Tanner Fokkens responds to Dr. Michael Zyda

In September 2022, hardware engineer and Donkey Kong expert Tanner Fokkens published a 48-page technical analysis, which was featured here at perfectpacman.com. This analysis went into detailed depth, dissecting exactly why Donkey Kong transition screens appear differently on MAME and on original arcade, and why the transitions seen on Billy Mitchell’s contested tapes could not have originated from an original, unmodified Donkey Kong arcade machine.

Recently, an out-of-court settlement was announced, bringing an end to the years-long legal battle between Billy Mitchell and gaming scorekeeper Twin Galaxies. A detailed review of that settlement and its ramifications is forthcoming here on this site. Per the agreement, a brief opinion document from Billy’s paid technical expert, Dr. Michael Zyda, was published. In his piece, Dr. Zyda speculates that the “girder finger” may be caused uncleaned VCR heads, aging components, or signal cross-talk from different components. (Dr. Zyda seems to be under the impression that every reasonable angle hasn’t already been explored by people who are willing to do actual tests on real Donkey Kong hardware.) Dr. Zyda also displays ignorance of basic conditions of the situation, which he also didn’t bother examining.

Dr. Zyda’s lazy theorycrafting has earned rebuke from casual observers and from technical experts alike. Zyda’s assertions have also inspired Tanner Fokkens to author a new rebuttal, addressing each of the points Zyda raises. Fokkens’ new analysis, which is reposted below, can be found at this link:

https://docs.google.com/document/d/14dvJrH_LYOXVa3sHpa-ZMQGLSrgNn7MiDVgLhNFQR8g

Also for more background on the totality of the evidence against Billy Mitchell’s tapes, see this site’s evidence compendium, which will be updated in the near future:

https://perfectpacman.com/evidence/

---

Tanner Fokkens
In Response to Dr. Zyda’s Technical Analysis

Recently, Billy Mitchell’s lawsuit with Twin Galaxies was settled. There has been a lot of discussion involving the circumstances and what was said in the statement, but I’d like to sidestep that for this update. In the statement, there was an additional technical analysis by Dr. Zyda, a Professor Emeritus at USC (University of Southern California), which Twin Galaxies said it could acknowledge, but not endorse or agree with. I would like to go point-by-point through his analysis, seen here: https://mikezyda.com/resources/Press/2024.01.15-ZydaExpertReportFinal.pdf

a.) When Billy is recording from the arcade machine running at 60 fps (frames per second), he plugs that 60 fps stream into a Two Bit Converter box (the converter) and the converter outputs a 30 fps stream to a VHS VCR for recording. There is no apparent sync signal from the 60 fps device to the converter box outputting a 30 fps stream. The purpose of the sync signal would be to make sure there is some kind of frame alignment. Without the sync signal and frame alignment, we actually don’t know what the converter is really doing. If we had frame alignment, we would expect that two frames from the 60 fps stream would be used to create one frame in some fashion. If we don’t have perfect frame alignment, then we are going to see visual artifacts such as the anomalies at issue. And all of the analyses of the Billy Mitchell tapes rest on anomalies recorded in this fashion.

• Point a is based on the false idea that there is no sync going to the converter board; there is. The following is the pinout of the Two Bit score converter board, provided to David Race from Billy Mitchell (pin 5 in our wiring, “composite negative synch” is difficult to read):

Donkey Kong PCB schematic, with connector P7 showing sync on pin 6:

• If there was no sync, video would not be able to sync at all, on any television or monitor. As explored in my paper, the vertical and horizontal sync pulses are needed for the monitor to know when the start of each frame and line is located. Additionally, given that the video output from the two-bit score converter is not interlaced (see here: https://www.youtube.com/watch?v=XKbkvqrzL5k), we know that the output has to have remained 240p 60fps. The change that Dr. Zyda is suggesting would require a framebuffer. Examining the block diagram for the converter IC (Integrated circuit) most likely on the one on the two bit score converter, we see it takes sync and does not have a frame buffer block:

Here is the block diagram of the encoder IC on the Two Bit score converter. Although the markings of the IC are scratched out in the photo, we can work out that the CXA2075 is the IC that is used on the device by comparing the functions of the IC to the circuit of the device and seeing that they match up exactly. (Per http://nedopc.com/catalog/CXA2075M.pdf)

Above, we see the Two bit score converter outputting to a CRT monitor. This proves that it meets the standards of NTSC video, meaning 60fps 240p. (Per https://www.twingalaxies.com/feed_details.php/875/twin-galaxies-explores-capturing-equipment-used-in-billy-mitchell-score)

b.) Another issue not well addressed is that of component aging. The Nintendo Donkey Kong boards are old and their various electronic parts are aging at different rates. As those components near failure, the potential for visual artifacts increases. This means there is no way to compare different Donkey Kong boards as their component failures most likely are in different parts. This includes the power supplies that power the Donkey Kong boards. As those power supplies age, we have no way of measuring how that aging will impact the Donkey Kong board’s operations and visual display. As such, there may not be any way to compare recordings from the same board done at different times. In my opinion component aging could produce the anomalies at issue.

• No component is suggested for what could cause such a failure. This is an untestable statement without providing a specific component – an unfalsifiable fallacy. Such a statement is akin to saying “the sun could explode tomorrow, we have no way of knowing”. A hypothesis is not evidence. Such a component failure, which would have to cause the scanning mechanism to go from the line-by-line scanning we have in arcade to the “snapshot” behavior seen in MAME, would certainly violate the NTSC video standards and result in the video for the player not working anymore. This is why even if components did fail in this highly implausible way, it would make it impossible for the player to even carry out a game. Therefore, the argument for component failure is invalid.
• Additionally, there are some important timelines on the board that make this hypothesis highly improbable:
  • 2004: Billy submits 1.047M score, supposedly using his board. MAME transitions shown on-tape.
  • 2007: Billy submits 1.05M score, in which the board he supposedly used was verified by “a Nintendo engineer” to ensure the board was in full working order. MAME transitions shown on-tape.
  • March/April 2018: Billy gives Carlos Pineiro the “Enzo machine” used in the first two scores to test at Arcade Game Sales. No MAME transitions are observed.
  • July to September 2018: Billy sends the board supposedly used in these scores to David Race (https://youtu.be/gxDNFeI2zS8?si=SSv3eOTfGbY6bNLB&t=299, June 25, 2018. See the appendix at the end of this letter as well). No MAME transitions are observed.
• If there were somehow components causing incorrect transitions in 2004 and 2007 due to aging, why were these anomalies not seen in 2018? The only thing we can conclude is that this hardware could not have been used to generate Billy’s tapes because it definitively generates video that matches arcade and never once displayed any signatures like the ones generated by MAME. Additionally, as a paid expert, Zyda too could have had access to the equipment to do testing on Billy’s behalf. This did not happen.

From this frame, we see the Two Bit Score converter has no issues outputting arcade transitions, with Billy’s exact board. (https://www.youtube.com/watch?v=XKbkvqrzL5k, per Carlos Pineiro)

c.) Another issue brought to mind is that of copies of copies of videotapes. As videotapes are copied and handed out to someone who then makes copies, visual artifacts often appear as second and third generation and beyond copies are made. It is important to note that it is unknown how many generations of copies were created to arrive at the current videotapes. Another issue with respect to this is whether the videotape heads were cleaned before the copies were made, including whether Billy Mitchell was cleaning his VCR’s heads before he recorded the Donkey Kong play. It is my understanding that he was not. Further, it is my understanding that Billy Mitchell recorded his tapes at long play (LP) to save on the cost of videotape. LP recordings are lower quality recordings and often record with much on tape visible noise. It is also my understanding that Billy Mitchell reuses videotapes without re-blacking the tapes beforehand. With dirty record heads, this often means that previous recording information will bleed through and be visible. In my opinion, this could produce the anomalies at issue.

• Point c says VCR artifacts could cause the MAME signatures. Nobody has ever been able to use VCR artifacts on arcade to show MAME transitions. David Race tested this extensively. Once again, this is an untested hypothesis and conjecture. Veritas did an extensive video on this topic, at one point trying to remove a video recording of Steve Wiebe’s previous world record many generations from the original video tape. This can be found here: https://youtu.be/PySD_FiEl5c?si=N13S2bj6rFZlUZlN&t=3266 (timestamped). Even with all these extensive steps to make a transition that doesn’t look like arcade, he was unsuccessful in doing so. This situation is a far worse case than the steps Billy’s tapes had undergone, so if this did not cause MAME transitions, it is unlikely we would ever see such a phenomenon occur.

d.) The wiring of test systems also seems to have the potential for causing visual anomalies. Questions like, “do the video carrying cables drape over or near a power supply?” and “are the cables carrying the video shielded cables or unshielded cables?” do not seem well answered. These scenarios could also produce the anomalies at issue.

• Point d says power supply crosstalk could cause MAME signatures. For the record, I am a signal integrity engineer, I work with crosstalk all day. The scenario Zyda describes could not have produced the anomalies he cites. The absurd odds of getting the exact same interference on the exact frame across many years and different setups is fantasy. The magnitude of crosstalk to actually alter the image would be absurdly high. We are talking about *volts* of received coupled crosstalk, that have coupled at the exact same frame with the exact correct signal needed to alter the NTSC video for *years*. This simply did not happen.

NTSC Color burst signal. IRE corresponds to brightness, the phase of the waveform corresponds to the color. Interference would have to perfectly modify this signal’s phase and amplitude to cause MAME signatures, and destructively interfere such that it removed sections of the girder from the arcade rendering.

e.) Component aging with respect to the Two Bit Converter is also a large issue. Recently, Neil Hernandez obtained a Two Bit Converter from Billy Mitchell’s garage in Florida and attempted to get it working with a Donkey Kong machine. It is my understanding that the converter had been sitting in Billy Mitchell’s un-airconditioned garage for many years. The initial connection showed completely wrong colors so Neil decided to see if he could repair the converter. Neil’s instincts as an arcade game technician told him he should start with the capacitors. Neil changed out the capacitors in the converter and as a result had an almost perfect picture output with proper colors. With replacing just the capacitors, Neil saw the “girder finger” on the monitor receiving the output from the rebuilt converter. Neil took one still picture of that girder. A copy of that picture is attached. Neil then replaced the old resistors on the converter board. When he completed that replacement of the resistors, the “girder finger” was no longer present. So, component failure from the old resistors produced the “girder finger” in the converter in this case. Neil hypothesizes that the decaying resistors were outputting the wrong voltage and the new resistors fixed that. In my opinion this is a reasonable hypothesis.

I asked Neil Hernandez to rebuild the converter with the old resistors to see if we could get a consistent “girder finger” output from that converter to the connected monitor. That experiment failed likely because the combination of heat from the soldering gun and the fragility of the resistors finally made them completely stop working, with the converter failing to work at all after that experiment.

However, Mr. Hernandez’s experiment, which I saw recordings of, depict numerous anomalies, including different numbers of girder fingers, Mario teleporting around the screen, barrels rolling the wrong directions including uphill, pies disappearing, and the hammer not vanishing when it is dropped. These resulting anomalies are consistent with my expert opinion that the game play on the subject video tapes, including the anomalies depicted, could be from original Donkey Kong hardware, and that the anomalies are the result of component degradation in the Two Bit converter.

• Below we see the frame from Neil Hernandez that Zyda is referring to:

• It is clear what we are seeing in this frame. We have the rolling shutter effect present, given the multiple frames etched together on the camera’s sensor. As for why we see the “3 girder pattern” (not really, it’s still wrong), it is because Neil seems to have modified the PCB to draw the video upside down [See section VIIa. “Flipped Orientation Effects on the Transition”, from my original paper.] This is disqualified by Twin Galaxies rules as it is a modification. The method for doing this modification is seen here: http://www.brasington.org/arcade/tech/flip/dk.shtml. Essentially, the video hardware is reading the data from VRAM in reverse order. Again, I will attach Figure 24 from my paper:

• We clearly see the 3 girders in the photo are from the 4 girder arcade cocktail mode drawing pattern, or that which results in removing a jumper on the board to get it to draw in reverse. The girders that are blocked out at the bottom in the photo are blocked due to the rolling shutter effect:

• It is worth noting that the rolling shutter effect is not even present in Billy’s tapes, because it is a direct feed game (whether it be MAME or Arcade). This is a product of a video camera capturing light from a monitor, which is not synced to the V-sync and H-sync pulses from the video feed, resulting in these meshed together frames. Essentially, this finding from Neil/Zyda is completely irrelevant to Billy’s tapes, because it is not a situation that was even possible in his recordings.
• The gameplay anomalies mentioned from Neil to Zyda seem to have been the result of altering power supply voltage to a very low level, where you get game-breaking behavior. However, still no MAME transitions. I have a game where I achieved 20M points on Donkey Kong due to a power issue, and I still saw arcade transitions (https://www.twitch.tv/videos/469199132). This situation is also irrelevant as there are no Arcade transitions that look like MAME presented as a result of this power supply experiment.

The final statement in the analysis, “However, it is clear that the video tapes could depict game play on original Donkey Kong hardware despite the anomalies depicted.”, is demonstrably false given that everything that has been said in this analysis is false. Dr. Zyda did not address any of the key points of the dispute, or the fact that Billy’s 3 tapes never show arcade transitions at any point. He also seemingly does not reference or acknowledge other key signatures that are present in arcade, like the “rivet ramp” when Jumpman clears the rivet stage.

Even if the artifacts he described did somehow cause MAME transitions, the highly unlikely chance of these artifacts causing a MAME transition would surely mean arcade transitions would happen occasionally. 

An example of one of these high-unlikely ideas was the hypothesis of frame-perfect crosstalk of the perfect frequency and amplitude. If this hypothesis is correct, why were these girder artifacts only seen during the transition, and not other parts of the gameplay? Billy’s gameplay tapes only had girder irregularities during the transition, matching exactly to a period version of MAME running at a vertical refresh rate of 60.60606 Hz.

Given the outlined issues in Dr. Zyda’s technical analysis, I believe he should retract his statement and conclusions.

(Original paper: https://perfectpacman.com/2022/09/06/new-technical-analysis/)

Appendix – Texts between Billy Mitchell and David Race, discussing sending David Race the board supposedly used in the 1.047M and 1.05M scores 

(From court filings, pp.121-130 https://drive.google.com/file/d/1RsgdOpRnh-yPcKjjveiLOxTchFl0oW4R/view?fbclid=IwAR2Ag6OpD2CtP-uoztlUEXfuVO3VhYbgOU2jGo1AxB5hhq_cLmsrWLMsL2o, from searching CACE21007130 on https://www.browardclerk.org/Web2/CaseSearchECA/, filing from 01/13/2022, 47 pages):