Original already posted at #233.
Only minor changes, such as value of the R_iv to 360R as per datasheet.
AK4499EQ - Best DAC ever
Cheers,
Patrick
Only minor changes, such as value of the R_iv to 360R as per datasheet.
AK4499EQ - Best DAC ever
Cheers,
Patrick
Have to agree on that.
... if I am fortunate enough to have your attention for a moment
I wonder if I might ask about another topic having to do with kitchen table dac building: For those of us not equipped to measure phase noise at low offset frequencies, do you find specially selected NDK SDA clocks to be similar to or better than, say, Crystek 951 parts in 22/24MHz frequencies?Thanks!
Hi Guys,
Just took a quick listen to D90 with a track I use frequently for dac testing. Listened using Audeze LCDX and Neurochrome HPA-1 headphone amp. D90 was set to DAC mode which disables volume control (presumably DSD bypass in effect for that mode).
The following statements are entirely IMHO only!
The most striking impressions are the same for all sample rates and for PCM/DSD. Basically, the sound is (1) a little dynamically compressed (likely due to LDOs powering loads that would fare better powered as in the eval board), (2) a little IMD sounding across all frequencies (partly blurring together and partly closing in on what should be some empty space between various instrument/vocal sounds), and (3) a little fuzziness (possibly related to using 45/49MHz clocks and then dividing them by 2 in an FPGA for most dac modes). Taken all together, sonically the shortcomings are rather disappointing compared to what I know is possible from AK4499 (and no reason to believe others aren't getting better out of it than I can).
D90 did sound a little better with some HQ Player DSD settings. Didn't try them all, and the laptop couldn't have run some of them anyway. What did sound somewhat better was ADSM 512 + modulator combined with the poly-sinc-ext2 filter. While there was some sound quality improvement with decent upsampled DSD, all the previously described D90 audible shortcomings remained in effect. In other words, using HQ Player and DSD bypass mode doesn't audibly help problems arising from other causes.
All the above is IMHO only!
Just took a quick listen to D90 with a track I use frequently for dac testing. Listened using Audeze LCDX and Neurochrome HPA-1 headphone amp. D90 was set to DAC mode which disables volume control (presumably DSD bypass in effect for that mode).
The following statements are entirely IMHO only!
The most striking impressions are the same for all sample rates and for PCM/DSD. Basically, the sound is (1) a little dynamically compressed (likely due to LDOs powering loads that would fare better powered as in the eval board), (2) a little IMD sounding across all frequencies (partly blurring together and partly closing in on what should be some empty space between various instrument/vocal sounds), and (3) a little fuzziness (possibly related to using 45/49MHz clocks and then dividing them by 2 in an FPGA for most dac modes). Taken all together, sonically the shortcomings are rather disappointing compared to what I know is possible from AK4499 (and no reason to believe others aren't getting better out of it than I can).
D90 did sound a little better with some HQ Player DSD settings. Didn't try them all, and the laptop couldn't have run some of them anyway. What did sound somewhat better was ADSM 512 + modulator combined with the poly-sinc-ext2 filter. While there was some sound quality improvement with decent upsampled DSD, all the previously described D90 audible shortcomings remained in effect. In other words, using HQ Player and DSD bypass mode doesn't audibly help problems arising from other causes.
All the above is IMHO only!
Hi Mark,
That IMHO quote seems a very good idea! It might prevent others to immediately ask foor proof etc. etc.
Maybe we should do this more often, so that superfluous discussions (outside the lounge) are avoided.
WT? ECPU? Really?
Seems to me ECPU is unacceptable for Audio. The dielectric is Acrylic, which has worse DF (~1%) than PET. PPS or polyprop should be used.
COG or NPO ceramics are fine to use. Never X7R ceramics.
Read Bateman much?
Seems to me ECPU is unacceptable for Audio. The dielectric is Acrylic, which has worse DF (~1%) than PET. PPS or polyprop should be used.
COG or NPO ceramics are fine to use. Never X7R ceramics.
Read Bateman much?
Mark,
I'm quite weary about NDK parts (although I use there Temp compensated VCXO parts).
A while back we measured the phase noise of new MEMs type parts and remeasured about a year later only to find a significant increase in phase noise with age - so it appears that MEMs oscillators wear out over time...
I've never read any details about this - so stay WELL away from MEM's parts.
What worries me about NDK parts is that its not clear if the part is a real Crystal based clock or MEM's - I've seen the MEM's parts described as Crystal oscillator modules, when the are clearly not - (MEM's is not a Quarts Crystal).
While the NDK SA parts might be really good parts (never measured or used them) - I'd recommend Crystek 90.3M / 98.3M parts and Divide down by 4 to 22M / 24M, this results in a 12dB reduction in phase noise (and buffers the Oscillator circuit from external loading effects).
The Crystek 90.3M / 98.3M XO's will also be a higher Q overtone circuits - again very good for phase noise.
Last edited:
John,
I've never seen Crystek clocks in those frequencies, but Accusilicon makes some...
The NDK parts are clearly crystal oscillators. I'm not sure why you would think otherwise. It's MEMS not MEM's, also. SiTime claims no aging mechanism and a quick search could produce nothing indicating that will happen. Forgive me for believing them for now over your uncontrolled test of sample size N=1.
Last edited:
Some OT moved.
Depends what your trying to achieve.
For SURE ceramics / Acrylic should never be used DIRECTLY in the signal path - but the lossey nature of Ceramics is very useful in digital decoupling as it helps to dampen transient current induced ringing on power rails...
Low loss caps perform worst as they tend to resonate more with the load step... sometimes being lossey is a good thing...
Its a really hard subject, I'm constantly fighting the sonic signature of capacitors - I try to avoid ANY in the direct signal path, but then PSU capacitors are also very much in the signal path...
Last edited:
Crystek makes CMOS output parts to 130MHz:-
Crystek
The Accusilicon parts I've measured are PLL based production programmable parts - this allows Accusilicon to cheaply manufacture small batches of custom frequency parts.
PLL based designs have much poorer close-in phase noise (where it matters for audio) then even a simple well designed XO circuit - I recommend to avoid Accusilicon unless you are certain they are not PLL based parts - even better if they provide PN plots...
Ltspice is well worth persevering with if you don't already use another decent simulator. Very powerful and handy tool.
TCD
John, which parts have you measured?
I believe their fixed oscillators are good. It's difficult communicate with them
and get any info from them. If anyone has a reliable contact please PM me.
TCD
You might try a combination of 2 polymer smd caps. Kemet have a decent
cap modelling tool and the right combo will go out to 1GHz. The board lyt,
as you know, will dominate.
TCD
Terry,
Sorry confirming with my notes I've confused projects / manufacturers - I was asked to measure an Accusilicon TCXO IC (Die package) for a large 3rdf party "Crystal" manufacturer - I've not measured there larger 14pin DIL type XO's... my contacts are within Asia...
The PLL based XO part I mentioned was a 7050 SMD part from Abracon - we where working with both devices at the same time...
Working indirectly with Accusilicon I know they provide PN measurements and appreciate Close-in phase noise requirements for audio, but they have REALLY poor sales support outside of China, in fact from what I understand they really only have the engineering resources to work on a one to one basis with key accounts...
I apologise for my confusion..
Speaking of AK4499 dac implementations, just did a bit of looking around inside of the Topping D90 case. Saw what looks to be four tiny voltage regulators, possibly as a source of 5v for Vref (one per channel). Don't recognize the chip from the marking code, but perhaps someone else around here will...
Attachments
This is a dangerous slope. Everybody knows that the only way to verify is something is audible is to ask Amir at ASR to measure it ;-)
