We love the new Panasonic TX-CZ952 OLED TV and have invited Gordon Fraser – one of Europe’s finest video calibrators – to fully analyse the performance on offer. His report is below and covers the picture performance in extensive detail. If you have any questions regarding the report or the television please contact me.
To read the full report as a PDF please click HERE
I had the pleasure to look at and calibrate one of the new Panasonic OLED displays last week for one of my Lumagen dealers. This TX-65CZ952B unit uses the curved LG OLED panel but with Panasonic electronics. There is no doubt, in my mind, that this is the best television I’ve ever seen. Not only that, when partnered with a Radiance Pro 4K scaler it was able to deliver the most accurate colour I’ve ever measured or seen in any domestic display.
The LG OLED’s can be great and the latest flat 950 series are indeed better than the previous models but they do seem to suffer from poor consistency and each model within the range has different sets of problems to deal with. I have had several clients send units back for refund or replacements. On the 65” sets that use the same panel as this Panasonic I have come across uniformity issues, low luminance level blockiness and serrated vignetting at edges of the screen that varied considerably from set to set. I see none of these issues on the Panasonic TX-65CZ952B. Indeed the only thing that I saw with test patterns was a sort of dither type noise in large patches of solid colours but it was not unpleasant like some folk see on older plasma panels and i could not detect it at all while watching actual tv, bd or streaming 4K content.
Before I put up some of the measurements from this set I’d like to mention a little about delta error. Delta Error is the metric we use to measure whether we can perceive a colour is not correct. The methodology and formulas for calculating this error have been developed through empirical perceptual testing over several years. You see comments about delta error measurements in reviews, usually suggesting that any measurement under three units is considered visibly perfect. This is, unfortunately, misleading. For a start, there are actually several formula’s for measuring delta error. So we need to know which one is in use. Just as an example. On this OLED with the default cinema preset, one of the coloured patches I measured was a 72% saturated RED at 100 percent brightness. This is the sort of level you’d probably see in one of those saturation sweep charts you see in reviews. Well, using the 1994 formula that is generally used in the reviews that patch measured 1.7de. Sounds good? Using the 2000 formula it measured 1.7 as well. However, using the 1976 LUV formula it was just above 15 units of error….not so good. So what’s going on?
Well the answer is that the below three units threshold is probably based on the 1976 research and formulas along with what I believe is hearsay from calibration courses. These formulas have been revised in 94 and 2000 and with those revisions you should really be considering the threshold for visible difference to be closer to 1 unit. For the “standard observer” or “ average viewer” anything under 1 is likely to be imperceptible from target but there will be some folk who can see a smaller difference than that for some colours. When I am calibrating I aim to get as many colours measured as possible under 1 for de2000 formula. Manufacturers of test equipment used in calibration vary in what they suggest.
Calman for instance, the software the most reviewers appear to use, suggest making sure that the error is less than 4 units no matter what formula you use.
While Lightspace has their threshold set at 2 units for 76luv and 2000de.
Chromapure, another domestic calibration tool provider suggest under 4 for 1976luv and lab formulas and under 1.5 for 1996 and 2000 ones. When you take the above in to consideration you can see quite easily how you can have reviews of displays claiming reference colour accuracy (de under 3 units for 1994) yet the reviewers will say they prefer the colours from one set over another with equally perfect (to them) measurements. So with all that in mind…lets go!
I am now starting a regime of measuring all new displays I come across to get some data on their operational characteristics. To that end I have made a custom colour patch sequence that I use with Lightspace by Lightillusion, to characterise each display. This patch set consists of 50 greyscale points plus the corresponding r g b points along with lots of different saturation and brightness levels for many other colours spread throughout the potential colour gamut. In total there are currently 680 points measured. I may further tweak this but for now it’ll do. This is what those points look like in a 3D chart.
Here is the typical range of measurements in a saturation sweep graph you see in reviews. 10 point white balance plus rgbcmy saturation sweep (25/50/75/100%) at 100%luminance.
To start I decided to just measure a couple of the out of the box presets on default. THX cinema and Cinema settings before going on to calibrate and then apply the Lumagen LUT file. As well as measuring these presets I also had a look at the in built colour gamut presets and also how linear the operation of the display is. With Lightspace I can take some very quick measurements of the RGB channels and white balance in order to see what’s called RGB separation. If a display is well engineered then it will have good RGB separation. This will mean that adjusting one channel will not affect another. Light illusion define it like this
“The definition of ‘good RGB separation’ is when the grey scale is an exact sum of the individual RGB values, combined with colour channel independence, where changing the value of one colour channel has no effect on the other two colour channels – there is no cross-talk between the colour channels.”
This is important in any display but particularly one where you have some form of CMS adjustment. With a non linear display with non optimal rgb separation you can make adjustments using the Colour Management System to make the coloured patch you are measuring be correct, but it’s likely you just changed everything else that uses that colour in a non uniform way, perhaps making it better but quite possibly making it worse and if you don’t bother measuring a large subset of colours again you wont know but your chart for the single colour you did measure will look lovely.
Here is what very good RGB separation chart should look like. This one is for an Epson LS10000 Laser light source projector.
Here are the same charts for the LG960 and the Panasonic TX-65CZ952B 65” OLED sets.
65” LG960 OLED
This sort of response is pretty typical of a display that has ABL circuits in it where there is constant on the fly dynamic adjustments going on. The Panasonic is better than the LG but not by a huge margin. When carrying out calibration with an off board processor like a Lumagen you want to find the preset on the set where rgb separation is at its best and where the colour gamut you want to target is fully covered already. Usually this is the native colourspace of the display if it has such an option. You can find that the preset colour gamuts in displays are often created using inferior colour management systems that work in HSL colorspace and that are not linear making them harder to correct even though they may appear closer to the target gamut in the first place.
So on to how we ended up after calibration. Remember this is not a detailed review of the operation of the set. You can read them elsewhere. This is just the first of a series of reports I’ll put up about what i am finding with new displays as regard to their colour accuracy both with and without external processing help and my general comments as to whether I like them.
First Preset I looked at was THX Cinema.
You can see how different the levels of error are depending on what formula you use. As I stated earlier, i aim to get as much as possible as close to or under 1 unit of error DE2000. I have further breakdowns of where these colour errors are, i.e. what hues and what luminance levels which can be useful when you are actually doing calibrations as you want any errors you are left with to be in saturation and luminance levels that rarely occur in actual viewing content. For now though I’d rather not confuse the issue.
Next one was the Cinema preset itself, non THX flavour.
You can see this is getting better. More percentage of errors in the lower ranges. I had limited time so I didn’t bother with the Bright presets or Dynamic and just went straight to calibrating the Professional 1 preset using the multi-point greyscale and gamma and the on board CMS. This was done using the REC709 colour gamut preset in the advanced menu.
Professional 1 preset.
This is a very good result for a calibration of a display with no external help. We still have 55% of the measured patches above my holy grail of 1 delta error 2000 and 32.5% still above 4 units of 1976luv error. This set will ultimately be going to one of my dealers showrooms where it may be used with a Lumagen Radiance Pro so to that end I decided to see just how good we could get it. Using Pro2 with different settings as a base for the LUT profiling I measured around 1350 coloured patches with Lightspace. If I was doing this for a client i’d measure more but the was ticking by this point. The more patches you measure the better the “map” of errors the software has in order to create the re-mapping LookUpTable LUT that is put in the Radiance to get ultimate accuracy. So with this information I made a correction file and then remeasured to see how close we now were.
Panasonic TX-65CZ952B with Radiance 4446 Pro running 4900+ correction point LUT
There you go. Nothing above two de2000, only 1.5% errors above four de1976luv and just over 80% actually under one unit for de2000. If I’d had time to measure closer to the full 4900 points this would undoubtedly have been even better but as it stands it’s the most accurate colour I’ve come across in any domestic display system. I am now saving my pennies and am hoping Santa is good to me.
I’d like to thank Nick from HomeSound in Edinburgh and the lads at HiFi Corner in Falkirk for allowing me the time to do all this research on their demo model. I believe it is currently in Falkirk if you want to see it for yourself. Also I should thank Mike Nagel at www.displaycalibrationtools.com for creating the online analysis tools I use in conjunction with Lightspace to analyse display performance.
And just to end. Here is what the CIE chart and gamut look like with all my measured points plotted in 2D like you see in reviews.
Thanks for reading,
Convergent AV LTD
Please do not copy or reproduce this material without my permission.