Itunes Ethernet>USB bridge: Why isn't modding the Airport Express popular? (plus story of unreleased Hovland player)

[Superdad]

Long, long ago, on AA in 2005, Gordon Rankin revealed that the DAC and S/PDIF portions in an Apple Airport Express are on a removable board and fed by a USB signal. He reported (but never posted pics or diagram) that it was a cinch (once he broke open the case!) to solder the leads from a hacked USB cable to the header with the USB signal. In an instant, he pointed the way to what was then--and still is AFAIK--the ONLY Ethernet/WiFi to USB bridge that can be used with iTunes!

 

[Now for a long digression: I was hanging around in those earlier days, in that thread and a few others, asking questions about elimination of S/PDIF, I2S possibilities, and USB stuff. Gordon, Nugent, Swenson were all still in early discovery mode on some of these topics. That's actually when I tapped John Swenson (a very talented audio and chip designer near San Jose who has blossomed into a patient and informative writer--read him!) for contract work for us at Hovland Company (doors closed this July after 10 yrs.; http://www.stereophile.com/news/hovland_calls_it_quits/) on an ambitious, multi-year development project which, in the end only showed in partially functioning proto form at CES. I recently tallied up that we poured about $50K into the project, with about $22K of that being stipends for Swenson. The technical and feature brief for the product was ahead of its time, and in some ways still is, although one can see elements of other products in our plan and vise versa. It was to be called the Hovland Digital Music Source (DMS-1), a full-on assault (read expensive) at a bridge between computer audio and ultimate audio. While we included a slot-load DVDROM drive so it could come out of the box and just play CDs, that was mainly to satisfy our many conservative (read older) overseas clients who wanted a matching CD player from Hovland. The real fun--aside from the fact that we spent over a year on the DAC portion (settling on a quad of BB PCM 1704K run NOS into an all-discrete output stage) and multiple power supplies--was that we used a really small computer-on-module (something like this http://compulab.co.il/x270cm/html/x270-cm-datasheet.htm) loading tiny LINUX. It was shielded plus isolated in the "dirty" lower section of the box by a 1/4" plate. Besides being used for driving the 4.3" color LCD, doing IDE for the DVDROM, ethernet, etc., John had it interfaced via GPI lines to a custom FPGA--into which he was programming all sorts of magic. The special-for-audio USB port (the computer module's standard USB was to be used for NAS, flash drive, whatever) was initially the Centrance-firmware, adaptive TI-USB1020A unit being produced @ Empirical, but there was other fancy reclocking on the DAC board. Later on, Centrance got strange about Steve's plan to sub-license, but by then John hatched a better plan so we could move closer to an async approach. (Swenson's latest idea before we had to stop funding was for skipping USB receiver chips altogether and embedding a USB version of what are called "cores" into the FPGA). Custom/selectable digital filters were also planned for the FPGA to try to better the already wonderful sound we were getting from the NOS 1704 (such would also keep Atkinson from trashing us in review measurements).

I am sure I am leaving out some cool planned features. Oh yeah, since the box had ethernet, online radio station streaming was a no-brainer, as was pulling down the track list for any CD that got inserted. Of course it was being programmed to work like a memory player: rip the disk in (still playing immediately as it had to function just like a standard CDP) then spit it out and play from ample RAM. At CES everyone thought it was to be a music server and asked about the hard drive. But that was never our intent. Who wants to manage a music library on a tiny screen? We had no interest in writing a bunch of library navigation code for a 4.3" display (though there is probably some LINUX s/w out there for such). With the computer in it, I imagine some geeks would probably even write their own apps or hacks to the Hovland DMS-1. But we just wanted to deliver the very best sounding Ethernet/USB DAC, and have people manage their library with whatever player/library management s/w they chose.

And that my patiently reading comrades, brings me to one of the most exciting yet frustrating parts of this development saga (which continues somewhat in John Swenson's home): ETHERNET STREAMING.

 

For the Hovland product, the trick USB input was going to sound great, but USB still requires the music-browsing computer to be close by. I (and many of you) want to put the computer elsewhere (at least not be stuck having it right there next to the DAC/amps) and maybe select music with an iPhone or some such. And we wanted people to be able to use whatever nice library s/w they like (iTunes, Media Monkey, Jriver, Foobar, whatever). Sure others have done proprietary things, and yes, LINUX Slimserver/Softsqueeze could be loaded onto the computer module with tunes "pushed" from the host PC. Yuck. What I think many of us are waiting for is the ability to fool the OS into thinking a given IP address on our local network is a sound card. Basically an open standard Airport Express (in that the AE is, among other things, an Ethernet DAC). The best hope right now appears to be netJACK (a work in progress building on JACK http://jackaudio.org), which is available free on multiple platforms, and Swenson recently posted (on AA) some of his experiment results.

 

By the way, despite some misconceptions to the contrary in a recent CA forum thread, there are reasons to believe that properly done ethernet streaming could sound better than even async USB or Firewire. Maybe John Swenson will join us over here at CA and explain his thinking on that matter sometime. I am NOT an engineer, so I just try to keep up.

Thanks for listening my story about the product that never arrived. Lots of people heard the prototype at two year's CES (including Connaker, Harley, Atkinson, Fremer, etc.--along with music lovers who mattered more!). A couple of pics are included below (in CES shot the player is the one on the left). Also shown is my listening studio system rack (before bringing out my turntable). On the shelf below the Mac mini is the DAC as described above but with no computer module or display. So it is just a Hovland/Swenson USB DAC.]

 

All of the above FINALLY brings me to the original reason for my post! Right now I control the iTunes session running on the headless G4 mini (which feeds the USB DAC; 500 gig. FW400 on the floor storing AIFF library) using JollyFastVNC on my G4 Powerbook (w/24" ext. LCD). Both machines are quiet but long-in-the-tooth processor-wise, and I can't even go beyond the 1gig RAM in the mini. I'm thinking about an Intel mini and some big NAS (for the whole family), but I'd have to give up both the above G4 machines to afford it (well, daughter would get the laptop).

I don't want to run a 20 foot USB cable (esp. since the input receiver is not async., just a 2706 with custom reclocking).

 

So my mind went back to Gordon's hack of the Airport Express so many years ago. Since the AE is still the only Ethernet device that iTunes supports streaming to, I am amazed that my web searching this morning did not turn up ANY results for what should be a VERY attractive option for computer-audiophiles with USB DACs. What gives?

 

So Gordon or others: Could you please kindly post some pics or a diagram of a hacked Airport Express board with USB output cable attached? I really want to try this. And I would likely bypass the tiny, switching power supply with a linear supply, so does anyone know the operating voltage for the AE.

Lastly, I intend to use it as a wired ethernet>USB bridge--no WiFi. I suppose someone could even figure out how to cut the AE's WiFi power trace...

 

Looking forward to some lively replies :0.

Ciao,

Alex Crespi

 

 

 

 

[The Computer Audiophile]

Hi Alex - I enjoyed reading this post. I remember first meeting you at CES 2008 and discussing some of these very topics. And, continued the discussion long after CES was over. It's really too bad the DMS-1 never made it to market.

 

There are a couple new products coming out that have Airport Express technology built into them. The companies received Apple's OK (similar to companies pulling the digital stream from an iPod) to do this. It's kind of like a half-step forward in terms of what your ultimate goal is. I was at a manufacturer's place several months ago and we were looking at audio interfaces that use Cat-6 cable. These are built specifically for audio and are add-in cards to a computer. This eliminates all the Ethernet/IP traffic on the audio interface.

 

There are many roads the lead to Rome. There are many ways to get great sound quality. I really dislike when people write off things like Ethernet DACs and other technology. I have a feeling that the knowledge to make a great Ethernet DAC work and the software to make it visible to an audio client is few and far between. Otherwise we would see some of these DACs on the market today.

 

Thanks for the post Alex.

 

[watchnerd]

Just to make sure I understand, did you hack the AE so that digital audio comes out the USB port on the AE?

 

If so, what are the data rates? Is it still limited to 16bit/44.1khz?

 

And does making audio stream come out of the AE's USB port mean it doesn't come out the SPDIF port any longer?

 

[watchnerd]

Chris, do you know if any of these companies licensing AE will be producing digital outs that are better than the 16bit/44.1khz the AE is limited to?

 

 

[Superdad]

Gordon @ Wavelength did the AE USB-output hack, not me. I am anxious to try it as I strictly use iTunes (which of course supports the AE directly), and all my digital music is just 16/44.1 AIFFs ripped from my large CD collection. So if Gordon or someone else does not post instructions/pics here, I may have to just buy an AE, crack the case and either do it myself or send it to my friend John Swenson.

 

To answer your second and third questions:

I am pretty sure that a hacked AE would still be limited to 16/44, though it is possible that higher rates may be sent given that neither the AE's DAC or S/PDIF ports function after the mod (those are both on a board that gets wholly removed).

Just to be clear, the USB port already on the AE is NOT the port that gets the output. That exposed port is just for printer attachment (and maybe USB drives these days?--not sure). The USB lines that the iTunes/OSX/Airport Express system streams audio to are internal. It just seems like divine luck that Apple chose USB signals as the means to feed the SPDIF/DAC daughter board inside the AE. It is even possible that they don't do it this way any longer, but it would be easy to buy a used early generation AE for hacking.

 

By the way, Rogue Amoeba sells an app called Airfoil (versions for Windows or Mac) that lets you direct any audio stream to an Airport Express. http://rogueamoeba.com/airfoil/mac/ I can not vouch for what OS audio layers the bitstream has to go through before it goes out to the AE, but I am pretty sure it would still work with a USB-hacked AE unit. If Airfoil is clean with the bits, then that would be a way for some of you to use player software other than iTunes. (And no, I have not contacted Rogue Amoeba to find out about sample rate limitations.)

 

I just searched further and found a page where a guy did an annotated teardown of an Airport Express. http://www.vonwentzel.net/ABS/Dissection-Express/index.html Wow, there is a LOT crammed in there! You can see the DAC daughterboard, the power supply, etc. Plus the main board is loaded on both sides. Once one takes this thing apart, there is no putting it back together. If I were going to do the USB hack, then of course it would be silly not to take the opportunity to wire up to it a clean linear power supply (the pic of the tiny built in shows it needs +3.3V and +5V).

So now I just want arrows to the pins for the USB signals!

 

AJC

 

 

 

 

[barrows]

Thanks for sharing, and I must mention that I am very sorry to see Hovland go.

I share the enthusiasm for ethernet streaming; the only reason there are not more products that can do this is because it requires more R and D than most high end audio companies are able to afford. Still, I look forward to more ethernet connected audio interfaces in the future.

As to the Airport Express, I think this approach may be too limited for a lot of audiophiles. It is my understanding that the AE requires files to be transmitted as Apple Lossless, and then decompressed at the AE itself. So I do not think that any files over 16/44.1 could be accommodated by the AE. I could be mistaken though.

 

[watchnerd]

So if it's not higher than 16bit/44.1kHz, what's the benefit of modding the AE to do USB data out?

 

The optical interface already puts out standard 16bit/44.1kHz data.

 

[cfmsp]

 

"It is my understanding that the AE requires files to be transmitted as Apple Lossless, and then decompressed at the AE itself. So I do not think that any files over 16/44.1 could be accommodated by the AE. I could be mistaken though."

 

Yes, Airtunes compresses files prior to transmission, and they are decompressed at the AE.

 

An entirely new chip would seem to be required for anything above RBCD, as well as software.

 

 

clay

 

 

 

[Superdad]

For years I have assumed that the on-the-fly Apple Lossless encoding and decoding of ALL iTunes/AirTunes/AE audio was a myth. Considering the AE first came to market in 2004, when computer throughputs were much slower, it is hard to understand how they could count on the host machine for realtime conversion to ALC (I know that decode is easier at the other end). Then again, I just ran some timed tests on my old 1.67mhz G4 Powerbook: converting an entire 5 min. tune from AIFF to ALC took 7.5 sec.; from AAC to ALC the same tune took about 9 sec. (little extra for decode from AAC before encode to ALC).

 

So assuming that AirTunes/AE still does always convert to ALC before transmission (I searched the web and still only find parroting or conjecture of this as received wisdom, never any true confirmation in recent years), then that would eliminate my interest since even the thought of listening to my tunes from a computer doing encoding turns me off. (I don't even like the sound of ALC or FLAC files that much--and I think it is because, at least on my machines, the decoding overhead leaves a sonic fingerprint.)

 

Given that the current Airport Express offers fast, 802.11n, I really don't see why they are (or might not be?) still clinging to on-the-fly ALC, especially since it creates latency issues (bad for video/audio sync in some applications). Besides, my intent was to hook it up via wired CAT5e, not via WiFi. Yeah, I know my application and interest puts me in a tiny minority.

 

Lastly, to watchnerd's question: Although yes, the limit of 16/44.1 is there, I am (was) interested in USB output--instead of the TOSLINK--because it would drive my custom USB DAC nicely.

 

So I guess I'll just have to wait for some other Ethernet>USB bridge option to come along. Please let me know if you think of any. Thanks to everyone here for all the comments and input.

ALEX

 

[cfmsp]

 

"...Then again, I just ran some timed tests on my old 1.67mhz G4 Powerbook: converting an entire 5 min. tune from AIFF to ALC took 7.5 sec.; from AAC to ALC the same tune took about 9 sec. (little extra for decode from AAC before encode to ALC)."

 

Before Amarra was capable of playing ALAC files, I amused myself a couple of nights by converting batches of CDs from ALAC to AIFF while the music was playing - and in a fraction of real time, it was almost 90-100xx, as I recall. This had no effect - that I could hear - on the music being streamed from the same instance of iTunes which was doing the converting.

 

Prior to that I had the same opinion as you share below.

 

"So assuming that AirTunes/AE still does always convert to ALC before transmission (I searched the web and still only find parroting or conjecture of this as received wisdom, never any true confirmation in recent years), then that would eliminate my interest since even the thought of listening to my tunes from a computer doing encoding turns me off."

 

As someone who desires simplicity in all things, I can sympathize, and agree with you, although there seems to be less and less rationale for doing so.

 

Cheers,

Clay

 

 

 

[Bob Stern]

In Stereophile over a year ago, John Atkinson found that the Airport Express had worse jitter than the Logitech Squeezebox because the

Airport Express has no flow control. The computer running iTunes streams the music at a rate controlled by the computer, not by the Airport Express.

 

[Superdad]

That is interesting. All I can find (on Stereophile.com) is JA's 2005 review of the Airport Express. While he does mention the problem of clocking, he implies that it is only an issue when the AE's internal DAC is used (DAC w/o a clock?). Because at the end of his measurements, when discussing measured jitter, he appears to have VERY different results between with the DAC used versus the S/PDIF output of the AE. To quote: "The noise floor has dropped by 4–5dB, the word-clock jitter to a respectably low 258ps, which is actually better than the case with the standalone D/A processor driven directly by my PC's S/PDIF output (provided by an RME PCI card)."

 

Can you point me to further discussion of the AE's jitter performance--in his review of the Squeezebox? I did a search at their site and skimmed what appears to be their last Squeezbox review--September 2006--but found no further reference to the Airport Express. I have the paper magazine issues, so maybe you can point me there if not on-line.

 

What I guess I really want to understand is: What sort of clocking, if any, does an ethernet>USB device need? Admittedly, having it output the audio bitstream from the USB port (the internal bus one via removal of the DAC/SPDIF board) is a bit different than sending data to a printer. But the USB is sending data out, not receiving it, so perhaps an async USB DAC would work well with it. Yet I have a hard time believing (with my non-engineer's brain) that the Mac or PC sending iTunes tracks to an AE over ethernet is really controlling the flow. Isn't that all part of the ethernet packet data protocol?

 

[Bob Stern]

JA's 9/06 review of Squeezebox said "There was a cleanness to the sound that I didn't get from the Airport Express's digital output feeding the same D/A."

http://stereophile.com/digitalprocessors/906slim/

 

The explanation about flow control I read elsewhere, but I cannot find it.

 

[watchnerd]

...why not just store your files in ALAC in the first place?

 

Then, presumably, it won't have to do any conversion at all.

 

[Superdad]

Read in Stereophile that the resurrected Micromega company (France) is incorporating a wifi receiver licensed from Apple in their new Airstream WM-10 DAC. Works wired or wireless, Mac & Windows, but just with iTunes. Sounds like they are putting the ethernet>USB board of an Airport Express in the box. I think Resolution Audio does this too (but not sure if they do it with full license), yet Apple may have given some assistance/technical details to Micromega.

 

Will be interesting to find out if it goes beyond 16/44. It is almost certain that they just take the USB signal from the AE's PCB (doubt Apple designed something else for their low-numbers product), so how Micromega then implements their DAC's internal USB receiver is unknown (adaptive or async?). As for their DAC/analog circuits' musicality, that too is unknown.

 

I mention this new product because it proves that Apple is willing to open things up a bit. It is even possible that conversion to ALC before transmission (over wire or wifi) is no longer necessary (maybe they turned it off?) because the 802.11n is plenty fast enough for full bandwidth.

 

Which brings me to my final comment about what Bob Stern said above (that AE's digital output has been reported to not sound as good as an SB):

I think that must be due not to the ethernet transport (nobody here complains much about ethernet transported tunes from their NAS drives, and the AE is first just an ethernet end point), but due to a combination of substandard internal USB>SPDIF/TOSLINK circuit being powered by a tiny switching PS while pressed up against a wifi transmitter/receiver.

 

If someone (maybe me) were to do as I suggested earlier and put just the Airport Express's ethernet>USB board in a new box fed by an external power supply (might as well use a small linear), then I bet the resulting USB feed would sound wonderful feeding a really good USB DAC. Essentially what I think Micromega has done. I may call them to see if the 16/44 limit is being lifted!

 

ALEX

 

[JunkScience]

Micromega has arrived:

http://en.micromega-hifi.com/airstream.html?PHPSESSID=67f7b5843b55103d3677a4ba661a3724

Looking at the specs, they say

"Sample rate (inherent to the iTunes software)................16Bits / 44.1 kHz"

 

Sadly no hi-rez ;-{

 

Larry