Philips 37PFL9632D Calibration

I calibrated the Philips 37PFL9632D, and assessed various benchmarks and eventual image quality after updating the firmware on the television to the latest version available at the time of writing (000.064.014.015).

Greyscale

Thinking that the [Movie] picture preset – [Tint] "Warm" as default – should yield a greyscale that is closest to D65, I measured the out-of-the-box correlated colour temperature (CCT) and RGB tracking on the Philips 37PFL9632D:

CCT of [Movie] preset	RGB tracking of [Movie] preset

As you can see, from 60% stimulus upwards the CCT was not doing too badly, hovering around 7000K (i.e. a touch blue). But for 50% stimulus and below the CCT dramatically jumped to around 9000K. Meaning? The darker areas of the image on screen will look too blue (relative to the brighter areas) with some compensatory colour shifts.

The culprits for this phenomenon are [Dynamic Contrast] and [Active Control], the default settings of which in the [Movie] picture preset are "Medium" and "On" respectively. Once I switched them both off, the results became much more respectable:

[Movie] CCT with [DC] & [AC] disabled	[Movie] RGB with [DC] & [AC] disabled

The ensuing CCT tracked fairly consistently across almost the entire luminance range, although there remained a slight bias towards blue. Using the [Custom Tint] controls to calibrate the greyscale, I managed to achieve the following:

Post-calibration CCT	Post-calibration RGB tracking

I probably could have improved the greyscale further but for 2 issues: [R-WP] was already set to the maximum value of 127 from the beginning leaving no room for manoeuvre; and there was no control in the user menu to adjust blue cut/ bias. Still, after calibration the delta errors (dEs) were reduced between 20% and 90% stimulus, and greyscale inched closer to D65. The top-end red clipping probably could be eradicated by lowering contrast, but I did not wish to compromise the contrast ratio even more as the calibrated black level was not fantastic to begin with.

Colour 

Aligning greyscale as outlined above is probably the best thing you can do to calibrate the colours on the Philips 37PFL9632D, as the colour controls available in the user menu are limited. [Colour] affects the saturation and intensity for all colours; [Hue] is greyed out for non-NTSC source; and [Colour Enhancement] chiefly alters the saturation of blue and its derivatives.

Post-calibration CIE chart with reference to HD Rec. 709

With the exception of green which is skewed towards blue, the primary and secondary colour points were not too far off their respective reference coordinates for Rec. 709, the colour space standard for HDTVs. You can't tell from this chart, but from my tests colour decoding was very good on the Philips 37PFL9632D with no significant push.

Benchmark Test Results At A Glance

Black Level 

Calibrated black level on the Philips 37PFL9632D measured similar to that on a Toshiba 47Z3030D, i.e. above average for an LCD television, but nowhere near the class-leading (as long as we're not talking about LED backlighting) blacks found on the Sony KDL40W3000 and the Samsung LE40F86BD. That said, the presence of Ambilight acting as bias lighting did deepen the perception of blacks somewhat.

Had the Philips 37PFL9632D offered some sort of backlight control in the on-screen menu, black-level performance could have been improved even further, but as things stand, one can only achieve the minimum backlight possible through the [Dynamic Contrast] or the [Light Sensor] options, neither of which are particularly appealing for their own reasons.

Even though engaging [Dynamic Contrast] lowered the measured blacks to the level of a Sharp LC42XL2E, undesirable colour shifts, black level fluctuation and shadow detail obliteration ensued. Covering the light sensor with a gaffer tape with it enabled could theoretically allow for an improved and consistent black-level performance and shadow delineation, but then you would have to consider how this mars the aesthetics of the television, and the sensing field of the remote control (the infrared sensor is located right next to the light sensor).

Motion Resolution

Chapter 31 of the "FPD Benchmark Software" revealed that baseline motion resolution on the Philips 37PFL9632D was 300, a common figure for LCD televisions not employing MCFI (motion-compensated frame interpolation) technology.

However, engaging [100Hz Clear LCD] trebled the motion resolution to 900, the highest I've have recorded on any LCD TV to date. Unfortunately this also introduced some interpolation artifacts in the form of halos surrounding objects in movement.

Video Processing

By designating a particular input on the Philips 37PFL9632D as [PC] in the [Installation] > [Connections] submenu, one can disable an umbrella processing engine which includes [Perfect Pixel HD] and [HD Natural Motion] (both of these don't work in [PC] mode even though the options are still selectable), and some functions not covered in the user menu. Marking an input with any of the other labels such as [None], [Blu-ray] or [DVD] would let this processing engine operate as normal. This has some important implications as we shall see later.

Standard-definition video processing was exceptional regardless of whether [PC] mode was engaged or not. The quality of scaling was among the best I have seen on an LCD TV so far, portraying sharpish detail without much edge enhancement. Video mode deinterlacing over 480i and 576i was equally superb – almost no jaggies were seen in the rotating wheel, bouncing bars and waving flag HQV Benchmark test patterns.

Standard-def film mode deinterlacing was not as impressive, though still acceptable nonetheless. The Philips 37PFL9632D successfully detected 3:2 cadence over 480i albeit sluggishly. 2:2 pulldown over 576i presented more of a challenge... the Philips missed this cadence lock fairly frequently, but then again not many televisions do well in this regard.

Over 1080i things got tricky. The Philips 37PFL9632D – with [Picture Format] set to "Unscaled" – failed the "HD Video Resolution Loss Test" on HD HQV Benchmark disc, indicating an issue with video-based 1080i > 1080p conversion that leads to a loss of resolution. The LCD television did pass the "Film Resolution Loss Test" on the same test disc by efficiently engaging film-mode 2:3 pulldown over 1080i, and smoothened the jaggies wonderfully in the 1080i rotating wheel and bouncing bars patterns.

However, labelling an input as [PC] would reverse the results: the 37PFL9632D would pass the 1080i video resolution test (i.e. convert video-based 1080i to 1080p without any loss in resolution) but fail the film-based one. Furthermore, 1080i video deinterlacing would take a significant hit, resulting in rampant combing (a.k.a. feathering/ tearing) and jaggies.

I'll try to sum up the above findings in a single paragraph. One of the most advanced video processing engine/ algorithm (that deals specifically with full HD 1080 signals) on the Philips 37PFL9632D causes the TV to display less-than-full resolution when converting 1080i to 1080p for video-based source. This engine/ algorithm can only be disabled by setting the particular input as [PC] in the [Connections] submenu. But when this engine/ algorithm is disabled (i.e. when [PC] mode is selected), the 37PFL9632D will exhibit not only a loss of resolution for film-based material, but also a lot of combing and jaggies for 1080 source.

Otherwise mixed edits were handled competently on the Philips 37PFL9632D. The various [Noise Reduction] settings cleaned up digital noise fairly effectively, but some fine detail in the image were truncated inadvertently. Because of this I generally refrain from engaging [Noise Reduction], but if you have to use it, the "Minimum" setting should be sufficient to curb video noise without losing too much detail.

1080p/24 Capability

The Philips 37PFL9632D accepted 1080p/24 signal from the Sony PS3. Telecine judder was reduced as a result, but in its place I observed an alarming amount of deinterlacing artifacts in the form of moire, line twitter and jaggies.

But wait, I hear you ask, how can there be deinterlacing artifacts when the video signal fed to the Philips 37PFL9632D is 1080p (i.e. not interlaced)? Put simply, this suggests that the Philips was converting the incoming 1080p signal to 1080i, and then deinterlacing it back to 1080p incorrectly (hence the artifacts).

Sending a 1920x1080 progressive signal from my HTPC to the 37PFL9632D confirmed my suspicions (I made sure [Picture Format] was "Unscaled"). In any non-[PC] modes, when the desktop was static everything looked fine and dandy, but once I moved my mouse cursor the originally-crisp alphabets started flickering/ shimmering. Engaging [PC] mode would extinguish the flicker (or line twitter in deinterlacing terms). Remembering what I wrote earlier about [PC] being the only mode that successfully converts video-based 1080i to 1080p without any loss in resolution, this is more evidence that the 37PFL9632D actually converts 1080p source signal to 1080i before reconstructing it (erroneously most of the time) to 1080p.

Input Lag

I do not own a CRT TV or monitor that accepts HDMI, so I used a Samsung LE52F96BD set to [Game] mode as my point of reference. By sending a DVI-HDMI video signal from my HTPC into a HDMI distribution amplifier which was then fed to both the Philips 37PFL9632D and Samsung LE52F96BD, I followed the instructions in this forum post, and determined that the Philips was lagging behind the Samsung on average 100ms (across 10 readings) irrespective of resolution/ settings. Depending on your sensitivity you may experience this input lag during console gaming sessions, or as lip-synch delay when not using the LCD TV to process the audio signal from source.