Why does the computer matter?

[Windows X]

While tons of pro people in studio field out there are suffering such great ordeal from over 5ms latency, some audiophiles claim Linn DS with 2000ms latency is pinnacle of hiend music server because network streaming has big enough buffer to handle this perfectly.....Since when the hell is memory playback or even SATA inferior to network streaming? I can't follow this logic at all.

 

I'm not Esoteric fanboy but from what I've known, this company also made solid drive for other companies like DCS/Soulution/etc.

 

[Windows X]

After face to face comparison to few hiend CD transports in proper setup (Esoteric/Viola/Karan/Rockport with proper acoustic room), there's no way in this year computer music server can get close to to hiend CD transport because we have mainboard like this:

 

1. they use crappy parts even all caps are replaced with proper ones won't have enough space to put in so we can only do for some

 

2. Very thin PCB circuit line and not 2oz teflon coated with silver or anything that works in audiophile's PCB grade. I heard some modern mainboard has 2oz copper but for very high power ones which is so bad for making music server.

 

3. It use poor IC switching power supply with cheap power distribution design

 

4. Path is too roundabout. We need shortest design ones

 

5. Poor clock and low quality I/O parts. Remember $300 RCA/XLR connectors and $10k interconnects? There're tons of people willing to pay for that despite the fact that all cables and connectors combined in hiend equipment are under $100-1k budget for any price.

 

6. buffering...buffering...overhead...virtual path...so on and so forth

 

How in this world computer audio can sound as good as CD? Let's combine our strength and make group order of custom-made board for hiend audiophile

 

[Miska]

Just on a generic level... For me, good engineering practice requires that the things I do can be objectively measured and tested. Possibly through automated systematic unit tests.

 

Such, that when I make a change I can quantitavely measure and test the impact. And numerically show the results. Maybe part of it because of my strong background on measurement and analysis systems.

 

Subjective listening is also very important, but any decisions based on this have to be based on objectively measurable theory.

 

 

[PeterSt]

Can you please give a reason *why* you think latency is so important ? Or / and the other way around : what would be so wrong with a latency of 2 seconds for audio playback ?

 

I guess you really should come to some point, or otherwise you won't get anywhere just because nobody can really help (with a problem unknown as it seems).

 

You know I already said you were correct on what you strive for, but stating that any 2 second buffering device *thus* sounds like nothing, does not make sense.

 

I'm 100% sure that at least with this everybody will agree.

 

Or we miss your real point completely ?

So, no need to repeat it all, but try to explain why you want this other quality from a mobo with poor design which 100% officially does NOTHING to the sound.

 

Peter

 

PS: Yes, I said "officially", which is different from what may happen in practice. But now from empirical finding and some real proof please ... instead of only thinking it.

 

[Windows X]

Have you ever wondered how boosted up 100hz or 200hz can make TV looks better from 50-60hz? Why TV companies needs to advertise their response time?

 

Let me think as consumer who has no proper knowledge or experiences about hiend TV. "I can't tell apart between 4ms or 10ms it should look the same, c'mon. And 200hz is bizzare. bluray's framerate is only 24fps so anything below 40 latency (25hz) is more than enough 50-60hz is standard and can archive best experiences from my research."

 

Yet a lot of uneducated or have ear training people can tell apart clearly between 1ms and 2ms in my system. They aren't even audiophile .

 

Science is understanding the phenomenon not to object it without trying to understand. Have you tried? Do you own or used to own $100k system and set them up properly for listening? Or did you perform buffer time test between 5ms and 50ms in various systems before selling (Let alone Linn DS that uses 2000ms like DirectSound)? If you did all that and still have this conclusion, I shall respect your conclusion without any counteract since that's what you can perceive. I used to not being able to tell apart between stock cables and $$$ cables so I can understand to some degree.

 

[Windows X]

Do you believe that TV with lower response time can archive greater picture? Do you think PS3 and OPPO-93 or Pioneer offer all the same quality if they all meet the same HDMI standard? Do you perceive the same quality from stock HDMI cable and $$$ ones?

 

I understand your passion for great sound from music server and so do I. It's just what we can perceive are from different perspectives. I starve for highest possible fidelity that I can experiences from this and it still lacks a lot to what $10-20k CD transports can do, especially good ones.

 

The best practice is real listening and A/B comparison in good setup. You should be able to tell the difference if your ears aren't too worn out. Otherwise, this won't be done because all explanations I can make will be objected from your mind anyway.

 

[Miska]

While tons of pro people in studio field out there are suffering such great ordeal from over 5ms latency, some audiophiles claim Linn DS with 2000ms latency is pinnacle of hiend music server because network streaming has big enough buffer to handle this perfectly...

 

The latency matters when output of the system depends on some change of state in the related surrounding physical world.

 

- For the case where you are playing a software-based synth instrument with physical world keyboard, you expect to hear the sound when a key is hit.

- For the case where you have ADC input of analog audio from some source, do some DSP and immediately output it through DAC, for example effects processing in live sound input from vocalist microphone to live concert audio.

(- For the overdub case, when number of pre-recorded tracks are playing and a new track is being recorded in parallel. However, this can utilize latency compensation for any known value.)

- For the case in playback, that some user control (like volume) is applied in real time to the data and output to DAC and the effect is expected to heard immediately.

 

Thus the only latency related to digital file playback is the last one the list, controls applied by the playback device/software in relation to the played back material. And the acceptable latency is defined in delay between user interacting with the player and when the result is expected to be heard. When playback is just flowing without listener interaction, there's no "latency" between event source and output response.

 

For pro audio, all three/four things matter.

 

The slightly related latency is induced by the computer and operating system on playback software submitting new playback buffers to played. This is a hard limit defined by the buffer length in time and when exceeded there's an audible glitch in playback. Typically one buffer length of silence played back by the device driver. Thus using large enough buffer/latency ensures glitch-less playback.

 

[1audio]

These arguements are borderline bizzare, if they didn't show an enormous misunderstanding of the base facts.

 

There is no meaningful way to compare frame multiplication and motion estimation on video to "latency" on an audio playback system. With video your eye is able to see the flicker and some of the other effects of slow refresh rates. Bluray is stored at 24/25 FPS (the same as the camera recorded) and the display system will either flash the same frame several time to get 72 to 240 flashes per second or the better systems will interpolate the motion between frames. This can be successful but some of us can spot the artifacts very easily. However there is no connection except confusion between that and latency on audio playback. The best analogy would be re-sampling to a higher sample rate, again with potential artifacts.

 

Latency may be a misnomer here, first used to describe the variable delay in accessing disk and tape data. In this case its really not a variable delay, but a fixed delay, the time through the process from accessing the data from its stored location to passing it to the playback system. Aside from increasing the activity in the host a shorter delay should make no difference in the bits coming out (or their timing) if everything is working right. However that increased activity may translate into a lot of noise. If you have poor noise isolation its conceivable that the shorter latencies are moving the noise spectrum up out of the audible range, or they may be more euphonic. How can you tell without meaningful measurements what is happening? Prettier sound is not always more accurate.

 

[Windows X]

My TV don't need to be below 10ms response time for best picture quality, right? I can look away from 100-200hz Motionflow since they don't really bring better picture? Let me tell you this

 

Linn DS feeding 2000ms meaning they're going to process a very large chunk at very long time for processing sound output. This can take a long time and you hear continuous big chunk for each passing second. This sound you hear is sound processed from 2 second time kept from buffer, so dull and lifeless. But for 5ms latency, you have 200hz like current Motionflow that can archive less chunky and more "room like" sound which can't be corrected from cheap stuff like DSP (like what Amarra tried to do and end up with artifact).

 

Yet I'm not satisfied with 200hz Motionflow-like sound. It's not real enough. A lot of inner detail from sound are missing, the weight and impact can't be controlled like real recording. Why? Because large buffer caused veiled sound and duller performance like CD Walkman. It sounds better when buffer turned off. If you don't know why buffer cause bad sound, go find it yourself. I don't have enough knowledge to make unknown experienced becoming understandable for you.

 

[Miska]

Do you believe that TV with lower response time can archive greater picture? Do you think PS3 and OPPO-93 or Pioneer offer all the same quality if they all meet the same HDMI standard? Do you perceive the same quality from stock HDMI cable and $$$ ones?

 

OK, now we are getting contradictory. Do you know how many video frames PS3 or Oppo have decoded already before sending over HDMI to the TV? Many of the DSP operations in modern video playback require at least two frames of latency, this is 83 ms for 24p video.

 

In "memory player" case it would mean that the playback "latency" is entire movie, like 2.5 hours, as it would have been decoded to a memory buffer before actual presentation.

 

So what buffers are counted for, where and what buffers are not?

 

 

[Windows X]

But in reality, each time signal passes from place to place always come with interferences. Transferring large chunk reduce performance to make the circuit become rock solid. Fetching 2000 samples from file to CPU cache, processed and send to card make worse sound than 1-5 samples each time because larger chunk being disturbed for longer time. You have to get your data out of your crappy machine asap for best sound.

 

In hiend audiophile world, even resonance control can affact the sound a great deal. Just making tip-toe weight balancing can make system sound a lot more refined and controlled. Do you believe that? Some do and people who do believe aren't ones with placebo effects. I can guarantee that.

 

[Windows X]

For reason why buffering is bad. Imagine CD reader can send data 500-1000 times per second. But not all designs can send 1000 so buffer is needed for some and have to receive data and fetch data as pool from CD meaning we need some compensation of its greatest performance to be something like receiving at 500 times per second and fetch out for 250-500 depending on design. Each step in pooling design caused delay. We can't make perfect I/O matching without some sacriface. Not to mention timing error from clock (something like jitter)

 

See? CD reader that can send up to 1000 samples per second gotta leave it at less samples per second making audio stream less continuous.

 

[ziggyzack]

'Can you please give a reason *why* you think latency is so important ?'

 

But no answer.

 

X, you're confused. Fundamentally, the doubling of refresh frequency from 25 frames per second (as filmed) to 50 stemmed from interlacing - so a CRT would pass each row of phosphors twice per frame, painting even rows in the first pass, then odd in the second pass. Refresh rates were increased to maintain the illusion of consistent screen luminosity and reduce the perception of flicker, which people could often notice from peripheral vision. More recently, with digital screens, there can be interpolation between frames - I find it a bit of a gimmick really - calling for higher refresh rates. Also, plasma and LCD screens may need more frequent refreshes than CRT phosphors, to remain illuminated.

 

This has nothing to do with audio latency, it's more analogous to sample rate, and even then, it's not a great analogy.

 

Latency is an issue in all kinds of digital systems. But it doesn't affect sound quality (I can't speak for implementations, but if it does affect SQ, something must be wrong!) The music data is stored on a hard drive, SSD, CD - why does the DAC care if it spends an extra 5us, 5ms or 5 seconds in a buffer (not to mention the rest of the digital audio chain) before it arrives? A very bizarre suggestion.

 

ZZ

 

[Miska]

Fetching 2000 samples from file to CPU cache, processed and send to card make worse sound than 1-5 samples each time because larger chunk being disturbed for longer time.

 

Actually with larger chunk, the total time is less. since it has lower overhead.

 

There's a fixed overhead of initiating a transfer of anything, let's call this "t", then there's the time taken by the transfer itself, which depends on the size, let's call it "y". As there's a fixed total amount to be transferred it has to be divided into "n" blocks of "y". And then there's total number of actions required to transfer the total amount of data, let's call it "X". The formula is:

 

X = (y + t) * n

 

Thus, smaller the y, bigger the n and higher the (t * n) part is.

 

This "t" part is especially strongly emphasized on modern computer architectures. But the topic itself is worth of an entire book, already available from number of authors...

 

And as I said before, DAC is being fed with a single sample at a time based on word clock and the above is unrelated to this process.

 

[1audio]

" Fetching 2000 samples from file to CPU cache, processed and send to card make worse sound than 1-5 samples each time because larger chunk being disturbed for longer time."

 

Makes no sense. How does taking a larger chunk disturb the audio for a longer time? Are the bits agitated and trying to escape the computer? Analogies are fine if they are based in fact, but when they are pure fantasy based on poor understanding of technology they can be very misleading. A larger chunk of data should disturb the system less simply because the system is fetching the data less often. If the bits are disturbed by sitting in the PC's memory longer how would they show it?

 

BTW, 100 Hz IS 10 mS, or the number you mentioned. Sony's motionflow is pretty good, but has some easily visible artifacts on certain images, like the demo video they were using at CES. . .

 

Keith Johnson mentioned to me when I proposed transferring disks to digital in a high performance sound isolation chamber that most audiophiles would be disappointed. The extra vertical modulation from the sound in the room helps contribute to "air and space", even though its not in the original recording. How can you be sure that the "vibration control" you are using is actually reducing the effects of room vibration? They may be enhancing it for a good euphonic effect. If you are serious you would move the system to an isolated room away from the speakers, which may obviate the need for the tweaks.

 

[Windows X]

You know? Everything can't do its job perfectly right. Higher latency introduce longer delay in sound. It's like refresh rate. You get 2000ms latency meaning you have 0.5Hz refresh rate per second (and that's refresh rate for receiver/controller). You get 5ms latency meaning you get 200Hz refresh rate per second and that applies to all electronic equipments not just TV. Go study more about how receiver in AD/DA works and implementation so you can get some ideas why lower latency can make sound better if you don't want to admit real listening to make conclusion.

 

[Windows X]

Here's good site for explanation.

 

http://www.pcmus.com/latency.htm

 

If you have an idea about recording has nothing to do with playback, I suggest you to revise this idea. Recording is getting in from the gate A to main controller and playback is sending out from controller to gate B. Similar principles, huh?

 

[1audio]

Where is the delay in sound audible? are you hearing an echo? Please try to understand what you are talking about. I work with the chip designers for audio and video systems on a regular basis and understand and teach others on how these technologies work. I have been working with computers since memory was magnetic cores and know a lot about the insides of all of these technologies. There is no parallel between refresh rate on a TV and latency on audio. None.

 

If you hear something describe what you hear and the circumstances under which you hear it. Make changes and describe the differences. See if it survives a double blind test (very hard). But ascribing cause to an effect is hazardous at best. The assumptions you are making, reduced latency makes better sound requires the following, a) the sound is better, not just different, b) that the change you are making is the cause of the difference, c) that the change you made actually took, and that the change had no other impact. This is all real work and requires serious objectivity.

 

[ziggyzack]

'You get 2000ms latency meaning you have 0.5Hz refresh rate per second'

 

X, this is so deeply flawed that I think there's no point continuing. All the best.

 

[Windows X]

Then explain how it's flawed. You guys kept asking for explanations and always object it or silence. I'm starting to wonder if any of you guys actually own hiend and set it well. All guys I brought into test can tell'em all. I just don't get it.

 

Why don't you guys try ranging up different buffer time for checking if latency affacts this? Make sure you explain your test equipments, setup and test methods before concluding something.

 

Don't Computer audiophile also welcome highend computer audiophile listener? To debate, you can object with reason or proof to say where it went wrong. Not just bashing own's ego from beliefs to each other. That's destruction.

 

[Windows X]

I did a great experiences about music server in various models like I said before, almost all possibly the best from firewire and PCI/e ones. With all optimizations, they ended up being totally different from great to crap for hiend from just change.....buffer time or latency. It's not self-testing but it involves a great amount of third-party people to audit and confirm it.

 

Higher latency introduces delay and cause audio stream less continuous with some details lost is what I found from this lengthy experiments in all setup ranging from few hundreds to few thousand of grands. How increasing latency cause this? If you know buffer's downfall for solid performance of data and signal transfer, that's it .

 

 

 

See if it survives a double blind test (very hard).

 

Actually, it's pretty obvious. I set out 1ms,2ms,5ms,10ms,50ms,100ms,500ms,1000ms. Most people if not all can clearly tell the difference from each change without telling what I did. But I did this in very highly resolving highend system under $100k budget. Maybe it's too easy because I let them listen to $10k CD transport beforehand. If I do this in under $10k system, it maybe hard for people who can't perceive what highend really is.

 

For anyone to be able to perceive, setup proper highend system and make a test. Highend isn't about money but dedication to archive higher fidelity (who can buy $5k for equipment stands, spending $10-50k for making great acoustic room)

 

[Miska]

TV analogy for latency... Is the delay from news reader reading the news to the TV receiver displaying and playing back the audio. How is this delay from the news room to the receiver display perceived? What is the latency of 60's movie shown on a satellite broadcast dedicated to old movies? The latency just up to the satellite and down back to the earth is already quite high do the limits of physics...

 

 

[CG]

X,

 

Some of the people on this thread who you're suggesting may not be all that knowledgable about "hiend" audio may have actually designed some of the equipment that you list in your "hiend" audio system.

 

Perhaps they are too modest to say so.

 

Just thought you should know...

 

[Windows X]

You need to put timing error and induced jitter that cause picture less vivid and has more grain too

 

http://www.cicsmemoryplayer.com/index.php?n=CPlay.SoftwareInducedJitter

http://www.cicsmemoryplayer.com/index.php?n=CPlay.ASIOLatency

 

These links are good start places if you're willing to learn about latency and induced jitter. I'm not trying to prove you wrong here please keep that in mind. If you think I maybe right, go read stuff there and learn plentiful information with proper documents. If you think I'm in wrong and don't try to right me with good information to try experimenting and conclude, that's not a good way to exchange our information.

 

I do see a lot of people here resemble me when I was unpolished in computer audio and highend system. I did tried all you've mentioned but they aren't true at all. The more I study how highend CD transport/DAC/Pre are assembled, the more I take this inefficient setup more seriously and take $$$ cables less seriously from what I saw inside these magical boxes.

 

[Windows X]

I do know so that's why $6k DIY preamp made by dedicated highend listener with proper knowledge can out-class $30k commercial preamp made by people who don't really listen to highend crap at all. And that preamp doesn't even have input buffer stage, he said it's nightmare for preamp to have this stage. Adding this will degrade your sonic performance to extent you can't keep on listening to it with this purist design.

 

I know well from what I'm talking about. It's not all talk but experiments and be convinced from combined results of my experiments.