Asynchronous I2S FIFO project, an ultimate weapon to fight the jitter

[roger57]

Originally Posted by iancanada

They are 0603. You can reduce those resistors to get bigger LED current. Schematics of this section posted in previous post. 620ohm is OK. How much current will be up to what kind of LED you are using. Try to use resistors as big as possible to reduce power consumption.

Ian

It's difficult to find very efficient panel LED's at 2mA. I think it relies to a large degree on a tight dispersion angle, DK had just one (3mm) less than 40 degrees. In any case, the LED's I've selected are bright enough at 2mA; to get that requires a 620 ohm resistor with the 3.3V supply.

Thanks,
Gary

[regal]

Originally Posted by necplusultra

I need my crazy pills

Me too. Its amazing how we can pick a GMR designed for TV's to prevent failures and think it has anything to do with audio quality.

I think we want galvanic isolation in audio for a different reason than most other applications. We want it to prevent ground bounce from interferring with the digital to analog conversion. I don't see how a GMR can do that since this GMR duplicates the EMI it sees on both sides of the ground up to 110MPS.

Best bet with the FIFO is still to use the SPDIF board in front of it, the addition of the GMR board isn't going to change that IMHO. Interesting attempt though.

[qusp]

what you talking about? we can use sillabs, AD, or IL type devices, all use different technologies.

[hochopeper]

Regardless of isolation on the input there is still some benefit in further attenuation of the EMI generated by the FPGA.

Yes there are EMI filters on the board but an isol board like Ian has presented seems logical in this application unless I've missed something.

edit: heh I missed seeing qusp's post before I wrote this.

[regal]

Correct me if I am wrong but a GMR as spec'd here (IL260E) works a lot like a little transformer. They provide galvanic isolation and these at 110mbs transmit EMI, to both sides of the taps.

Believe me I wish I was wrong but the isolation board would really just stop current surges not EMI to the ground plane. I am just looking at the physics.

[hochopeper]

Originally Posted by regal

Correct me if I am wrong but a GMR as spec'd here (IL260E) works a lot like a little transformer. They provide galvanic isolation and these at 110mbs transmit EMI, to both sides of the taps.

Believe me I wish I was wrong but the isolation board would really just stop current surges not EMI to the ground plane. I am just looking at the physics.

My understanding is that the isolator chips each work slightly differently but for sake of convenience that is how we generally chose visualise their operation.

If we continue with that visualisation then I believe that the efficiency of the high freq noise transmission is not going to be perfect. We could visualise this lack of high frequency efficiency as a low pass filter I guess. It is for this reason that we often associate GMR with adding jitter, isn't it? So we select a GMR that is able to pass sufficiently high frequencies that the data is cleanly passed, but noise at a higher frequency may be attenuated. We then are able to rely on EMI filters applied by Ian in his design for addressing the lower frequency EMI (they may also address higher freq too, but higher the freq the harder these things become to address generally speaking) and the GMR may eliminate some of the higher frequency noise created by the transport and FIFO FPGA chip.

No one has tested these yet. No one has said they will change the world of digital music as we know it. It is simply an experiment to try to squeeze the last peice of musical juice from the digital data-stream. Considering that by the time a transport, dac and IV are to the stage of development and integration comparable to the FIFO design, we're generally well past the point of diminishing returns. This, in my opinion, is a nice experiment that may allow the last 0.5% or so of improvement, for a lot less money than other improvements that many are making.

I for one am interested to see the outcomes!

[iancanada]

I have to say the isolator is really working for my system, especially followed by a re-clock stage(to eliminate additive jitter from isolatro board, but MCLK still local always ) . This is not a new technology, we could find it in many high precision high speed data acquisition system. It even being widely used in medital electronic system. They made signeficant improvement to the uV level medical signal processing performance beasuse of the isolation of common mode and EMI noise, as well as the safty.

Ian

[iancanada]

Just some digital audio solutions from TI application note for reference:

"Figure 11(b) shows an improved method for high performance,
mixed signal board layout. This method adds digital
isolation between the DF1704 and the audio DACs, and
provides complete isolation between the digital and analog
sections of the board. Texas Instrument’s ISO150 dual
digital coupler provides excellent isolation, and operates at
speeds up to 80Mbps."

Ian

[iancanada]

All of the FIFO isolator PCBs for evaulaiton shipped today. A longer 10P FFC/FPC cable and a 7P PH2.0 cable included just in case you don't have. Please refer to the schematics I posted before for the BOM. Let me know if there is any question.

Ian

[glt]

Wow, thank you so much. Let us know the cost including shipping.