REPORT: Input lag Part 3
A delay in image output on the LCD would arise for analogue playback because the analogue signal must first be digitalised before the monitor can use it to control the individual transistors.
Here, the first problems encountered are the correct display of the digital signal and the defining of a fixed point of reference to act as a time reference.
Whilst in the past, the digital image signals generated by the graphics chip were transformed into analogue signals and usually transmitted from the PC to the monitor via D-Sub HD15 or BNC PC, it was necessary to establish a new and more cost-efficient standard for the high bandwidths required by digital data transfer at increasingly high resolutions. Only thus could additional losses in quality be avoided which would otherwise have arisen with repeated digitalisation of the signals in the monitor.
The Digital Display Working Group or DDWG was founded in 1998 by the companies Compaq, Fujitsu, HP, IBM, Intel, NEC and Silicon Image. It was the task of the DDWG to pour the new transfer standard into a binding specification which covers all relevant elements for the signal transfer, beginning with the change to the graphics card and stretching to the decoding in the monitor, and naturally including the plug, cable, protocols and key electrical data of the signal. In April 1999, this specification for the Digital Visual Interface was published and it has been valid since then, under the abbreviation DVI.
In order to create a fundamental understanding of what actually happens during the measurements, it is necessary to present at least some of these elements in more detail.
T.M.D.S. and 8b/10b coding
T.M.D.S, developed by Silicon Image, is used as a protocol for data transfer. Here, the abbreviation TMDS stands for Transition Minimized Differential Signalling, a special form of differential signal transfer which uses 8b/10b coding and should allow for especially stable data transfers via cables which can be acquired inexpensively.
Next comes a brief explanation of what 8b/10b coding is and why 10 bits are generated from 8 Bit reference data, which gives rise to a higher bandwidth requirement and thus seems to be counterproductive.
For each single pixel, 24 bits are produced in parallel by the graphics card, consisting of three 8-bit blocks, which carry the colour information for each of the three primary colours. Then, the parallel 8-bit blocks are serialised and thereby the number of transitions which occur is monitored. Each of these high-frequency bit changes from 0 to 1 or 1 to 0 leads to electromagnetic radiation and this to cross talk between individual connections, i.e. the induction of interference signals or EMI for short. Thus, minimising the number of bit changes leads to lowered radiation and thereby higher signal quality, which allows for higher bandwidths and cable lengths.
Figure 0: Generation, conversion, transfer and reconversion of the image signals.
As an example, we selected 8-Bit colour information for red: 01010101. Thus, there are obviously bit changes. The TMDS algorithm transforms every bit using XOR or XNOR, which means that fewer changes are left over. "01010101" is changed to "00110011". In addition, a ninth bit is added which shows that a transformation has been carried out. This ninth bit is also referred to as "encoding bit". The aforementioned 8-bit sequence of seven bit transformations is changed into the 9-bit series: 001100111 and thus demonstrates a remainder of just three bit transformations.
The tenth bit is intended for the containment of an additional problem that arises when too many objects of the same time are being transmitted: the build-up of electrical potential. In order to prevent this, for long, monotonous bit sequences, e.g. if only the object "1" is to be transmitted, the first eight bits of the 10-bit word are occasionally inverted so that the potential in the temporal means is neutral.
What we have described above is easier to reproduce using an example:
The following bit sequence is to be transmitted (without bits nine and ten for reasons of simplification): 11111111111111111111111111
The original 8 bits of a bit word to be transmitted are thus: 11111111.
The number of bit transformations is 0, which means that no minimisation is needed. Thus, the encoding bit is a 0: 111111110.
Since the inverting must be carried out for the second bit word, the last bit is a 1 and the first eight are inverted. The 10-bit wide result for this originally 8-bit wide word is therefore: 0000000001 and thus, the aforementioned result, for simplification, bits nine and ten are removed: 111111110000000011111111.
From this process, 460 valid 10-bit combinations arise for conventional 8-bit colour display, since many of the 256 possible gradients can be represented by two valid values, but others can only be represented by one.
Four 10-bit combinations are used to allow clear markers for the horizontal synchronisation, or H-Sync for short, and vertical synchronisation, or V-Sync for short, to be embedded in the signal so that a new synchronisation and thereby also the refresh rate can also be restored in the case of loss of synchronisation or a signal interruption.
© 2002 - 2017 PRAD ProAdviser GmbH & Co. KG | All rights reserved! | Imprint | Privacy Notice
Registered trademarks and brand names are property of their respective owners.
PRAD assumes no liability for the content of external pages that we link to!
|General information: Our website is not partially displayed correctly with adblocker enabled!|