ackoDAC based on ES9018

[acko]

Originally Posted by hochopeper

I am not familiar with any of the asynch clocking options since I have never investigated them. Would be worth asking Acko about the 'Turbo Master Clock' though. I am not sure if it is able to work in asynch mode, its description has reference to modifications needed on transport end, which may or may not suit your application.

The AKD12P DAC module has provisions for an on-board XO or an external unit. I am leaning more towards the external one as it give options to experiment with different clock modules. Also, if the transport provides MCLK, this can be used and saves the cost of a dedicated Clock module.

As for the Turbo Clock, see how it is connected to the DAC as in the pic. The Master Clock output of the unit is connected directly to the DAC Master Clock input. There is no interaction with the EXA2UI transport at this stage, so in effect this is Async mode. Of course, you could just whack in a much cheaper clock module for this case.

The SQ of the setup is really good, so it goes to indicate that the NDK XOs and signal processing of the Turbo Clock are working well. I haven't done direct comparisons yet but based on memory, appears sounding better the Crystek types used previously with this DAC. So standard Async mode with low jitter XOs does sound very good. The other thing I have noted is that the DPLL BW of the DAC has to be at least Medium-High @192KHz and possibly Highest BW for 384KHz for stable lock.


Based on the above results, I am still undecided as to whether to go further and mod the EXA unit for synchronous operation. For this case, 2 slave clock outputs from the Turbo Clock goes into the Clock inputs of the EXA unit. The two on-board XOs need to come out. Then I have got to tap the embedded Fs signals of the EXA to switch the Turbo Clock accordingly. Fairly, intrusive process and may cause damage or void any warranty of this fairly expensive EXA unit, so treading carefully

Anyway, the benefits of going forward will be to get the DAC to lock at the lowest BW in the Synchronous mode. Plus DAC operating at 98.304MHz would make the jitter reduction even better compared to just using the simpler 24.576MHz MCLK directly from the EXA unit.

Also, the Master Clock from the Turbo unit is inverted w.r.t. the Slave Clocks to reduce noise further as indicated in one of the ESS App note, so all lined up!

Will all these translate to noticeable improvements?
Will find out soon ...

[acko]

In response to enquires, this 9016 DAC will be used in mid-range products with more integration like IV and controller to follow soon. This chip itself is only a tad lower in performance than the big brothers 9012/18 but has simpler power supply requirements. Other differences are no dual mono capability or lock indicator

Also planned, USB DAC based on Amanero.

[qusp]

just wait for your fifo for sync mode mate. only thing you lose is multichannel and DSD and you can then have it on everything

strange, I can get stable lock at 192 running async with fifo in 'lowest' DPLL but as with anything, the dac needs some time to warm up, perhaps you havent given it this time in your testing? much research has been undertaken to get to the bottom of this, it seems environmental effects, RF, spill from mains, ground bounce etc all affect this, as it exists still in what should effectively be jitter free conditions.

To this end, one thing you might consider mate is putting a digital isolator on the MCU, i'll be running the MCU USB connection and its i2c through an isolator, a spare isolator position was put on the fifo ISO board for just this sort of thing.

are you planning multichannel versions of the 9016?

[hochopeper]

Thanks for the info on your work Acko!!

I am a bit confused if dahlberg has been asking about AKD16 or AKD12 though. Earlier questions were about the differences between the two, and my responses were about trying to implement an external clock with the AKD16. Then it seemed like he was referring to the onboard osc, which is only on the AKD12.

Though I am a bit puzzled, what is the implication of having no lock indicator? Is it purely a user interface feature or does it have some technical significance that I'm not noticing having only had limited experience of hearing 9012 DACs of others and studying their datasheets. What is the lock indicator typically used for?

Regardless I think Acko's covered most bases in the discussion above!

[qusp]

nah just user interface and it also gives visual indication of dropped connections etc, its a bit annoying, I wonder if it can be deduced/sniffed some other way? of course most transports used with it will have some form of lock indication.

in case part of your confusion was related to my reply, I was referring to this line from post 511

Originally Posted by acko

The other thing I have noted is that the DPLL BW of the DAC has to be at least Medium-High @192KHz and possibly Highest BW for 384KHz for stable lock.

[acko]

Actually, I found it useful.

In my earlier post on lock issue with low DPLLBW @192KHz, it is the tiny intermittent blink on the indicator that tells it. No dropouts actually heard, possibly smeared as noise and affecting the SQ.
Even if transport has lock indication, it is the DAC that does the final processing.
BTW, the Hiface USB connected this way did better with BW down to 'low' and stable lock @192KHz. Some complex interaction is happening! There is this main board shield surrounding the DAC that is still floating. I will try to ground or earth this and see.

Thanks for all your feedback and suggestions

[dahlberg]

I was asking about the "akd16".
There are pads on the board marked "mlck", I was assuming they where
supposed to be populated by "the simpler 24.576MHz MCLK" or some
"semi high end solution".
Not at all if an external master clock is used. Maybe a misstake on my part or... ?

I will try to make a description of how I'm going to use these things.

The center of all functions will be the dsp filter that "Chapark" is working on
and at the I2S input I will have "Loriens" usb reciver. I will also be in need of
both digital and analog inputs of the dsp. If I'm not misstaken that will
disqualify the possibility to use the Wave IO as a masterclock, right? The
plan for outputs are to connect two stereo dacs to I2S and use ESS digital
volume control before sending this to four channels of John Broskie's "Unballancer".

People are giving the ess volumecontrol god reviews, I really hope that this is true.

I have also read somewhere that if syncronus mode failes it will be quite
nasty, so I'm leaning to asyncronus mode for now.

If I'm going for the "akd16" do I really have to get an external clock to get a
decent result? And if so, can I run both "akd16" on the same master as
asyncronus or will I need two external clocks?

[qusp]

haha so do I, I meant its a bit annoying its gone, sorry for the confusion =)

with the fifo with the isolator and running synchronously it makes it so the clock belongs to the dac again, not USB ground, or subject to the high slewing FPGA, but the reclock stage is after the isolator, so no jitter added by the isolator like normally.

[qusp]

you cannot run multiple dacs asynchronously and expect the phase relationships to be accurately matched like you need when using them in a digital crossover

[dahlberg]

Originally Posted by qusp

you cannot run multiple dacs asynchronously and expect the phase relationships to be accurately matched like you need when using them in a digital crossover

As I suspected then.