On improving the video quality from the HA-8-3 color graphics card
Terry Smedley
Recently, Glenn and Les have expressed disappointment at the quality of graphic images produced by the HA-8-3 color graphics card. While prototyping a clone of the HA-8-3, I observed the same image quality issues – severe digital artifacts on edges, smeared colors, and static images that oscillated between “OK” and “bad”.
The TMS9918A video chip that’s used in the HA-8-3 was also used in several commercially available game consoles or computers – ColecoVision, TI-99/4, Tomy Tutor and others. I found volumes of postings about the video quality of those devices, along with some interesting ideas on how to improve the display.
To summarize what I learned about the TMS9900 family:
Composite video produces crappy images. Period.
The upscaling processor used in many LCD displays makes the situation worse by interpreting the 240 line progressive scan frame from the TMS9918A as an interlaced 480 line frame and incorrectly applying de-interlacing algorithms to it. The unnecessary de-interlacing can produce the oscillating images that have been observed with the HA-8-3.
External boxes are available to replace the display’s upscaler with a scan doubler. The doubler simply repeats each scan line to form a 480p image that can be displayed without any of the de-interlacing artifacts.
The TMS9928 chip outputs a component video signal (YPbPr) that offers much improved image quality compared to the composite video output of the TMS9918 used in the HA-8-3. The two chips execute the same software instructions and are pin-compatible with each other excepting three video output pins.
There is a very interesting FPGA project that replaces the TMS9918A chip pin for pin. It directly produces VGA output running the same instruction set as the TMS9918A. It is, unfortunately, not available now, although a second generation of the FPGA may be available “soon”.
Based on those findings, I am starting down this path with the HA-8-3 clone prototype:
1)
I should have hands on a composite video scan
doubler early next week. Maybe that
makes the display “good enough” without any further design work. https://www.retrotink.com/product-page/2x-mini
2)
TMS9928 chips are on their way. The component video output from the TMS9928
needs buffer circuits to drive a 75 ohm display. TI published a document describing in detail
the circuits required. I’ll add the
buffer circuits onto the clone prototype and see if the component video
solution produces enough improvement to warrant the additional circuit complexity.
http://spatula-city.org/~im14u2c/vdp-99xx/e3/SPPA004A_9928-29_9118-28_Interface_to_Monitors.pdf
That’s a start.
Here is a fascinating look at the development of the FPGA replacement for the TMS9918:
https://www.eetimes.com/creating-the-f18a-an-fpga-based-tms9918a-vdp/#
Here is a site that discusses modifications to many game consoles, some based on the TMS9918A, to produce high quality RGB video (not the same as component video). Glenn mentioned he had seen some other vintage computers being displayed with the OSSC scan converter that uses RGB inputs to produce very high quality images.
https://videogameperfection.com/products/open-source-scan-converter/
Glenn Roberts
On Sep 16, 2021, at 4:59 AM, Terry Smedley <terry....@gmail.com> wrote:
--
You received this message because you are subscribed to the Google Groups "SEBHC" group.
To unsubscribe from this group and stop receiving emails from it, send an email to sebhc+un...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/sebhc/7010c1da-3ce2-4637-bddd-c06aa68d9f80n%40googlegroups.com.
