I2S: the one box solution..spinout from gold BNC thread

[grataku]

That looked like a good breaking point to open the discussion on how to entirely bypass the hated spdif interface and thereby avoiding all its impedance mismatch problems.

For starters, if it's not too much to ask, can anyone draw up a quickie block diagram of how it is with SPDIF and how it would work without it, or point to some datasheet diagram...

Hopefully all the people that are versed on the topic will contribute.

[jean-paul]

Hi Grataku, avoiding SPDIF is not that complicated certainly if I2S ( only 3 datalines + ground ) is the used protocol in the cdplayer. But even with the Japanese protocols it isn't very hard to do. After all the cdplayer has a DAC in 99% of the cases.

Just connect the datalines of the new digital filter/DAC to the right pins in the cdplayer and remove the old digital filter/DAC/outputstage components. Of course this is only possible if the cdplayer in question has enough room to house the new DAC-board and an extra power supply. Did it in the past and it wasn't rocket science. Most time went to designing a PCB that fitted in the case of the player and the usual connecting-pins-to-what-other-pins issue.

It all depends on checking datasheets of the used chips in the player and the ones of the new DAC combo and the design of a quality PCB and supply. Or am I misinterpreting your words ?

[Jocko Homo]

I'm not sure they ever expected us to use SPDIF, unless, of course, it was with the dreaded TOSLINK on the back of CDPs. TOSLINK was the ideal high-tech compliment to "perfect sound forever".

Which probably explains why not matter how you do it, it stinks.

I belonged to the AES when all this was being hashed out. In brief, it is the reason that I no longer send them any money. The whole CD scheme was just a big con to make money, with little regard as to whether or not it actually worked. Signoro J. A. Pan was the driving force behind it, and the AES just nodded their head.

These are the same twits who then came up with the AES/EBU disaster, which had to be re-written as the first interation was dreadful.

Block diagram?? I don't think there is enough stuff to merit one. Maybe I'm wrong.............

Jocko

[tbla]

...respect...i allways enjoy your posts..hehe..

[Elso Kwak]

Originally posted by grataku
That looked like a good breaking point to open the discussion on how to entirely bypass the hated spdif interface and thereby avoiding all its impedance mismatch problems.

For starters, if it's not too much to ask, can anyone draw up a quickie block diagram of how it is with SPDIF and how it would work without it, or point to some datasheet diagram...

Hopefully all the people that are versed on the topic will contribute.

Hi Grataku,
Actually the same signals that were coming from the CS8412 , MCK, DATA, LATCH & BCK are now coming from the DSP. MCK (masterclock) is not used in NON-OS application.
Nice idea to start a new thread on this topic.

[Jocko Homo]

In the typical Philips platform of the mid - late 80s, the SPDIF signal would first go through a serpentine maze of PCB traces, to an IDC connector. Here, it would mate with a cheap coax cable, of dubious quality, of many cm in length. Back through another IDC connector and through more PCB traces until it reached the output transformer.

(I hope Signoro Tent is reading this, and can shed some light on this part of the story.)

The transformer has very poor bandwidth. It can not pass anything that remotely appears to be a square wave. I can only assume that this transformer was picked so as to minimise EMI, because it has no other redeeming features. Other than the obvious one: it is probably cheap.

The digital output on these CDPs sounded very poor.

Jocko

[Jocko Homo]

A certain high-end audio designer (we will call him "designer", as he is not an engineer) decided to improve the digital output on the CDPs referenced in the previous story.

He put a 75 ohm resistor in series with the digital output pin on the SAA7220. Hey, it had to work, right? The right impedance, extra hot output level for no good reason, famous name on the CDP. Just had to be great.

No, it stunk; for many reasons.

First, he failed to take into account the output impedance of the chip, which is around 25 ohms. So, the impedance is now around 100 ohms. Boy Wonder designer also decides to make a cable to sell with this transport. Lo, and behold......he comes up with a bizarre "ham radio" connector, known as a UHF Type. It is a 93 ohm connector.

And surprise of all surprises, he finds out that this kludge now sounds best with a........(drum roll.......) 93 ohm cable.

Only problem, the D/A boxes that this was connected to sometimes had 75 ohm inputs. And it sounded like crap then. Especially when hooked up to units designed by a certain foul-natured, foul-mannered, foul-mouthed troublemaker of Italian heritage. Who took great delight in pointing out how much it sucked to all the dealers who carried both product lines.

So far, we have wrong impedance, level too high (EMI City, Arizona..........), but there is yet one more snafu he designed in.

Boy Wonder wants to have the ability to reverse absolute phase. OK, good idea; only problem..............reversing phase on the TX end does nothing. It has to be reversed after the SPDIF RX chip separates the data signal out.

I'm sure you all know who also pointed that out with great delight to all the same dealers. Some were had to convince, but eventually they all had to........

[BrianL]

Yes, yes, the one-box solution is best,

BUT..........................

What do you do when the CD transport dies?
All your work wasted.

What do you do when you end up with multiple
digital audio sources of various formats?
Create a new set of "upgrade" guts for each?

I've talked with Jocko about one of his spiffy
B-B DAC boards for my aging CD player, but
how long until it craps out?

But how do you clean up S/PDIF short of bunches
of digital circuitry to buffer the bits
in a FIFO and play them out at a consistent
speed?

[guido]

Hi,

When i took apart a very old cdrom player, there was a pcb with some RS422 senders/receivers. I tried them and i got a 11 MHz clocksignal through (clock back from DAC to CDP). Now going to
try to use them to sent I2S from CDP to DAC and clock from DAC to CDP. I am using a fifo in the dac to get rid of sync problems and jitter coming from the CDP.

The chips are AM26LS31 and AM26LS32 if i recall correctly. These carry a diff. signal (4 channels). At the receiver, the + input is terminated with 240 ohms to vcc and 240 ohms resistor to gnd.
The - input is terminated with 120 ohms to gnd (again, if i recall correctly).

Connector is DB15 and i am going to try a PC joystick cable first. If all works, i am going to get me a nice shielded cable.

So i am working with something i had lying around and with trial and error.

Any comments ??

GuidoB (not Tent :-)

[dggs]

I2S is a slow bus. It's not very difficult to send it to another box. I would use LVDS rather than RS422.

I2S -> LVDS driver ->CAT5 cable -> LVDS receiver -> I2S

CAT5 cable has twisted pairs inside. It's good for differential signals like LVDS. It's easy to get and cheap. You can go for CAT5e, CAT6 or CAT7 if you like.
LVDS driver/receiver are easy to get too. For example, the National Semi DS90C031/032 from DigiKey. These devices can support data rate in excess of 155.5Mbps. More than enough even for 192k/24bit.