I was hoping to avoid making another post on this topic, but given sosolar’s most recent posts, I feel it’s necessary to summarize and clarify the statements and claims made.
Sosolar surfaced in the Massdrop RDAC discussion at the beginning of the week, claiming the RDAC was a copy of his Hibiki project. A lot of people got excited by this, maybe not understanding all the details, but seeing the layouts of the PCBs looked the same, and accepting that as proof of sosolar’s claims.
Seeing that post, we (Massdrop) were surprised to see the similarities as well. We were unfamiliar with the hibiki DAC, and we didn’t play a role in the development of that top PCB, so we reached out to William at Airist. He let us know he wasn’t familiar with the project either, and was reaching out to his engineering manager to get more information (thus the post from William explaining reference designs and saying he was unfamiliar with the hibiki DAC).
At this stage, we needed to get details from the engineering manager and the contract engineers he hired to work on this project. The layouts looked the same as the RDAC, but the arrangement of some parts was different, and the spacing between parts was also different. That’s all we knew at the time. Sosolar continued to feed the discussion, making further claims, and posting details of his project on various discussion boards, talking about releasing a product and gathering feedback. At this point, I had already provided my contact information as a direct response to his comment on our discussion, and received no response.
While this was happening, we were able to speak with the engineering manager from Airist and his contract engineer. The engineer explained his process; initially he was provided with a set of specifications, a mandate for a ladder arrangement that would meet the specs of William’s overall design (the top PCB represents ~25% of the overall RDAC design). Given that mandate, he began searching for reference material, and discovered this reference design posted for free use (says that on the post) in 2014:
http://bbs.hifidiy.net/forum.php?mod=vi ... tid=994769
From the basis of this reference design, he looked for examples of layouts that could provide a good baseline, while allowing him to build the underlying structure to serve the purpose of William’s design. Important note here, PCBs have multiple layers, the top board of the RDAC has four. The top layer is where components are positioned, but the content of the other layers determine how those components interact.
At this point, the engineer found the Hibiki DAC, and while the same configuration of parts wouldn’t yield the desired result, the general layout would work for this project, so he used it as a reference and creating the RDAC top PCB layout. This is why it looks the same, and why we subsequently posted as much in our discussion. My wording could have been more direct, but it’s not an exact copy, so “inspired by” seemed like the best option, but ultimately it was more confusing than clarifying. The position of parts is nearly the same, but sosolar continued to post, accusing Airist’s team of reverse engineering the board, which isn’t the case.
From there, still having received no contact from sosolar, I reached out to the email address associated with his Massdrop account, explaining that I’d like to understand exactly what he’s saying, and to understand more about the Hibiki project in general. Now I want to pause here, and make clear an important understanding in professional communication. Emails are intended for the specified recipients, and it’s bad form to break that trust. I won’t be posting screenshots or exact wording from his message, but following sosolar’s most recent posts I feel it’s important to share some of the general themes in our communication.
He responded to my messaging with a development timeline for the Hibiki DAC, starting in June 2015 and ending in July 2016, and demanded that we post a comparable timeline for RDAC development. His timeline had great detail, with lots of links to posts and things of that nature, which makes sense given the Hibiki was developed publicly, with sosolar sharing details with the hifidiy.net community. It’d be nice if posting a timeline would answer the questions at hand, but the RDAC development was private, there are no public posts to verify the statement. Given the many responses from people saying they didn’t believe our previous statements, I didn’t think an unverifiable timeline would answer questions for anybody.
The most important part of Sosolar’s response, was around some parts in the design he said were unnecessary, essentially saying they were stylistic choices on his part, that wouldn’t exist on another board unless the functionality was being copied through a reverse engineering service.
All of these parts which are unnecessary for the HIbiki, but are critical to the function of the RDAC. We answered all of Sosolar’s questions directly, here is the explanation we provided for each of the sections sosolar called out:
For the op amps:
The op amps on the RDAC top board are not just buffers, they are necessary for the operation of a sign-magnitude ladder. The output from a network of so many resistors is very weak, so we have to bring that signal strength up or else further processing will drown it in noise. These op amps do that amplifying, merge the outputs from the positive and negative ladders into one output waveform, and filter out the switching artifacts.
For the large transistors:
Because the R-2R ladder is basically a 48 resistor network on each side, you need a fair amount of current to actually drive the whole thing, not to mention extra overhead.
For the logic switches:
These are the same switches used in the open source schematic posted in 2014, and that is how the engineer picked this part.
Sosolar’s response to this message said that he was waiting for us to post the timeline, followed by a sentence that I think was trying to suggest he wasn’t accusing Airist of reverse engineering and copying the DAC design, but the layout alone (not trying to knock his english, I don’t write a character of chinese, but I couldn’t understand his wording) which would be a change from his previous statements, but I can’t be sure. Hopefully we can talk about this on the phone where mandarin/cantonese speaking team members can remove the language barrier in this technical discussion.
I responded to his message early this morning, hoping to catch him still awake so we could have a call, and in that email I proposed a solution that would give clarity to the community. We could both provide our gerber files to an unbiased third party (I suggested the new editor of IF, guy has no skin in this, and I’ve been asking folks for his contact info today in an effort to reach out and see if he could do this), they can examine the files and tell the community if the RDAC top pcb is a copy of the Hibiki pcb. I haven’t received a response from sosolar, but it’ll be morning in China soon, so hopefully sosolar is game for what seems like a pretty amenable situation.
TL;DR
Overall the layout of parts on the hibiki DAC and the top board of the RDAC are nearly the same. Neither Massdrop or William Tse of Airist was aware of the Hibiki DAC before sosolar’s post on Massdrop. The layout is not 1:1, but it’s clearly close enough where folks feel it’s essentially 1:1. We’ve explained how the hibiki layout was used by Airist’s contract engineer, and hats off to sosolar for creating a layout that works for the RDAC’s execution needs. If he had interest in compensation for this, he has plenty of avenues available to express that, as it stands, it seems he’s more interested in posting about the balanced hibiki he’s releasing this August. The similarities end at that top layer, and we’re glad to provide RDAC files for verification by an unbiased third party.
We’re doing everything we can to resolve this amicably, and we’ll keep trying, hopefully sosolar will engage with us.