Fulton County 2020 Absentee Ballot Results Were Physically Impossible and Files Were Modified
The scanning of ballots in Fulton County has been the subject of tumultuous debate for over a year. Most of the controversy centers around suspect scenes from the late-night State Farm Arena surveillance video and whether the actions captured were right, wrong, proper or…something else.
Aside from the questions that persist about what could be seen in that video, there are new questions about what happened there that couldn’t be seen. The arena was used as the Fulton County ballot processing center and is where all 148,318 absentee ballots were scanned. After the process of voter authentication and signature verification, the ballots were separated from their envelopes and arranged in batches of approximately 100. Each batch was then scanned using one of the five Cannon DR-G2140 high speed machines in which an image of each ballot was simultaneously created and saved to be tabulated later.
As a result of a pending lawsuit, Fulton County produced those scanned absentee ballot image files which were then made public and have been under review by various individuals and groups. One of those concerned citizens, RonC, (who is non-political) found something he couldn’t explain. The image files have timestamps that are recorded as the ballots are being scanned. Those timestamps reveal that the image files were created faster than the physical capacity of the scanner. In many cases, several times faster.
The manual for the Cannon DR-G2140 lists the specifications on page 23:
The manual shows the machine has a maximum capability of scanning up to 140 document sheets per minute, but this rate is based on scanning standard 8 ½ X 11-inch paper. The absentee ballots are almost twice the length at 18 inches long.
Using the State Farm Arena surveillance video, we measured the actual realized scan times which averaged approximately 80 seconds per batch (under perfect conditions with no paper jams or complications). Which equates to approximately 800ms or a little faster than 1 ballot per second. However, the image timestamps show something different. Below is a spreadsheet of sample data that has the timestamp of each ballot image as it was created:
Image data of actual ballots scanned by unit 5162 on 11-05-2020
The timestamps in the red boxes above show 8 ballots scanned in the same second, then followed by 6 ballots in the very next second. Fourteen ballots in 2 seconds!?!? This shows that the ballot images were created at a speed that is physically IMPOSSIBLE.
The same scenario is replete throughout the image files and affects thousands of ballots but does not mean that just those ballots scanned at impossible speeds are the only ones affected. The impossibility simply reveals that the ballot image files were created outside of the natural scanning process. Is it nefarious? It certainly isn’t expected or “normal”.
We’ve talked to several experts about the anomalies. Given the specifics, such as the files being saved on ethernet-connected, high-performance network drives and other specific technical parameters of the arrangement, there remains no clear and justifiable cause. Even if one considers lag-time, buffering or some type of computer processor or network-caused “ebb and flow”, it does not explain the impossible scan times, even when averaging all images in each batch.
There’s more. If we expand the same dataset to include the modified file timestamps, we reveal additional problems:
The only instance in which a ballot image file should be modified is if it has gone through the adjudication process. Adjudication is when the machine cannot determine the voter’s selections for various reasons and the ballot is sent to a bi-partisan team who look at the markings and attempt to interpret the voter’s intent. The team votes and if they agree, the original image is supplemented to reflect the decision.
The modified timestamps shown above reveal that all of the files were modified. Again, the ballot images should not be modified unless processed for adjudication (1 out of the 50). Perhaps there is a reasonable explanation or was caused by a technicality only affecting those few ballots. Considering the images seem to have all been modified at the same time, and within only one hour of scanning, the benefit of the doubt should be extended to election officials. That is, if it was isolated.
If we zoom out to the 30,000-foot level and look at all batches from the same scanner as shown below, we are able to see the scope of a much larger problem. First, all of the batches have been modified as shown on the far right. The scan times for each batch are in yellow and the numbers shown in red is the amount of time (hours) between when the batch was scanned and when the batch was modified.
For example, batch 001 was finished being scanned at 1:51 pm on 10/22. That same batch was modified 192 hours later at 1:38 pm on 10/30. The batch was scanned and then modified 8 DAYS LATER.
On scanner 5150, a total of 44 batches were modified between 3 to 8 days after being scanned. Not only is such a large window of time used to process ballots inexplicable, but it is also unacceptable. Unfortunately, these findings show extreme variations in the process, are wide-spread and cannot be dismissed or ignored.
Important is the last 4 batches shown on the page had 0 hours between the scan time and modification time. In fact, the last 4 batches scanned were modified before the previous 44 batches. This removes the possibility of some type of consistent procedure or forced process causing the modification of all batches on the same date.
Remember, the findings above only show the anomalies associated with one of the five machines used to scan absentee ballots, a process which continued until November 5th. (More to come soon on the significance of the time between scanning and file modification time.)
Attempting to seek understanding and answers for his findings, RonC went to the logical government entity charged with such matters- the Fulton County Board of Registration and Elections (BRE).
RonC summarized his findings in an email that detailed the scan speed and image file modification issues. The BRE then sought answers from the County Elections Director, Rick Barron. The email below shows Mr. Barron’s response to the board:
There are 4 basic takeaways from Mr. Barron’s email.
First, it’s obvious that Mr. Barron feels the same about the impossible scanning speed and went so far as to describe RonC’s finding as “bizarre”. The problem is this was not simply a “claim” as characterized by Mr. Barron, rather a factual observation of Fulton County’s data.
Secondly, Mr. Barron directly states: “You can’t modify ballot images that aren’t adjudicated.”
Again, he affirms our understanding and expectations, yet the Fulton County image files that were not adjudicated are shown to have been modified.
Third, he states that the equipment is proprietary, and the security manuals are not available. In his defense, Mr. Barron was probably referencing the election system as a whole or in a general connotation; however, the scanners are Cannon consumer-off-the-shelf machines that can be purchased through numerous retailers.
Finally, we did follow Mr. Barron’s advice to observe the scanners in operation by watching the State Farm Arena surveillance video. He was correct, the scanners are slow. In fact, they are so slow that the ballot images produced by the scanner are created faster than the machine can scan the ballots, and that truly is “bizarre”.
More to come soon.
“RonC” contributed to this report