 01:12 PM

> So what would be the optimal choise for imager's size
> one considers the resolution, aperture, size, weight, price,
> etc.?

Very keen question, Toke. I have always thought that --since economics of scale and competition in the field of 35mm photography have for decades provided us with increasingly high quality and low cost lenses-- the logical choice is to use a 35mm sensor. Especially if, as you say, sensor size is not such a cost factor. Even considering --as I do-- that sensor size does affect the price of a camera to a large degree, there is such a vast array of cheap 35mm lenses sharp enough for good ol' chemical photography, that surely at least some of them must be good enough for 1920x1080 video.

There are still a few technical problems to fix for this to be possible. The most important one is dealing with the way photo lenses focus light onto film, because where film doesn't care about the direction of light, electronic sensors do, so their will probably have to be something like a HAD sensor, but instead of a lens in front of each pixel it will have a white translucid element to diffuse light. Of course this will mean losing some light, but at such a large sensor size that should not be a problem.

The other important technical problem for HD at consumer prices has been dealing with all that data, but this problem has already been solved by using MPEG2 at 25 Mbps. It would work even better at 50 Mbps with a hard disk instead of tape. And using the MPEG4 AVC can only be better still, not to mention QuickTime-native so anything can edit it.

So the question is, which company will take this opportunity and hit the nail on the head with a consumer 35mm 1920x1080 video camera with an SLR lens mount?

For one thing, it has to be somebody with no significant pro line to kill. This rules out Sony, Panasonic, Ikegami and perhaps Canon because they provide the lenses for high-end video.

I really hope people from JVC, Nikon, Minolta and other companies are reading this.

Kevin Dooley · January 13th, 2005, 01:20 PM

Maybe a good alternative (if sensor size does affect cost--which I think it would) would be a 16mm sensor size--after all, there are lots of different 16mm motion picture lenses on the market (I imagine both zoom and primes) that are older and fairly cheap...

Ignacio Rodriguez · January 13th, 2005, 01:28 PM

I see your point Kevin, but is it not even easier and cheaper to get good 35mm photo lenses than it is to get a hold of good 16mm lenses? Besides, at 16mm, there is not too much light getting in there and I believe we are aproaching the thoretical limit of small sensors with Sony's HAD implementation on the FX1/Z1. Heck Sony is even using non-square pixels to suck every possible photon falling into the imaging area.

Toke Lahti · January 13th, 2005, 10:37 PM

<<<-- Originally posted by Ignacio Rodriguez: Besides, at 16mm, there is not too much light getting in there and I believe we are aproaching the thoretical limit of small sensors with Sony's HAD implementation on the FX1/Z1. -->>>

1/3" sensor is 4.8mm wide and super16 gate is 12.4mm wide, so s16 sized sensor would be 3 stops more sensitive...

Ignacio Rodriguez · January 14th, 2005, 08:41 AM

> 1/3" sensor is 4.8mm wide and super16 gate is 12.4mm
> wide, so s16 sized sensor would be 3 stops more sensitive...

Yes. Surely better, but still not enough to reach PD170 sensitivity with a single sensor at a high resolution, I think. Besides, if --as you claim-- cost is not so dependant on sensor size, the bigger the sensor the more control we have of depth of field and we want that. More light, more control, more choice of lenses for less money... so my vote goes for 35mm single sensor that can use photo lenses. Yours?

Toke Lahti · January 14th, 2005, 12:19 PM

<<<-- Originally posted by Ignacio Rodriguez: ... so my vote goes for 35mm single sensor that can use photo lenses. Yours? -->>>

I have some canon ef lenses, so using them with moving image sounds tempting, but I believe focus pulling in documentary style shooting with 36mm wide imager might be too hard. Using 25mm wide image in 35mm cine film for fiction work is hard enough.
But those nikon and canon cameras that have 1.5-1.6 focal length multiplier imagers might be in reasonable range.
Olympus and panasonic just released co-operation with 4/3" cameras.
If they would just develop also a videocamera based on that size.

If 4/3" imagers and lenses would become common in still world, I see no reason why those lenses wouldn't be cheaper than modern kino lenses.

So for overall camera I'd choose 4/3". After all that's 90% of academy aperture...
Preferably with 4k-resolution (12Mp?, 16Mp?) in univiusium format...
If just somebody would come up with casing that would be handy for both still and moving pictures.

Juan M. M. Fiebelkorn · January 18th, 2005, 02:52 AM

Some little facts:

1) A 2/3 inch sensor is in fact the nearest to a 16mm film equivalent.
They are about 9.6 to 10mm wide sensitive area (16mm film 12mm aprox).

2)Sensitivity is not so easy to calculate based on sensitive silicon surface.
A ProcamHD3560 (2/3 inch) is far more sensitive than the 36x24 mm sensors used on Kodak's latest digital SRL camera.
A DVX100 has higher sensitivity than many other cameras I've seen with bigger sensors.

3)Any decent modern SRL lens (call it Canon, Nikkor,Zeiss,Leica or even Pentax :) ) easily exceed an "aerial resolution" of 100lp/mm.
Limitations are not in fact related to the lenses themselves but to contructive characteristics of camera body, shutter mechanism, ambiental/weather conditions and negative itself.
(think about that an ultra prime lens is supposed to resolve beyond 600lp/mm "aerial" and you will never get much more of around 80 lp/mm on 35mm movie film negative no matter what you try to do)

Also remember that Century optics takes apart Canon SRL zooms, disassemble them, put all those nice lenses inside different mechanics and sell them for cinema use at 50 times the original price.

Also I would trust more on modern SRL lenses resolution if using them for digital HD sensors than getting old 16mm lenses wich were not design for higher resolutions than the one available for 16mm film at their time (you know old coatings, etc)

these are my two cents, hope nobody got offended by this.

January 10th, 2005, 01:14 PM	#46
Ignacio Rodriguez Contributor Join Date: Sep 2003 Location: Santiago, Chile Posts: 932	> I think it is as follows... the 3CCD is noticeably more sensitive - > somewhere in the neighbourhood of 3 times because the > sensors are much smaller in the 1CCD camera - less light gets > to them (a bit better than 1 f-stop sensitivity improvement) Correct. Perhaps somewhat less than three times, because some light is lost to the color dividing optics, but three sensor cameras are usually more sensitive than same-size 1 sensor cams. Actually I think this is why Sony had to make one of it's new palmcorders 3-CMOS. One high-res CMOS at such a small size probably doesn't have enough sensitivity for your average birthday party. __________________ Ignacio Rodríguez in the third world. @micronauta on Twitter. Main hardware: brain, eyes, hands.

January 10th, 2005, 03:28 PM	#47
David Kennett Major Player Join Date: Jul 2003 Location: North Ridgeville, Ohio Posts: 407	For years we have underestimated the progress of inovation. When I first saw the JVC HD-10, I was impressed with the balance of priorities. The single CCD has a separate white and green pixel areas. Since green is more than half luminance info, this already gives a pretty high percentage of the CCD area to luminance. The yellow and cyan (or is it magenta?) elements also contribute a high percentage of the luma (especially the yellow). Such a combination allows a better luma S/N at the expense of the chroma. This is exactly what happened in the HD-10, as the chroma S/N suffers more. I thought this was an excellent compromise, as there is seldom much chroma present in very low light. And there is more to come! It's risky to make predictions that a particular technology will not be superceded. I specifically remember that I thought that CCDs would never replace Plumbicons. __________________ Dave

January 10th, 2005, 05:04 PM	#48
Toke Lahti Major Player Join Date: Oct 2004 Location: Helsinki, Finland Posts: 158	Re: Re: Color accuracy is another major reason for 3CCD <<<-- Originally posted by Toke Lahti: Example: white light coming in. 3ccd: 1/3 of the light is green and that gets into green ccd. 1ccd: 2/3 of the light is filtered from green pixel, so 1/3 of the light gets into green pixel. Where's the difference just considering sensivity? -->>> So I need to continue a little bit: Let's assume that 1ccd and 3ccd has same size pixels. Then they will get same exposure. If you want same resolution from both then 1ccd should of course have more pixels than one chip from 3ccd. And that means bigger chip for 1ccd. If I had a choise I would take a camera with 2/3" 1ccd than 1/3" 3ccd.

January 10th, 2005, 06:39 PM	#49
Ignacio Rodriguez Contributor Join Date: Sep 2003 Location: Santiago, Chile Posts: 932	> I would take a camera with 2/3" 1ccd than 1/3" 3ccd. But you would get a better image with a camera with three 1/2" CCDs, and that would probably be less expensive than a single 2/3" CCD. Sensor size is critical to the cost of a camera system because it directly affects the size of the lenses and because larger sensors are much more expensive to make because yields are much lower, this means more CCDs have to be made to get a good one, thus more money is spent. This is how things have been for the SD world for some time now with CCD technology. With higher resolutions, aligning three sensors accurately is probably more difficult and costs more money, hence the single sensor concept makes more sense than in the SD world, but large CCDs are expensive. However, large CMOS sensors are common now in digicams. And CMOS technology has gone a long way in the last few years. I think we will end up using giant (i.e. 35mm) single CMOS sensors for HD video, but not just yet. It will take some time, those thing are still expensive at large sizes, even if not as expensive as CCDs. __________________ Ignacio Rodríguez in the third world. @micronauta on Twitter. Main hardware: brain, eyes, hands.

January 13th, 2005, 08:21 AM	#52
Ignacio Rodriguez Contributor Join Date: Sep 2003 Location: Santiago, Chile Posts: 932	> 200 dollars for a 1920x1080 one Really? But at what size? 1/6", surely not 1/2" and definately not 35mm, right? At 1/6", even with three sensors, sensitivity is low, such is the case with my PDX10. Even though it doesn't do HDV, it has a high-res sensor which accounts for it's great downsampled SD image, better than the PD170 under certain conditions. It's sensitivity seems similar to that of the FX1/Z1, although Sony has done a better job with gain and noise and the sensor is larger. I doubt that the sensor array in this new cam costs anything below US$800. And as Toke hints, the larger the sensor, the larger the optics and they are also a very expensive component, especially if you want the 800 lines or more of res for HDV. __________________ Ignacio Rodríguez in the third world. @micronauta on Twitter. Main hardware: brain, eyes, hands.

January 12th, 2005, 11:00 PM	#50
Juan M. M. Fiebelkorn Major Player Join Date: Jun 2004 Location: Buenos Aires , Argentina Posts: 444	In fact the sensor itself isn't the biggest cost of a camera these days. Think that a normal sensor costs around 200 dollars for a 1920x1080 one. The cost itself goes on housing mechanics, IC and the like. The sensors used on a consumer camera indeed cost just tens of dollars.... ICs cost more than them.Not to mention all the motors and mechanicals involved in tape mechanisms.

January 13th, 2005, 07:53 AM	#51
Toke Lahti Major Player Join Date: Oct 2004 Location: Helsinki, Finland Posts: 158	not to mention the lens...

January 13th, 2005, 08:39 AM	#53
Toke Lahti Major Player Join Date: Oct 2004 Location: Helsinki, Finland Posts: 158	<<<-- Originally posted by Ignacio Rodriguez: But you would get a better image with a camera with three 1/2" CCDs, and that would probably be less expensive than a single 2/3" CCD. Sensor size is critical to the cost of a camera system because it directly affects the size of the lenses and because larger sensors are much more expensive to make because yields are much lower, this means more CCDs have to be made to get a good one, thus more money is spent. -->>> 3x1/2" use more silicon than 1x2/3" so if quality of production line is ok, then 3x1/2" should be more expensive. These hd chips have been made many years now, so the yield issues should be in the past. But like Juan, I don't think that ccd/cmos costs are so critical anymore or at least in the future. And then the critical part will be the lens. You can already see that in consumer still cameras. Because megapixels are advertised, you can buy a cheap 8Mp camera with lens that has resolving power less than half of the chip. A good lens for consumer dslr costs many times more than the body. Zeiss digiprimes are the sharpest lenses in the world and their resolving power is just the needed 100lp/mm for 2/3" hd. But they would be very expensive to make even for mass market. For 1/3" lens to have resolving power of 2Mp, it needs to have 200lp/mm and that would be even more expensive. So to design a successful new hd camera to pros/prosumers with optimal price/quality ratio, one needs to find the sweet spot in lenses. So what would be the optimal choise for imager's size if one considers the resolution, aperture, size, weight, price, etc.? I believe that 4/3" would be ideal format for overall use. Biggers are too bulky for document use and smallers lack of sensivity and resolution. If it would become mass market product maybe lenses for that size might have even 4k resolving power for reasonable price.

January 13th, 2005, 01:12 PM	#54
Ignacio Rodriguez Contributor Join Date: Sep 2003 Location: Santiago, Chile Posts: 932	> So what would be the optimal choise for imager's size > one considers the resolution, aperture, size, weight, price, > etc.? Very keen question, Toke. I have always thought that --since economics of scale and competition in the field of 35mm photography have for decades provided us with increasingly high quality and low cost lenses-- the logical choice is to use a 35mm sensor. Especially if, as you say, sensor size is not such a cost factor. Even considering --as I do-- that sensor size does affect the price of a camera to a large degree, there is such a vast array of cheap 35mm lenses sharp enough for good ol' chemical photography, that surely at least some of them must be good enough for 1920x1080 video. There are still a few technical problems to fix for this to be possible. The most important one is dealing with the way photo lenses focus light onto film, because where film doesn't care about the direction of light, electronic sensors do, so their will probably have to be something like a HAD sensor, but instead of a lens in front of each pixel it will have a white translucid element to diffuse light. Of course this will mean losing some light, but at such a large sensor size that should not be a problem. The other important technical problem for HD at consumer prices has been dealing with all that data, but this problem has already been solved by using MPEG2 at 25 Mbps. It would work even better at 50 Mbps with a hard disk instead of tape. And using the MPEG4 AVC can only be better still, not to mention QuickTime-native so anything can edit it. So the question is, which company will take this opportunity and hit the nail on the head with a consumer 35mm 1920x1080 video camera with an SLR lens mount? For one thing, it has to be somebody with no significant pro line to kill. This rules out Sony, Panasonic, Ikegami and perhaps Canon because they provide the lenses for high-end video. I really hope people from JVC, Nikon, Minolta and other companies are reading this. __________________ Ignacio Rodríguez in the third world. @micronauta on Twitter. Main hardware: brain, eyes, hands.

January 13th, 2005, 01:20 PM	#55
Kevin Dooley Major Player Join Date: Dec 2003 Location: Fort Worth, TX Posts: 414	Maybe a good alternative (if sensor size does affect cost--which I think it would) would be a 16mm sensor size--after all, there are lots of different 16mm motion picture lenses on the market (I imagine both zoom and primes) that are older and fairly cheap... __________________ Kevin Dooley Media Director, Pantego Bible Church

January 13th, 2005, 01:28 PM	#56
Ignacio Rodriguez Contributor Join Date: Sep 2003 Location: Santiago, Chile Posts: 932	I see your point Kevin, but is it not even easier and cheaper to get good 35mm photo lenses than it is to get a hold of good 16mm lenses? Besides, at 16mm, there is not too much light getting in there and I believe we are aproaching the thoretical limit of small sensors with Sony's HAD implementation on the FX1/Z1. Heck Sony is even using non-square pixels to suck every possible photon falling into the imaging area. __________________ Ignacio Rodríguez in the third world. @micronauta on Twitter. Main hardware: brain, eyes, hands.

January 13th, 2005, 10:37 PM	#57
Toke Lahti Major Player Join Date: Oct 2004 Location: Helsinki, Finland Posts: 158	<<<-- Originally posted by Ignacio Rodriguez: Besides, at 16mm, there is not too much light getting in there and I believe we are aproaching the thoretical limit of small sensors with Sony's HAD implementation on the FX1/Z1. -->>> 1/3" sensor is 4.8mm wide and super16 gate is 12.4mm wide, so s16 sized sensor would be 3 stops more sensitive...

January 14th, 2005, 08:41 AM	#58
Ignacio Rodriguez Contributor Join Date: Sep 2003 Location: Santiago, Chile Posts: 932	> 1/3" sensor is 4.8mm wide and super16 gate is 12.4mm > wide, so s16 sized sensor would be 3 stops more sensitive... Yes. Surely better, but still not enough to reach PD170 sensitivity with a single sensor at a high resolution, I think. Besides, if --as you claim-- cost is not so dependant on sensor size, the bigger the sensor the more control we have of depth of field and we want that. More light, more control, more choice of lenses for less money... so my vote goes for 35mm single sensor that can use photo lenses. Yours? __________________ Ignacio Rodríguez in the third world. @micronauta on Twitter. Main hardware: brain, eyes, hands.

January 14th, 2005, 12:19 PM	#59
Toke Lahti Major Player Join Date: Oct 2004 Location: Helsinki, Finland Posts: 158	<<<-- Originally posted by Ignacio Rodriguez: ... so my vote goes for 35mm single sensor that can use photo lenses. Yours? -->>> I have some canon ef lenses, so using them with moving image sounds tempting, but I believe focus pulling in documentary style shooting with 36mm wide imager might be too hard. Using 25mm wide image in 35mm cine film for fiction work is hard enough. But those nikon and canon cameras that have 1.5-1.6 focal length multiplier imagers might be in reasonable range. Olympus and panasonic just released co-operation with 4/3" cameras. If they would just develop also a videocamera based on that size. If 4/3" imagers and lenses would become common in still world, I see no reason why those lenses wouldn't be cheaper than modern kino lenses. So for overall camera I'd choose 4/3". After all that's 90% of academy aperture... Preferably with 4k-resolution (12Mp?, 16Mp?) in univiusium format... If just somebody would come up with casing that would be handy for both still and moving pictures.

January 18th, 2005, 02:52 AM	#60
Juan M. M. Fiebelkorn Major Player Join Date: Jun 2004 Location: Buenos Aires , Argentina Posts: 444	Some little facts: 1) A 2/3 inch sensor is in fact the nearest to a 16mm film equivalent. They are about 9.6 to 10mm wide sensitive area (16mm film 12mm aprox). 2)Sensitivity is not so easy to calculate based on sensitive silicon surface. A ProcamHD3560 (2/3 inch) is far more sensitive than the 36x24 mm sensors used on Kodak's latest digital SRL camera. A DVX100 has higher sensitivity than many other cameras I've seen with bigger sensors. 3)Any decent modern SRL lens (call it Canon, Nikkor,Zeiss,Leica or even Pentax :) ) easily exceed an "aerial resolution" of 100lp/mm. Limitations are not in fact related to the lenses themselves but to contructive characteristics of camera body, shutter mechanism, ambiental/weather conditions and negative itself. (think about that an ultra prime lens is supposed to resolve beyond 600lp/mm "aerial" and you will never get much more of around 80 lp/mm on 35mm movie film negative no matter what you try to do) Also remember that Century optics takes apart Canon SRL zooms, disassemble them, put all those nice lenses inside different mechanics and sell them for cinema use at 50 times the original price. Also I would trust more on modern SRL lenses resolution if using them for digital HD sensors than getting old 16mm lenses wich were not design for higher resolutions than the one available for 16mm film at their time (you know old coatings, etc) these are my two cents, hope nobody got offended by this.