Have you measured the propagation delay between q and q prim of that flip flop? Most flipflops are constructed in such way that one inverter delay is incured between the outputs, this is usually not noted in the datasheets.
Could you give more details on the modded squeezebox? any links to the mod you did to feed the clock from the MK7 and take the I2S signal?
Any links in general to mod a squeezebox for I2S? I am new to this
What about direct fed i2s from Pace-Car reclocker?
I am doing something similar, using a Squeezebox to an Empirical Audio reclocker and then feeding i2s directly to the TDA1541A, with resistors in the path. The results seem very good to me. Is this a possible viable alternate solution?
From the TI datasheet page 4:
CLK to Q, 0.8 ... 1.2ns @ 2.5V supply voltage and 15pF load
CLK to /Q, 0.7 ... 1.1ps @ 2.5V supply voltage and 15pF load
Propagation delay increases with capacitive loading:
CLK to Q, 1.4 ... 1.8ns @ 2.5V supply voltage and 30pF load
CLK to /Q, 1.3 ... 1.8ps @ 2.5V supply voltage and 30pF load
Reducing load capacitance below 15pF can further reduce propagation delay. In the MK7 I run the flip flops at approx. 2.5V with approx. 2pF load.
The generated jitter depends on the combination of propagation delay and noise on power supply / ground. Propagation delay is affected by load capacitance and supply voltage.
In general it is best to run the flip flop at power supply voltage that gives lowest propagation delay, and lowest possible load capacitance.
When a divider or reclocker circuit is build using this very fast logic, simply connecting the in / outputs together, will lead to high ground-bounce levels and poor jitter performance.
I used resistors to minimize ground-bounce. I placed them on programming inputs like the D input and on clock inputs that are not critical for direct output signal timing. I also used a shift-register divider so I didn't have to add extra logic (gates) that cause more interference.
Also note that the power supply for these flip flops needs to be extremely clean, similar to masterclock power supplies.
Latest development on the MK7 DAC, improved decoupling for the TDA1541A.
In order to optimize TDA1541A decoupling I soldered all decoupling caps directly to the pins of the chip as close to the IC housing as possible.
This way the added extra length of IC socket pins can be eliminated. The active divider decoupling pins no longer need to be connected to IC socket nor PCB. This further reduces leakage currents.
Isolators aren't perfect, so even a IC socket or PCB adds some leakage current. Problem is that with the TDA1541A the leakage current must stay well below 61nA. With average voltage of approx. 10V on the active divider decoupling pins this translates to approx. 164 Mega Ohms.
I attached a picture of the decoupling caps soldered to a S2 chip. Pin #1 is at bottom right.
In order to optimize TDA1541A decoupling I soldered all decoupling caps directly to the pins of the chip as close to the IC housing as possible.
This way the added extra length of IC socket pins can be eliminated. The active divider decoupling pins no longer need to be connected to IC socket nor PCB. This further reduces leakage currents.
Isolators aren't perfect, so even a IC socket or PCB adds some leakage current. Problem is that with the TDA1541A the leakage current must stay well below 61nA. With average voltage of approx. 10V on the active divider decoupling pins this translates to approx. 164 Mega Ohms.
I attached a picture of the decoupling caps soldered to a S2 chip. Pin #1 is at bottom right.
Attachments
Hi,
Funny, when I made my first TDA1541 DAC around 12 Years ago, it looked a lot like your picture, except I wrapped the TDA 1541 into copper foil and used that as local ground-plane, as this reduced parasitic inductance... One could also use silver foil.
Ciao T
Latest development on the MK7 DAC, improved decoupling for the TDA1541A.
Funny, when I made my first TDA1541 DAC around 12 Years ago, it looked a lot like your picture, except I wrapped the TDA 1541 into copper foil and used that as local ground-plane, as this reduced parasitic inductance... One could also use silver foil.
Ciao T
0.1ns is still 100ps so for the first instant of every clock transition there will be a small glitch.
It would be possible to generate a true and inverted signal without incurring as much delay that might be worth perusing.
I don't doubt for a second that it's not an improvement but what intrigues me is how the hell you managed to do it without frying the chip
Hat's off to you for having the soldering skills....astonishing
I'll leave that one ' off ' the list of things to do
Andrew
Both these fets and the acrylic caps are in the category unobtainium...
I finally found the fets for 2.75 EUR (usually more, up to USD 8) at Reichelt.de, plus EUR 10 shipping...
The caps are from newark.com, so for me EUR 25 in shipping alone...
John, where did you get these babies? Maybe a GB would be worth while?
I finally found the fets for 2.75 EUR (usually more, up to USD 8) at Reichelt.de, plus EUR 10 shipping...
The caps are from newark.com, so for me EUR 25 in shipping alone...
John, where did you get these babies? Maybe a GB would be worth while?
Nice work! At one point in this journey, I believe the thought was to keep ground lines from each bank of decoupling caps separate, and have nothing else on them. Do you now see an advantage to grounding directly to pin 5, or is it of no consequence to sound quality?
What is the audible impact of using caps with leakage current above 61nA? I used some SMT 8.2u organic polymer bypassed with 220 nf film. They have a leakage current of 3uA. They sound very good compared with the 220nf film alone.
Bit0-Bit15, I count 16 bits resolution. The challenge is an output stage wehre the lower bits aren't swamped by noise.
Most music hirez or otherwise only has a few bits resolution, a dynamic range on a commercial recording of 40db ~ 7 bits would be exceptional. Probalably the reason most of us reading this thread understand the TDA1541A understand the old sayin "its the first watt that matters most."