Yes I had him sort and measure a bunch of them. But after we get the 5125a, we can do our own measurements in house. Also all kinds of testing that's cumbersome to send out very easily. He was surprised when I sent him a photo of our reel of 2520SDA's, as he didn't know they existed yet. Can't wait to measure them as even 1 randomly selected from the reel, sounded better to me than this 2520SD here:
This sounds like the higher-Fsw or lower-Fsw UcD dilemma. It can be demonstrated that it is a 2 slope optimization problem, and optimum point is where the product of the 2 optimization functions is minimum ("noise of the clock itself" * "noise of clock conversion plus noise derived from working at the resulting clock frequency"). Optimum clock interference level is going to be similar to electronic noise floor of the converter itself, thus contributing ~3dB more, unless improving clock by a few dB is affordable.
Last edited:
As a generalist, after talking to all of the specialists, it became clear that you can't get better than the master, or reference clock itself. No matter what kinds of tricks you do downstream, anything at all is going to introduce more jitter. Then I studied how the Xmos U8 USB chip handles the clock division internally. What I discovered was only if you output 11.289Mhz audio, do you get a straight clock division by using only the rising and falling edges of the clock, without any further jitter causing dividing, or PLL generation. This happened to be ideal as I wanted to use only that frequency for my DAC anyways. Then on the DAC end I optimized the discrete analog filter to only be the best at filtering 11.289Mhz. Set way up at 180Khz so we have no phase shift in the audible band. This means it will be poor for lower sample rates, but it didn't matter because all that filter will ever see is 11.289Mhz audio anyways, with modulators pushing the noise floor to sub -120dB levels up till around 100Khz.
So I used very low phase noise clocks like in the example plots I posted, 8 separate LT 3042 LDO's, with it's own dedicated to the clock and Xmos chip, and Belleson SPZ discrete pre-regulators for each of the voltages. Then to power that regulator board I used the lowest noise SMPS on the planet. For analog stage Weiss OP2-BP discrete opamps, powered by the Belleson SPZ discrete regs. Filter uses all the very finest Panasonic ECHU PPS caps and thin film resistors
Anyways the results were shocking. Beta testers also agreed. So developing our Ravenna board we specifically designed it to take this same concept to the best levels possible. Since we aren't bound to the limitations of a general use ASIC USB receivers, much higher quality results have been obtained. I'm quite confident we can now surpass the MSB Select DAC 2 using a $2 crystal, and free media player software for the SRC/SDM engine. No only that 4 channel boards that can be clock synced within 2 nanoseconds for active speaker use.
Last edited:
You'll need to learn Japanese, and buy minimum 1000 units. Also 7 month lead time! If you know a better way to get your hands on them let us know
I was thinking about offering my Purestream DAC as a DIY kit. If so I will offer the NZ2520SDA's as an option. You'll also be able to select your own 990 pinout discrete opamp for the gain stage, as well your own caps and TO-220 type regulator for the regulator board. SMPS optional for the feeder supply, but I recommend the Daitron HFD series in +-15v. Pretty impressive P-P ripple noise for a feeder supply. Under the Purestream load well under 1mV P-P ripple. No problem for the Belleson SPZ/LT-3042 combo to clean up with the more than 200dB of PSRR!
Attachments
Of course - but this is all irrelevant.
I was simply giving an example to try and show 6dB difference in phase
noise per halving / doubling of clock freq (all other things ie; jitter being
equal) which you didn't appear to be grasping.
Sometimes better to (re) clock directly off XO, sometimes clock buffer can
improve performance, depends on application.
So what are you driving with the clock, flip flops, shift registers?
How many elements in the OP network? Are you using a moving average
filter?
You are keen to talk about your product... now is the chance!
Yes, I have to agree with you here Mike.
Some of those early published AES papers on jitter have not done pro audio any favors.
Where as we both know and agree that reclocking off a VHQ fixed low phase noise oscillator is optimum for sound quality, this almost -never- happens in pro AD converters. Due to the fact that to be marketable they must have ability to lock to external word clock. As such they almost always use VCXO or DDS generated clock driven by PLL. Others just use ASRC.
AES says that Nano Seconds jitter is OK so audiophools concern over our pro audio bretherens practices are met with scorn and belittlement.
Life goes on - I guess it's all part of the territory.
I'm intrigued by this new offering with 4 I2S ports. Anyone know if there's a Linux kernel supporting it?
Firefly | Make technology more simple, life more intelligent. -- Focus on establish an open source platform based on cutting-edge technology , MiniPC, Single Board Computer
That's a pretty cool unit. Looks similar to the Solid run stuff.
https://www.solid-run.com/marvell-armada-family/clearfog/
I've done quite a bit of testing with similar devices. I'm going to be offering up the plans for my Superstream streamer design for free soon. Might be worth trying as I already have about 30 in the field and everyone is thrilled with them. It's all made from off the shelf components that are readily available. My client with the MSB select DAC says it sounds better than the $3500 MSB network renderer.
So I did
Very friendly Niels from Hypex told the following, and is OK with presenting the info here:
The recently presented NCAS500MP is an OEM product; not available for DIY.
Then, more interesting, the active DIY modules series AS and PCS2 will disappear and be replaced by a series of socalled Fusion Plate Amps; scheduled for the third quarter of this year.
The Fusion series will have 8 models in total; there will be 1-way, 2-way and 3-way models based on Ncore technology and DSP based on adau1701.
The Fusion series will be available for OEM and DIY (Fusion...).
Hope this helps to get things clear