freeDSP V2.0 (ADAU1452) developement thread

Boards are on the way

Boards have been shipped today. A little late, since I ordered on 30.9 :(.
Hopefully the shipping will be a little bit faster :confused:
 

Attachments

  • elecrow.png
    elecrow.png
    472.6 KB · Views: 1,063
Last edited:
I've ordered from Elecrow and they are very fast - the times they state (4-7 dats) are more or less correct, and they will take a picture of your pcb before they send it out (as 'proof' that they're not faking shipping time as production time). The longest waiting time is probably for the shipping, depending on what shipping option you chose, and where you live.

You may give pcbway.com a try. With DHL-Express it tooked 4 days all over. (manufacturing+shipping) :D
 
Hi Pitrsek
nice to see you finalized a board. great stuff.
I m afficionados of 1452 as it is amaizingly powerfull.
I have a bit of a project in mind too, with this ADAU and dual pcm1792.

anyway, if you want to be safe, I d recomend to find a way to not consider ADAU as a master clock... the jitter even with a good supply migth not be enough for the hifi, from some feedback provided on another forum in europe.
Also be carreful, all the nice ASRC inside the adau are great but they just transfer jitter into thd. My personnal approach is to use an SI570 for the MCLK and consider the adau as slave, with its cycle synchronised on LRCLK. Now you need a uP to synchronize the frequency of the SI570 with the incoming rate and by reading the ASRC ratio you can reasonbly do that and then organize to have a ration of 1 or 2 or 4. Enjoy the next steps, I ll stay tuned on this thread.

BTW how will you solder all these tiny pins of the adau ??? oven ?
 
ADAU ASRC Jitter

well reading my above statement about jitter and ASRC in the ADUA1452, I should give more fact. Here is a plot I did early 2015 by connecting the Minidsp xmos usb streamer with the ADAU1452 on the AD Evaluation board.
An externally hosted image should be here but it was not working when we last tested it.

the plot shows a 1khz tone going trough 2 internal asrc (on input and output), with the ADAU runing at 96khz cycle with its own onboard clock.
the Minidsp streamer works assynchronously but is know for its very bad jitter as the 24/22mhz clock is provided from a PLL chip (see a topic on this here:
http://www.diyaudio.com/forums/digital-source/244042-master-clock-isolator-minidsp-usbstreamer.html

So question is can you here the -140db effect of the dual asrc jitter, probably no:eek:. On the other end, REW is computing a THD+N of 0.009% and this 10 time more than a 0.001% which you can get out of a decent DAC implementation...

So another approach is to drive your DACs MCLK with a stellar clock, like the one describe on Herbert site here:
https://www.by-rutgers.nl/rutgerS-Clock.html
and use this clock as the XTALIN for the ADAU too. Then the ADAU can be master I2S to the DAC but you may need to consider reclocking of the I2S line if you dac is not already imune to the is2 jitter (seems the PCM1792 has its own reclocking :p) as the ADAU will provide more than 150ps rms (and some we can here jitter effect down to 1ps ish)

this way you avoid the burden of the clock sync between source and DAC, and trust the ADAU ASRC to make the conversion in a reasonably good way.
Also it s quite good to use ASRC at the output of ADAU, for their good band filter with sharp rolloff instead of the DAC internal digital filter (with a ratio of 1:1 or 1:2) (cant find the diyaudio post about that)
 
Last edited:
Hi maxidcx,
the idea indeed is to use ADAU as master clock for all - signal source(RPi in this case), and dacs. I've selected NDK NZ2520SD as a clock source, its connected to low jitter clock driver. You can connect differet clocks via u.fl - there's a multiplex on clk driver input. I have to reed up a bit about jitter measurements, so i can make my own. Thx for your comments and measurement.

Boards still haven't arrived.
 
Jitter at -140dB, zero probability of audibility.
Harmonic distortion at 0.0000%, zero probability of audibility.
Noise at 0.009%, -81dB, conditional with some probability.

If the noise is present on a continuous basis and the device is gain stuctured such that master attenuation is implemented prior to the device, the noise will be amplifed by the power amplifier(s) gain. Note that 0.009% (-81dB) noise is > 10dB greater noise magnitude than a mediocre DAC or CD player. A daughter board for the master volume attenuation - via optocouplers inline with output opamps' inputs -could be a good workaround. The microcontroller regulating the optos could be operated from a single potentiometer or remote data pot.
 
when it comes to volume, a R2R network based on DAC8812 at the input of an inverting amplifier seems to be a great solution in term of simplicity and performance. Mirand audio just released a preamp with this solution. Then a buffer based on the Ovation preamplifier (hifisonix site) seems extremly good and able to drive headphone.

Any news on your pcb assembly :) ?
 
Hello,
I still do not have boards. 25day shipping limit was a few days later than i stated(made mistake in calendar). But still, the boards should have been with me for roughly two weeks now. Right now I'm waiting to be refunded by elecrow.

Speaking about elecrow - I was not really impressed with their communication and attitude. They asked me to wait, and wait, and wait a little bit more. A professional thing to do would be to offer refund on 26th day of shipping(if you declare to your customers that the shipping will take 25days). Or make the boards once more and ship them via express. This I would consider appropriate action to remedy the screw up. If you consider elecrow, I would strongly recommend better shipping method than regular china post. And then I would reconsider once more if I really want to make a business with this company.

I'll get the boards made locally, which will take another week.
I did not ordered them sooner because I was hopping for the boards from elecrow to arrive :(
 
Well, you get what you pay for. Why blame Elecrow for the shipping you had selected?

China Post might be to blame for slow delivery but mostly it's the local customs inspection. Shipping stuff from China to Germany includes 7-10 days customs clearing on a regular basis.

Never had problems with Elecrow regarding shipping - China Post is just slow but they never lost a parcel.

If you want your boards super fast and somewhat cheap, try pcb-way.com
 
when it comes to volume, a R2R network based on DAC8812 at the input of an inverting amplifier seems to be a great solution in term of simplicity and performance.
I think this kind of thing is subject to preference. I'd be apt to recommend an analog volume control on the output of the board to prevent trunicating bits and reduce the decline of the SNR. It would allow attenuation of the noise floor in proportion to the reproduction signal for lower playback levels.

The PGA4311 is a 4 channel volume control with 31dB gain and -95dB attenuation. THD+N at 1kHz is stated as better than .0005%. It can be daisy chained between multiple (freeDSP) boards, allowing the volume of 4, 8, or 12 channels to be controlled from a single rotary encoder. This could make a worthwhile accessory for DSP boards, effectively converting them into a DSP preamp.

Seems like a good idea.
 
Last edited:
Hello,
I've recieved boards from a local manufacturer :)
Tomorrow I'm leaving for Germany to visit my FreeDSP friends, then I start to populate and test it. Hopefully I'll have something to share early next week.

At the moment, it's not published for a group buy, it's not even tested yet. I haven't made my mind about it. Either there will be a pcb group buy, or fully populate boards will be available. Dunno yet.

About Elecrow - In the end the communications returned to what I would consider "forthcoming and appropriate". I'll probably try them again in the future. Why blame Elecrow? Because I have made deal with Elecrow, I did not made deal with china post. I gave them(Elecrow) my money in exchange for my pcb, that would arrive by agreed deadline. That's the deal. If the shipping time on Elecrow site was "up to two months" I wouldn't say a word, and would be patiently waiting. In that case I would not probably selected china post either....
It's my first experience with china post, I'll avoid them next time. I can understand that it's not their fault, its fault of one of their business partners. But if this problem is not uncommon, they should warn their customers , extend the shipping time, or not offer china post shipping at all. My 2 cents.
 
Update time. Board is almost populated - only DSP and clocks are missing. Power supplies are working nicely. I just need to make some output impedance measurements to make sure that my design matches reality(to see if simulations fits). Fit to RPi is nice. Next time I probably use shorter header, so I can use 1cm screw columns. This way I need to insert a nut under the column to adjust columns height. After impedance measurements I'll populate oscillator, measure clocks and install rpi to see if it works or not :).
 

Attachments

  • WP_20151206_16_28_46_Pro.jpg
    WP_20151206_16_28_46_Pro.jpg
    615.9 KB · Views: 708
  • WP_20151206_18_05_15_Pro.jpg
    WP_20151206_18_05_15_Pro.jpg
    566.4 KB · Views: 683
  • WP_20151206_19_03_56_Pro.jpg
    WP_20151206_19_03_56_Pro.jpg
    539.9 KB · Views: 672
  • WP_20151206_20_07_13_Pro.jpg
    WP_20151206_20_07_13_Pro.jpg
    480.8 KB · Views: 652
Hi,

I'm also working on a DSP project with ADAU1452.
Tried RPI as I2S source, but in the end the Pi isn't that good for this purpose.

Is there really a way to use it as a slave?

Beaglebone is maybe a better choice to act as I2S slave (MCLK IN available on header). Also the onboard flash is significantly faster than SD card, and also robust against hard shutdowns.



At the moment, the first rev:
AD -> DSP -> DA = 0.0006% THD+N single ended 0dBV, differential 0,0004% THD+N.
An externally hosted image should be here but it was not working when we last tested it.




The second rev preview.
An externally hosted image should be here but it was not working when we last tested it.