Modular multi channel DIY AD/DA Box

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.
But does it sound better at higher sample rates because of the added content or because of the converter working better at these sample rates? I've heard converters that sounded better at higher and others that sounded better at lower sample rates. I think it has more to do with the digital filter. We cannot hear anything beyond 25 khz, so the additional content won't be important. It will not show up in the finished product anyway, unless you release your music to an analog format or a high end format like SACD. But the artifacts caused by the digital filter in the converter will definitely be audible by humans, and I find them more harmfull to the signal than any content beyond hearing range taken out by an analog filter . We could use a switchable filter though, a simple switch at the input/ouput would be a cheap solution.

Biasrocks said:
Wow.  :eek:

Crippling higher sampling rates with subpar filtering, seems to me to defeat the purpose.

I don't want to get into a long debate on sampling rates, but I can't let this one go.

A short example

A crash cymbal has 40% of it's spectral content over 100kHz. Cymbals in particular sound like sushi(*) at lower sampling rates because of the limited bandwidth.

In my experience, many instruments just sound better at higher sampling rates, this is not subtle.  8)

Some background.

http://www.cco.caltech.edu/~boyk/spectra/spectra.htm

A summary of this paper's findings. Column one refers to the figure showing the spectrum in question. Column two identifies the instrument. Column three gives the sound pressure level measured at the microphone. Column four gives the measured frequency extension: For instruments with harmonics, this is the highest frequency where harmonics are still present; for those without harmonics, the highest frequency where the sound is still at least 10 dB above the background. (See text.) The last column tells what percentage of the total energy is contained in the range between 20 kHz and the limit given in the previous column.

Code:
Instruments With Harmonics

Fig.   Instrument                  SPL     Harmonics       Percentage
                                         (dB)     Visible To        of Power
                                                   What Freq.?    Above 20 kHz

1. Trumpet (Harmon mute)    96.      >50 kHz           0.5 
2. Trumpet (Harmon mute)    76.      >80 "               2. 
3. Trumpet (straight mute)    83.      >85 "               0.7 
4. French horn (bell up)       113.      >90 "               0.03 
5. French horn (mute)           99.      >65 "               0.05 
6. French horn                    105.      >55 "               0.1 
7. Violin (double-stop)           87.      >50 "               0.04 
8. Violin (sul ponticello)          77.      >35 "               0.02 
9. Oboe                                84.      >40 "               0.01

 
Instruments Without Harmonics

Fig.   Instrument           SPL   10 dB Above    Percentage
                                  (dB)   Bkgnd. to      of Power
                                           What Freq.?   Above 20 kHz

10. Speech Sibilant        72.     >40 kHz          1.7 
11. Claves                   104.     >102 "            3.8 
12. Rimshot                  73.     >90 "               6. 
13. Crash Cymbal         108.    >102 "            40. 
14. Triangle                   96.    >90 "                1. 
15. Keys jangling            71.   >60 "              68. 
16. Piano                     111.    >70 "               0.02

(*)Sushi as in sounds like ass. :)

Mark
 
living sounds said:
But does it sound better at higher sample rates because of the added content or because of the converter working better at these sample rates? I've heard converters that sounded better at higher and others that sounded better at lower sample rates.

I'm basing my opinions on a lot of different boxes over the years, but currently I'm running an Apogee AD-16X
along with a Benchmark DAC-1.

Mark
 
Have you tried downsampling audio that was recorded at a high sample rate using a high quality sample rate converter (the free Voxengo r8brain is the best one IMO)? I've found that a signal recorded at a higher sample rate with a converter that performs better at these higher sample rates retains this better quality even when downsampled (or subsequently upsampled again). This can only mean that the content above hearing range is not responsible for the quality difference between sample rates, but rather differences in the way the converter handles the signal at these rates, which again points to the filter.

Biasrocks said:
I'm basing my opinions on a lot of different boxes over the years, but currently I'm running an Apogee AD-16X
along with a Benchmark DAC-1.

Mark
 
Hello,

besides the discussion about the sampling frequency (I do prefer higher frequencies, the 20kHz limitation is an invention of the CD recording industry we do recognize higher frequencies unknowingly as some studies with real music instruments demonstrate) I want to focus on the project itself again.
I've now a DAC board but with I2S it is not really useful. ;)
Here I found a ADAT receiver board:
http://electronics.dantimax.dk/Kits/Digital_audio/index.html
Does anybody know this boards? I think I've seen the pcb already somewhere here on the board. However I think it is quite useful. You can combine it with two of my 4channel boards and you have then already an 8 channel DAC with ADAT input.
Besides that for another application I need a multichannel AES receiver/transmitter. Therefore, I've decided to change my plans a little bit and to start with the design of such an interface this weekend.

Gregor, I'm planning to visit Cologne for a sightseeing/weekend trip. The date is not fixed yet, but I can let you know when I'm in Cologne and perhaps we can then do a test.

Raphael
 
I would definately be interested in something like that. I would like to incorporate something like that in a GSSL Clone ping
 
Great project so far.
I am pretty impressed from the pace of progression.
And for sure I am very interested in this project to fire up a DAC with 32 or even 64 channels 44,1 or 48khz for downmix.
(complete mixer dac)

What I am not that interested in is 96khz.
I want to explain why....
First, I am in one row with living sounds, there is no evidence people can percept 'more' at 96khz sample rate. Double blind tests show always that there is no perception of those 'high frequencies', not even 'sublime' or something like that.
All people who claimed to hear above 22khz that I met, could definitely NOT (try with a signal generator! Let someone outside the room switch it on and off...if someone can tell when it is on or off, tell him he is a lucky basterd, like I was iin the age of 20. My army doctor repeated the test three times because he didn't believe the results.)
Second, living sounds is also right about the converter chips sounding better at 96khz. I guess the lower sample rates simply suffer from poorer internal hardware downsampling.- I made tests with high quality software downsampling and yes, I found the results much better than directly recorded in 48 or 44.1. This is especially true for cheaper converters, but I noticed this even on the mytek AD (while it still has exceptionally good quality in 44.1 and 48).
Dan Lavry oftenly states that the optimal sample rate for audio would be 60hz theoretically, and I find 48khz sounds pretty good to my ears, beeing the nearest industry standard in the range.

OTOH, I really like 96khz when tracking because of the low latencies that are possible, takes alot of headaches out of live monitoring.
But mainly, I am interested in a downmix DAC box so latency is not a problem for me....

Another thing to remind is, that I did not find a free/diy SMUX solution that is known working so far.
The wavefront chip has been successfully implemented in a single speed dac project on this forum already - remember the optorec boards from Mikkel.
I guess double speed brings quite some new problems of it's own without really adding that much of value.

The great benefit that I seein this project is:
We should be able to find a complete affordable downmix dac solution for the lightbridge that might even beat the SSL alpha link (at least in channel count....).
Personally I would even go for 64ch of single speed because I still have a (synced) 4 card sonorus daw that is capable of that (and I do not like the idea of mixed brand dac's due to different latecy issues or 8x Behringer ;D).

As for SMD soldering:
Don't be afraid of this.
It is actually very easy, in fact easier and faster than thru-hole. Even with the dac chip.
Raphael already explained the way to go. If you do not believe that it is no problem, take a piece of modern electronics from the junk pile and practice a bit of soldering and desoldering. You will be surprised how easy it is.
I do quite some work on computer mainboards and after some hours of practice you might be able to solder and desolder components that you can barely see without magnifiying glass. You should buy yourself a high quality pair of fine tweezers, though. Money well spent anyway.

Kind regards and many thanks Raphael,

Martin


P.S.:
Voxengo r8brain free is really great for downsampling, I totally agree. High quality does not need a Saracon always...


 
smallbutfine said:
Voxengo r8brain free is really great for downsampling, I totally agree. High quality does not need a Saracon always...

It's the best. There is a very usefull comparison here: http://src.infinitewave.ca

The thing is that most of the downsampling algorithms with good frequency and phase response seriously compromise impulse response. To my ears that's even worse than alising, as the sound loses definition. More or less the same happens with the digital filters in the converter chips. With R8brain free you loose a little at the top of the spectrum (the dog/bat area :)), but the signal stays untouched everywhere else. The current version doesn't suffer from the phase flip anymore. It's better than the pro version and better than Weiss and all the other expensive ones in the test.

In a nutshell a converter chip where the filtering could be turned off completely and be substituted with an analog one would really yield far better dynamics (= impulse response). From a sonic perspective that's far more important than phase or frequency response IMHO.
 
For the sake of clarity and usability, could we try and stick more to the project in this thread, and take the filter and sample rate discussions to the drawing board area?
 
With R8brain free you loose a little at the top of the spectrum (the dog/bat area :)), but the signal stays untouched everywhere else. The current version doesn't suffer from the phase flip anymore.

:eek: :eek: :eek:

going to test it.

1. your description not co-incides with math btw. That little r8brain ringing is misleading, because the ringing itself lies at lower frequency than the "long" ringing of other filters. One lies at 18khz, the others at >21khz.  

2. theres no phase shifting in r8brain free. Its linear phase. As such should be avoided because the pre ringing it embeds(!) is at lower freq's.  

3. so you hear the ringing in the dog / bat area, wow ... Too bad sampling theorem is something different.

It's better than the pro version and better than Weiss and all the other expensive ones in the test.
ha-ha-ha-ha ! My math is better than your matchbox.
 
playboss said:
1. your description not co-incides with math btw. That little r8brain ringing is misleading, because the ringing itself lies at lower frequency than the "long" ringing of other filters. One lies at 18khz, the others at >21khz.  

2. theres no phase shifting in r8brain free. Its linear phase. As such should be avoided because the pre ringing it embeds(!) is at lower freq's.  

3. so you hear the ringing in the dog / bat area, wow ... Too bad sampling theorem is something different.

It's better than the pro version and better than Weiss and all the other expensive ones in the test.
ha-ha-ha-ha ! My math is better than your matchbox.

1. Judging only from the impulse response pictures the ringing is considerably lower with R8brain free, it looks like the same frequency to me, too. But if you've done the measurements it might be at 18khz, but still of considerably lower amplitude.

2. There was a 180° phase shift in the version used in the test, as can be witnessed by the impulse response being "upside down" in the graph.

3. I never claimed I could hear anything in the dog/bat area. Obviously didn't realize the ringing was that high (if this is true). It might still get into the audible spectrum, via intermodulation distortion, for instance.

BTW, I wouldn't know much about the math, I go by observance and listening tests.
 
Total side-note:
What about a serious analog path to the DAC chips? Transformer balanced I/O?
With a bit of tweaking, this box might be able to go from AD16x land to Burl or JCF maybe?
I saw one of these JCF DACs at AES and it looks really cool!
jcfdac.jpg


On any other subject here, I am a noob and leave serious discussion to the experts....
 
OMG. I think I would never go for tubes in the DAC. Serious analog filtering? Sure. But please no tube effect sound.
I would stay clean, affordable and silicon for sure. Leaving 'sound' to the summing stages after the DAC. Much more versatile.

Raphael,
the dantimax.dk boards have been designed by Mikkel for a forum project to feed Apogee cinema DAC's that have been grabbed in a sale.
Very good off-the-shelf diy solution. No need for reinvention of the wheel.
But I guess if we go for higher channel count, design should/might be different because of proper clock distribution to all chips - am I right??
And so YES, I find the I2S DAC boards,*very* useful and versatile!
There are all options for feeding them like one would want to.
Consider 8ch Adat and 4ch AES/EBU or spdif with one clock in one case, all same sound!
Somewhere in future, one may even have an opensource/openhardware madi interface.
Or consider usb audio feed from some of the TI newer chips.

Other point, does anyone have a promising idea for proper s/mux (if 96khz is so desireable by many of us)?
When I researched the wavefront chips some years ago, I did not came up with a conclusion how to do it right. At least, not without some serious microprocessor programming.

An AES/EBU board might be a much simpler design than Mikkels ADAT boards.

I am very, very curious how the test boards compare at the upcoming listening tests with pro grade dac's that you planned.
If they can hold up to, say, benchmark or swissonic or something in that range, I would be in for the next batch of boards definitely.

Thanks alot for coming up with this.
Kind regards,
Martin 

 
Sorry to butt in at this stage, I think this project is uber-cool. Congrats on getting it this far! Hope to be able to put one of these together at some point!

I have one question about the technical side of AD/DA. I was reading a recent Lab thread comparing analog summing to ITB mix, someone suggested that ADAT was inferior to AES/EBU.  Most people (except for that one person who felt pretty strong in his opinion) said that it's all 0's and 1's and that there is no difference, and I am wondering what you all think (those of you who understand the underlying technology, not like me who just plug the digi stuff in and hope it works), especially rafael since he is (maybe) choosing to go down the ADAT road.  Is there any technical difference between these (or any) digital protocol, or is it all the same once it gets to digital?  I mean SSL makes all those format converters, so I always I assumed it was pretty much the same.

sorry if this is OT...

 
mitsos said:
Sorry to butt in at this stage, I think this project is uber-cool. Congrats on getting it this far! Hope to be able to put one of these together at some point!

I have one question about the technical side of AD/DA. I was reading a recent Lab thread comparing analog summing to ITB mix, someone suggested that ADAT was inferior to AES/EBU.  Most people (except for that one person who felt pretty strong in his opinion) said that it's all 0's and 1's and that there is no difference, and I am wondering what you all think (those of you who understand the underlying technology, not like me who just plug the digi stuff in and hope it works), especially rafael since he is (maybe) choosing to go down the ADAT road.  Is there any technical difference between these (or any) digital protocol, or is it all the same once it gets to digital?  I mean SSL makes all those format converters, so I always I assumed it was pretty much the same.

sorry if this is OT...


AES/EBU has advantages over ADAT in terms of jitter. This wouldn't be a problem however, if you clock the converter internally, which is the best solution as long as there is a clock of reasonable quality inside.
 
living sounds said:
AES/EBU has advantages over ADAT in terms of jitter. This wouldn't be a problem however, if you clock the converter internally, which is the best solution as long as there is a clock of reasonable quality inside.

+1.

ADAT is fine for moving bits from one place to another, as is AES/EBU. Neither should be used as converter clock source if you can help it at all, with the standard Wavefront ADAT receiver being somewhat worse than most integrated SPDIF/AES receivers.

For a given amount of money/design effort a good free-running crystal oscillator will always beat a slaved clock jitter-wise. If for some reason you must slave your converter to an external clock, use either WC or AES11.

JDB.
[whether WC or AES11 offers better jitter performance depends on the application. On paper a competent AES11 implementation will beat a competent WC-based PLL, but getting AES11 right is harder]
 
Cool, thanks for the replies guys. I assume then this will be internally clocked? 

I'll go read the rest of the thread now.
 
For this to work in my setup and to provide much need flexibility, I would like the ability to have the option of an external WC source clocking the converter. Unfortunately, without WC this converter will be pretty much useless since I'm clocking my entire studio with an external WC source. I would also like the ability to run @ 96Khz via SMUX.

I totally understand if it's too ambitious for this project, but without this it will severely cripple my ability to integrate it into my current setup.

Mark
 
WC Input is definitely necessary on ANY digital audio equipment.
I'm also watching this thread with great interest, would love to build 32 Channels of AD/DA in good quality, maybe even more.
 
Hehe, I somehow knew this kind of discussion would come up.
Surely, internal clocking wins hands down against WC, IF the clock is done just properly (not even stellar).
And I understand most of you would implement WC. If I remember it right, mikkels boads even have the option for an external WC (at least, WC can be easily added).
But for my side, I would be very glad to build a solution that can internally clock alot of channels leaving external clocking only as an option. In this case, the less external clcking, the better. One 32 channel DAC with one internal clock. This is something I would be after.
(I *guess* that clocking 64 channel would bring some more problems laying out clock signals to the chips. Maybe I am wrong.)
Mainly because this would eliminate the need of WC busses between 8 channel boxes.
I guess this makes boxes like DA16(x) and the SSL alpha link so attractive as well to many people. Less possible clocking/jitter probs, less external cabling etc also means less service and more production.
At the moment, 24channel was the biggest on the market I am aware of. (I do not know if the SSL has more than one clock internally...anyone? Maybe clock routing length is a problem at above 16/24 channels I/O?)
Maybe it is even fine to just spread one clock to all Mikkel optorec boards to get a fairly good result. Well, surely not...

Kind regards,
Martin
 
Anyone have any thoughts on interfacing these DACs with a PC via firewire? I've done a little research, but I know very little about digital communications. Would a FPGA be necessary? a DSP chip?

Here's an interesting firewire/ FPGA card that might be applicable:

http://www.orsys.de/322c6713.htm

The idea is that it would communicate via firewire between a pc DAW and the DAC unit.

It seems like SPI would be more useful than I2S for multi channel use, although if you're going from AES to the DAC, I2S might be better suited.

Anybody else looking into this, or are the majority of people interested in ADAT lightpipe interface?

I can't see using ADAT for more than eight channels- the number of cables/ ports you'd need to do high channel counts would be annoying and/or expensive.

-Dave
 
Back
Top