Yes, it should work by default. However, 100MHz Crytek might be perfectly fine if you bypass and power it better than the default. Some people including me find that Crystek seem to like 805 SMD film bypass caps, not X7R. IME they also like a dedicated, low output impedance, fairly low noise voltage regulator. The also like the outputs buffered depending on loading. The same sorts of things may be true for the NZ2520SD as well. Clocks oscillators are analog, more than they are digital. IMHO they should be treated accordingly. That's what works for me anyway. Others may disagree regarding the above of course.
Last edited:
I, for example, strongly disagree. There is absolutely no good reason to use SMD film power supply bypass capacitors (in general, expensive and limited range of values) instead of X7R ceramic capacitors, as recommended in the regulators data sheet.
Example: the low noise LT3042 regulator datasheet (page 15) recommends:
Say 10uF for a decent margin, now anybody can try to find a film SMD cap with the required spec, good luck. The best I can find is this F161ZS106K050V KEMET | Mouser Canada at a cost of about $20 a piece. The size is 16.5mm x 14mm x 7mm and it's DF is specified as 3% @10KHz, which immediately leads to an ESR of about 50mOhm, at the very limit of the LT3042 requirements. The ESL is not specified for this film cap, but the rule of thumb for PCB is about 5nH/inch. Given the size of this cap, it would be impossible to make the to/from traces shorter than 1 inch, which brings us straight to an ESL of 5nH, already more that twice the LT3042 requirement of 2nH.
I'll let everybody decide if using such a monster cap at the output of a low noise regulator makes any sense. X7R ceramic capacitors are far from ideal, I would not use one in the signal path, but for power supply decoupling they are ideal. They cost 10 cents or less a piece, and come in all SMD sizes.
Of course, some may say why not using an X7R cap at the LT3042 regulator output terminals, then decouple the crystal with one of these SMD film monsters. That would be indeed an interesting audiophile approach. Combining the high price of a SMD film cap with the disadvantages of an X7R ceramic cap
.
Last edited:
Optical (TOSLINK) is handled the same as SPDIF. For an ESS dac that means it needs to be routed into one of the DAC chip's SPDIF inputs. Some of the dac chip inputs are configurable by I2C programming to serve different functions at different times. I2C programming can also select which input to use as the digital audio source. If no I2C programming then options may be limited.
Also, if using an I2S isolator, the trade off is that isolators tend to increase jitter. Of course ESS dac chips have internal ASRC for that, but they tend to sound better if DPLL_Bandwidth can be set to the lowest stable value, as ESS recommends. To be able to set that to a low value, it might help to reclock after the isolator. More circuit complexity would be the downside of doing that.
Also, if using an I2S isolator, the trade off is that isolators tend to increase jitter. Of course ESS dac chips have internal ASRC for that, but they tend to sound better if DPLL_Bandwidth can be set to the lowest stable value, as ESS recommends. To be able to set that to a low value, it might help to reclock after the isolator. More circuit complexity would be the downside of doing that.
IME it means there are probably still other problems with the dac implementation, and or possibly with the clock power and or bypass implementation. May I ask how you are powering the dac board, and how are the power supplies wired up? What kind of power transformer? IME everything matters for getting a dac to perform well. Can you post pics?
Also, may I ask about the rest of your system (power amp, speakers, preamp, etc.)?
EDIT: There should be an SMD bypass cap in the location outlined in red in the pic below. One electrolytic cap to bypass the clock is not a proper implementation. You could use X7R, but IME a .1uf or .22uf film cap in SMD size 805 works best for clocks. Digikey carries some good ones from Cornell-Doublier or Rubycon. Up to you though.
Also, may I ask about the rest of your system (power amp, speakers, preamp, etc.)?
EDIT: There should be an SMD bypass cap in the location outlined in red in the pic below. One electrolytic cap to bypass the clock is not a proper implementation. You could use X7R, but IME a .1uf or .22uf film cap in SMD size 805 works best for clocks. Digikey carries some good ones from Cornell-Doublier or Rubycon. Up to you though.
Attachments
Last edited:
I, for example, strongly disagree. There is absolutely no good reason to use SMD film power supply bypass capacitors (in general, expensive and limited range of values) instead of X7R ceramic capacitors, as recommended in the regulators data sheet.
Example: the low noise LT3042 regulator datasheet (page 15) recommends:
Say 10uF for a decent margin, now anybody can try to find a film SMD cap with the required spec, good luck. The best I can find is this F161ZS106K050V KEMET | Mouser Canada at a cost of about $20 a piece. The size is 16.5mm x 14mm x 7mm and it's DF is specified as 3% @10KHz, which immediately leads to an ESR of about 50mOhm, at the very limit of the LT3042 requirements. The ESL is not specified for this film cap, but the rule of thumb for PCB is about 5nH/inch. Given the size of this cap, it would be impossible to make the to/from traces shorter than 1 inch, which brings us straight to an ESL of 5nH, already more that twice the LT3042 requirement of 2nH.
I'll let everybody decide if using such a monster cap at the output of a low noise regulator makes any sense. X7R ceramic capacitors are far from ideal, I would not use one in the signal path, but for power supply decoupling they are ideal. They cost 10 cents or less a piece, and come in all SMD sizes.
Of course, some may say why not using an X7R cap at the LT3042 regulator output terminals, then decouple the crystal with one of these SMD film monsters. That would be indeed an interesting audiophile approach. Combining the high price of a SMD film cap with the disadvantages of an X7R ceramic cap
Most will want to try a film cap for sonic reasons, this is DIY... remember
If you would like to replace an MLCC with a film cap, for whatever reason,
probably the best option is Rubycon PML cap. They will come closest in size
and ESR. For example your 10uF jobby above, at 16V would be (mm)
4.5Lx3.2Wx2.6H. As expected they are fairly expensive.
If people wish to experiment with these, Digikey sell a PML cap kit. You get
42 caps of various sizes for around $25 which is really good value considering
the cost of these caps.
If you do not possess a CRO not really recommended
TCD
Maybe okay, probably not ideal. Could have ground noise problems too, IME it is not uncommon. Hard to say more without hearing it.
It would still be helpful to see pics of the dac with all the power supplies and interconnection wiring. Maybe some opportunities for easy improvements. Again, hard to say without seeing it.