Blackmagic Forum

focuspulling wrote:

Fabián Aguirre wrote:

focuspulling wrote:This is a technical forum.

Cinematography is purely a technical pursuit?

Emotionally,

Fabián

It's (almost?) universally understood (and evident if you've paid attention) that this forum topic called Cinematography never ever discusses the style or technique or art of cinematography at all. It means "Cameras," as Blackmagic literally subtitles it, "The place for questions about shooting with Blackmagic Cameras." Stay on topic?

You seem like a really fun dude to be around on set. Not emotional at all.

Fabián Aguirre wrote:You seem like a really fun dude to be around on set. Not emotional at all.

I am. But stay on topic? This is a thread about the PYXIS 6K (see title).

No more please.

timbutt2 wrote:
So I’m not saying that I’m not happy Blackmagic did this design. I’m simply stating it’s a pass at the moment for me as I’ll wait for the next model due to having satisfaction with what I currently have.

Tim, I too feel the same way so for now I am certainly loving the cameras that I am using for my projects and looking for the PYXIS G2 when it’s available.

It's five grand with an evf and a bit of rigging it would take to make it minimum config for me. It would take a couple of pounds off my shoulder in comparison the my minis. It has no special place that I haven't already covered. I also can't find any reviews where anybody has used one. There are always "ya-buts" with BM cameras or any camera for that matter. Always interesting to see new products from BM but for now that's as far as it goes.

I don't like the Pyxis (just personal taste) but I very much like the direction BMD are heading with it.

wemrick1 wrote:It's five grand with an ev... It has no special place that I haven't already covered…

Hmm. Not a perfect camera and not suited to all current shooters, but may broaden their market share.

Something borrowed, something new:

It is a 6K camera that is designed to oversample for 4K delivery in a box form designed to be rigged and easier to mount on various stabilizers which many people wanted in an entry-level cinematic full frame camera (although the ‘box’ might be larger than most imagined).

It does have the dual ISO ranges that the Pocket camera shared making it better suited to somewhat lower light levels and BMD appears to have determined that this is beneficial to continue in their entry-level box camera.

Strangely though one XLR port was sacrificed perhaps to support dual CFexpress recording for continuous recording at all 6K resolutions which was requested by many and is a very welcome addition to keep media storage more cost-effective for their target audience compared to the high-end URSA Cine 12K.

The argument that this single XLR port encourages off-camera recording is only satisfactory for certain use cases. Most times I only use one of the two XLR ports on the UM4.6K (as a disaster recovery option) as the MixPre-6 II ports suffice and provide 32bit audio.

You have the option of using one of three fixed ‘native’ mounts which means many shooters can continue to use their EF photo lenses and many PL lenses for cine without adapters. The L mount was also requested by many as it is adaptable to several other lens mounts.

You know there are many other features, such as integrated cloud and proxy support, that may be considered innovative for BMD.

Certainly the pairing with the new EVF and the extensible EVF are highly desirable for many shooters. I do believe a new generation of Video Assist monitors will follow within the year. These two options make this camera more flexible than many other BMD cameras.

I do think it’s an appealing option that may take some time to build an appreciative audience. And there will be many who hold back on Pyxis at this time. The original Pocket camera was quite a sensation and exceedingly popular. I loved its imagery. But I didn’t purchase it because I was waiting for something with more flexibility and robustness. Fully rigging the original UM4.6K met those needs and continues to work for me and my clients, but in several ways these 2024 cameras offer an opportunity for growth, continued learning, and greater expression. That’s their appeal to me, the sum of the technologies as creative tools is greater than the parts.

… I also can't find any reviews where anybody has used one...

Perhaps BMD hasn’t circulated any prototypes to influencers and the production version is not quite ready for prime time, but it is expected soon.

It's been a while since I posted in the cinematography forum. I've been working mostly as a colorist for the past 5 years. Hello to my old friends.

I have a question regarding the Pyxis L-mount and its level of sturdiness. Can it stand up to the focus motor without being anchored to a mattebox and rails? I still have my EF SLR Magic APO MicroPrime Cine lens set. EF mounts are notoriously wobbly and must be anchored down with a mattebox to rails to use a focus motor - not ideal for gimbal/minimalist work. I'm thinking about getting the L-mount so I can use EF and PL lenses with adaptors (an adaptor with EF locking) that anchor the adaptor to the cage or rails to support the L-mount. Curious about people's thoughts: 1) Go with the aforementioned L-Mount adaptor route; 2) Part with the SLR Magic set (but lose that look) and buy new PL primes (costly); 3) Go with EF Locking Pyxis and call it a day? Thanks!

Scott Stacy wrote:I have a question regarding the Pyxis L-mount and its level of sturdiness. Can it stand up to the focus motor without being anchored to a mattebox and rails?

No, unfortunately. The L-Mount in the Pyxis is only pin registered so it doesn't have the kind of grip that you'd need to avoid rotational lens shifts during focus pulls. The way around this, like you mentioned, is either by anchoring your lenses to some kind of rod support or by anchoring your lens mount adapters to a rod support or to a camera cage. The Pyxis is new and so there isn't a thriving ecosystem of accessories yet, but I have no doubt that we'll be swimming in cages, lens adapters, and other support accessories soon enough.

I mean, Hell, if I can build a perfectly serviceable BMPCC4K rig in 2024 then the Pyxis is gonna be easy as pie to work with.

rick.lang wrote:

wemrick1 wrote:It's five grand with an ev... It has no special place that I haven't already covered…

Hmm. Not a perfect camera and not suited to all current shooters, but may broaden their market share.

Something borrowed, something new:

It is a 6K camera that is designed to oversample for 4K delivery in a box form designed to be rigged and easier to mount on various stabilizers which many people wanted in an entry-level cinematic full frame camera (although the ‘box’ might be larger than most imagined).

It does have the dual ISO ranges that the Pocket camera shared making it better suited to somewhat lower light levels and BMD appears to have determined that this is beneficial to continue in their entry-level box camera.

Strangely though one XLR port was sacrificed perhaps to support dual CFexpress recording for continuous recording at all 6K resolutions which was requested by many and is a very welcome addition to keep media storage more cost-effective for their target audience compared to the high-end URSA Cine 12K.

The argument that this single XLR port encourages off-camera recording is only satisfactory for certain use cases. Most times I only use one of the two XLR ports on the UM4.6K (as a disaster recovery option) as the MixPre-6 II ports suffice and provide 32bit audio.

You have the option of using one of three fixed ‘native’ mounts which means many shooters can continue to use their EF photo lenses and many PL lenses for cine without adapters. The L mount was also requested by many as it is adaptable to several other lens mounts.

You know there are many other features, such as integrated cloud and proxy support, that may be considered innovative for BMD.

Certainly the pairing with the new EVF and the extensible EVF are highly desirable for many shooters. I do believe a new generation of Video Assist monitors will follow within the year. These two options make this camera more flexible than many other BMD cameras.

I do think it’s an appealing option that may take some time to build an appreciative audience. And there will be many who hold back on Pyxis at this time. The original Pocket camera was quite a sensation and exceedingly popular. I loved its imagery. But I didn’t purchase it because I was waiting for something with more flexibility and robustness. Fully rigging the original UM4.6K met those needs and continues to work for me and my clients, but in several ways these 2024 cameras offer an opportunity for growth, continued learning, and greater expression. That’s their appeal to me, the sum of the technologies as creative tools is greater than the parts.

… I also can't find any reviews where anybody has used one...

Perhaps BMD hasn’t circulated any prototypes to influencers and the production version is not quite ready for prime time, but it is expected soon.

There are some features included that I do not fully understand, for example oversampling. Is that really preferable? It hits me a little like dithering which can be hit or mis for a finished product. I may fall into that audience that will appreciate the unit more over time. It's not that I necessarily dislike anything about the camera, its just that I have cameras that do what it does in my possession. In other words circumstance.

I'm wanting more information on the various formats it shoots in, and how they would change angles of view of the lens.

Sent from my SM-S908U using Tapatalk

Oversampling will compensate for the fact that no Bayer-filter sensor is delivering the resolution its number of photocells suggests. About 20% more is the minimum you need to use the full resolution of the target screen. Not that it matters much on a cellphone
And then there’s stabilization or reframing.

Uli Plank wrote:Oversampling will compensate for the fact that no Bayer-filter sensor is delivering the resolution its number of photocells suggests. About 20% more is the minimum you need to use the full resolution of the target screen. Not that it matters much on a cellphone
And then there’s stabilization or reframing.

That helps a lot. 20% would be significant. Lol, the cell phone. It is amazing the images they can produce given it's a phone but no, I won't be selling my cameras and buying the latest Iphone anytime soon.

I was thinking about their screens, actually.
Some have very high numbers of pixels, but how closely do your eyes focus?

Red claimed for their very first camera, that the 5K photocells it had were needed to resolve 4K. While we all know about the value of claims by Red, audiences could spot the difference when downsampled on a good 4K projection. A bit more won't hurt, so I think that those 6K cameras by BM hit quite a sweet spot, leaving a bit of leeway for the other two points.

My understanding is a 40% oversampling is ideal. A 6K camera gives you more real estate to reframe or stabilize in post and deliver 4K. We’re at the point where 4K has arrived with 8K knocking at the door, but I assume it’s a long way from being a required delivery for most members of this forum.

wemrick1 wrote:It's five grand with an evf and a bit of rigging it would take to make it minimum config for me.

I was one of the handful here who constantly asked and wanted BMD to deliver a box camera and they sure did. I remember a post I wrote stating to take everything they’ve done with the BMPCC6K G2 and put it in a box, less the monitor, and sell it for < 2K USD (it’s been years but I think I said $1895 or $1995, probably less). The Pyxis came close but no cigar. If it cost 5 grand to rigged it up with the EVF, that’s not a good deal at all. Instead of that silly side monitor, they should just have priced it with the EVF. But of course, that’s not how they’re thinking, ever, and regardless of what is suggested here in the forum, I think it just goes on deft ears and for us forum members to debate, make wishful request, etc. BM products came notably to my radar since they came out with the Pocket OG and BMCC 2.5K was how reasonable their prices were. It’s changing which makes me wonder if they can still be making affordable quality products with competitive features that is par with a RED/Nikon, ARRI, or Sony. Maybe there’s a better light at the end of the tunnel with the Pyxis G2.

Do you think we can come to anything close to a consensus about the attributes of the Pyxis G2? There have been items mentioned in prior posts but not much detail.

Nothing is ever enough ...

Sent from my SM-S908U using Tapatalk

rick.lang wrote:Do you think we can come to anything close to a consensus about the attributes of the Pyxis G2? There have been items mentioned in prior posts but not much detail.

This community is truly cooked. Consensus about the successor to a camera that has not shipped to a single customer yet? There's not enough grass in the world to touch at this point. If only we all put this much energy in to using these tools instead of posting about them... yeesh...

Alex Mitchell wrote:

rick.lang wrote:Do you think we can come to anything close to a consensus about the attributes of the Pyxis G2? There have been items mentioned in prior posts but not much detail.

This community is truly cooked. Consensus about the successor to a camera that has not shipped to a single customer yet? There's not enough grass in the world to touch at this point. If only we all put this much energy in to using these tools instead of posting about them... yeesh...

Totally agree with this. People seem to never be happy and complain about something that is not even out yet..

The problem with external monitor, even using a SDI with the VA 12G on the UMP G2 there is a noticeable latency. I don't know if that is improved with the EVF using USB-C?

Ellory Yu wrote: BM products came notably to my radar since they came out with the Pocket OG and BMCC 2.5K was how reasonable their prices were.

The BMCC did cost $2995 when it was released. For the exact same price you now get the Pyxis 6K. What changed your perception of a reasonable price?

Bunk Timmer wrote:

Ellory Yu wrote: BM products came notably to my radar since they came out with the Pocket OG and BMCC 2.5K was how reasonable their prices were.

The BMCC did cost $2995 when it was released. For the exact same price you now get the Pyxis 6K. What changed your perception of a reasonable price?

Yes but it wasnt until after April 2013 that more people actually got their hands on the BMCC2.5k and by then the price has dropped to $2k. However if you factor in inflation cameras in general has gotten cheaper over the years.

BMD still is the epitome of great price/performance ratio, the Pyxis price is phenomenal.

Florent Piovesan wrote:

Alex Mitchell wrote:

rick.lang wrote:Do you think we can come to anything close to a consensus about the attributes of the Pyxis G2? There have been items mentioned in prior posts but not much detail.

This community is truly cooked. Consensus about the successor to a camera that has not shipped to a single customer yet? There's not enough grass in the world to touch at this point. If only we all put this much energy in to using these tools instead of posting about them... yeesh...

Totally agree with this. People seem to never be happy and complain about something that is not even out yet..

I disagree in the case of Pyxis. It feels like a missed layup due to 2-3 polarizing decisions (which I won't repeat, they are out there everywhere in the internet) and the group of potential customers who clearly see room for improvement does not seem to be small.

The Pocket 6K line showcases that BMD listen and adapt to wishes and criticism. Constructive criticism is an important driver for improvement in my opinion, something both parties (manufacturer/customer) ultimately benefit from.

Michel Rabe wrote:The Pocket 6K line showcases that BMD listen and adapt to wishes and criticism

Indeed.

There have been no less than 5 sensor families housed within the Ursa Mini: Cmosis, Fairchild, Onsemi, Sony, and Blackmagic. If this camera is a success then a Pyxis Pro is inevitable and it's features relevant.

Good Luck

WahWay wrote:Yes but it wasnt until after April 2013 that more people actually got their hands on the BMCC2.5k and by then the price has dropped to $2k.

Ellory Yu wrote:BM products came notably to my radar since they came out with the Pocket OG and BMCC 2.5K was how reasonable their prices were.

Cheers.

Howard Roll wrote:… If this camera is a success then a Pyxis Pro is inevitable and it's features relevant…

Hopefully “inevitable” means 2025 when I hope to be able to purchase the Pyxis Pro.

Holy cow you guys! BMD just unveiled a whole line of Pyxis Pro bodies!



Of course not really—we don't even have Gen5 colour in the 4.6K—but c'mon. Y'all are asking for a ton of market segmentation.

Alex Mitchell wrote:Holy cow you guys! BMD just unveiled a whole line of Pyxis Pro bodies!

pyxis pro.jpg

Of course not really—we don't even have Gen5 colour in the 4.6K—but c'mon. Y'all are asking for a ton of market segmentation.

I'd love an URSA Mini Pro G2 sensor in the PYXIS PL. I'd take it in a heartbeat. The 4.6K G2 was an incredible sensor. Even if it only had 60 FPS max frame rate like the UMPG1, but was the G2 sensor then it would be amazing. And would make a great little B-Cam to the UMPG2.

Tim, I can’t see BMD putting an URSA 4.6K Pro G2 Super 35 sensor in a camera designed to be full frame.

Quick question for CaptainHook. Can you tell us the distance between the two top 1/4 20s on the Pyxis?

rick.lang wrote:Tim, I can’t see BMD putting an URSA 4.6K Pro G2 Super 35 sensor in a camera designed to be full frame.

I know. Simply saying I'd still buy it in a heartbeat as a B-Cam for the UMPG2.

Honestly, I'd be happy to go to VistaVision. Maybe if they found a way to get that 4.6K G2 sensor to VV size at 6.5K with that amazing rolling shutter and high frame rates. Then that would be a nice smaller alternative to the UCine12K for the more corporate and lower end projects.

rick.lang wrote:Tim, I can’t see BMD putting an URSA 4.6K Pro G2 Super 35 sensor in a camera designed to be full frame.

But why not? They did the reverse thing with the Pocket line. And replacing a larger with a smaller sensor should be less hassle.

I can see (and wish) the Pyxis body sport a s35 sensor in the future.

But for those who want to shoot Super 35 on the Pyxis, the 9K sensor window is pretty much a traditional Super 35 film gate in terms of the 24.9mm width. The rolling shutter at 9K won’t be as bad and I can see people shooting 9K with their Super 35 cine lenses such as the Sigma 18-35mm and Tokina 11-20mm. I would. So I think there’s not a lot of incentive for BMD to offer a camera with a Super 35 sensor open gate. I could be wrong, but BMD still has some other Super 35 offerings for those shooters.

Let’s say if BMD did include a smaller sensor, it would be interesting to see how well that is adopted. After all shooting full frame does require more care and higher quality than shooting mirrorless for example.

timbutt2 wrote:

Alex Mitchell wrote:Holy cow you guys! BMD just unveiled a whole line of Pyxis Pro bodies!

pyxis pro.jpg

Of course not really—we don't even have Gen5 colour in the 4.6K—but c'mon. Y'all are asking for a ton of market segmentation.

I'd love an URSA Mini Pro G2 sensor in the PYXIS PL. I'd take it in a heartbeat. The 4.6K G2 was an incredible sensor. Even if it only had 60 FPS max frame rate like the UMPG1, but was the G2 sensor then it would be amazing. And would make a great little B-Cam to the UMPG2.

The UMP G2 does 4.6k full sensor at 120fps. I believe the sensor is different than the G1 which shares the same sensor as the UMP 4.6k OG.

rick.lang wrote:But for those who want to shoot Super 35 on the Pyxis, the 9K sensor window is pretty much a traditional Super 35 film gate in terms of the 24.9mm width. The rolling shutter at 9K won’t be as bad and I can see people shooting 9K with their Super 35 cine lenses such as the Sigma 18-35mm and Tokina 11-20mm. I would. So I think there’s not a lot of incentive for BMD to offer a camera with a Super 35 sensor open gate. I could be wrong, but BMD still has some other Super 35 offerings for those shooters.

Let’s say if BMD did include a smaller sensor, it would be interesting to see how well that is adopted. After all shooting full frame does require more care and higher quality than shooting mirrorless for example.

I think you're confusing the sensors in the PYXIS. The PYXIS has a 4K S35 mode. It's the URSA Cine 12K you're referencing there.

WahWay wrote:

timbutt2 wrote:

Alex Mitchell wrote:Holy cow you guys! BMD just unveiled a whole line of Pyxis Pro bodies!

pyxis pro.jpg

Of course not really—we don't even have Gen5 colour in the 4.6K—but c'mon. Y'all are asking for a ton of market segmentation.

I'd love an URSA Mini Pro G2 sensor in the PYXIS PL. I'd take it in a heartbeat. The 4.6K G2 was an incredible sensor. Even if it only had 60 FPS max frame rate like the UMPG1, but was the G2 sensor then it would be amazing. And would make a great little B-Cam to the UMPG2.

The UMP G2 does 4.6k full sensor at 120fps. I believe the sensor is different than the G1 which shares the same sensor as the UMP 4.6k OG.

Probably it is an updated sensor from the G1. It does do 120 FPS full sensor 4.6K, which I love! If they got that in a PYXIS limited to even 60 FPS I'd be fine as long as it had the same rolling shutter as the UMPG2.

timbutt2 wrote:I think you're confusing the sensors in the PYXIS. The PYXIS has a 4K S35 mode. It's the URSA Cine 12K you're referencing there.

Absolutely correct. I think I was still asleep with that post!

rick.lang wrote:My understanding is a 40% oversampling is ideal. A 6K camera gives you more real estate to reframe or stabilize in post and deliver 4K. We’re at the point where 4K has arrived with 8K knocking at the door, but I assume it’s a long way from being a required delivery for most members of this forum.

Relevant: https://en.wikipedia.org/wiki/Nyquist%E ... ng_theorem

Let's say there was a sensor that could actually deliver all information at one photosite. i.e. 1 photosite = 1 pixel ...even then we need 2x the capture pixels to eliminate artifacts like aliasing, moire, etc.

But the reality is that a bayer pattern sensor, or even and RGBW sensor, requires 4 photosites to equal one complete pixel. The rest is inferred. It's inferred really, really well, in a way that takes advantage of how we see, but it's still missing a ton of "real" information.

That's a long winded way of saying that there is most definitely real world benefit to supersampling at capture. 40% is barely adequate. 100% would be ideal. Beyond that, it gets wasted.

With a bayer colour filter array, a 2x2 matrix at a minimum may be used by the image processor, reading the signal from one red and one blue and two green photosites.

When using BMD’s custom filter array RGBWWW (which BMD and everyone else confusingly refers to as RGBW), my understanding is that 36 photosites are used to build each pixel, using the signal from 18 white, 6 red, 6 green, and 6 blue photosites. So if you are planning on an 8K (URSA Cine 12K) or 4K (Pyxis) delivery in which one is oversampling, I’m not sure what the ideal and minimum oversampling rates are considered to be for the new URSA Cine 12K and Pyxis 6K when the best colour and enhanced details are desired.

So, after testing a few new-ish cameras a week ago, I've decided to cancel my Pyxis pre-order.

Not because I'm choosing one of the cameras I tested, but rather because it served to remind me that currently none of the sub $10K cameras offer a meaningful improvement in PQ over my current camera. The only element of picture quality left for meaningful improvement is dynamic range, but they're all stuck in the 12-ish usable stops range. I keep hoping the next generation of sensors focuses on dynamic range instead of resolution or low light, so I keep waiting.

So, the only reason to choose the Pyxis would be for the form-factor. It seems like BMD considers the Pyxis to be their answer to the box camera, but the more I look at it, the more it seems like an Ursa Micro, and the Ursa bodies are a form factor I have never liked.

Does this mean you’re going to look at acquiring the URSA Cine 12K with advertised 16 stops dynamic range?

rick.lang wrote:With a bayer colour filter array, a 2x2 matrix at a minimum may be used by the image processor, reading the signal from one red and one blue and two green photosites.

When using BMD’s custom filter array RGBWWW (which BMD and everyone else confusingly refers to as RGBW), my understanding is that 36 photosites are used to build each pixel, using the signal from 18 white, 6 red, 6 green, and 6 blue photosites. So if you are planning on an 8K or 4K delivery in which one is oversampling, I’m not sure what the ideal oversampling rate is considered to be for the new URSA Cine 12K.

When we're synthesizing information, there's no one answer as to how much is enough, because one method of doing so could be materially better than another. And one method might be better in one scenario and worse in another. We can safely say the required samples will be no LESS than what Nyquist–Shannon requires for actual captured information, and will usually be more.

rick.lang wrote:Does this mean you’re going to look at acquiring the URSA Cine 12K with advertised 16 stops dynamic range?

No, as, again, I strongly dislike the Ursa form-factor. I'm also afraid it wouldn't rent well.

Right now, the only currently shipping camera that offers both a do-almost-anything body and a material PQ improvement is the RED V-Raptor X. Maybe something from Kinefinity would as well, but I've never tested one and they seem stuck in CODEC hell.

joe12south wrote:But the reality is that a bayer pattern sensor, or even and RGBW sensor, requires 4 photosites to equal one complete pixel.

As an aside a single RGB pixel derived from Braw on the 12K RGBW sensor references ~30-40 photo sites. The color filter kernels are 7X7 but not all pixels are sampled, the R and B Braw pixels alone use 16 photo sites each to derive a value so we're already at 32 before we talk luminance.

Good Luck

rick.lang wrote:my understanding is that 36 photosites are used to build each pixel,

The CFA arrangement is a 6X6 pattern. The filter kernels are 3x3, 5x5, and 7X7, but not all pixels are sampled for each component. Braw is 4:2:0 at the end of the day, so it's one pair of R/B color samples for every 4 luma samples.

Good Luck

Thanks for the additional information lads!

Howard Roll wrote:

rick.lang wrote:my understanding is that 36 photosites are used to build each pixel,

The CFA arrangement is a 6X6 pattern. The filter kernels are 3x3, 5x5, and 7X7, but not all pixels are sampled for each component. Braw is 4:2:0 at the end of the day, so it's one pair of R/B color samples for every 4 luma samples.

Good Luck

Interesting. i've never really looked into the sensor much as I couldn't find any PQ benefit to it when tested against bayer pattern cameras. What is the rationale for so many white sample points when arguably two green samples work similarly well for determining luminance?

There are bayer pattern cameras with more dynamic range. There are bayer pattern sensors with better light gathering. And I've seen no demonstrable improvement in color accuracy. The only tangible benefit I'm aware of to this structure is spatially "scalable" RAW. But surely that's not enough of a reason to go with an entirely new sensor. I'm just so curious...why? Why pursue this path vs adding phase detect pixels or dual gain output...tech that has obvious real world benefits?

I know we are off topic in terms of this Pyxis thread. Perhaps we should start a thread to discuss the benefits of the BMD’s proprietary CFA and sensor in their high-end cameras? I think 16 stops of dynamic range in the URSA Cine 12K is one answer. There have been other discussions about this when the URSA Pro. 12K was announced.

joe12south wrote:What is the rationale for so many white sample points when arguably two green samples work similarly well for determining luminance?

Sensitivity. If you ever had the pleasure to shoot B&W with a sensor without Bayer filters, you would have got up to 1.5 stops more. It's surprising how much light those filters swallow.

And then:
Are you sure that BRAW is only 4:2:0? I get much better keying and colour selection with it than any 4:2:0 sources.