Hi,
The nature of these problems is that you typically cannot get a steady performance for several seconds. If you measure for several seconds, you average most single or random problems out.
Oh, it is not that straightforward, nor is actually much in audio, but there are many who prefer to ignore such inconvenient truths.
Ciao T
The nature of these problems is that you typically cannot get a steady performance for several seconds. If you measure for several seconds, you average most single or random problems out.
Oh, it is not that straightforward, nor is actually much in audio, but there are many who prefer to ignore such inconvenient truths.
Ciao T
Ah,
That's quite allright then. I need to fid the time to connect mine to the I2S on my Async USB Gizmo, instead of using SPDIF...
Ciao T
That's quite allright then. I need to fid the time to connect mine to the I2S on my Async USB Gizmo, instead of using SPDIF...
Ciao T
In my experience this "bite" on transients is the ultra-realism that I put down to jitter as every time I have reduced what I believe is jitter, the sound has become more silky & analogue-like without losing any detail. I likened it in the past, to MSG i.e an exaggeration of what's their but ultimately fatiguing.
I guess the question of whether it's on the recording could only be answered by those that are involved in recording & comparing the mic feed with the recorded sound as per this thread http://www.diyaudio.com/forums/everything-else/191488-source-problem-souless-sound.html
I guess the question of whether it's on the recording could only be answered by those that are involved in recording & comparing the mic feed with the recorded sound as per this thread http://www.diyaudio.com/forums/everything-else/191488-source-problem-souless-sound.html
Ah the 'bite' I'm talking about isn't that kind of bite, as its not in any way fatiguing over the longer term. Its not really only on transients either. The 'bite' you're talking about - the fatiguing kind - I spent time getting rid of on my AD1955 over many months.
I do have a recording which has this fatiguing kind of 'bite' recorded on it, I play it sometimes to demonstrate what an average system sounds like. 😀 So my system can certainly deliver your kind of 'bite' when its on the recording.
I do have a recording which has this fatiguing kind of 'bite' recorded on it, I play it sometimes to demonstrate what an average system sounds like. 😀 So my system can certainly deliver your kind of 'bite' when its on the recording.
Well, to me it's fatiguing but I don't think it is to most people - I've even heard people complain about this lack of bite when it's gone & actually preferring it at times. A bit like Pano said that the ESS DAC was a bit too smooth sounding for some - a reduction of jitter?
Yeah, the fatiguing kind of 'bite' I associate with problems in the I/V stage of DACs. My explanation for Jocko's story of when he used CFB (which mainly don't have it) amps for I/V was 'customers didn't like the sound' is exactly what you're saying here. Some people like their music supplied with MSG all the time 😛
But a particular mechanism in use might mean, for example, that behaviour is normally perfect but you get one glitch every minute, or that the playback pitch slowly oscillates/switches between +/-0.1% error. It might not even be audible and it wouldn't necessarily show up in certain FFT measurements, but a purist would not like to know it was there - just as a purist would object to inaudible but uncorrectable errors on a CD.
What this thread tells me is that an outboard DAC is not just a DAC, but an adaptive system which may, or may not, be stable with all sources, and that its performance may vary over seconds, minutes or even hours. It may be affected by how close the source's sample rate is to some nominal value. 10ppm error is fine maybe, but 200ppm causes frequent glitches. Two people arguing over the same DAC's performance may be talking about, effectively, two different systems just depending on the fine trim on quartz oscillators in their sources.
Just a note: Don't limit your testing to 1KHz. A lot of the nasties start showing up north of there. 5KHz is a good test point for this. Lower freqs, also.
Of course! It's what they're used to. Yum! Burnt pancakes, just like mama use to make. 😉
Hi,
Yes, the problem is that if use several complex devices interconnected in a system the interactions are complex and as often as they are gross, they are subtle. The gross ones are easily spotted and eradicated by competent engineers (or maybe not, judging by some other threads). The more or less subtle ones are arguably what the "High End" is all about, or at least should be...
Ciao T
Yes, the problem is that if use several complex devices interconnected in a system the interactions are complex and as often as they are gross, they are subtle. The gross ones are easily spotted and eradicated by competent engineers (or maybe not, judging by some other threads). The more or less subtle ones are arguably what the "High End" is all about, or at least should be...
Ciao T
Nor steady-state sines into multi-averaged FFTs. Stuff tends to get missed that way.
I'm always very sceptical of my own ability to hear the difference between sources, believing much more strongly in the 'sighted bias' idea.
However, about 12 years ago, back in the days when I had money to spend, I bought a Marantz CD player - a CD6000 I think, not sure what variant. It certainly had the looks I liked, but when I tried it I was convinced that it was too 'smooth' and lacking the bite I was used to, 'pulling its punches'. I figured that I was never going to get over the mental block that was convincing me of this, so I took it back and got an Arcam Alpha 7SE which seemed far more punchy.
Would the differences between the internals of these two CD players lend any credence to my one and only 'audiophile moment'?
http://wikhifi.wikispaces.com/Marantz+CD6000
I think the Alpha 7SE uses a PCM1716 DAC.
Yeah that was my recollection from a conversation with the man a few years ago, so it may be hazy !
Apparently they've found a trick to remove this limitation. If someone finds the gory details, I'd like to read it.
If the "A" is "Asynchronous", not necessarily, for instance we can resample 44.1k to 48k (which is a non-integer multiple) say in software on a PC. In this case both clocks (well, virtual clocks) are not equal but it could be argued that they are synchronous since there is zero drift between them. It is just like one clock was derived from another using a PLL or something. You always know that on average for every 441 input samples you get 480 output samples. Not the case if the clocks come from two different non-locked oscillators, then when some drift occurs, 441 input samples will correspond to 480.0001 output samples for instance. That would really be asynchronous since the drifts change as time passes.
Yeah
If you oversample a signal to a very high frequency, you get a high sample rate signal, but with a very small part of its bandwidth which is used. The slew rate of this signal will be pretty small.
I've done a little check : I took a sinc filter which oversamples the input 907 times (like 44.1k -> 40 MHz).
Let's suppose we use linear interpolation to get the value at the output of this filter at some instant t. I've compared the value of the real sinc(t) function, to a linear interpolation between the two closest points of the 40 MHz sampled sinc function.
Maximum error corresponds to 8 LSBs of a 24 bit word (or 0.03 LSB of a 16-bit word). That looks like it's well buried in the noise floor...
At 10 MHz output the error is much higher, about 0.5 LSB of a 16 bit word.
Approximately, the error is proportional to the inverse of the square of the output sampling frequency so the faster the better !
Anyway, if input and output frequencies are almost identical but not quite (say they come from different crystals) you'll get a small frequency difference between the two. Since the ASRC interpolates between different sets of coefficients, probably using ROM and hardware with as few bits of precision they could get away with to save some silicon, that would probably be a good recipe for rounding errors showing up. How many "steps" are there in this linear interpolation, and what happens when those steps are stepped pretty slowly ?... That's why some recommend to use an output sample rate that's totally unrelated to the input one, at least it wipes the crap under the rug...
You bet, I've reread this and still can't figure it out.
Hm, aren't you breaking some NDA ? lol. I'd like to know why you say "limited".
Indeed, I wonder what the true value is...
Which chips would you recomment ?
The Sabre is really expensive.
However using several DAC chips opens a can of worms for clock distribution.
Sorry, Coppertop, - can't answer - it all depends on implementation & empirical investigation.
PeuFeu & Thorsten - it's my understanding that the NDA has been lifted - was this not reported on the TP thread some time ago?
Anyway, the Sabre DAC jitter reduction is not as perfect as claimed as feeding it a synchronous MCLK sounds far better!
PeuFeu & Thorsten - it's my understanding that the NDA has been lifted - was this not reported on the TP thread some time ago?
Anyway, the Sabre DAC jitter reduction is not as perfect as claimed as feeding it a synchronous MCLK sounds far better!
@Peufeu
If I understand you correctly, you still have to effectively slide the oversampled waveform past the fixed output clock and linear interpolator at the right speed, otherwise you run out of data eventually, or are constantly reading out too slowly and slipping against the incoming data. Setting this rate would pose just the same problem as the conventional PLL-based systems wouldn't it? The same sort of levels of jitter suppression, for the same size input buffer? No 'paradigm shift' in other words..?
If I understand you correctly, you still have to effectively slide the oversampled waveform past the fixed output clock and linear interpolator at the right speed, otherwise you run out of data eventually, or are constantly reading out too slowly and slipping against the incoming data. Setting this rate would pose just the same problem as the conventional PLL-based systems wouldn't it? The same sort of levels of jitter suppression, for the same size input buffer? No 'paradigm shift' in other words..?
Precisely. Jitter sidebands and peak spreading, as well as distortion. Not inconsiderable information.
Hi,
At some point my CDP was a heavily modded Philips CD-720 (new clock and supply for said, new supply for DAC and analogue stage, analogue stage direct coupled and with CFB Op-Amp correctly applied). I also had an old silver Jap-Amp also heavily modded (forgot which brand/model, probably a JVC Super AA) and wanted something better looking and more "Audiophile".
So I did something a little stupid. I bought unheard, unseen, from my local Dixons on special order a Marantz CD-67SE and Marantz PM-66SE, based on the rave reviews they where getting.
The fact that these where MUCH WORSE than what I had led to massive amounts of upgrading, cumulating in batter supplies for both Amplifier and CD-Player analogue stages.
In my analysis, the DAC in the CD-67 (and also a rang eof others including CD-63/6000) was a first generation NPC Delta Sigma type. This had massive ultrasonic output, enough for it to be more than the low grade analogue Op-Amp's could handle. So it was "tuned" to sound soft and rolled off, yet it still had that annoying hissy sibiliance that I hate.
The upgraded units, even when running on internal AC power, had a level of subjective sound quality that I dragged them around London dealers for quite some time, trying to find something better. Imagine how the dealers faces looked when this apparently cheapo combo beat up major gear of the time in ways that where not pretty.
So there may be something to it.
In my view the PCM-1716 is a plainly awful DAC taken in it's own right, but it has enough analogue stage stuff on board that no ham-fisted designer can mess it up. So this may have helped.
Ciao T
At some point my CDP was a heavily modded Philips CD-720 (new clock and supply for said, new supply for DAC and analogue stage, analogue stage direct coupled and with CFB Op-Amp correctly applied). I also had an old silver Jap-Amp also heavily modded (forgot which brand/model, probably a JVC Super AA) and wanted something better looking and more "Audiophile".
So I did something a little stupid. I bought unheard, unseen, from my local Dixons on special order a Marantz CD-67SE and Marantz PM-66SE, based on the rave reviews they where getting.
The fact that these where MUCH WORSE than what I had led to massive amounts of upgrading, cumulating in batter supplies for both Amplifier and CD-Player analogue stages.
In my analysis, the DAC in the CD-67 (and also a rang eof others including CD-63/6000) was a first generation NPC Delta Sigma type. This had massive ultrasonic output, enough for it to be more than the low grade analogue Op-Amp's could handle. So it was "tuned" to sound soft and rolled off, yet it still had that annoying hissy sibiliance that I hate.
The upgraded units, even when running on internal AC power, had a level of subjective sound quality that I dragged them around London dealers for quite some time, trying to find something better. Imagine how the dealers faces looked when this apparently cheapo combo beat up major gear of the time in ways that where not pretty.
So there may be something to it.
In my view the PCM-1716 is a plainly awful DAC taken in it's own right, but it has enough analogue stage stuff on board that no ham-fisted designer can mess it up. So this may have helped.
Ciao T
- Status
- Not open for further replies.