GBS-8200 and GBS-8220 experiments part 1
GBS-82XX experiments part 2 now posted here (26th Jan 2015)
You may have seen on ebay and quite a few arcade accessory sites, a number of “GBS-8220 RGB / CGA / EGA / YUV to VGA Arcade Games HD Video Converter Boards”, which for the low cost promise a cheap solution to connecting a modern LCD monitor to retro consoles and video arcade systems. Based on user experiences posted on various message boards, the device does not always deliver. Various issues have been reported, from a complete lack of signal to poor video to speckles on the display and problems with moving video.
I was initially drawn into the investigation after answering a few questions on display issues on the EAB forum, http://eab.abime.net/showthread.php?t=75451 and subsequent conversations in another topic, http://eab.abime.net/showthread.php?t=66922, both threads are quite long. Since 2006 I have been involved with developing video systems for my day job and have gained some valuable experience in interfacing ‘non-standard’ video to modern displays. With this in mind I ordered a GBS-8200 board from Ebay and waited.
There are numerous forums posts that state to get the best out of this board you need a 5V high current, good quality, power supply and you should replace the electrolytic capacitors with higher quality ones. Being an Electronic Engineer with 16 years experience, I was dubious about this and set about investigating the board. A list of 21 items to test was created and depending on the results, I hoped to improve the basic video connections and obtain reliable, proper colour video images!
Whilst researching the devices on the board, I came across a thread the the SHMUPS (an abbreviation of Shoot-em-ups) board who has played with the software, http://shmups.system11.org/viewtopic.php?f=6&t=52172, this provided a link to the programming guide, which I did not have. I did have the TVIA-5725 datasheet. Playing with the software settings is for the next post on this topic, this one deals with the hardware.
There are two versions of the board available and two similar models. The GBS-8220 is a dual output board, the GBS-8200, which I have, has a single VGA output. The GBS-8220 is typically a V3 board, the GBS-8200 a V4 board. Easiest to show with some photos:
The highlighted areas show a crucial difference, the V4, GBS-8200 board, has a switchmode power converter, whereas the V3 board has a linear regulator on the the V4 GBS-8200 board only has 1 x VGA output! 😉 throughout this article, I will refer to the two boards collectively as the GBS-82XX.
My test setup
I use the cable supplied with the GBS-82XX board with the following connections:
Amiga signal Amiga pin Wire GBS marking
Red Video 3 Red R
Green Video 4 Green G
Blue Video 5 Blue B
Ground 16 Black GND
CSYNC 10 Yellow S (via 680 ohm resistor)
Do not use the YPbPr input, it will give you incorrect colours and you can not adjust the on-screen video settings as easily.
The following items were tested ‘in the lab’:
- Measure the +5V current using current probe and DMM
Measured 0.4A constant.
- Measure the on-board 3.3V supply ripple, FFT DC to 200 MHz.
Around 50mV from the 3.3V switcher. Noise TBD
Around 20mV from the LDO.
TI app note slva115 LDO regulator stability.
- Feed in 3.3V LVTTL sync signal, see what happens, does noise go away. No effect.
- Change 1.8V LDO regulator output capacitor from 22uF ceramic to Tantalum. This worked well, device no longer susceptible to susceptibility test.
- AC couple video, 220uF capacitors before termination/potentiometers.
Possible improvement, will test with a game
- Test LM1881 see what happens.
Note outputs are not compatible with GBS82xx. 4.3V+ for logic 1 exceeds 3.6V max for Vih. Will be Ok with 680R resistor.
- Tweak LM1881 Rset resistor, see what it does. Will not do anything on the Amiga video.
- Try CGA from the Amiga.
CGA works, there is a distinct left/right shift. Needs reduced sync amplitude, 680R from CSYNC of Amiga to GBS board.
- Feed in composite PAL video signal to Y of YPbPR input, is the image stable?
Yes. Colourburst affects picture as expected, sparklies on screen. It has issues with black level clamping, most likely due to the colourburst reference.
- Feed CVBS into Sync input. CVBS works, providing video is un-terminated.
- Check for 160 ohm resistor by pin 155, IREF generation.
- Check for filter circuit on PLL, 3K R + 33n and 3.3nF
There is a lot here, let me summarise some of the results:
- You do not need a high current power supply. The GBS-8200 board drew a maximum of 0.4A, this was measured with a DMM and a current monitor circuit, which provided current traces on an oscilloscope.
- The power supplies and noise levels were acceptable. Proper probing techniques together with a FFT plot were used to look at the onboard power supply noise on the 3.3V and 1.8V supplies. There was nothing unexpected, with FFT plots upto 200 MHz. There is absolutely no need to replace all the electrolytic capacitors on the board as they remove noise effectively.
- The 1.8V supply on the V4 board I have was unstable. A ceramic capacitor was used on the output of a linear regulator, one which was not stable (AMS1117) with this capacitor type. See http://www.ti.com/litv/pdf/slva115 for more details why. I replaced the capacitor with a 22uF tantalum bead. Prior to replacement, switching on/off my soldering iron transformer, caused the GBS-82XX board to glitch, now it does not. A simple but effective EMC susceptibility test, which this board failed. The original capacitor was a 10uF part with an ESR of 0.02 ohms, much too low for the regulator. Must check this on a V3 board.
- Do not connect the sync output of an Amiga or a LM1881 sync stripper direct to a GBS-82XX board. Both devices have a logic 1 level of >4.6V! The maximum input voltage for the TVIA-5725 device is 3.6V. If you exceed the absolute maximums, you start activating the ESD protection diodes on each input, never a good idea. The excess voltage is clamped to the device internal supply by these diodes, so the 3.3V internal (to the TVIA-5725 device in this instance) gets spikes every time the input is exceeded, pulling up it’s supply.This does affect the normal operation of the device and reduces reliability. Simply connecting the CSYNC output of an Amiga or the Composite sync output of an LM1881, to the GBS-82XX board, via a 680 ohm series resistor, removes this problem. I measured a signal amplitude of 3V.
Observations on the GBS-82XX boards
Whilst testing this board, I did make a few observations, which will focus future work:
- There are some scan-conversion artefacts, when connecting a 50Hz source to a 60 Hz monitor. This was most noticeable as a band of noise slowly moving up the display when viewed against a grey background. Changing the Amiga video to 60 Hz NTSC, removed the noise band.
- The de-interlacing method is not ideal, though this may be signal related. This was most noticeable on scene demos.
- Connecting the Amiga (and an Atari system) via the RGB inputs provided a good stable image but had some noise, this might be removable with ADC filtering.
- The TVIA-5725 can cope with a sync-disturbance and some non-standard video. A good sign.
- Using the GBS-82XX menu, reducing the sharpness from the default of 10, to 3 improved the display.
- I used the following video settings, Brightness=10, Contrast=90, Saturation = 50, Sharpness = 3.
I wanted to test something other than the Amiga, to see how well the GBS-82XX coped with another retro system. The Atari 7800 was nearby. My unit has been modified to output S-Video (Y/C). I then connected this to a RVA dev board, which converted the S-Video to YPbPr. The YPbPr video was fed into the GBS-8200 board. It sort of worked. I had trouble with the RVA board outputting black and white video from the Y/C source. I tried tweaking a few settings but this is a task for another time. Feeding in composite video, using a S-Video to composite lead from the Atari 7800 worked. Atari 7800 up-scaled to VGA:
For a 10 minute quick test, it showed the future potential of this board. There were some white speckles on the video though.
Though the RVA scan-doubler, when I get it working, may surpass it 😉
So I have played with the hardware, what’s left to do and what is next?
I want to check the SDRAM interface in more detail to ascertain if this contributes to the noise issues.
Play with the sampling and conversion settings of the TVIA-5725 device.
Experiment with the scan-converter.
Try 60 Hz games on the Amiga, see how they perform in terms of noise.
The I2C port now has a pin header soldered onto it and I observed that it operates at 87.3 KHz and has regular traffic. A Pickit serial, will be used to inject I2C commands for testing, this is no simple task, decoding what to try but will be fun!
Test a V3 board.