Jump to content
IGNORED

Airport Express performance: objective and subjective


Recommended Posts

Eloise, I see what you mean. Some differences are easy to measure, other things are quite difficult. I read in another post on this forum that measurements concerning DAC's can become really tricky. A person who mods and builds DACs said they could change several parts such as swapping out capacitors etc. on the board which does absolutely nothing to the measured specs but can change the sound dramatically.

 

david is hear[br]http://www.tuniverse.tv

Link to comment

"So for my hyphothesis: two different speaker cables are put on a test rig one is a generic 186 strand OFC copper cable; the other a Nordost Valhalla. I think it's generally accepted as fact that both will measure exactly the same (given normal lengths and bandwidth). Certainly any difference will be marginal. Yet 3m length of the basic copper cable will cost you maybe £5-10 where as the Valhalla cable is in region of £2-3000 (hope this are reasonably accurate figures but I'm lying in bed with iPhone).

 

Now instead of using the cable between a signal generator and spectrum analyzer, we put them into a real world system of a high end audio system with suitably revealing components (all in anachroic chamber of course)

 

If we use a calibrated microphone and suitably accurate analogue to digital converter can we then measure any (repeatable) differences or are those differences just too minute to measure?"

 

Let me tell you that cable differences are extremely difficult to measure, and yet easy to hear. If you are using a microphone to pick these up, you wont get any differences. You must use much more sophisticated gear and test signals.

 

Here is a difference that I was able to measure, but the effect is only visible at MHz and higher frequencies, well beyond audibility.

 

http://www.empiricalaudio.com/computer-audio/technical-papers/direct-immersion-lno2-study

 

And yet, the difference between the sound of these two cables was radical. The one that was dipped in liquid nitrogen was really "dead" sounding and very narrow soundfield. Like listening through a plumbing pipe. The other characteristics, L, R and C were unchanged up to 100kHz.

 

If you cant see the graph, I have contacted my websmaster. It should be there.

 

Here is another study that I performed with both simulations and measurements:

 

http://www.empiricalaudio.com/computer-audio/technical-papers/clarity7-electrical-performance

 

As you can see the differences are minute at best, and yet the sound difference was easily distinguishable.

 

Some more interesting reading on an often contentious subject:

 

http://www.empiricalaudio.com/computer-audio/audio-faqs/short-versus-long-cables

 

As you can see, I have probably done more analysis, measuring and computer modeling than any other cable company out there.

 

Steve N.

 

Link to comment

 

Ears more accurate than electronics.. Granted measuring output from speakers is going to be hard. But ...

 

"the difference was beyond audibility" yet easy to hear. So you're saying there was no measured difference within the range of human hearing yet a difference could be heard ...

 

If you were blindfolded I wonder whether a difference could be heard ...

 

 

 

HTPC: AMD Athlon 4850e, 4GB, Vista, BD/HD-DVD into -> ADM9.1

Link to comment

Read that all with much interest, especially after what I achieved today : http://www.phasure.com/index.php?topic=692.msg6183#msg6183

 

In fact the subject turns out to be the same : "can it follow". Well, it can't.

 

In addition to what I've written in there, this is about normal music data (not wild on transients at all) over SPDIF and a digital cheapish interlink of 75 Ohm 60 cm and 70 cm higher grade video cable. What I wrote did not take into account any cable issues, but it possibly influences.

 

As you can imagine - at looking at that data - it is not only a tough job to get the samples aligned, but it is even a tougher job to get the lot normalized, because there's really no clue on what reality must be for amplitudes. On the latter I found that the DAC (this one anyway) is not stable for a longer burst of positive (or negative) samples, and the longer the (fed) burst stays, the lower the amplitude will get. This might be electrically normal, but evenso it is practice in normal music. The other way around, feeding the DAC with even positive and negative shorter bursts (or single samples) doesn't cut it either, and the average amplitude may be under less than 10% of the intended. By analyzing the data closely I was able to find an algorithm anyway, and the conclusion is : we listen to soup. Tomato or curry, I don't know, but I like it very much. It is nowhere near what the data tells though, and maybe we should be glad all is so analoguely smoothened. One thing is clear : no way that what I have with my DAC is coincidentally the optimum which will exist somewhere. That might be less rounded, or it might be more. But there's room in there for a whole universe, and this is where DACs will differ. Should be in the output stage.

 

Forget about jitter. Jitter must be 1000nds of times less relevant. A bit perfect feed ? don't make me laugh.

ONE THING : We could be looking at the ADC used. So for now that is the disclaimer.

 

Peter

 

Lush^3-e      Lush^2      Blaxius^2.5      Ethernet^3     HDMI^2     XLR^2

XXHighEnd (developer)

Phasure NOS1 24/768 Async USB DAC (manufacturer)

Phasure Mach III Audio PC with Linear PSU (manufacturer)

Orelino & Orelo MKII Speakers (designer/supplier)

Link to comment

Lets measure the brain wave patterns concerning the eye's actual response of people's reactions to pieces of art to determine whether they truly see something beautiful, there has to be some measureable difference determining what they are looking at is really attractive because obviously we can not possibly determine an object is more attractive than another. We are simply robots. ;)

 

david is hear[br]http://www.tuniverse.tv

Link to comment

"the difference was beyond audibility" yet easy to hear. So you're saying there was no measured difference within the range of human hearing yet a difference could be heard ...

 

Yes. Try it yourself. Go to your local welding supply and get a thermos of liquid nitrogen. Dip one of your analog cables in it and then compare against an identical one that is untreated. Especially obvious with a silver cable.

 

Warning: Once you do this, it will be worthless, you wil have to throw it in the trash.

 

Steve N.

Empirical Audio

 

Link to comment

Peter - I have a fix for the A/D jitter. I have the patent disclosure ready on this one:

 

When the recording is captured (sampled), you at the same time with the same A/D clock capture a second track, a "jitter" track. This is a special data pattern that is just for measuring/detecting jitter levels. This jitter track is stored and transported with the music track.

 

Then, on playback, the electronics monitors the jitter track and compensates on-the-fly for the record clock jitter.

 

This way a really inexpensive A/D can be used. The jitter is removed on playback, just like Dolby noise reduction removed tape hiss.

 

Steve N.

Empirical Audio

 

Link to comment

Peter - nice graphs, but I have some questions:

 

Is the digital data output of the D/A you are capturing on the pins of the D/A converter, on an active output stage or on an AC-coupled output?

 

What D/A chip is used?

 

Is your Fireface being driven from a Mac or PC, and if PC, what OS and how are you bypassing Kmixer/audio stack?

 

Steve N.

 

Link to comment

Steve - In this process no digital data is being captured. Only analogue from an in this case passive (discrete) I/V stage, RCA-out.

This is held against the digital file.

 

PCM1704-uK (x 4) in balanced setup. Btw note that I suspect something going on here (balanced), which is about the impossiility to sustain an e.g. positive burst. But that's easy to check with other DACs.

 

Fireface : WASAPI again. And not to forget : at capturing a loop back ... 100% bit perfect.

 

Peter

 

PS: Since I am actually in the process of automating these kind of measurements (and then bump into stuff like this) - when this is done I will be comparing various situations. Right now it would be too tedious and time consuming. But I will be there in one or two weeks now.

 

Lush^3-e      Lush^2      Blaxius^2.5      Ethernet^3     HDMI^2     XLR^2

XXHighEnd (developer)

Phasure NOS1 24/768 Async USB DAC (manufacturer)

Phasure Mach III Audio PC with Linear PSU (manufacturer)

Orelino & Orelo MKII Speakers (designer/supplier)

Link to comment

I really don't know the answer to this ...

 

If you put exactly the same musical waveform into an ADC twice ... will the digital output be exactly the same both times, or will there be some differences?

 

Eloise

 

Eloise

---

...in my opinion / experience...

While I agree "Everything may matter" working out what actually affects the sound is a trickier thing.

And I agree "Trust your ears" but equally don't allow them to fool you - trust them with a bit of skepticism.

keep your mind open... But mind your brain doesn't fall out.

Link to comment

Btw, sorry to be so OT. I know, this is my fault. But whereever I have a chance I want to continue and improve on this.

 

Steve, that looks interesting by itself. But I have my doubts. I mean, where I in here can only show some speaking snapshots, I can obviously scroll through that data, zoom into it, zoom out on it, and as you can imagine you can look hours at it, and find the "patterns" all over, being there for you to qualify - somehow.

 

What I want to say is : when you see how fragile everything is (like a moving cursor influences sound), I already don't see how to get that jitter data in without influencing itself. And in the end we wouldn't know what to look for anymore. Keep in mind that your approach will be based on jitter as the only influence being there, not taking into account the vast variances a DAC incurs for. On this matter it is not different from what I suggested earlier in It's Friday here, so here is the BIG one :-) : an external anomaly of whatever kind may create a peak in the DAC (may it be in the DAC chip or right after it) but this anomaly - possibly 0.00002 seconds short, creates an own life of thousands of samples. This is what you see in the graph I referred to in the other post, only now this is not about external anomalies, but it is just how the DAC inherently works -> give it a transient of, say, 10% of the maximum voltage, and it is bothered by it for half a second. In the mean time dozens of other transients passed, and in the end we must be glad to hear music.

 

Of course this particular discussion is about the influence of the ADC, but how to measure that without the influence of the DAC ? Right now this looks like a dead end. But that is only now :-).

Anyway, of course I am very much willing to take the attempt, but we should know that this jitter pattern has to be derived from somewhere. On this matter maybe you recall that I was very ery doubtful on jitter measurements, at that stage already knowing that the output of a DAC is so much different from what's put in. Well, now you've seen it, and I can tell you that with this particular jittery connection (just SPDIF, no SRC behind it) it measures just over 1ns. I think this is a great job of the measurement device, looking at the output compared to the input. Until someone can explain to me that with such an output this still can be done and how it is done, I claim this is just wrong measurement. Crux : how to get hold of a reliable jitter pattern for the ADC ?

 

A last one for now : recalling your cable measurement outlay, and actually being able to prove differences in the MHZ area (or whatever it was but far beyond the audible area), this is behind us. When we hear a difference (and I mean really hear it, like I would dedicate this to myself) we can see it by this means. However, where this is intended to be looked at as an absolute difference, the means I used in the link I just gave should be used. Because remember, this compared two analogue signals, and no matter the ADC is influencing, it will come out. If nothing comes out ... well ... I can do the same measurements by microphone, but I really think that won't bring more. But I think I have sufficiently proven that there's a whole sh*thole of stuff never seen before, and digital thinking obviously never has been based upon that.

 

Still very much interested in your jitter removal solution, let me know when it's disclosed.

Peter

 

 

Lush^3-e      Lush^2      Blaxius^2.5      Ethernet^3     HDMI^2     XLR^2

XXHighEnd (developer)

Phasure NOS1 24/768 Async USB DAC (manufacturer)

Phasure Mach III Audio PC with Linear PSU (manufacturer)

Orelino & Orelo MKII Speakers (designer/supplier)

Link to comment

Including the DAC in front of it (which is the only setup I used -> could have been an analogue device of course), so including the additional dimension of the D/A conversion and its possible issues, the output does not deviate more than one bit plus and one bit minus (which would be decimal +1 / -1). So, looking at the wild being off of the (my) DAC, two subsequent runs show exactly the same result, apart from the 1 bit being off here and there.

 

Although at first glance this looks like unavoidable random noise, it also looks like dither. This latter would be logic, knowing that the ADC (Fireface) will internally work with more bits than the 16 this is all about. So if all is right it dithers, and if that is so even random noise doesn't show. This, however, makes what we see for huge differences (at different players) even more important.

 

So, no matter how wild and strange the deviations are, they are completely repeatable up to one bit deviation which may even be caused by dither.

 

Peter

 

Lush^3-e      Lush^2      Blaxius^2.5      Ethernet^3     HDMI^2     XLR^2

XXHighEnd (developer)

Phasure NOS1 24/768 Async USB DAC (manufacturer)

Phasure Mach III Audio PC with Linear PSU (manufacturer)

Orelino & Orelo MKII Speakers (designer/supplier)

Link to comment

"In this process no digital data is being captured. Only analogue from an in this case passive (discrete) I/V stage, RCA-out."

 

If this is actually just the voltage across the I/V resistor, then it is probably the power to the PCM1704 that is being modulated, or the Vref or both. There may also be some ground-bounce mixed-in. Power delivery to the D/A chip is critical. Simple decoupling with a .1uFd cap for instance is grossly insufficient IME.

 

Steve N.

Empirical Audio

 

Link to comment

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now



×
×
  • Create New...