Has anyone ever demonstrated how different digital cables produce different levels of jitter? Jitter is directly measurable, so this should be trivial to do.


I'd like to know how different kinds of bit errors lead to variations in "transparency", dynamic range (bit loss???) and "crisper bass rhythms" and at what rate/level of bit error production these phenomena become audible. This should be readily audible, and probably easily simulated on a PC too!


I have a dragon in my garage...

Edited by user 03 May 2012 15:16:15(UTC)  | Reason: added an extra point

I and those I listen with have no trouble hearing the difference between dacs and/or digital cables. In saying this I am seldom contradicted by others who have listened to different cables but quite often by those who have not but who are convinced AU PRORI that such differences cannot exist. Measurements of certain parameters are just that; measurements of certain aspects of an object. They are NOT a complete description of the sound of it will produce. I wonder if magazines on gourmet cooking are plagued by demands that the writers show by chemical analysis why their recommended recipes are superior? Plato must have been right in saying that we know everything at birth and gradually forget; there are any number of things I use to know that I don't now. I marvel that I use to know that the only thing a TT had to do was to revolve at the correct speed; that speaker wire and ICs were all the same; that all amps of similar power sounded the same. The advocates of term limits over here must be on to something as the net result of 50 years experience in audio has been to addle my brain to the extent that I now believe that just about everything you do matters and that at least half of the sound quality of a system is due to careful set up and judicious selection of compatible gear and accessories. When I first started I would have subscribed to the current view of my mechanical engineering friend that you can tell what a component sounds like from reading the spec sheet. He has just selected an amp in this manner.

HIfistan, I want to add something about your friend. I must confess to bearing a degree of prejudice against him, as apparently do you in regards hifi matters, since I recall reading somewhere else here some anecdote about him going on about a ribbon tweeter vibrating at 3Hz, or some such nonsense.

In fact, he is quite correct about referring to spec sheets - he is, after all, a mechanical engineer and in the real world of engineering (rather than in the world of audiophilia) specsheets are everything. It's easy to understand why he would transfer this understanding to the mystical world of hifi... he doesn't realise the central role of voodoo in our hobby!

Back on topic, there is the matter of what specs you might hope to read for digital cables. For a co-ax cable you would expect to see characteristic impedance, for instance. What you don't see is any mention of RFI shielding (in decibels?) or - of course! - jitter. I don't know about shielding ( you could probably find out though, Belden and other realworld manufacturers do publish detailed info on their cables) but there is I think a very good reason that jitter is not part of the spec - cables don't produce jitter!

In general, we have a problem in our industry, well... audiophiles have a problem... that too many measurements might take the mysticism out of hifi. We DO need more and better and much more consistant specifications - it would be nice if we could better match amps to speakers without having to go through all the rigmarole of 'auditioning'. It's hard though, amp makers don't tell us much about what their amps are actually capapbel of, there are no really useful standards for power, energy reserves/PSU 'stiffness' etc. And try getting figuers on distortion or noisefloors, etc for speakers! The result is the industry expects us to convince ourselves via the hopelessly flawed but sales-friendly method of 'auditioning'...

I don't want to labour [ note British spelling] the point; I don't think specs are meaningless. I just don't think they tell the whole story. I had amps that were essentially perfect technically back in the 70s; at least as to measured distortion etc. If passive components like resistors and caps can sound different then complex assemblies of them should sound different however they measure. Many have remarked on how using different parts lead to a different sound but I don't recall anyone being able to account for the change by taking measurements. I am not a physicist nor even play one on here; I just say what I hear and leave it at that. My friend can hear that cables sound different but he attributes any difference to capacitance or resistance differences between them.


Plato must have been right in saying that we know everything at birth and gradually forget

Plato was abjectly wrong.

About this yes; I was attempting humor. But in both Science and Religion his influence is still huge. In one of his last interviews Werner Heisenberg said "The whole of Western Thought can be seen as a contest between Democrates and Plato; in out own time we have lived to see the victory of one of them [Plato]." John Barrow's "Pi in the Sky" is a good introduction to Plato's influence in physics.

Edited by user 03 May 2012 22:27:39(UTC)  | Reason: added material

You may have accidentally misquoted me. I wasn't suggesting for a moment that RFI would be respoinsible for bit errors - indeed, as I said in my post, when digital transmission of 1s and 0s stops working, the modern world will end...

I'm not an electronics engineer by trade, I was nominally an embedded firmware control systems engineer, though latterly I've earned my crust running projects. For the last 10 years, it has been my pleasure and privilege to work with and alongside one of the finest analogue/RF design teams in the country, probably in the world. I feel relatively safe saying this because what we were doing - building mixed-signal ICs incorporating multiple CPUs and one or more 2.4GHz radios onto a few square millimetres of CMOS - was declared to be impossible by the then-finest minds of the late 1990s. We also incorporated an FM radio - much harder than it sounds, and much harder (frankly) than any of us thought it was going to be :-) - and eventually got the noise floor of these contraptions down low enough that we could incorporate a GPS receiver as well! As this wasn't an academic IP development exercise, this was building a company that sells hundreds of millions of chips a year and has a market cap approaching a billion pounds.

Why am I saying this? Because what I've learned is that, to quote one of our lead gurus - Everything Bloody Matters! He and I - with him dreaming up experiments and me being his monkey (I was program manager, so if I wanted the chip shippable, the buck stopped here...) - spent weeks characterising one of our early chips, even spending a happy time moving SMD decoupling capacitors half-a-millimetre at a time up a track and measuring the resulting changes in radio performance - because, it turned out in this particular case, that said cap and track was forming a resonant circuit with the bond wire inside the chip package.

One thing I learned very early is that D-to-A converter circuits are remarkably adept at finding ways that even their designers - especially their designers - never imagined to convert any sort of noise into crap on their outputs. Noise on their power supplies, noise on their control lines, noise on the ground, (phase) noise on their clocks - you name it, it'll turn up somewhere you least want it, and a million quid's worth of test equipment doesn't always help you find it...

Sorry to rant, all I'm trying to say is that I have no trouble whatsoever believing that a D/A converter's performance can and will be compromised (aka altered) by incoming noise.




I'm sure that a chip's digital back end team who've been struggling for 3 months to get Cadence's vastly expensive tools to give timing closure on a few million gates over process/voltage/temnperature will be delighted to hear that.

Sorry, but if you join the forums and start beiung sarcastic, then this is what happens :-)


Look, we've done the bits-r-bits argument. Let's not start it again. Where we're at, at the moment, is that people whose ears we trust are saying - and in fact many people have been saying, for many years - that they can hear differences between SPDIF interconnect cables. I, too, found that very hard to swallow; like you, my first reaction was "no, it can't be true". But two things persuaded me that we should keep an open mind - one, there's a hell of a lot of people saying it; and two, my experience (both personal and professional) with RF suggests that there's a reasonable explanation hanging about out there that warrants further investigation.

We all need to remember the great un-named law of engineering, which has been invoked a lot over the last 10 years of my professional life as I've wandered over to an analogue engineer's desk with a graph I've just taken: If a thing happened, it must have been possible!

Edited by user 04 May 2012 20:18:50(UTC)  | Reason: Corrected quoting, should have previewed first, want to rush off to watch TV

I also studied perception as a graduate student in Psychology at the Univ. of Chicago and later in my studies in the History of Science when I was doing my PhD. Granted that our senses can be deceived; how is that relevant to our appreciation of something, sound quality, that appeals directly to these fallible senses? All sound reproduction is based of fooling the senses; there are not really a group of musicians in my listening room. Sound reproduction is the art of illusion; if I can't trust my ears to determine whether I like it or not what am I to turn to? A group of scientists in a lab chock full of test gear measuring carefully and awarding grades? Consumer Reports use to do careful tests of audio gear over here in a scientifically unimpeachable manner. I found the results usually contradicted my own impressions. On the other hand the equally objective tests done by the old HiFI Choice were quite helpful and I still refur to them decades later. Let me reiterate this point once again: LISTENING TO MUSIC IS A SUBJECTIVE EXPERIENCE. Trying to "objectify" it is a waste of effort.

Yes, absolutely; If digital logic didn't work, the world would end. It can be noisy as hell, but that doesn't matter as long as you stay in the digital domain. It basically works.



Well, this is where it all gets a bit surreal. Remember I said we'd stuck an FM radio into one of our chips? How hard can that be, we thought? Well, we thought that we had it working, so off it went to the customer for acceptance tests. This is all water long-passed under the bridge, so though I'm slightly uncomfortable talking about it, it's probably safe enough. Anyway, customer rang up and said "You're having a giraffe, aren't you? I can't sell this, it's bloody awful! Where's the real one?". He wasn't connecting it to some high-quality output transducer, or even some high quality measuring device, he was simply doing what he was supposed to - soldering it down in a phone PCB, and listening to playing Radio 4 over the standard supplied combined earpiece/aerial wire. Difference turned out to be (cutting out several weeks of investigative work at both ends) the levels of phase noise on his clock vs our lab one, and our PLL/VCO needing some "persuasion" with a firmware-based mallet :-). Our lab clock followed his published clock spec, but it's not just the sum of phase noise that mattered, it was where its energy was relative to the clock fundamental - sidebands close-in or far-out; the old wow-and-flutter spectrogram from the turntable reviews is the exact measurement that I've spent a lot of time staring at, it's just that the scales on the X-axis are in MHz...

That bit of surrealarity - which may not be relevant, but it might raise a smile - aside, you can get some very odd effects at the digital-to-analogue interface from the most unlikely sources. Trivial example: if you have a sinusoidal clock driving ICs with schmitt trigger inputs, then if you (accidentally) vary the amplitude of the clock, you bang a phase hit through the system. Mobile phone designers go spare about this - "how dare you switch an extra few fF onto our clock line?" How do you accidentally vary the clock amplitude? Put it though a simple buffer with a wobbly power supply... How do you wobble the power supply? Connect it to anything digital - a nice big FPGA with some SRAM in it will do nicely, watch it take a gulp of current every time the CPU reads a line of RAM :-). That'll give you a nice high-frequency source of phase noise for your nice clean clock... OK, don't use sinusoidal clocks, use square ones, but when you have crystals and PLLs in your system, they tend to turn up. My problem, talking about this, is that, as I said before, I'm not an electronic engineer. I've spent a long, long, time in labs staring at spectra on analyers, looking at real problems and explaining them to the circuit's designers, but I have no theoretical knowledge to back any of it up... :-(

The only people who can tell the circuit designer what an individual D/A chip will do with a particular noise signature on its clock are going to be the D/A chip designers - Wolfson, for example. How audible - what sort of degradation in performance - the results will be in the finished sound is something that you or I simply can't tell. All we can hope for is that the designers done the best job he can to clean it all up.




Yeah, OK, I'll buy that. Susan Greenfield in her TV series years ago summarised that view by saying something like that when we're asleep, we dream a reality. When we're awake, the dream continues exactly the same, it just occasionally bumps into and gets referenced to the external reality, so the content of the dream somewhat resembles external reality. I'm paraphrasing, but I think that was the gist, yes?



I see your view, I don't see on the evidence presented that I should agree with the conclusion. I'm not trying to sound snotty or elitist or in any way special, because I'm not, but I think that there's a skill that comes with practice. Though I drive every day, I can't drive like Michael Schumacher, I've never had the opportunity to learn. Though I drink <much less than I used to> I can't taste subtle differences in wine like (say) Jancis Robinson, though with practice I am getting slowly better, because I am taking every opportunity that I can to learn.

People can learn to play tennis, or the guitar, or do calligraphy, or take photographs, or paint. Why do you suggest that listening is the one skill that people can't learn?

I think I've got better ears than the average man-or-woman-in-the-street. That's not because of any sort of genetics, I'm certain, but because I've spent the last 20-odd years since I was a student practicing critical listening, backing up what I think I hear with second opinions and, where possible, blind testing (see my post a few months ago, for example, about blind-testing some RF filtering arrangements around my house).

The average person on the street, I have noticed, doesn't listen in the same way. They don't critically assess what they're listening to. On the other hand, you could sit me down with them in front of a football match and ask us both what we see. The other person may wax lyrical about some arcane bit of ball placement and the masterly use of spin to control yadayadayada, I'll just say "a bloke kicked the ball to the other bloke". Not interested.



Ah, now that was religion, don't get me started on mass brainwashing, and control of the illiterate poor, I'll be here all night.

A much better tongue in cheek counter-argument is the old Fiat Strada ads, if you're old enough to remember tham. Yes, we all said, millions of Italians can be wrong!


The reason I said we've been here before is 'cos we have. If you go back to the early 90s and find the archive of rec.audio.high-end, you'll find me flying a subjectivist flag saying "digital sounds crap" and engaging in some lovely flame wars with the (mostly American) objectivists who absolutely refused to believe that 2 CD transports could sound different to each other, feeding the same D/A converter. God, the rows we had, between the "it sounds different" brigade and the "it can't possibly sound different, the bits are being read with no errors and sent over SPDIF, they're identical, the THD and frequency response are the same, I don't need to listen to it to know that they're the same, you're just deluded idiots" mob. Ah, happy days. Anyway, to cut a long story short, we were vindicated, because eventually the idea that jitter is important came to the fore. A combination of badly jittering transports, and crap clock recovery in the DACs of the era meant that all was explained. The subjectivists said "we bloody told you", the objectivists basically coughed, looked down, shuffled their feet, and said "yeah, well, we knew that". Which they didn't, of course, else they wouldn't have designed that stuff in the first place. Moral: Trust your ears, not (just) your measurements. Your measurements will tell you what you've measured, but if you're not measuring the right things, you'll never see them. Never be afraid to design new tests and measurements. 99% of them will only tell you what you already know, you're always looking for the 1% that doesn't give you the answer you expect.

So that sort of gross jitter - a monstrous first-order sound killer - is now a solved problem. Mostly, though Martin/Chris's measurements or Paul Miller's (more instructive, sorry Martin) regular charts in HFN do still show that there's some incompetence out there, so we have to stay on our toes.

So with that out of the way, and some good entrenched opinions in the world of the merits or otherwise of FIR and IIR digital filters, we're now looking for second-order effects.

Anyway, I hope the preceding long paragraph about the 1990s jitter wars helps to explain why so many of us of a certain age react so badly when told that we're imagining things. We stood by our ears, and we were right to do so.



Well, this isn't half a dozen loonies. This is a bunch of people, with a reasonable track record on hearing differences, all reporting similar effects. I would say that's worthy of investigation, not of scorn.

Thank you, RK, much obliged. That's a great article. The opening few paragraphs describe the last 10 years of my working life to a tee. Patiently tracking the first wafers through fab, through assembly, through Fedex, watching them get stuck at customs in Stansted for 2 days (!), then soldering them down, plugging them in and powering them up and.... The moment of elation when you realise it's basically alive, followed by weeks of depressing testing as you chase round looking at all the things that aren't quite right, and hoping you find them before the customer does...

I could have short-circuited Kondoh-san's investigations, though. It's always the PLL :-). More generally, it's always the bits you weren't able to test beforehand, which in a mixed-signal ASIC means it's the analogue bits. If the digital bits don't work, then send yourselves back to school with a note saying "should do better".

Kondoh-san was perhaps on a older/larger technology than we've been using, or had vastly better facilities in-house, if he was able to probe it on the fly. For 180nm or 90nm (or less), we were able to probe the die pads but not often the chip circuit inside the pad ring; a fabless semi such as we were has to ship the parts to a third party for FIB work, whereby clever elves can solder wires onto the points you want to probe. The FIB elves are incredible when you ask them to modify bits of the circuit several layers down. They cut their way down, maked the change you want, then attempt to build their way back out again, sometimes successfully.

Plenty of discussion without a recommendation how about

"What is a good cable for say £100"?

Thanks Stan,

TFU was my initial thought