by acl » Sun Feb 09, 2025 6:28 am
A timing corrector for 60/120 line mirror screw
Karen Orton
Many members are familiar with my 'NBTV timing corrector' - a device which counts lines and imposes pre-defined delays on each line so that some image artefacts caused by mechanical errors are cancelled. The resulting improvement can be striking. For some time now I have been considering a timing corrector for higher line rates and bandwidths, specifically, to correct for small mechanical errors in Steve Ostler's mirror screw monitor. This is a 25 frames per second 60 line screw with the capability of 120 lines. The bandwidth requirement for these standards far exceeds the capabilities of my usual PIC-based approaches.
At one point I considered using a DSP (TMS320F2808). In the end however, I decided upon a combination of hardware for implementation of controlled delays, and a PIC for supervision and to manage the various interfaces. It is perhaps not the best decision - the chip count for the project went into double figures - something I have come to regard as a bad sign. However, much of the complexity is simply down to the need for three channels of timing correction (Steve's monitor is a colour system). The instrument I eventually built uses two 100x160mm circuit boards: a horribly complicated board (HCB) bearing analogue-to-digital converters (ADCs), first-in-first-out (FIFO) memories, digital-to-analogue conversion (DAC); and a much simpler PIC board (SPB).
The ADCs are of the 'flash' type and have a resolution of only six bits, which is okay provided a gamma close to 0.5 is used. The existing mirror screw monitor uses a gamma of 1.0 throughout, and this presented a slight problem. Fortunately, the Aurora World Converter can be set to output a gamma of 0.45 thereby presenting a signal suitable for direct input to the ADCs. Restoration to a gamma of 1.0 must then be performed on the reconstructed outputs, and for this I used a piecewise linear circuit based on diodes. This gave very satisfactory results, despite being discrete and having no adjustable components. I am doubtful however that this circuit would work so well over the full range of component spread. Ideally, a third board carrying analogue input and output processing would have been included but there was simply no more room left in the case! Such a board could use high speed op amps to mitigate the effects of component spread. More importantly, more attention to analogue processing might avoid the situation where the input gamma and output gamma are different. In respect of the latter, I regard the current design as a compromise.
Actual timing correction is carried out using FIFO memories. The PIC has the ability to suspend input to the memories from the ADCs, and to suspend output to the DACs (which in the process, will be forced to black level). This control is sufficient to time shift individual scan lines. In addition, the PIC must charge each colour channel's FIFOs with a 'head' of samples at the start of each frame, to reflect the spacing of the line-of-lights. Other duties of the PIC are to monitor button presses for setting adjustment, and to sense the setting of three potentiometers using the PIC's internal low speed ADC.
Actual adjustment is performed by a hand controller on a lead. This plugs into the font panel and allows the instrument to be sited some distance from the viewing position. While editing, it is arranged that the currently selected scan line flashes so as to make its identification easy. An internal pattern generator furnishes a 625 line compatible signal bearing a vertical bar. The intention is that this signal is fed into the Aurora World Converter so that the bar is displayed on the mirror screw monitor. Adjustment can then proceed with the aim of getting the right side of this vertical bar (which is almost central in the image) free of jaggedness. At my suggestion, Steve built a new line-of-light to use three separate rows of red green and blue LEDs rather than his previous colour mixing arrangement. Colour convergence is now achieved using small additional fixed delays to individual colour channels. These delays are specified by front panel potentiometers as opposed to an internal digital representation. This is to allow easy adjustment of the convergence following minor repositioning of the line-of-light, or following set-up at a new venue.
Klaas's IC610 was used in the 625 line pattern generator. It proved easy to find a signal (that on pin 10) which could be used to generate a vertical bar. The 2.5MHz clock required by this part is supplied by the PIC. Coincidentally, the resolution of the timing corrector is based on this frequency, leading to timing steps of 400 nanoseconds. Such steps represent: 13 seconds of arc of screw rotation; 0.06% of a scan line, or 5% of a square pixel* at 60 lines; and 0.12% of a scan line, or 20% of a square pixel* at 120 lines. The 2.5MHz sampling rate allows for a video bandwidth of 800kHz, which is wider than that of either standard by a good margin.
An internal serial port was a late addition to the circuit. This is only a matter of three components and may permit reading or writing of the settings by a computer. Future firmware enhancements may include auto-repeat on the adjustment buttons (it currently takes eighty presses to move between the adjustment limits). Also to be added is a means to shift all settings along by one line (in either direction). This latter feature is in anticipation of some small alteration to a system, resulting in the timing corrector changing its 'opinion' on which line is line zero. Such a change would instantly invalidate all of the settings - a situation that can be almost completely recovered using the mentioned shift facility.
* Based on the assumption of 80 square pixels per line at 60 lines, and 160 square pixels per line at 120 lines