Vibrational effects on vintage hifi components

[jimmc101]

Originally Posted by crackle

I seem to remember that a long time ago there was an article in Wireless World ...
Mike

You didn't imagine it. It was in the April 1970 edition.

Dynamic Range Versus Ambient Noise by George Izzard O'Veering

OK so emeritus and Craig beat me to it.

[barrymagrec]

Originally Posted by crackle

I seem to remember that a long time ago there was an article in Wireless World ...
Mike

I seem to remember this article, it was in an April edition - I think it also mentioned the possibility of obtaining adjoining property at advantageous prices due to the noise levels.

[Craig Sawyers]

Originally Posted by stevehertz

I stated quite clearly in the first post that vibrational effects vis a vis record decks was NOT the issue I was referring to, that is a well documented, well covered phenomenon. As I said, I'm talking about the use of so called specially designed chassis, mountings, dampening devices, feet etc etc wrt the effects of vibration (from speakers, whatever?) on solid state and mechanical components that make up the amp, tuner, CD, whatever, but NOT record decks. Just to clarify as a lot of the responses so far seem to be about record decks! I'd like to keep the thread on track, not about record deck feedback. Thanks.

Wow!

[stevehertz]

Originally Posted by Craig Sawyers

I too have been prone to fashionable audio tweaks. So I try to prove things one way or the other.

1. Vinyl and CD's are prone to vibrate in an acoustic field. Vinyl clearly so, particularly to oscillation at the cartridge compliance/arm mass resonance with an increased chance if this is too high in frequency. A lid, counterintuitively, can make matters worse - you now have air resonances in a closed box, and panel resonances in the lid structure to contend with. CD's less obvious; but if you feed the focus servo signal into a headphone amp, you hear a clear and distorted version of the music. This is the remaining error signal in the focus servo. So does regular listening level music have an effect on CD playback? Yes.

2. Valves - yes. Mullard published two papers in 1962 (attached) on the topic.

3. Semiconductors and passive components used in high quality audio. No. Take piece of equipment (a) without a signal going through it. Play music through equipment (b). Look at any evidence, with any test equipment I have at my disposal (audio spectrum analysers, 1uV/cm oscilloscope etc) at the output of equipment (a) and there is no evidence of anything at all. Smoke and mirrors.

The only thing that might be going on is equipment panels resonating and adding acoustic colouration. The old Naim advice of "don't site your system in a room with musical instruments" actually makes sense in the context of resonances.

4. There are certainly effects from other resonant objects in the room. We had an open cabinet with some cut crystal glassware. The trick to finding the effect of this sort of stuff is to stand in the room and clap your hands - and listen. In our case, there was a high Q chorus of glassware ringing. We moved them to another room.

5. Cables. Yes, with provisos. Triboelectric effects have been already raised, but it is easy to check. With no audio going through the system, turn the volume up and wiggle the cable around. Silence is good! and vice versa clearly.

Good answers Craig, thanks. I pretty much agree on all counts except re cables I would say don't turn the volume up, just wiggle them at your normal listening level, if you don't hear anything then there's no effect to worry about. Plus, actually picking them up and shaking them about is clearly much worse or onerous than them being subjected to relatively small sound waves whilst lying still. Tests have to be relative. You can 'prove' lots of things by taking them to extremes. That serves no point and just feeds paranoia.

[mhennessy]

Originally Posted by Craig Sawyers

CD's less obvious; but if you feed the focus servo signal into a headphone amp, you hear a clear and distorted version of the music. This is the remaining error signal in the focus servo. So does regular listening level music have an effect on CD playback? Yes.

This proves that vibrations from speakers are vibrating the disc enough to require minor corrections from the focus servo, which really isn't all that surprising given the minute dimensions involved, but does it prove that it has any effect on the electronics that are reading out the 0s and 1s from the FIFO buffer and turning them back into an analogue audio waveform?

[Radio Wrangler]

Originally Posted by Craig Sawyers

3. Semiconductors and passive components used in high quality audio. No. Take piece of equipment (a) without a signal going through it. Play music through equipment (b). Look at any evidence, with any test equipment I have at my disposal (audio spectrum analysers, 1uV/cm oscilloscope etc) at the output of equipment (a) and there is no evidence of anything at all. Smoke and mirrors.

I can back this up.

I quite often test RF and audio electronics strapped to a vibration rig. High class spectrum analysers and distortion meters are in use and the rig is producing vibration levels considerably in excess of those measured in single engine piston-prop aircraft and one set of tests evaluates the equipment under the levels of vibration expected in damaged helicopters.

The shaker table is about the size of a mini, the voice coil needs oil cooling and applied amplifier powers have been up to a quarter of a megawatt in one extreme case in the past. The frequency response allows the shaker to reproduce music when the test lab has its public open day.

I've also shaken state of the art test equipment (spectrum analysers, precision receivers) while monitoring their noise floors. Once, the previous test job on a shaker had been the guidance platform for the Ariane 5 launch vehicle, so the rigs can realistically simulate a space launch.

Microphony problems: Nil.

On one occasion the coax cable for the master accelerometer on the table went open circuit. The control of the table went open loop and the table repeatedly smashed into its endstops until someone hit a panic mushroom switch. The gear on test was an HP spectrum analyser and an Icom R9000 mounted in a Schroff mini-rack. When things came to rest, I saw the analyser boot itself up and the Icom came on. Both worked to spec. Post incident investigation showed a couple of cracked spot welds in the Icom and both units had broken their memory back up batteries off of their mountings and connections.

Some testing can get a bit hairy

Even without equipment malfunction, the levels of vibration go far beyond what even the true aficionados get up to in their listening rooms.

As far as CDs go, vibration can induce servo positioning errors and focus servo errors. The data off of the laser undergoes a lot of delay and an awful lot of processing before it leaves the CD player. There is extensive error correction as well. From the error point of view, it should either be perfectly normal, or it should be showing major effects. The same thing with DVB-T television, there are no cues as to problems, then it abruptly comes apart big time.

This is a normal characteristic of digital systems employing error correction. So the audiophiles say that what they hear is the effects of the servos fighting the vibration affecting the power supplies and getting into the audio path. Hmmm
well that sort of thing can be designed out by competent designers, and the incoming mains is dirtier than what the servos will impose on it with washing machines, heating in the same house and arc welders in some neighbour's garage.

If it is a problem, the solution lies in good electronics design and maths, not in ultra-sharp spiky feet.

David

[Craig Sawyers]

Originally Posted by mhennessy

This proves that vibrations from speakers are vibrating the disc enough to require minor corrections from the focus servo, which really isn't all that surprising given the minute dimensions involved, but does it prove that it has any effect on the electronics that are reading out the 0s and 1s from the FIFO buffer and turning them back into an analogue audio waveform?

That makes the mistake of assuming that digital is 0's and 1's. It is an analogue waveform, with finite transition time, being passed through a series of non-linear amplifiers.

In any event, the signal coming out from the laser assembly on an oscilloscope is very definitely analogue with a classic eye diagram (that is how you set a mechanism up, by optimising the window in the eye diagram). Any systematic errors in the focus and tracking servos as a result of disc vibration thicken the eye diagram transitions.

Disc vibration is something that was understood very well back when Meridian were making vaguely affordable CD transports. The CD mechanism was attached to a glass panel which sealed with a felt gasket against the front of the sealed casework.

Now of course, as you argue you can clock the jittery data stream into a FIFO, and then clock it out again and voila nice data. But wouldn't it be nice if it were not necessary?

ultra-sharp spiky feet.

Have one use, they stop the carpet getting a squash mark, the wee holes made soon go with a good Hoovering (beats as it sweeps type as implied)

[Radio Wrangler]

Of course vibration of the player will have an effect on the signals being read from the laser interferometer reading the disc. There will be added noise in the voltage dimension and there will be jitter in the time dimension. The question becomes one of whether it matters.

The signal from the interferometer is a photodiode current which is soon turned into a voltage. Voltages are voltages, currents are currents. Neither are digital or analogue. Digital and analogue refer to the assignment of meaning to the values of voltage or current. As Craig describes, the waveform from the photodiode can be viewed on an oscilloscope. As the places and lengths of the pits vary with the music in its encoded form, the waveform will appear almost random. It will be trying to switch between two levels, but both those levels will be noisy and there will be slower wandering due to disc flatness and servo tracking characteristics. It won't zoo instantly from one level to the other, but will move across at a finite speed. It will take a small but significant fraction of the time assigned to a minimum length pit in each direction.

If we sync-up the scope to the sampling clock, and the CD player is working well, the sync time should always happen at a clear spot in the picture, about half-way between the transitions. Multiple traces will build up, many at the low level many at the high, and many making transitions in all of the possible combinations.

The picture looks like an eye with a blank centre. If the noise isn't too big that the high and low levels get close enough for confusion, and if the transitions are clear of the sampling clock, then at the time of the sampling clock, we could make a good decision as to whether there is a pit there or not. The player is sophisticated enough to try to move the sampling clock and the voltage of the decision threshold to best fit the eye. When the laser head was set up someone viewed the eye and twiddled the adjusters to get the eye opening as wide and tall as possible.

So now we've turned the signal into a digital form, above the threshold and below the threshold are considered to purely indicate whether there is a pit there or not. Occasionally, it will get it wrong and an error bit goes down the pipeline to the digital systems.

But the player was also fiddling about with the timing of the sampling clock trying to fit the timing of the eyes coming off the disc. This isn't good, this wandering could affect the sound. So it has to be fixed. We know that the timing the pits represent was supposed to be dead regular, so any deviation from dead regularity is wrong. So if we have a dead regular clock from a reliably stable source, we can delay the data by stacking it up, and taking it out at regular intervals. If our stack gets too full, before it overflows, we can tell the disc motor to slow a little. If our stack gets too empty, we tell the motor driver to speed up. So now we have data coming out of our stack at nice regular intervals and the CD is turned at whatever speed is needed to keep the stack topped up to average about half-full.

Any jitter in the audio coming out of the player is that of the stable crystal oscillator, not that of the stream coming off the disc.

The tracking servos moving and focusing the laser are working all the time, following variations in the thickness of the disc, eccentricity of the hole and just thermal electronic noise. The motor speed servo is hunting up and down keeping the stack within its happy bounds. Adding vibration from outside doesn't start the servos having to act, they're doing plenty already, it will change the patterns they are working in. The player already needs to be carefully enough designed to keep normal servo action out of the audio path.

So there is already a fifo stack in CD players. They need it. In the extreme you could read the disc, do all the error correction, stash it all in a currently modest amount of electronic memory, shut down the servos, stop the disc, and then read the memory to play the music.

Now, let's just rewind back a bit... what about that crystal oscillator used as a dependably regular clock to correct the timing to?

Crystal oscillators ARE affected a little by supply voltage variations. If you buy a decent quality oscillator, how much is in the spec sheet. Good design is needed to keep the variations of the supply voltage low so that the effect on the oscillator is trivial?

A few parts per million per volt variation is routine, I've had special oscillators with fancy regulation where it's in the few parts per billion league. Is this low enough? Well, yes if you're pragmatic and have a sense of proportion, but if your belief system says there is no limit to the perception of your ears, then nothing can ever be good enough.

Oh, and Quartz crystals are gravity sensitive. Run one, measure its frequency, then turn it over and measure it again. Damnation, it's different! keep on repeating and you'll find it's repeatable. A 2g change in gravity due to the turnover may give you a few dozen parts per billion. So the crystal will act as a transducer for the acceleration value of vibration in certain planes. Is this a problem for you? well, it depends on your beliefs. If your hearing capability is infinite, nothing will ever be other than a problem. For normal people, it's probably OK. If you want to take precautions, a low phase noise crystal oscillator on a vibration mount ought to put it well below perception by anyone who believes in ear tests

David

Vintage HiFi didn't (and of course doesn't) get affected. I find this out by reading the advertising blurb (and as we know advertising sucks the most out of anything) and not seeing a single mention. Only the modern stuff suffers (unless you spend lots), seen by the advertising mentioning it. Must be true...

[mhennessy]

Originally Posted by Craig Sawyers

Now of course, as you argue you can clock the jittery data stream into a FIFO, and then clock it out again and voila nice data. But wouldn't it be nice if it were not necessary?

I've never seen a CD player that doesn't have a FIFO buried away inside its chipset - they are essential for all the reasons David explains. And even if you could eliminate it, what would be the point besides marketing?

[samjmann]

I've seen the honeycomb design on many designs as a chassis stiffener. Doesn't
it mean that you can get away with using thinner metal and reduce cost? Just think of the silver paper on your Tunnocks Tea Cakes, smooth it out and it's much less 'strong'.

The FIFO buffer in Davids post is a good point, I worked on a Panasonic compact music system recently, and on this you could actually stall the disc for a good 15secs before the memory 'ran out' SJM.

and on this you could actually stall the disc for a good 15secs before the memory 'ran out' SJM.

And the "walk around" type boasted 60s or more of buffer, now we have MP3 players, it's all buffer.

[mhennessy]

Originally Posted by samjmann

The FIFO buffer in Davids post is a good point, I worked on a Panasonic compact music system recently, and on this you could actually stall the disc for a good 15secs before the memory 'ran out' SJM.

Traditionally, the FIFO was always very small - bear in mind the cost of RAM back then! The classic SAA7310 used an external 8KB RAM chip. It was only ever intended to store a tiny fraction of a second (about 1.5ms IIRC), and was there to elegantly side-step the wholly impractical goal of making a cheap mass-produced plastic disc spin with no jitter at all. In any other area of engineering, the FIFO would be considered the height of elegance - unfortunately, one of the main tenets of being an audiophile is a distrust of good engineering - though this is the first time I've heard anyone suggest that a FIFO ought to be rendered unnecessary (at presumably huge expense).

As RAM prices fell, it meant that FIFOs could be bigger. For a portable or in-car machine, this means you can implement an anti-shock feature, whereby you can re-read data from the disc if the machine was disturbed on the first attempt.

This was part of the MiniDisc standard too. Here, the buffer wasn't just to allow portable use, but also to allow for editing (so the laser can be seeking to a different part of the disc without interrupting the music).

The fact that long buffers appear in home CD players as well is simply because the chipsets are cheap and easy to use, and a long buffer doesn't do any harm, providing the logic controlling it doesn't stick at a fixed delay, which would introduce noticeable operational latency. You simply have to allow the "water level" to vary, which is easy to implement. Many modern transports are actually DVD mechanisms (based on computer drives), so these read the data off in bursts every few seconds.

To return to the original question:

In terms of vibration in a CD player, the master clock crystal is the only vaguely vulnerable part of the machine, and once Paul Miller of Hi-Fi News started promulgating the whole "jitter" nonsense*, manufacturers were obliged to react - hence you quite often find measures like bits of foam rubber damping material stuck to crystals, slots all around the oscillator section of the PCB to mechanically decouple the oscillator from the rest of the PCB and dedicated power supplies too. It's good practice; it doesn't do any harm and it doesn't really add much to the cost, but make no mistake: it's 99.99% Marketing.


* Jitter in digital audio is actually a massively important subject, and there exists a great body of work available to study. However, the established reality bears only a very loose connection with what the hi-fi fraternity have come to believe.

[Radio Wrangler]

Hmm, a CD player without a FIFO, relying on mechanical factors to control jitter....

I first thought about phase locking a massive turntable to a crystal reference oscillator, but that smacks of the Japanese quartz lock direct drive record turntables, and the hifi magazines would lay into that idea.

So it would have to be done without a crystal oscillator, and the clock derived from the signal off the disc.

This would all be fabulously expensive, dramatically heavy and with a long wait for the thing to come up to speed and it would be properly sensitive to vibration. AND it wouldn't work very well, so there would be audible differences from model to model.... Think of the opportunities for manufacturers to use esoteric materials. Oooh!

All in all, audiophile heaven!

David

[G8HQP Dave]

How would one do CD decoding and de-interleaving without a data buffer of some sort? Some people seem to imagine that 16-bit audio comes straight off the disc; it doesn't.

[G6Tanuki]

I remember at least one battery-powered portable CD-player which had sufficient buffer capacity that it actually shut-down the disc motor (to save battery power) once the buffer was full.

A 10-track CD and 64Mbytes of RAM - you could typically load an entire track into buffer then turn the motor off for several minutes.

The DAC doesn't care [or know] if the bits it's converting are coming from a few tens of Megabytes of RAM or direct-from-disc: so long as there are bits there when the DAC calls for them it's happy.

[John10b]

Wow, how very interesting and complicated!
Can I go back to post 4, “ the vibrations affecting the air spaced variable tuning capacitors, particularly on FM”.
I did not experience this during my service days, could you expand please.
Cheers
John

[Herald1360]

Well, if an airspaced capacitor controlling the LO is being shaken about, it will introduce unwanted FM into the IF signal. If this is high enough it will be detected as unwanted noise.

Unless applied vibration happens to coincide with mechanical resonances of the capacitor in the audio range, it's unlikely to be an audio problem. Metal fatigue in the capacitor, maybe......

[Radio Wrangler]

Originally Posted by G8HQP Dave

How would one do CD decoding and de-interleaving without a data buffer of some sort? Some people seem to imagine that 16-bit audio comes straight off the disc; it doesn't.

Sorry, what I meant was without a re-clocking buffer.

One is still necessary for de-interleaving and the FEC process.

DAvid