Russ White said:
Thanks, changed those jumpers to the settings you suggested.
SCK from the receiver was not used (not clearly visible on the picture I posted).
Russ White said:
If the source is 16-bit, leave everything on 16-bit.
Using the metronome to create a 24-bit word length srews up the sound.
Just try Hallelujah performed by Jeff Buckley (from the album "Grace") and you will hear his guitar generate mid-bass that is really non-musical and certainly not produced by his excellent guitar.
Or try Into my arms, written and performed by Nick Cave & The Bad Seeds (from the album "the boatman's call", not from his "best of", because they screwed up the mastering on that album).
In that case you will here the double bass merging into the piano and generate a really messy synthbass.
None of these artefacts appear when you stay 16-bit end to end, I tried this at 128fs and 256fs (wich are nearly identical to me, 256fs sounding the most natural, 128fs a bit more lively).
It strikes me that the datasheet of the SRC4192 is stuffed with technical information, while, to understand that chip, you have to read the datasheet of the AD1896.
Would it be possible to assemble one metronome with AD1896 instead of the SRC4192? I'd be happy to order one.
I suspect the SRC4192 has some serious tradeoffs to be able to generate 192kHz signals...
OneyedK said:
Doing the oversampling itself creates data that needs more than 16 bits to accurately represent it. If you throw away these extra bits by running at 16 bit, you simply add needless distortion.
I appreciate the value of subjective tests, but it's also improtant to compare them with objective understanding.
OneyedK said:
I am sorry, but I am not finding this in the least...
So it makes me wonder if there is something else going on.
The bit clock is generated, it really makes no difference at all if it is 16 or 24 bit. especially since the metronome is the master.
The I2S spec is pretty straightforward. The only difference between 16bit an 24bit is the number of bit clock cycles per sample. Either of these frequencies is very easy to generate from the XO.
I have done quite a lot of switching from 16 to 24bit just to check. And both sound clear as a bell.
Cheers!
Russ
Spartacus said:
Errr... Come again? Why would oversampling imply a greater number of bits?
My objective understanding tells me that if you have 8 samples instead of 2 to represent an 20kHz sine wave,
there is no reason why the 6 new samples couldn't be represented by the same amount of bits.
Russ White said:
There might be something else going on... but I don't know what...
same amp (Musical Fidelity A3 dual mono)...
same speakers (Tannoy system 600)
same CD-player (Denon DCD-1550AR without audio board)
What I suspect is that the original 16-bit signal is beïng stuffed with zero's instead of being requantisised to 24-bits.
That's why I'd like to give the AD1896 a go...
Spartacus brings up a very good point, there is really no difference internally between 24bit output and 16bit. Let me state that better, the 16 most significant bits of the 24 bit output are identical to the 16 bit output version. 
From the Datasheet (page 1)
What that means is, the sample goes through the same 28bit process and conversion regardless of how it is output 16,20,24bit.
In fact the reason its there at all really is just allow you to use DACs etc which can only take 16/20 bit PCM.
I have also tried the other output modes left and right justified. They all seem to work well.
Cheers!
Russ
From the Datasheet (page 1)
What that means is, the sample goes through the same 28bit process and conversion regardless of how it is output 16,20,24bit.
In fact the reason its there at all really is just allow you to use DACs etc which can only take 16/20 bit PCM.
I have also tried the other output modes left and right justified. They all seem to work well.
Cheers!
Russ
I have studied the AD1896 and the SRC4192 very well, here are some key facts:
1) The AD1896 appears to only support a 24bit data path, this is usually less effective when implementing a digital filter than 28 or 32bit. SRC4192 is 28bit.
2) The AD1896 does not support 128fs master mode output.
Aside from the fact that the same sort of transformation occurs for 16bit data on the AD1896 as it does for the SRC4192. The 16bit data is dithered in both chips the same way, just from an (apparently) smaller sample in the AD1896.
If you want to do 192khz output from the AD1896 you will need a XO the running double the frequency as XO on the stock metronome, and you will need to be sure your DAC can handle a master clock that fast. Opus can't.
Now I have no idea which is "better". What I do know is the SRC4192 suites the Opus better because it supports 128fs output.
Cheers!
Russ
1) The AD1896 appears to only support a 24bit data path, this is usually less effective when implementing a digital filter than 28 or 32bit. SRC4192 is 28bit.
2) The AD1896 does not support 128fs master mode output.
Aside from the fact that the same sort of transformation occurs for 16bit data on the AD1896 as it does for the SRC4192. The 16bit data is dithered in both chips the same way, just from an (apparently) smaller sample in the AD1896.
If you want to do 192khz output from the AD1896 you will need a XO the running double the frequency as XO on the stock metronome, and you will need to be sure your DAC can handle a master clock that fast. Opus can't.
Now I have no idea which is "better". What I do know is the SRC4192 suites the Opus better because it supports 128fs output.
Cheers!
Russ
Russ White said:Spartacus brings up a very good point, there is really no difference internally between 24bit output and 16bit. Let me state that better, the 16 most significant bits of the 24 bit output are identical to the 16 bit output version.
The steps for 24-bit are quite a bit smaller than the steps for 16-bit...
The output of the Wolfson varies from 0 to 5V.
With 16-bits, there are 65536 steps of 76,3nV each.
With 24-bits, there are 16777216 steps of 59,6µV each.
If the 16 MSB's of a 24-bit signal were truly identical with the 16-bit signal, would the Wolfson's output still vary from 0 to 5Volt???
OneyedK said:
It does not vary from 0-5V in any case
The first 16bits of of a 24bit PCM signal are exactly the same magnitude voltage wise as those of the 24bit sample.
But a short answer to your question, the WM8740 treats the sample the same it was input as 24bit or 16bit. The 24bit sample is just 8 bits more precise.
Cheers!
Russ
OneyedK said:
Ok, think of it like this: We have some data, the number 5. 5 takes one digit to represent it. But now lets do some maths, lets divide our data (5) by 4. The result is of course 1.25 .... but now we need three digits to accurately represent our data.
Oversampling is just maths. Once you do some maths to 16 bits of data, you need more bits to keep accuracy.
The AD1896 is much the same.
Hi, Russ!
It's been very interresting following the evolution of the Opus dac, and I will definitely be getting one myself (as soon as my ear infection passes
)
In the mean time I wondered if you could help me with a more general question about the WM8740 dac, since you have quite a bit experience with it.
I'm modifying a cd-player which uses the WM8740 in differential mode (mono mode that is ). I tapped the signal from VOUTRP on both dacs, and fed it through 4,7uF coupling caps, to a Yamaha amp of unknown input impedance and capacitive load. The sound is generally much more airy and open and with no apparent HF noise, but the bass seems to roll off pretty high. Any idea what this could be? Should I use bigger coupling caps? Is the load from the amp to heavy? Or should I tap the signal another place? Any ideas?
Best regards from Norway
It's been very interresting following the evolution of the Opus dac, and I will definitely be getting one myself (as soon as my ear infection passes
) In the mean time I wondered if you could help me with a more general question about the WM8740 dac, since you have quite a bit experience with it.
I'm modifying a cd-player which uses the WM8740 in differential mode (mono mode that is
). I tapped the signal from VOUTRP on both dacs, and fed it through 4,7uF coupling caps, to a Yamaha amp of unknown input impedance and capacitive load. The sound is generally much more airy and open and with no apparent HF noise, but the bass seems to roll off pretty high. Any idea what this could be? Should I use bigger coupling caps? Is the load from the amp to heavy? Or should I tap the signal another place? Any ideas?Best regards from Norway
I have successfully installed my opus into my car audio system and it sounds really really good . Enough that I've decided to actually put it in instead of just testing it. However, I've found out that using my crossover to adjust the volume is less than ideal given it's in my trunk. Any suggestions on what I can use/build to provide a nice volume control that will be accessible from the dash? One consideration is that I will need to control two opus boards, both running in stereo to provide for front and rear channels (I'm not much of a fan of using cable splitters)
At the moment I'm using the opus single ended using only the + and G, though in the future I plan on throwing in ballsie's after I figure out how to do the volume control
. Enough that I've decided to actually put it in instead of just testing it. However, I've found out that using my crossover to adjust the volume is less than ideal given it's in my trunk. Any suggestions on what I can use/build to provide a nice volume control that will be accessible from the dash? One consideration is that I will need to control two opus boards, both running in stereo to provide for front and rear channels (I'm not much of a fan of using cable splitters)At the moment I'm using the opus single ended using only the + and G, though in the future I plan on throwing in ballsie's after I figure out how to do the volume control
Painkiller said:
Thanks.
Very likely the input impedance of the stage after the DAC is too low for 4.7uf caps.
To give you an idea we use 22uf on the output of the Opus so that it can drive as low as 1K input impedance with a corner of 7.2hz. of course it is much better with an impedance of > 2K.
4.7uf into 1K has a corner freq of 33.9 hz... ouch...
Cheers!
Russ
ok guys, I found the problem, the group delay jumper should be open or high,
not low, as I set it :-(
The artefacts disappeared with that setting...
Russ, you were right from the beginning, but this fairly undocumented group delay feature really set me on the wrong foot...
Sorry...
Yeps, I can hear that now with my own ears...
Thanks for the explanation Spartacus. I thought that was covered by the internal 28-bit for filtering,
didn't occur to me that you actually could loose accuracy when switching the output of the ASRC to 16-bit.
But yes, you do, it's audible now...
Thanks to everybody who cared to share some knowledge!
not low, as I set it :-(
The artefacts disappeared with that setting...
Russ, you were right from the beginning, but this fairly undocumented group delay feature really set me on the wrong foot...
Sorry...
Russ White said:
The first 16bits of of a 24bit PCM signal are exactly the same magnitude voltage wise as those of the 24bit sample.
But a short answer to your question, the WM8740 treats the sample the same it was input as 24bit or 16bit. The 24bit sample is just 8 bits more precise.
Yeps, I can hear that now with my own ears...
Spartacus said:
Thanks for the explanation Spartacus. I thought that was covered by the internal 28-bit for filtering,
didn't occur to me that you actually could loose accuracy when switching the output of the ASRC to 16-bit.
But yes, you do, it's audible now...
Thanks to everybody who cared to share some knowledge!
BrianDonegan said:Ballsie will be tough in a car, as it requires a split rail supply (-, gnd, +). You could try an SMPS for it.
I have ideas for volume, but need to drink more coffee before replying...
Ah you're correct, guess I'll just stick with simply using the + and G
Enjoy your coffee, I just woke up and need some myself Something like this would not take much effort:
12V IN +/-12V OUT
Oh and hi Russ btw long time.. And yes I will be getting me an OPUS sometime
/Tobias
12V IN +/-12V OUT
Oh and hi Russ btw long time..
And yes I will be getting me an OPUS sometime /Tobias
Something like this, available in the US and for much less money, could be a good choice as well:
http://search.digikey.com/scripts/DkSearch/dksus.dll?Detail?name=PT5062A-ND
http://search.digikey.com/scripts/DkSearch/dksus.dll?Detail?name=PT5062A-ND