DS and jitter rejection with Toslink


#1

Hi @tedsmith

I’ve always wondered but never gotten around to asking (and if my assumptions below are incorrect, please don’t hesitate to correct me):

With the DS toslink input, my understanding is that the toslink source’s clocking info is completely ignore by the DS.

So how does the DS’s FPGA actually know how to correct the timing of the 1’s and 0’s and therefore how to eliminate the jitter, if it ignores the source’s clocking info? Or is there still some info coming from the Toslink source that allows the DS’s FPGA to correct this timing accurately?

Thanks in advance


#2

All digital audio inputs are treated the same way by the FPGA code - there’s a fast internal clock that’s used to sample the inputs. It’s fast enough that no incoming edges are missed. Then the FPGA “just stares at the patterns” to find the 0’s and 1’s:

DirectStream-PPT-7-aes-decode.jpg

After the data is decoded you know where the sample boundaries are and you can see how often they come - if there are about 96000 samples (x 2 for stereo) per second you know that the sample rate is 96k and therefore you know how to upsample.

The control of the output clock is just monitoring the FIFO to make sure it doesn’t get too empty or full of samples… That ends up with the output clock being the long term average of the input sample rate, but it’s not fooled by instantaneous sample rate changes (like either jitter or changing input selects.) But nowhere is there a regenerated input clock or anything that would pay attention to it if it existed.

If you want to know more about how to decode the bits themselves checkout:

https://en.wikipedia.org/wiki/AES3

https://en.wikipedia.org/wiki/S/PDIF

and

https://en.wikipedia.org/wiki/Differential_Manchester_encoding


#3

Fascinating and incredible stuff, thanks Ted!


#4

I just installed my first ever glass Toslink cable, from Lifatec, to use with another source with my DS.

Until then I’ve been using an Analysis Plus ‘medical grade’ plastic cable and it reliably passed 24/192 without issues but the sound was a little soft and lacking top end details, compared to other inputs, to my ears in my system. I always make sure I unplug other cables when comparing.

The difference this glass cable made is not subtle - wow. More bass, more detail and increased desire to turn up the volume more and more without listening fatigue.

I wonder what the differences between glass and plastic cables, could be due to?

Both cables support bitperfect playback at 24/192 and there’s essentially no EM/RMI, no groundloops, no leakage current loops involved.

Many others have noted positive improvements with using glass too, not just with the DS, so I wonder what it could be.

The plastic cable is 1m in length and Lifatec is 20 inches, so the glass cable is around half the length. It’s possible a shorter plastic cable would be an improvement on the 1m length.

Even though the DS does a class leading job with jitter reduction, I guess less jitter with the Lifatec glass fibers still helps even more?


#5

If both pass the bitperfect test and both are driven from the same source hardware with the same source software then it might seem that there’s little else left but jitter - but just like the UpTone Audio USB Regen for USB, the power an optical receiver draws depends on the quality of the signal it gets. Either way it would be nice if the DS were a more perfect device, but, in the end with audio, it always seems that everything matters.


#6
Ted Smith said everything matters.
Absolutely Ted. I now appreciate this so much more because I wasn't expecting this with Toslink.

But the DS’s already stellar sound is even more phenomenal with this short glass cable.

I’ve tried all the USB conditioning devices and well known cables under the sun and for <24/192 music, while some of those combinations were phenomenal, none of those combinations comes close to the DS’s Toslink input with this short glass cable, to me, in my system.

I don’t know where all this bass and high end details suddenly came from. Wow.

If somebody made a short hard adapter (like a few inches) with glass fibers, to go between the source and the DS input, that will probably be even better than this 20 inch cable. I don’t know if this exists though.


#7

Mi2016 - can I ask what the source is? I have a couple of the Lifatecs and find the optical to be “cleaner” than the average USB, but nonetheless sorta hyped and, um…“glassy” sounding. But I put that down to the 50 cent optical output on the source, or noise from the source. Or maybe just personal preference.

Heck, I still haven’t decided about Huron ; ). I suspect when I try Torreys again, I’ll figure it out. : )


#8

I thought optical cables had to be a minimum length that was 1.25 or 1.5m. Something about the whole wave being passed?

I was running a DH Labs toslink between my PWT and DAC. I thought it sounded better, so I left in for about a month, then eent back to my long time Music Metre Fidelus digital cable. It is close, but I kept the Fidelus in.

I have heard a lot of good things about the Lifatec cables. Maybe when the budget allows I’ll try one.


#9
badbeef said Mi2016 - can I ask what the source is? \
Hi badbeef,

This is a cheap HiFiBerry board and cheaper Rasperberry Pi as the source - but the kicker is each board is independently powered by an Uptone UltraCap LPS-1 and the Toshiba TX179P TOSlink transmitter is powered by this ultra low noise 5V power supply.

It has a couple of low jitter and low phase noise oscillators feeding the Toslink connector too.

So while the overall source is cheap, the ultra low noise power supplies are pretty substantial (compared with the stock SMPS power supplies).

But it wasn’t a fair comparison between cables because the 20-inch glass Lifatec is half the length of the plastic Analysis Plus cable, so that’s possibly a factor too.

I finally can see/hear why Ted loves the DS’s TOSlink input for <24/192 music, now that I have a half decent source and short glass cable.

But I definitely agree, personal preferences can differ significantly too. On Huron, two of my closest friends don’t really like Huron’s sounds but they also dislike our other friends Playback Designs player. And unsurprisingly the latter friend and myself love Huron (we find they sound quite similar now).

Many thanks


#10

Not just limited to Toslink but fiber optic cables is general, length and bend radius affect the amount of back reflection within the cable.

The best Toslink cable that I have used is a patch cable by Sys.Concepts Inc.

They use a 1300 strand glass cable for their cables.

Very reasonable pricing and delivery time.

You can customize just about everything, connector type ( standard, mini, straight or 90 degree) , color and length.

Their Toslink cables will pass 24/192 without issue.

EDIT: they can also make custom short “hard” patch cords for your gear, you need to provide the connection types, angles and measurements.


#11
jeffstarr said I thought optical cables had to be a minimum length that was 1.25 or 1.5m. Something about the whole wave being passed?
Hi Jeff,

Ted is much more knowledgeable than me on this (and everything) and I haven’t ask him this before but Rob Watts loves Toslink with his DACs too and he told me to find the shortest Toslink cable I could find, especially for 24/192.

He uses these himself: https://kabeldirekt-store.de/index.php/audio/toslink-kabel/kabeldirekt-optical-toslink-digital-audio-cable-pro-series.html


#12
MrDerrick said

EDIT: they can also make custom short “hard” patch cords for your gear, you need to provide the connection types, angles and measurements.

Thanks MrDerrick! I just saw this on their website. Very interesting indeed - I really don't need to add another cable to my huge collection but I'm tempted to hear what an ultra short length sounds like

#13

I’m so happy with Toslink performance after getting this glass Toslink cable (and more importantly I’m so happy with Huron), I’m thinking of getting a 2nd DS to go in my 2nd system and I’m going to replace the HiFiBerry board with one of these:

http://www.pi2design.com/uploads/4/8/5/3/48531975/visio-502dac_datasheet_p4.pdf

Nice NDK clocks powered by the nice low noise LT3042 linear regulators, feeding the Toshiba Toslink connector.

And the Toslink connector is powered externally by the Uptone LPS-1 at 5Vdc, and the Pi3 by it’s own separate LPS-1 at 5Vdc.

<img src=“http://www.pi2design.com/uploads/4/8/5/3/48531975/20170621-141409_orig.jpg” width=“1100” alt=“20170621-141409_orig.jpg” " />

<img src=“http://www.pi2design.com/uploads/4/8/5/3/48531975/20170621-141337_orig.jpg” width=“1100” alt=“20170621-141337_orig.jpg” " />