Raspberry Pi I2S to SPDIF Hat

I think this is pretty critical to have right to get near perfect spdif. This might cover up all the benefits in the overall design, and make it tough for us to evaluate the sound--

 

I read the data sheet yesterday and was going to point out what I thought was a typo. Sadly I read it is not the case. After all the discussion to get a perfect 75 ohm impedence and the fact that we let go the RCA connector for spdif in speed 2 I would have never expected a 50 ohm BNC connector.
I hope it doesn't spoil too much the benefits we were trying to achieve.

 

Yes, that was my mistake. I do not think it will make a big difference, but we will certainly correct it in the final version. If you have soldering skills it would be simple to get a 75 ohm BNC and replace this. If not we can fix it here after you've done the initial testing.

Remember, these are protos and you are my test dummies for a reason!

Cheers,
Michael

 

No problem Micheal. I was alredy thinking about getting one and solder it myself. As soon as it arrives I'll see if I can get a suitable one.
What about the 50 ohm transformer? How can the output be 75 ohm?
In my ignorance, so please excuse in advance any BS I might be writing, the cap should not change the output impedence. Isn't required a "compensative resistive load" (rough italian to english translation) to do that?

 

The transformer is specificied by in/out ratire (1:1 in this case) and it's inductance. As long as the primary is configured with a suitable capacitor value to get 50 ohm at ~25Mhz (SPDIF frequency) it will "pass" this impedance to it's output.

Cheers,
Michael

 

From my SW engineer:

CSB503SPD2 hardware release notes:

The 5V external power jack on the CSB503SPD2 board is for the application of a separate 5V power. There is a removable jumper for this configuration. The filtering and protection we put on this imput will not allow enough current to power both the Pi and the SPD board from the external Jack.


For the Pi3 and the SPD only (no other shield)
You should either power from the pi3 usb with the spd jumper in
OR
Power the pi3 through the usb, have the jumper out and use an external 5V supply connected to the SPD board. You would need to apply power to the SPD first, then the Pi.


For the PI3 and the SSD/SPD combo:
Power through the 12V SSD jack and leave the spd jumper in, no power to pi3 USB power input

Never apply 12V input to the 5V power input of the SPD board.


GPIO 5,6,and 13, in addition to the I2s pins are reserved for use by this board. Any software that uses these pins for another purpose could cause a conflict.


CSB503SPD2 software release notes:

The CSB502SPD2 will run with the hifiberry-digi driver without any modifications. This will default to 27MHz oscillator PLL based MCLK mode. For reduced jitter, we have included two additional oscillators to drive the MCLK directly depending on sample rate. For this configuration we have a custom driver.

We have created a modified Volumio 2 RC1 SD image with our custom driver in place of the Hifiberry Digi driver for easy installation and testing while our drivers are being pulled into the kernel. This image and instructions for use are available at this link:

https://github.com/Pi2Design/csb503spd2/releases

It works well on the Pi2. It also works on the Pi3 but we have seen kernel crashes. This is a Pre-release image for testing purposes.

There are some bugs with this release of volumio 2 (which we do not believe are related to the board or driver), but we chose it beacuse it is the latest and (mostly) works on Pi3.
Bugs include:
Software volume control has no effect.
Pi3 needs bluetooth disabled or the sound is garbled -> See note in instructions at download link above.
We have seen kernel crashes on the Pi3.

 

For those test dummies worried about the effect a 50ohm connector vs. a 75ohm one would have; don't be too worried as the effect is really minimal. Here is some simulations taken from another site:

 

@Mban What about near the end of the paper on 75Ohm/50Ohm impedance mismatches and standing waves...

 

From my understanding it all depends on the length of the mis-matched signal path. I.e. for a 3 meter long cable, if you were to use a 50ohm cable there would be a larger effect on signal modulation/attenuation. For a 1.5cm piece of wire going from a 50ohm connector to the circuit board (as is this case), there should be very little signal modulation. That article stating "5% signal loss" is a major oversimplification, and is more relevant to the case where you have a whole cable that has mismatched impedance (make sure not to use a 50ohm coax cable!).

 

I think the biggest issue is perception. Audiophiles will jump all over the impedance mismatch and use that as a reson to not buy the SPD. We all know at times its hard to reason with us audiophools.

 

Also, it is critical to understand that signal loss is not equal to data loss. 100% data recovery can easily be maintained in the face of 10%, 20% or even higher signal loss. As I have said before, the clock jitter is much more critical to data recovery than signal strength. As long as signal strength is sufficient to allow the receiving end to "recover" the data stream and clock, the absolute signal level does not matter. The output level from the transformer out the cable is ~600mv. The minimum level per the specification for 100% data recover at the receiving end is 200mv. That's just 1/3 the output! That means 30% signal loss would still provide 50% more than the minimum.

Proper impedance matching is important in theory, but in practice there is plenty of margin to insure 100% data transmission.

Cheers,
Michael

 

Since the 50ohm will be changed to 75ohm in production it is a moot point. All will be happy. Has anyone listened yet?

 

Just want to discuss it a little while changes can be made, I do understand it's a moot point since the production version will have the correct part. However a few people are already testing them! Glad we did kick it around a little, looks like we're just splitting hairs. And by no means meant it as another endless audiophile debate . Thanks!

 

By all means it is important to look and check everything out properly.

Months back your ideas actually got me interested in the PI and the digi+ just hoping that this will really be killer when everybody provides their input and Michael produces the final product.

Even though the BNC is wrong, the listening will give valuable feedback for the overall product. The XLR is correct so people with DACs that can use it will not be impacted. This is a great project and looking forward to the fruits of everyone's labor.

Regards
Bob

 

What do people really use the BNC connection for? Very few DACs have BNC connections. is it primarily for pro gear or measurement gear?

 

Only DAC I have had with BNC was the Benchmark DAC1. Ditto for XLR. Since we are now giving feedback here is one more. Could there be 2 versions of this? One with just the basic outputs (RCA, Toslink, etc) and then this one?

 

We have Toslink and RCA out on our 503DAC1, which tbooui has reviewed the first revision. The final revision adds clock switching. As a DAC board it also has the PCM5122 DAC with Line and headphone outputs.

Cheers,
Michael

 

I have looked at that one too, but do not need the DAC, nor all the other outputs, so I continue to follow this thread to see how this turns out. I really like this board design and the most important fact - seperate PSU for the HAT and PI.

 

Didn't we have in program 2 versions of the board? Speed 1 & speed 2? BNC\XLR & RCA\TosLink? What am I missing here?

 

Hah! Might be I missed something. Please correct me if I am mistaken.