Looking for upgrade from Apple TV, would like to keep using iTunes and wifi


My wifi-based sound has never matched my dear departed Arcam Alpha 9 CD player.  I've read forum after forum but I don't see a simple solution.  One possibility is that an audible problem occurs when the bit rate is changed from the CD native 44.1 kHz bit rate to the more traditional 48 kHz.

Here is my wifi system: iMac (iTunes) -> wifi -> Apple TV -> optical TOS -> PS Audio Digital Link DAC. 

The 300GB of music has been ripped from CD and is stored using Apple Lossless compression.

Yes, I know it's a lot of Apple software and gear.  I want to keep the iMac and iTunes and the wifi link.  I like the playlist features of iTunes, including the Genius feature. 

The Apple TV comes with a remote control and I can access the iTunes playlists directly, using my TV as a monitor, even though the iMac is in a separate room.  Once the playlist is playing I turn off the TV.

I am an electrical engineer and I've even worked measuring jitter on serial bit streams.  I understand that converting from 44.1 kHz to 48 kHz will cause some sort of interpolation error whereas converting from 48 kHz to 96kHz can be done with no trouble.  However, the Apple TV outputs 48 kHz so the bit rate must have been changed somewhere along the chain.  My PS Audio DAC won't even lock to the native CD rate of 44.1 kHz.

Are there any suggestions for a wifi receiver and DAC that will match the sound of a good/great CD player?  The budget is about $600, with an upper limit of $1000.

Thanks
pmiguy
There's always issues doing sample rate conversion it's not as simple as saying that 2x is going to be better than 1.8x.  Some algorithms are better sounding than others though.

I'm not sure I understand the problem with your DAC though, 44.1 should be easy for any DAC.

Wyred4Sound Remedy is an asynch. sample rate converter and it samples everything to 96/24, around $300-$400 new depending on the sales.  I have one, and used it for a couple of months, works peachy.

Are you willing to change to a local music server? Then you could use USB 2.0 to feed  your DAC which is often the superior connection.  I built a Linux server for $600 and got Web, Android and iPhone apps. :) I have it feeding a USB 3.0 hub, which is then connected to my DAC, but there also an optical output.  Double check your DAC is USB 2.0 and driverless on Linux. If it is, then you are all set.
Wifi through Apple TV and optical to the dac does not sound very good.  No idea why not, but it doesn't.  You will have much better results doing a usb connection direct to the dac.  The cheapest option is to buy a Mac Mini and use it as a dedicated server, if you are sold on iTunes etc.  It also allows you to use third party software like Pure Music or similar, which will not work over wifi.

It does not have to be dedicated when using network (USB, WiFi, Ethernet etc) since computer timing is irrelevant.  I use computer for other tasks while playing music over WiFi.

While 48kHz to 96kHz conversion is easy (PLL - even ratio) 44.1kHz has to be interpolated to 48kHz and it won't be accurate because frequencies are too close.  I'm surprised that your DAC would not work with 44.1kHz - a basic frequency of CD.  

I use WiFi with Airport Express and jitter suppressing DAC.  Adding jitter suppressing async. rate converter like  Wyred4Sound Remedy,  as suggested by eric_squires, is a good solution.  It won't fix your 44.1/48 problem, but will make your DAC less sensitive to jitter (that converts to noise on analog side).

Airport Express is very basic but there might be better choices.  You could also switch to async. USB DAC or USB/Spdif converter. Async USB DAC receives music as data in packets.  New timing is added with low jitter internal DAC's clock. Computer timing becomes irrelevant.
@kijanki

I'm afraid there's some misconception. Interpolation is the process of mathematically guessing missing samples. It is still interpolation if the new sample rate is 2x or 2.18 and there will still be precision issues. The latest rate converters can handle either equally well.  Whether that is good enough or not is another story.  Music is rarely that linear, so guessing a sample between to adjacent samples, even if dead in between, is still guesswork.

There are some advanced algorithms which use splines, but rarely do manufacturers disclose what they are doing.  Wadia was a rare exception, but it was their claim to fame.  At least if you use Linux you can find some of these and apply them to do SRC for you.

Best,

Erik
Erik, you're probably right.  I assumed that interpolation with integer ratio, commonly done in oversampling, will be more accurate. Perhaps Apple TVs are just poorly designed, since many people have complained about their performance with 44.1kHz source.
I don't see an easy solution for me yet.

This is getting pretty technical but here goes:

My DAC has two lights on the front, 96 kHz and 192 kHz.  That's why I assume it will lock to 48 kHz (double rate upsampling) but not lock to 44.1 kHz.  An interesting experiment for me would be to run a CD player output into my DAC and see if the LOCK light comes on (assuming the CD player would output 44.1 kHz).

Any upsampling DAC will reclock the data using a PLL generated clock.  The PLL has natural jitter reduction.  I'm not sure that adding another jitter reduction device would help much.  Most jitter reduction circuits are probably just using PLL reclocking anyway.

Interpolation for a 2x sample rate should be fairly easy, maybe just average the adjacent samples.  Interpolation for a 48/44.1 ratio would be computationally complex.  I doubt that the AppleTV does the interpolation well.

I do have a 10 year old Mac Mini that I can pull out of storage.  It has USB 2.0 and also an optical output.  Maybe the local server is the way to go.



Pmiguy, PLL reduces jitter but not up to the levels of async rate converter. My Benchamark DAC1 rate converter AD1896 does very fancy operation to achieve fantastic results. From the datasheet:
(full datasheet: http://www.analog.com/media/en/technical-documentation/data-sheets/AD1896.pdf)
The output rate of the low-pass filter of Figure 5 would be the interpolation rate, 2^20 x 192000 kHz = 201.3 GHz. Sampling at a rate of 201.3 GHz is clearly impractical, not to mention the number of taps required to calculate each interpolated sample. However, since interpolation by 2^20 involves zero-stuffing 2^20– 1 samples between each fS_IN sample, most of the multiplies in the low-pass FIR filter are by zero. A further reduction can be realized by the fact that since only one interpolated sample is taken at the output at the fS_OUT rate, only one convolution needs to be performed per fS_OUT period instead of 2^20 convolutions. A 64-tap FIR filter for each fS_OUT sample is sufficient to suppress the images caused by the interpolation. The difficulty with the above approach is that the correct interpolated sample needs to be selected upon the arrival of fS_OUT. Since there are 2^20 possible convolutions per fS_OUT period, the arrival of the fS_OUT clock must be measured with an accuracy of 1/201.3 GHz = 4.96 ps. Measuring the fS_OUT period with a clock of 201.3 GHz frequency is clearly impossible; instead, several coarse measurements of the fS_OUT clock period are made and averaged over time.
Output D/A converter operates at 110kHz only (could operate at 192kHz) to obtain lower THD distortions. I posted this technical excerpt from the data sheet only to show how complicated operation of just one small chip can be. Jitter artifacts are at very low level, but are not harmonically related (like THD) to the root frequencies thus more audible. Noise produced by the jitter is proportional to signal level and not present without it.
The issue isn't just of finding the right "sample count" but also consider what happens with amplitude math. What do you do when your interpolated step is 1/3 of the way between the two bits?

To be honest, I don't like the idea of most ASRC's not because of the math but the algorithms. Schiit has a great white paper on this, about closed-form rate conversion. With an ASRC you give up (completely) on the idea of bit-perfect transcription in exchange for timing (jitter) improvements. Still, the Remedy does a very fine job, and from what I've read, is especially good with non-audiophile sources. That was certainly my experience with Internet Radio, but I did not try it with AppleTV.

I will also say that I upgraded my DAC and the improvements went away.

Best,


Erik
I would like to add to comparison between PLL and reclocker that reclocker does not replace PLL but works in addition. Any timing imperfections left by reclocker will be further repaired by PLL.

Any upsampling DAC will reclock the data using a PLL generated clock. The PLL has natural jitter reduction. I’m not sure that adding another jitter reduction device would help much. Most jitter reduction circuits are probably just using PLL reclocking anyway.
Jitter reduction circuitry used in reclockers, like REMEDY, is not based on PLL but rather on Asynchronous Rate Converter (does not use PLL).  PLLs have limitations.  Since based on phase comparison they can actually amplify phase noise of the reference clock.  The best scheme IMHO, that is both bit perfect and performance limited only by the quality of the internal reference clock is Asynchronous USB DAC.
By the way, I've never heard of a DAC that couldn't do 44.1. I think you need to find your original docs. :)

Best,


Erik
I like the idea of Async USB.  Why rely on the computer to do the timing?  It's better to let the DAC do the timing. 

I'll see what I can do with my old Mac Mini and the DAC that I have.  Eventually I may be in the market for a USB to SPDIF convertor or a new DAC.  I may look at the Remedy solution also.

Thanks everyone for the quality discussion!