[Answering your questions in order, there will be some redundancy/overlap…]:
The transition between the zero marker on the end of the last chunk and the one marker on the beginning of the next (current) chunk means that we know what to expect for about two bit times. Ideally that transition is exactly lined up with the upgoing clock transitions. If it is lined up exactly then it will be probabilistically one or zero depending on fuzz/chaos/noise whatever. If the down going and upgoing clocks are a little out of phase, then either the zero on the end of each chunk will be seen over and over or the one at the beginning each chunk will be seen over and over.
I don’t need biphase mark since the two clocks I’m dealing with are very close in frequency and phase. The deserializer uses the local clock to sample in the middle of each down coming bit. That’s reliable if the two clocks don’t drift relative to each other more than a half of a bit in each 10 bit chunk. Not a hard goal to meet. For more details here’s a link to a similar pair of SERDES chips. SN65LV1023A data sheet, product information and support | TI.com
I implement the serializer in the FPGA so I don’t need that chip. But using a canned chip in the analog box will keep the digital noise localized and the signal from the optical receiver to the deserializer is balanced and the DSD is transmitted balanced so there’s not much noise from them.
All that’s needed to keep things aligned is one zero to one transition in a consistent place in the bitstream. For example, a one in bit time one then, say, 8 data bits, then a zero in the last bit time. You can either do an explicit phase alignment signal of (in the above case) 5 ones, then 5 zeros. After the sync up happens then the middle eight bits can be anything and the lock will be kept. In my case the DSD is random enough that I just put a DSD bit after the initial one and another before the final zero so if things aren’t lined up correctly the lock will bounce around until the constant pattern of … zero, one … is seen every ten bits.
The optical signal is at 112.896MHz, close enough to 100Mbit Ethernet that parts designed to transmit Ethernet are perfect (100Mbit Ethernet uses a 125MHz signal so there’s room for error checking, framing, etc.)
Got it, thanks! Though my mind is a little blown by the last part about the frame alignment just finding itself. Will keep pondering that one
Edit: Oh!! If the bit being sampled is random (like DSD) then the digital box clock will stay exactly as it is and the alignment will drift towards the boundary bits, at which point they become non-random and are a self-correcting feedback loop. Did you come up with this yourself or is there prior art? It’s gorgeous.
The key is to make sure that your data never has two bits in a row that are a zero followed by a one. (If they were always zero and one, why are you sending them…) So if things look locked but aren’t then there will be some change in the bits in the presumed zero/one transition. Then you just have to slip one bit and see if that works…
I had already implemented something similar in ECL for the first design when I encountered a SERDES chip that had the right payload size and rate that would work. I’m not usually one to forgo an interesting project, but debugging and fixing ECL running at a such a rate while writing the software to use it seemed like too much fun so I saved a lot of money, development time and product cost by using a canned chip.
I haven’t perused all 970 posts, but even if I had I doubt the answer is still the same as one in 2018, 2019, etc.
Ted, or anybody who has heard this and the DSD Mk II, how does the two-chassis design change the sound qualities vs the Mk II? Is it a subtle difference or more significant? Is it more noticeable in soundstage recreation, or is it some other quality?
One thing I’ve noticed is that live cymbals sound incredibly delicate, and that seems to rarely come across in recordings. I’m not sure why that is - it’s probably something in the recording chain or the mastering - but if the the dual-chassis solution offers better nuanced decay and spatial recreation, it may interest me.
I haven’t listened to them side by side and I’m incorporating all of the Mk II tricks that I’ve learned into the TSS design.
There’s a sense of ease that the TSS has that first struck me. Also instrument tones are a little more realistic. Having the digital separate from the analog also gives a blacker background which can let you hear into the music a little more.
Like the DS and the DS Jr. I don’t think that hearing the Mk II and the TSS in separate times or spaces you’d say that one is clearly better than the other, but in the same room the difference should be obvious.
I’m thinking of waiting for the TSS, instead of purchasing the MKII. I like the idea of the Jensen transformers. I wouldn’t constantly think about trying the Lundahl’s in the TSS like I would in the MKII. I’m also going to get the AirLens when it comes out, so I’ll be all setup for the TSS.
What kind of cores, shielding and winding material do the Jensen’s use? I tried looking it up on their web site, but couldn’t find anything.
I’m not so interested in how they are built, but more in “How do they sound?” and “How do they measure?”. I value the Bessel response since it’s the closest to a pure delay in time and hence preserves the input waveshape the best. When I talk to their competition, they say that it doesn’t matter, but it does in my experience and hence I’m not as interested in transformers that don’t spec their frequency response and try to keep it under control.
My experience is that there are diminishing returns as the cost goes up. The Jr to the Mk II is a very big step. I’d like to think the TSS will be a big step up from the Mk II, but I think fewer people will hear that difference in their systems and be willing to pay for it. The TSS will still outshine the Mk II but somewhere up there the differences all become personal preference.
I just can’t get over the transparency I’m getting from the DS Sr. with the LL1948Ag transformers and mods. It seems like the musicians are performing in front of me. I’ve never heard anything like it before from my stereo.
For more details here’s a link to a similar pair of SERDES chips.
