I wonder as I have been building oscillators for years . COG are about 30 ppm and standard crystals 30 ppm . Some say it is not that , it is the moment to moment changes we hear . I also suspect absolute frequency less important than fluctuations as long as 0.1% correct . An oven is well worth doing . My old frequency meter had one . The cheap 30 ppm resistors seem to match COG well and result in 5ppm when lucky . Ironically more expensive resistors sometimes are not the best match .
As for the 4060 . I was wondering how DAC's work . Do they lock onto the reference given or do they establish a timing sequence when powering the crystal ( i e . the oscillator input fires up the crystal in sequence or receiving the output of the crystal sets the wheels in motion , the later I hope ) ?
If the DAC will accept the 4060 I will run it on a 9V battery for starters . I will pot down the output to start ( to avoid damage to DAC input ) . I will try a bit of shaping as the crystal into such oscillators is not a square wave I seem to remember , I doubt it matters but ...
The inspiration for this came from a very early HI Fi Choice review which was astonished to find a cheap CD player that had poor absolute speed was well liked sonically ( Sansui + 0.4% ) . The investigation found an RC clock . They hypothesized that RC might sound better . some years later the clock became the most suspect component .
Cheap test idea . Record 1kHz on a CD . Test before and after . If no frequency meter handy leave generator running to use as a beat frequency test . It probably will be OK if the generator is warmed up for an hour first . Even reading a clock into a microphone will work as a CD time reference . After one hour see how close . 3 seconds out would be OK .
Shouldn't be too hard. Mine uses 8 Philips TDA DACs, their output being a linear signa. This is then taken to the other half of the board, which has another rectifier, pair of electrolytics and 3 pin stabilizers, all of serving only the AD 847, which is in a socket.
Well, you can take the signal from the socket and feed it to a pair of tubes, or a set of FETs and MOSFETs, or whatever.
I don't have the link on this PC, but I'll get it for you. I must also mention that since then, I have see literally a myriad of Chinese copies at 10% of the original price, which makes me suspicious, but thaat you will have to tackle on your own. I prefer to feed Ozzies rather than the Chinese. I find I can trust Ozzies, but not the Chinese.
I was going to order a class D PCB from China ( inc chip ) . My reckoning being they would struggle to clone a Philips chip of that complexity . Wanted to test the specs more than listen 200W for $15 seemed a bit of fun . Want to drive motor with it . Don't give a stuff if it is any good or fails . A little valve amp Separo SE88i seems well made , the exception to the rule .
For me a simple "answer" is that digital circuitry, and audio analogue circuitry are a bad mixture. Always. Digital loves sharp transitions, beautiful square waves. Analogue hates these, it yells out, "Interference, interference ...!"
Frank
This is impossible to argue. I could say in all honesty that my Yamaha CDX 993 does very well indeed on its own, with three separate stages at the output - I/V converter, buffer and output amp proper. The I/V stage uses an op amp (AD826), the other two are fully discrete with dual trannies, current mirrors and the rest of the bells and whistles.
Someone might say that while it does fine, it could do better still in some other way.
I believe - right or wrong - that it's the I/V stage which is critical. Get that right, using a very fast op amp, and you should be home and dry. I find it symptomatic that standard op amps (as Yamaha original had, I pulled it out and stuck the AD 826 inside) seem to lose quite a bit of focus and detail. It appears to me a very fast op amp will by default improve the sound.
My friend John builds magnetometers . His advice to me was to use film caps and an oven . He said COG also acts as quartz . They use a small bunch of various capacitors to match a resistor for near perfect stability . From what I understand this rivals quarts if so ( < 5ppm ) .
John said that some DAC's will be happy with what I am doing ( read very happy ) others not . He said if not happy it will be obvious .
I thought a true square-wave to be best . To use the full 2 to the power 14 divisions might be best . Or less if the caps dictate that . I have a hunch that polystyrene are worth trying . I have some 33 nF lead foil types ( Philips ) . The lead has good aging qualities with the lead out wire ( interesting English there , lead and lead ) . I also have 10 nf standard type ( and values in nf to pf ) . What I hope to hear is less harshness and openness .
John said that some DAC's will be happy with what I am doing ( read very happy ) others not . He said if not happy it will be obvious .
I thought a true square-wave to be best . To use the full 2 to the power 14 divisions might be best . Or less if the caps dictate that . I have a hunch that polystyrene are worth trying . I have some 33 nF lead foil types ( Philips ) . The lead has good aging qualities with the lead out wire ( interesting English there , lead and lead ) . I also have 10 nf standard type ( and values in nf to pf ) . What I hope to hear is less harshness and openness .
People when talking of CD player clocks always seem to confuse long-term stability with short-term jitter. Quartz crystals, when used properly, are fairly good at both. CR oscillators can be made to be good at long-term stability but are almost always completely useless at short-term jitter. This is because of their very low effective Q; I think I read somewhere that a Wien bridge has a Q of 0.25. Quartz is more like 10000.
It is jitter which matters for audio. Long-term stability, unless really bad, is not noticeable. Just use a good quartz oscillator: no need for ovens, GPS syncing, temperature compensation etc. The circuit techniques for low jitter may have little to do with long-term stability and vice versa; an exception is using good quality crystals, which can benefit both.
It is jitter which matters for audio. Long-term stability, unless really bad, is not noticeable. Just use a good quartz oscillator: no need for ovens, GPS syncing, temperature compensation etc. The circuit techniques for low jitter may have little to do with long-term stability and vice versa; an exception is using good quality crystals, which can benefit both.
You could use an OXCO, bit pricey tough..
http://nl.farnell.com/jsp/search/br...k=gensearch&Ntt=oxco&Ntx=mode+matchallpartial
http://nl.farnell.com/jsp/search/br...k=gensearch&Ntt=oxco&Ntx=mode+matchallpartial
I think this will be some fun for the winter months . Rotel were very kind to offer upgrading advice . Were it not for the RC reference in that old copy of Hi Fi choice it wouldn't have occurred to me . The good news with this one is I can only use my ears as I doubt my test gear will show any difference ?
Opamps are not optimal for I/V conversion, but until someone designs an integrated optimized device they are about the best we have, if we are constrained by size and cost. Barrie Gilbert once remarked to the effect that an opamp was the worst thing one could use for this task!
The problem is that opamp input Z, open loop, is high, but for fast and accurate I/V conversion we want to start low and reduce further. Patrick has shown an approach using common-gate FETs which is a good start. I augmented it a bit by adding a loop around the input device to reduce the input impedance, and additional enhancements are desirable.
If so-called "current output" DACs had outputs, as the name suggests, that were high impedance, the problems would be alleviated to begin with. But "current output" is almost a misnomer. A better description is "output with a code-dependent output impedance and limited voltage swing capability, which has to be terminated in something that behaves like a much lower impedance."
Brad
Depends on your gear. Measuring jitter is now commonplace, but still requires specialized equipment.
DF96 is spot on. Think of a quartz crystal as a big flywheel with low-loss bearings. If frequency change over a wide fraction of center frequency with agility were a requirement, it's a poor choice --- that might well be an application where an RC oscillator would be effective. But as he points out, that's not what we need for digital audio.
Among other things, an R/C oscillator is also much more sensitive to power supply variation than a quartz crystal oscillator, and that will add to jitter. I'd want to power it with one of those super duper active parallel regulator circuits I've seen around here to get extra low noise on the power supply line.
Yes, and the devices we would use to provide the gain to overcome the RC network loss to sustain oscillation have varying amounts of noise. CMOS in particular is famously noisy at low frequencies. Using lower impedances and hence more power will help, along with low voltage noise devices. But again, where RC oscillators have been appearing lately is in cost-challenged designs with modest requirements for accuracy, not for high stability and low jitter.
Quartz may be challenged by micromechanical resonators at some point.
OK, but how many times has your music gone through op-amp based A2D, D2A conversions before you got it?
Want stability? An atomic clock is only about $6K. Of course, the cable from clock to transport will cause more jitter that the rubidium clock is gaining you anyway. An integrated crystal/oven module is in the hundreds so a more practical choice. Nano machines are a very cool technology to watch. Just not here yet.
I'd like to see more effort on the recording end.
Want stability? An atomic clock is only about $6K. Of course, the cable from clock to transport will cause more jitter that the rubidium clock is gaining you anyway. An integrated crystal/oven module is in the hundreds so a more practical choice. Nano machines are a very cool technology to watch. Just not here yet.
I'd like to see more effort on the recording end.
Could someone say how jitter is caused . I read about it so many times . Not sure anyone has gone inside the problem . People speak of the oscillator working with high Q . I remember the in port receiving a distorted sine wave from the crystal when I had a look . I doubt the scope was corrupting it greatly . I will try to recreate that tomorrow if I can , my new scope is very high impedance and floating which must help . If of any value I will post it . Seems to me a sine wave is a good starting point . It passes to a Schmidt trigger next I guess . High Q or no I see no connection to a sine wave ? If the quartz is producing a square wave into Q Q dash port the then that to me seems different and yes Q matters . I suspect it isn't . No doubt it was a ghost of my measurements . I will try to find out . I did load the crystal with 1 M so that might be why I saw a sine wave ?
I've got a gut feeling that down the track the obsession with low jitter will be seen to be off the track a bit: I've never specifically worried about reducing jitter to get good sound, it's always been other matters that were more important in making the sound more "analogue" - whatever that means, more real.
Of course, if one puts a great deal of energy in improving jitter specs, then what you're doing is frequently improving other aspects of component performance, which is what I feel is doing the real job ...
Frank
Of course, if one puts a great deal of energy in improving jitter specs, then what you're doing is frequently improving other aspects of component performance, which is what I feel is doing the real job ...
Frank
The trick is not getting jitter, it's how to stop it happening! Nature abhors a perfect beat, at least at our scale of size, just like you never see perfect straight lines in the great outdoors. The slightest imperfection, bit of randomness, anywhere, will cause the striking of the beat to be a touch off ...
Frank
Rubidum clocks may be very accurate but not necessarily low jitter - in fact, a lowly integrated oscillator has probably less jitter than a rubidum clock, they are optimised for different purposes. For A/D, D/A you don't require perfect accuracy but you would like zero jitter. Two different things.
Jitter comes for instrance from the uncertainty of a zero crossing of a square wave. At some point, the receiver has to decide whether a bit is a '1'or a '0', and that zero crossing wavers a bit with power supply, temperature, noise, what have you, so that then causes the zero crossing to waver to and fro and that's jitter.
Malcolm Hawksford showed that even the data pattern (the music) determines some of the jitter, so jitter depends on the actual series of ones and zero that make up the music.
jan
Jitter comes for instrance from the uncertainty of a zero crossing of a square wave. At some point, the receiver has to decide whether a bit is a '1'or a '0', and that zero crossing wavers a bit with power supply, temperature, noise, what have you, so that then causes the zero crossing to waver to and fro and that's jitter.
Malcolm Hawksford showed that even the data pattern (the music) determines some of the jitter, so jitter depends on the actual series of ones and zero that make up the music.
jan
Last edited:
I do wonder if many cloaks that worked well just had better power supplies ? One thing I will try is improving the power supply all around the DAC . That is a spectrum analyzer job so can be measurement based . I remember doing this years ago and getting all sorts of nasty things . In the end I measured across the decoupling caps , I feel that is reliable .
- Status
- Not open for further replies.