Drobe :: The archives
About Drobe | Contact | RSS | Twitter | Tech docs | Downloads | BBC Micro

Hifi buffs told Iyonix audio is good enough

By Ian Chamberlain. Published: 10th Sep 2005, 00:01:15 | Permalink | Printable

Castle investigates glitches, audiophiles scramble for gold plugs

Iyonix logo over some loud speakersAn Iyonix user has recently discovered an issue with his computer when playing audio above a certain frequency. Barnsley based Lee Shepherd first noticed a "crackling" noise when he played a version of the song Blue Savanah through his HiFi to test the RISC OS 5 powered computer which he bought from CJE earlier this year. He told Drobe that it seemed to get worse with higher frequency audio.

The self-confessed HiFi buff contacted Castle and arranged to receive a replacement motherboard. After installing the new board, Lee was disappointed to find that the sound output was worse than before and the machine sometimes even failed to boot. It was decided that Castle would return the original board, which had by now been repaired to remove a soldering fault.

Lee continued to run tests on his Iyonix, which uses an AC97 chip - a budget quality part used in many PCs - to deliver sound output. When playing a sample known as an "audio sweep", which starts with a low pitch and raises the pitch to test different frequencies of sound, Lee claims he could still hear "crackling" artifacts.

Earlier today, Castle's John Ballance said, "We were not aware of the issue and we have been investigating Lee's observations".

Mr Ballance believes that the subtle interference is caused by a timing error in the Iyonix hardware which may or may not be resolved by a software update. Apparently the problem can be replicated on all Iyonix PCs.

He went on to say that although Castle will continue to investigate, ultimately HiFi quality sound cannot be expected from the on-board sound chip. He also remarked that it is an opportunity for a 3rd party to provide drivers for a PCI sound card that would suit Lee and other users who require high quality audio output.

The issue has not been noticed by the majority of other users, despite the machine having been in the hands of customers since the end of 2002. Iyonix user Michael Drake said, "Since I got the motherboard fix, I've been happy with the sound output of my Iyonix."

Castle have recently supported efforts by Simon Wilson to bring 3D acceleration to RISC OS. "What's disappointing about this market is that fewer and fewer people are prepared to have a go," John added, referring to the lack of drivers developed for the Iyonix's PCI interface. He invited users to contact Castle if they need help developing drivers to bring new hardware to the Iyonix.

Despite the issues with audio output, Lee said he will be keeping his Iyonix, as he says he would be "lost without it".

Links

Iyonix website

Previous: Patch released to solve Oregano 2 socket puzzle
Next: VirtualAcorn drops VAT and prices

Discussion

Viewing threaded comments | View comments unthreaded, listed by date | Skip to the end

About PCI for the RISC OS platform: I think it's fair to say that its intended introduction in the Risc PC 2, back in 1998, was probably two years too late. Even back then the number of people interested in doing their own stuff (either with original hardware or with commodity equipment) was dropping below a certain "market vibrancy" threshold.

 is a RISC OS Userguestx on 10/9/05 12:41AM
[ Reply | Permalink | Report ]

The onboard audio from PCs is utter crap 99% of the time. My RPC actually provides quite a good quality audio output by comparison, and indeed the MkI motherboard with separate 16-bit sound card produces higher quality sound than the later MkII with onboard 16-bit sound.

A simple solution to audio quality problems is to provide an SPDIF optical or coaxial output, preferably one that outputs at the sample rate of whatever's being played and doesn't resample everything up to 48kHz. Then connect this digital output to something with a digital input, for example an offboard DAC, AV amplifier, or even use a minidisc deck as an external DAC.

Cambridge Soundworks (amongst others) make a set of speakers with an SPDIF input and an internal DAC - this gets sensitive audio equipment away from electrically noisy digital circuitry.

The solution I've come up with (this is on a WinXP PC, but is potientially a solution for the Iyonix) is to use an external USB soundcard, Creative MP3+ Blaster, and hook the SPDIF optical output to an Arcam Black Box 50 offboard DAC. MP3s played on this actually sound better than playing a CD on the Technics CD player sat next to it.

Of course, playing the CD through the digital out on the CD player into the external DAC sounds better than the MP3, but the point remains.

Iyonix has USB, does it not support a generic USB audio device?

 is a RISC OS UserSquirrel on 10/9/05 10:02AM
[ Reply | Permalink | Report ]

"What's disappointing about this market is that fewer and fewer people are prepared to have a go"

Right. What I find disappointing is that Castle itself does not take a more active role in bringing PCI card drivers to the Iyonix. I believe it would help sell new Iyonixes - specialized Iyonixes for audio work could be an option, an Iyonix for DTP (or whatever) with the help from someone like Spellings, for example.

It has been known by 'audiophiles' for a while now that Iyonix' standard audio output is rather bad. Castle should have known this. It's true most standard PC's provide this AC97 standard on-board sound chip, but what I gather is that the Iyonix's quality is even poorer than that of a RiscPC. When the Iyonix surpasses its predecessors in almost every respect, why then is audio not among them? In this age of multimedia one would assume adequate audio quality should be a priority? Perhaps Spellings might provide drivers for a decent quality PCI or USB sound card with SP/DIF in & output. At least there is one pretty decent audio editor available for our platform which would make good use of it - SampleEd. Find it here: [link]

 is a RISC OS UserhEgelia on 10/9/05 11:52AM
[ Reply | Permalink | Report ]

As I said in my previous posting, the simplest solution would be to stick an onboard SPDIF connector on the Iyonix motherboard, allowing you to connect the machine to an offboard DAC of some sort.

One of my RISC PCs (the one I used for music and audio) has a DMI-50S card (complete with optional XG card), and does sound rather good. Beats the onboard sound anyway, and sounds a lot better than the SB Live card on my PC.

 is a RISC OS UserSquirrel on 10/9/05 12:22PM
[ Reply | Permalink | Report ]

"The issue has not been noticed by the majority of other users, despite the machine having been in the hands of customers since the end of 2002."

Yes, but let's be fair, most people can't tell the difference between a 99 cheap and nasty mini system and a few grand's worth of proper hi-fi separates.

And these are probably the same people that say 128kbit MP3 sounds 'as good as CD'. Yes, if you're using the analogue outputs of your CD-ROM drive to play CDs it probably does - in fact it probably sounds a good deal better, as the analogue outs on CD-ROM drives are positively dire, and even a basic soundcard will sound better.

 is a RISC OS UserSquirrel on 10/9/05 12:27PM
[ Reply | Permalink | Report ]

So if 128k mp3 isn't very good. What bit rate is the best to convert cd's to.

 is a RISC OS Usersa110 on 10/9/05 1:08PM
[ Reply | Permalink | Report ]

160-320kbit VBR, lame options -q1 -mj -b160 -V0

If you must use CBR, use 192kbit, lame options -q1 -mj -b192.

Even at 320kbit though MP3 sounds lacking in detail compared to the original CD or WAV file. FLAC is your friend!

 is a RISC OS UserSquirrel on 10/9/05 1:18PM
[ Reply | Permalink | Report ]

It all depends on what you're listening to. My ears aren't particularly sensitive, and usually 128k MP3s sound fine to me (although I've not got any high quality output to compare them to), but some things sound dire when encoded with the same encoder at the same rate.

 is a RISC OS UserSimonC on 10/9/05 1:46PM
[ Reply | Permalink | Report ]

Squirrel: "As I said in my previous posting, the simplest solution would be to stick an onboard SPDIF connector on the Iyonix motherboard, allowing you to connect the machine to an offboard DAC of some sort."

With 'simple', do you also mean a user-friendly solution? I'd still recommend a PCI or USB solution, although most regular users don't really need digital audio output, unless wanting to experience a DVD movie in full, for example. A well isolated and tuned/grounded analog board would suffice in most cases. The DMI is practically the best analog sound board for Risc PC's. Heck, it's the best expansion if you want to do anything audio or MIDI related with a RiscPC! I use an Irlam i16 which is supposed to have digital I/O, but that does not function sadly, though its analog sampling quality is pretty decent.

"Yes, but let's be fair, most people can't tell the difference between a 99 cheap and nasty mini system and a few grand's worth of proper hi-fi separates."

I disagree there, but ofcourse it all depends on the experiences one has here. Also, not all CD-ROM drives have bad analog outputs! Please don't generalise. I definitely agree that 128kbit mp3's (even with optimisation) sound far worse than their CD originals, however most people perhaps lack the trained ears and decent monitor equipment to distinquish properly between them. Personally I would be horrified if anyone compresses my release-quality music to an mp3 / 128k file - I always rather give them a CD.

 is a RISC OS UserhEgelia on 10/9/05 3:57PM
[ Reply | Permalink | Report ]

In general PC sound quality is poor. The AC97 chip (AC97 *anything*) is simply a standard set by Microsoft and is the standard used by most PC's. I generally listen to music on my Iyonix by listening to the output from the CD-ROM's output socket - so can't honestly say I've heard any problem but then as this does by pass the Iyonix sound system that's not really a surprise.

It would be relatively simply to add a TOSLINK (S/PDIF) optical connector to the expansion port on the Iyonix (an idea I've being toying with) - in theory this *should* give the best quality sound - and at the least expense (the Optical Transmitter has absolutely no electrical connection to the Digital To Analogue converter you've attached to (typically a Hi-Fi) so no "noise"/RFI should get across).

 is a RISC OS UserAMS on 10/9/05 8:17PM
[ Reply | Permalink | Report ]

A thought just occurred to me - someone with IYONIX Linux installed might like to try audio playback with that OS to see whether the crackling still occurs. If possible I'd suggest playing WAV files from memory rather than decoding and playing MP3 files, to keep things simple.

I wonder whether it might be IRQ latency that's causing the audio to break up momentarily; I've certainly heard the audio stop/break up very badly when an application fails and the OS maps the whole task into memory.

With RO being structured as it is, there is - sadly - also the chance that the MP3 decoding routine isn't be called promptly enough, merely because the OS is spending a lot of time in SVC mode. Hence my suggestion that playing a WAV file from RAM be used as the test.

 is a RISC OS Useradrianl on 11/9/05 3:52PM
[ Reply | Permalink | Report ]

Adrian.. as far as I can see the issue is related to dma latency over the pci bus. the ali chup runs a circular buffer that is 2 sound buffers long. there is an instant once per 2 sound buffers where 1 sample is not correct. Depending on the relative phasing of the ac97 sound framing and the dma this can show or not show to varying degrees. (tests at 48KHz, the native rate of the stac decoder chip hung off the ac97 bus).

there are 3 possible solutions I can surmise at present.

1: since the ali chip is on the secondary bus, and since it bus masters to get data, arrange for some memory to be present in pci space on the secondary bus .. it might also be possible to pinch a little nvidia memory for this.. a little further away, but still probably quicker to respond than the 80321

2: find some way of synchronising the phasing of sound dma and ac97 framing

3: use another sound card but on the primary bus. I'm more than willing to guide others in doing this, if contacted 'off line'

I am persuing option 1 mainly.

Cheers

John

 is a RISC OS Userjb on 12/9/05 7:59AM
[ Reply | Permalink | Report ]

jb: sorry.. option 2 .. dma phasing

 is a RISC OS Userjb on 12/9/05 8:01AM
[ Reply | Permalink | Report ]

hEgelia:

We have a build of Cino that can output sound from a Soundblaster Live PCI card, although theres no SharedSound OS driver for this at present. It is on our "to do" list - but we have enough ongoing projects at present without starting another!

Cheers,

/Neil/

 is a RISC OS Userspellinn on 12/9/05 8:34AM
[ Reply | Permalink | Report ]

Squirrel:"Even at 320kbit though MP3 sounds lacking in detail compared to the original CD or WAV file." Hmm, maybe you should be careful what you claim or you might loose legitimacy! I've not heard of anyone being able to make such a distinction in a proper "ABX" test. ( E.g. try making that statement on [link] )

 is a RISC OS Useradamr on 12/9/05 8:57AM
[ Reply | Permalink | Report ]

Does anyone fancy writing an introduction to PCI stuff on the Iyonix? The PCI SWIs documented on the Iyonix site are fine for reference, but aren't really intended as a hand-holding introduction. Other PCI information on the 'net tends to be tied to x86 hardware.

If a friendly guide to how PCI works (specific to RISC OS and the Iyonix) existed, people might be encouraged to work on projects to drive extra hardware. I'd love to be able to write (and give away) a SoundBlaster driver for the Iyonix! But experiments I've undertaken so far to mess around with extra PCI cards on the Iyonix have been heavy going, just because it's so easy to lock up the entire computer when you don't really know what you're doing! Make a change ... compile ... run ... sigh ... reboot machine ... reload source code ... edit ... recompile ... run ... sigh ... reboot ...

Perhaps I should start collating coding notes on a Wiki, then just let anyone chip in a bit. It might help encourage people to dabble.

 is a RISC OS Userkrisa on 12/9/05 9:10AM
[ Reply | Permalink | Report ]

spellinn: Nice to hear that and yes, please do finish an existing project before taking on another ;)

adamr: Sorry to say, but there really are people who just pick up the codec's artifacts! Most of the time it is not your average electronic dance music which is compared, but very hi-fi acousticly sound takes of classical music to specialised audio recorded for just the purpose of researching audio compression. Ever heard of people clinging on to their old vinyl records, in stead of replacing them with so-called 'superior' CD's? It's the 'feel' of it, although I must admit somewhere around 256k (on LAME -q0) I loose touch and can hardly distinquish between the (properly recorded & mastered) CD original and MP3 conversion. The same happened with Sony's superior ATRAC3 format - certain people just distinquish between that and a CD or DAT original. Certain Buddhist monks do stuff not easily explainable by modern science, but they still manage to do it... :) Let's just say certain people are truly blessed with extraordinary acute hearing abilities.

krisa: Excellent idea! If Castle wants people to "have a go", as John Ballance suggested, they should publicly encourage it more and provide some space on their/another website for just that purpose. Releasing the Iyonix / RO5 PCI spec would be ideal.

 is a RISC OS UserhEgelia on 12/9/05 10:04AM
[ Reply | Permalink | Report ]

SA 110, Squirrel :

Try using Ogg Vorbis! ;@) seing as a comparitive representable sound can be achieved at 64 Kbs Ogg Vs 128 MP3 (In most tracks)

mind you, you need an SA to run it well.

 is a RISC OS Userem2ac on 12/9/05 10:04AM
[ Reply | Permalink | Report ]

hEgelia:I don't claim to be an expert in this area. All I know is that in places like the hydro forums, where the experts /do/ hang out - no-one would claim to be able to distinguish "lame -extreme" from the non-encoded source.

Adam

 is a RISC OS Useradamr on 12/9/05 10:19AM
[ Reply | Permalink | Report ]

adamr: Yes, I know what you mean. I know LAME's encoding options and it is very good at what it does. For general purposes, it is excellent. I merely acknowledge the fact that the MP3 format is based on lossy compression, which means it discards aspects of the to be compressed sound material via a highly sophisticated psycho-acoustic model to cut down dramatically on filesize and these discarded aspects can be noticed by experts (via technology or even the human ear). In most music, especially modern electronic music, these artifacts are extremely difficult to isolate and for most people and purposes are negligible.

 is a RISC OS UserhEgelia on 12/9/05 12:17PM
[ Reply | Permalink | Report ]

Lame is about the best MP3 encoder out there at present, and yes, the MP3 files it produces do sound very very good - in isolation.

However, I've tried an A/B comparison of MP3 VBR, MP3 320 CBR and WAV uncompressed, played through a Turtle Beach Audiotron (sadly no longer available) into an Arcam AVR-250 with Mordaunt-Short Avant 908 speakers, using Audio Innovations Silver speaker cable (and yes, the cable does make a difference). Can't tell between VBR and 320 MP3s (except the VBRs are smaller), but play the WAV and suddenly there's a lot more detail and depth to the music. Go back to the MP3 and you can hear slight artifacts on things like hi-hats etc.

Mind you, you're talking to the person who had an MP3 player in his car and could clearly hear compression artifacts whilst belting down the M40 at 130mph... I now have a Sony MEX-1HD, which uses ATRAC compression instead.

 is a RISC OS UserSquirrel on 12/9/05 4:56PM
[ Reply | Permalink | Report ]

"Mind you, you're talking to the person who had an MP3 player in his car and could clearly hear compression artifacts whilst belting down the M40 at 130mph..."

I guess you can also hear the sirens from miles off too.

 is a RISC OS Userjonix on 12/9/05 6:14PM
[ Reply | Permalink | Report ]

Not with the music up that loud! Can see the blue & red flashing lights though, and brake hard accordingly!

 is a RISC OS UserSquirrel on 12/9/05 6:16PM
[ Reply | Permalink | Report ]

krisa: What kind of information are you after? The vast majority of PCI related information will be device-specific. The only things you can really generalise about are PCI configuration space and RISC OS PCI API specific things which are documented on iyonix.com.

Here's some basic stuff, but most of this is just what's documented in the PCI API (note that although this is in BASIC, you really want to be writing a module - especially so if handling interrupts):

Finding a PCI device on the bus:

SYS"PCI_FindByID",vendorID%,deviceID%,-1,0,-1 TO,,,fh%

Mapping in PCI device memory:

SYS"PCI_HardwareAddress",flags%,bar%,,fh% TO,phys%,size%,log%

Interrupt handling for PCI devices:

REM read device vector number DIM buf% 8 SYS"PCI_ReadInfo",&800,buf%,8,fh% TO,,len% vector% = buf%!0

REM claim vector SYS"OS_ClaimDeviceVector",vector%+(1<<31),handler%,pw%

REM enable interrupts for device SYS"OS_Hardware",vector%,,,,,,,,0,1

hEgelia: The PCI spec has been available since the Iyonix was released:

[link]

 is a RISC OS Userjmb on 12/9/05 7:24PM
[ Reply | Permalink | Report ]

@Squirrel: The only reason you can notice a difference is because you know wether lossless or lossy compression is used. I am reasonably sure that no human beyond the age of 16 with undamaged ears can distinguish between a 48 kHz, 16 bit WAV and a lame-encoded 320 kbps mp3 generated from that WAV, when they have no way of knowing wich one is being played.

Because of a bet I tried this with a blind sound engeneer (blind people have far better trained ears than we do) who claimed that he could hear a difference. He could indeed make out 128 kbps mp3s, but anything above 192 kbps he could not. We also tried this with a few other audio experts and even children (who have a lot better hearing than grown-ups) and got the same results.

The mp3s were played through the same equipment, set to the the same parameters, as the WAVs, in a professional recording studio: www.clickmusic.de -> English -> Equipment.

For anyone who still is not convinced: [link] (German, Google language tools are your friend ;-)

 is a RISC OS UserJGZimmerle on 12/9/05 9:09PM
[ Reply | Permalink | Report ]

jmb: "The PCI spec has been available since the Iyonix was released"

Thanks for pointing that out - I feel like an idiot and will check before posting next time ;)

JGZimmerle: I agree with your point there, but I should add that, apart from the encoding options, bitrate and monitor equipment used, the resulting MP3 quality is also dependent upon the type of audio which is encoded. Certain types of music for instance can sound perfectly true to the source at a lower bitrate than other types. (In my experience Classical or ethnic music would require a higher encoding bitrate as Techno.) However, at about 256kbps and higher, I definitely agree anyone would be hard pressed to make out any differences.

I should also point out that context matters very much here; encode a song at 160kbps, then encode a individual acoustic sound at the same bitrate/config. Chances are you'll make out a difference in the latter case, whereas it would be very hard to distinquish in the case of the song. That's why we record musicians without any lossy compression in a studio.

 is a RISC OS UserhEgelia on 13/9/05 11:39AM
[ Reply | Permalink | Report ]

JGZimmerle:

I just did a blind test here. I ripped a track, then encoded it as FLAC, OGG Vorbis and MP3. I then threw all 3 into Winamp, along with the original WAV file. Hit the random play then repeat buttons, closed my eyes and hit play. Without looking I could hear a huge difference between lossy (MP3/Vorbis) and lossless (WAV or FLAC) formats. OGG Vorbis does sound better than MP3, but still lacks detail when compared to a lossless format such as FLAC.

I then repeated a test with a friend selecting a version of the track and playing it, same result. I was able to distinguish immediately between lossy and lossless versions.

Of course FLAC sounds identical to the original WAV, as it's mathematically lossless.

 is a RISC OS UserSquirrel on 18/9/05 6:21PM
[ Reply | Permalink | Report ]

Theoretically lossy doesn't need to sound worse than losseless it could even sound better on cheap amplifiers (as the amplifier has an easyer job not amplifying the sound you don't hear anyway).

The point is that when you listen to MP3s a lot you will learn to hear the artifacts after a while. At first 128Mb/s or even 64Mb/s MP3pro sounds fine but in the end you might need > 192Mb/s as the artifacts (which you didn't notice before) disturb you.

To avoid needing to make better quality MP3s after a while better chosse the high quality right away.

 is a RISC OS UserJaco on 6/10/05 3:27PM
[ Reply | Permalink | Report ]

Certain lossy encoders (eg Sony's ATRAC) do actually clean up the sound by removing stuff that although you can't hear does "muddy" the sound. When I first invested in a minidisc deck a few years back I was quite surprised at this, when some MD copies actually sounded cleaner than the original CD.

However, MP3 doesn't fall into this category, and MP3 encoded material does sound noticably worse.

Also it depends very much on the listener. I seem to be cursed with being able to hear minute differences that most people wouldn't, eg the slightest trace of MP3 compression, even the difference between OFC and silver speaker cable (the silver cable sounds better!). I say "cursed" because it means I end up spending a lot more on audio hardware than on CDs to play on it...

 is a RISC OS UserSquirrel on 7/10/05 8:57AM
[ Reply | Permalink | Report ]

Please login before posting a comment. Use the form on the right to do so or create a free account.

Search the archives

Today's featured article

  • CDs Available
    Drobe Special Projects
     2 comments, latest by piemmm on 21/10/03 12:23PM. Published: 23 Jul 2003

  • Random article

  • Zero to OnePointNought
    Blame the sysadmins, they're not even a real country anyway. iconbar.com is back.
     5 comments, latest by moss on 23/9/02 2:09PM. Published: 3 Sep 2002

  • Useful links

    News and media:
    IconbarMyRISCOSArcSiteRISCOScodeANSC.S.A.AnnounceArchiveQercusRiscWorldDrag'n'DropGAG-News

    Top developers:
    RISCOS LtdRISC OS OpenMW SoftwareR-CompAdvantage SixVirtualAcorn

    Dealers:
    CJE MicrosAPDLCastlea4X-AmpleLiquid SiliconWebmonster

    Usergroups:
    WROCCRONENKACCIRUGSASAUGROUGOLRONWUGMUGWAUGGAGRISCOS.be

    Useful:
    RISCOS.org.ukRISCOS.orgRISCOS.infoFilebaseChris Why's Acorn/RISC OS collectionNetSurf

    Non-RISC OS:
    The RegisterThe InquirerApple InsiderBBC NewsSky NewsGoogle Newsxkcddiodesign


    © 1999-2009 The Drobe Team. Some rights reserved, click here for more information
    Powered by MiniDrobeCMS, based on J4U | Statistics
    "Every once in a while I get the impression that sometimes the things published on Drobe are not 100% accurate"
    Page generated in 0.3287 seconds.