DAC64 & Torodial power supply

Are you seriously suggesting that signal reflections can be bad enough to turn 1s into 0s and vice versa?

It would be the first time I've heard anyone suggest that a DAC might not be receiving the correct sequence of bits from a transport and I do find it somewhat hard to believe.

Signal reflections and the like aggravating jitter sure, but not changing the data stream. In any case, I would imagine that randomly changing some 1s to 0s and some 0s to 1s here and there would cause terrible glitches and very audible problems and not be of the fairly subtle nature of jitter related issues.

Michael.
 
Originally posted by timpy
Then when you add in cable reflections your digial bits can be blurred to the point where it can't tell the 1's from the 0's reliably any more.
Click to expand...

Really? Are you sure? Now electrical is almost certainly better than optical in this respect, but even my humble CD transport showed no sign of anything like that level of reflection when I measured it's digital output recently (I'd post a graph but I don't have a means of hosting it), on a BNC connection FWIW. In fact, looking at it there was no immediately visible sign of a reflection at all, whereas one big enough to confuse 1's and 0's would have to be huge.
 
Hang on, guys hang on, I said can be. :p Perhaps I'm overplaying the hand in the case of SPDIF.

Seriously though, yes in logical extremeis they could be bad enough to turn ones into zero's in theory. It's unlikely for cdps though sure, too low frquency and too low a power. It can be a serious problem in the higher frequency and power used for digial comms, and that is one reason their systems use more robust methods.

It depends on the way the receiver works. The receiver will work to a threshold, they won't be looking at the aiming values but somewhere near the top of a signal will be 1 for instance and somewhere near the bottom will be 0. I doubt very much that they use the above 0.25V is a 1 and below 0.25V is a zero, as the problem would surely be critical. More like 0.4V and above is 1 and 0.1V and below is zero with a deadband in between.

The normal problem will be confusion caused around the threshold point where the reflection alters the signal enough for the threshold point to apparently seem to move. This would potentially only be a mechanism to cause jitter though. External noise would have the same effect in the same way.

Of course if the DAC64 doesn't employ a PLL based on the clock frequency of the transport, would the noise / reflection that made it cross the 1 threshold twice look like two 1's because it discards the time base in between :D:D Now I'm just being silly (I hope) :p

It depends on whether the SPDIF stream has a return to zero in the stream, ie. does it need to perform a transition before another character in the stream is recognised? In which case it'll be more robust even than tightly defined voltage thresholds.

Cheers
 
Last edited by a moderator:
Bugger beat me to it :grrr: Timpy. I go out the office for 5 mins and he's at it again.
Electricaly quiet transport that arn't battery powered, it's possible to get very close to this Mike, with total reverse isolation of incoming A/c and other bits 'n' pieces, but then these through bring up they're own problems. But it works
Far as I can see, the buffer won't induce any jitter once in the ram as it just uses it's incoming signal to ref off. , however it's ref sync signal is the problem, I can see the logic behind the toslink, I've managed to blag one for saturday, so (I can open this one up as well, and we'll measure all the formats at the incoming points, reciever chip/<>Buffer<>after the fifo buffer to.)
Anther Fifi buffer user Teac D70/P70, that uses a PLL/Fifo (ten times that of a dac 64), plus a dual word bit clock link, thats total bib and braces. Wm
 
Last edited by a moderator:
Right I'll get back to 64 psu's, the off board one we constructed psu ultilising the TRIC, also used a special umbillical, that connected to a matching socket fitted into the wall of the camcase, now inside (where the smps used to reside were some 'quatum filters (foo foo dust for ranson)' these placed after the bridge,and do wonderous things for noise :) just by doing the psu first made quite a healthy increase in preformance and the trait that I must be only one to hear (although John Carter did smile when I mentioned this ;) ) hardness had gone.
The sound was more open dynamic and far less relentless, a more rewarding sound was left.
Now we had to drill into the camcase & do some internal relieving also, so it wasn't a 5 minute job by any means. This also makes a 2 box dac as well, and once modded unless your a skilled tig welder, a bit irreversable too
We have Further modded the dac, but I have to admit we haven't used or tried the Fibre optic.
Now the Fusion Tech player ?????, uses the dac 64 and valves anyone care to comment on the sound of the player. Wm
 
Last edited by a moderator:
Originally posted by MartinC
Really? Are you sure? Now electrical is almost certainly better than optical in this respect, but even my humble CD transport showed no sign of anything like that level of reflection when I measured it's digital output recently (I'd post a graph but I don't have a means of hosting it), on a BNC connection FWIW. In fact, looking at it there was no immediately visible sign of a reflection at all, whereas one big enough to confuse 1's and 0's would have to be huge.
Click to expand...

Here's the graph that Martin was talking about:
DigiCableGraph.gif


...and he wanted me to add this text to it:

The waveform was measured using a LeCroy 9310A digital oscilloscope, sampling at 100MHz. The BNC output of the Arcam Delta 250 CD Transport was terminated with a 75 Ohm load, in parallel with the 1M Ohm 'scope input impedence. The graph shows the first start of a 500us file, gathered whilst the transport was playing a CD (Massive Attack's "Blue Lines" IIRC). I have 25 or so example waveforms that I've measured but they all look very similar.
Click to expand...

Michael.
 
Edited to remove response to Paul's now edited post.

I took this measurement and others to look at the effect of different mains cables on the output, and from a quick look I could see no visual difference, nor from a quick and rough look at the spectra. I've been meaning to spend more time on this but haven't got round to it yet. You have though perhaps brought up a valid point that the sampling frequency I used might not have been good enough for detecting jitter.
 
Last edited by a moderator:
Not only but also

Remember you cannot measure jitter accurately using an oscilloscope that's self-referenced (triggered) to the signal you are trying to measure.

At some frequencies jitter will be invisible, at others it will be double the actual value.

I'm not sure how Martin is measuring his.

Andy.
 
Sorry Martin, completely bogus and pointless post...

In terms of reflections and actual data corruption I think it wouldn't work at all rather than just mess the odd bit up.

Paul
 
Re: Not only but also

Originally posted by Andrew L Weekes
Remember you cannot measure jitter accurately using an oscilloscope that's self-referenced (triggered) to the signal you are trying to measure.

At some frequencies jitter will be invisible, at others it will be double the actual value.

I'm not sure how Martin is measuring his.
Click to expand...

[Right, I'll just prove how little I know about this...]

You're right, I wasn't as such, but I'd be interested if you could explain a little more why I couldn't measure jitter this way? From my naive view point I thought this would manifest itself as very slight variations in the detected time widths of each bit, which in principle I'd have thought I could look at from the sort of data I have? That is, rather than getting a stream of 1s and 0s arriving at exactly equal time intervals, there is some spread of these time separations. This spread I'd assumed was probably roughly Gaussian (Normal). What, no doubt large chunks of info., am I missing?
 
Originally posted by wadia-miester
Martain, you could well be right about your sampling frequency rate.
Click to expand...

Yeah, thinking about it jitter values are usually of the order of a few hundreds of picoseconds aren't they (although I'm not sure how that time is defined), so potentially you'd need to be sampling at ~10GHz, which is a pretty expensive scope. That's a seriously wide bandwidth for digital cables to be dealing with too isn't it...
 
Martin, an external clock starting @10ghz would be a basic requirement, you would need at least 1000 times the capacity of the 'sample' being taken, from an indepenant clock.
Cables are a very small part of the business Martin, instrumentation & implemenation of digital intergration systems for areospace & industry & civil.
The Clocks which run the dsp chips in the Wadia's dac section start at 25 Meg.

www.omigatech.co.uk

Wm
 
Originally posted by wadia-miester
Cables are a very small part of the business Martin, instrumentation & implemenation of digital intergration systems for areospace & industry & civil.
Click to expand...

Sorry, I was merely thinking out loud that it was a wider bandwidth than I'd appreciated before...

As I said, I wasn't trying or claiming to measure jitter before, which is a good job frankly :D .
 

Latest posts

Back
Top