[jumpingjackflash5]

You can write your reaction under his blog. I think he is open to comments/discussion.

 

[Jawed]

He's softened his recommendation, but I still can't see why he would recommend something that's wrong (deliberate aliasing) under any circumstances. Perhaps his reasoning is that when you think the recording is good use the linear phase filter. That doesn't explain why making "bad" recordings worse is productive. It just seems like a superstitious belief.

Maybe he's compensating for a DAC that just has a bad chip and so prefers a setting that lessens a particular combination of ills?

 

[jumpingjackflash5]

I don't wanna act as his advocate here but I think he just wanted to explore the options available, since many audiophiles prefer minimum phase and slow-roll of filter with big amount of aliasing (to make it very long and do not pre-ring much under any signals). So he tried to find acceptable compromise with aliasing that cannot be heard by itself and the slightest phase shift possible and showed them.
I think that the knows that linear phase is the standard of digital audio and with proper recordings it works very well.
And I discussed under his blog that I think in the light of knowledge about steep filtering that it is not necessary to allow aliasing in majority of scenarios.

 

[jayz]

Another question for Rob. ...

Some high end DACs feature femto clocks these days, a few even go as far as simplified atomic clocks. Obviously the more precise and stable the tick I would imagine matters to the digital side but my question is, considering the cost involved, is it your opinion that better logic on the digital side gives more return on investment than the purist approach of going to any length to create ultra stable clocking, dcS comes to mind with their super expensive master clock. Not trying to bash any manufacturer but just like to get some perspective of the marketing mess out there.

 

[Rob Watts]

External clocks are a disaster.

Yes I really mean that!

Why is 10 MHz or 10.00001 MHz at all important? it makes no difference at all, but having an external clock creates huge problems.

The clock needs to be transmitted (jitter and noise +++) then 10.000MHz is useless as an actual clock for the DAC operation - so a analogue PLL is used locally to create the actual clock the DAC needs. And a PLL can give very low cycle to cycle jitter - but they give horrendous levels of low frequency jitter - and low frequency jitter is much more important subjectively than cycle to cycle jitter (particularly for jitter immune pulse array DAC's). Low frequency jitter causes skirts on fundamentals, and these skirts sound exactly like noise floor modulation - so it sounds brighter. So when connects an external clock, a local PLL is engaged, low frequency jitter increases, and it sounds brighter - and people are conned into thinking that the brightness is due to more transparency, which of course it is not.

The only way of reducing low frequency jitter is to use a crystal oscillator that has been specified for low frequency performance - and keep it very close to the active switching elements, preferably with only one logic level (buffer) in the way.

And that is precisely what I do with all my DAC's.....

 

[Dobrescu George]

External clocks are a disaster.

Yes I really mean that!

Why is 10 MHz or 10.00001 MHz at all important? it makes no difference at all, but having an external clock creates huge problems.

The clock needs to be transmitted (jitter and noise +++) then 10.000MHz is useless as an actual clock for the DAC operation - so a analogue PLL is used locally to create the actual clock the DAC needs. And a PLL can give very low cycle to cycle jitter - but they give horrendous levels of low frequency jitter - and low frequency jitter is much more important subjectively than cycle to cycle jitter (particularly for jitter immune pulse array DAC's). Low frequency jitter causes skirts on fundamentals, and these skirts sound exactly like noise floor modulation - so it sounds brighter. So when connects an external clock, a local PLL is engaged, low frequency jitter increases, and it sounds brighter - and people are conned into thinking that the brightness is due to more transparency, which of course it is not.

The only way of reducing low frequency jitter is to use a crystal oscillator that has been specified for low frequency performance - and keep it very close to the active switching elements, preferably with only one logic level (buffer) in the way.

And that is precisely what I do with all my DAC's.....

What happens if I like a brighter signature?

 

[Mython]

What happens if I like a brighter signature?

Devil's advocacy aside, then there are far cheaper ways of increasing perceived brightness, via increasing noise floor modulation...


"To each, his/her own!"

 

[Dobrescu George]

Devil's advocacy aside, then there are far cheaper ways of increasing perceived brightness, via increasing noise floor modulation...


"To each, his/her own!"

I am a bit serious, this might seem unpleasing for a question, but if @Rob Watts latest works all erase brightness and make music smoother, guitar picks will eventually sound smoothed out and unnatural, will lose that metal tinge that makes me listen to acoustic guitars with metallic strings in the first place, so it will make the device less transparent and less realistic, no?

Acoustic guitars with metal strings need to bite.

Same for cymbals, if a cymbal crash is not bright and not energetic, and if it doesn't have bite, then it is less transparent and sounds non-real compared to a real live cymbal crash, which can and should be deafening at close proximity.

 

[Mython]

I am a bit serious, this might seem unpleasing for a question, but if @Rob Watts latest works all erase brightness and make music smoother, guitar picks will eventually sound smoothed out and unnatural, will lose that metal tinge that makes me listen to acoustic guitars with metallic strings in the first place, so it will make the device less transparent and less realistic, no?

Acoustic guitars with metal strings need to bite.

Same for cymbals, if a cymbal crash is not bright and not energetic, and if it doesn't have bite, then it is less transparent and sounds non-real compared to a real live cymbal crash, which can and should be deafening at close proximity.

No, your question isn't displeasing, Most of us, here, are quite happy to exchange different ideas and opinions without taking them personally.

Your question seems a little illogical to me, rather than displeasing, but I respect your point of view, all the same.

Without intending to speak 'for' Rob, I believe he was referring to false brightness.

Why do you feel that reducing or eliminiating false brightness will also reduce realism or transparency?

Or is it that you prefer a less-realistic portrayal of acoustic music, in order to provide a greater subjective impression of 'pseudo-transparency'?

 

[Dobrescu George]

No, your question isn't displeasing, Most of us, here, are quite happy to exchange different ideas and opinions without taking them personally.

Your question seems a little illogical to me, rather than displeasing, but I respect your point of view, all the same.

Without intending to speak 'for' Rob, I believe he was referring to false brightness.

Why do you feel that reducing or eliminiating false brightness will also reduce realism or transparency?

Or is it that you prefer a less-realistic portrayal of acoustic music, in order to provide a greater subjective impression of 'pseudo-transparency'?

Good question

Sometimes, I wonder, I am used to live instruments and music, but the more I analyse some of the recordings I'm listening to, the more I realise that some of the mastering was actually done considerably smoother than the thing sounded live

I'm talking only about rock / metal music, with large dynamic compression and lots of guitars and things, lots of it sounds different when recorded. I fear that with metal music, often the masters erase some of the brightness that should be there, leading to music that sounds falsely smooth. Makes me wonder if revealing that recording alone isn't enough and I actually end up needing more and more EQ simply as a result of their fault

Which kinda makes me wonder what is the most recommended way to add some brightness without introducing noise. As in, what is the cleanest EQ supported by a software, and such?

 

[Rob Watts]

I am a bit serious, this might seem unpleasing for a question, but if @Rob Watts latest works all erase brightness and make music smoother, guitar picks will eventually sound smoothed out and unnatural, will lose that metal tinge that makes me listen to acoustic guitars with metallic strings in the first place, so it will make the device less transparent and less realistic, no?

Acoustic guitars with metal strings need to bite.

Same for cymbals, if a cymbal crash is not bright and not energetic, and if it doesn't have bite, then it is less transparent and sounds non-real compared to a real live cymbal crash, which can and should be deafening at close proximity.

Actually this is a serious issue, and it is something I spend a lot of time over - the yin-yang balance. So let's say you reduce noise floor modulation, and eliminate the skirt issue(actually another form of noise floor modulation - there are actually several), then you can end up with a sound that is way too soft and warm. And for sure it is much more accurate; but if it means that you end up with a presentation that is too dark and warm, then it could end up being overall unsatisfactory. My standard answer to this is to make the system brighter - such as speaker location, etc. But with product design, there is another way of improving the balance - by improving the innate transparency. And there are a number of ways of doing this.

I recall with the Dave development, at one point it was sounding very very dark and warm - much warmer than most people's impression of neutral. Then we have the issue of what actually neutral really is - and it is simply an average of bright or dark products. But nobody has heard a perfect DAC - as that does not exist - so what is truly neutral? Now when you are designing state of the art, and you are looking for the DAC that neither adds or subtracts to the sound, then I was happy to accept that Dave being too dark was the correct approach. But then I discovered another aspect, and this was definitely more accurate - it was the 256FS WTA filter - and suddenly the sound shifted from too dark to something more average (I won't use the term neutral).

Given that Dave is now pretty much accepted as state of the art - the most natural and transparent DAC available now - and it's certainly my reference (rather BluDave!) - then that makes it easier when designing more inexpensive products - in that the overall tonal balance should not drift too much from the BluDave reference. And that means that at every price point I have to ensure that the degradations in refinement and transparency is evenly matched - and I actually do do this at the design stage. I have to take cost out of a particular design, and so make sure that cost savings impact both refinement and transparency fairly evenly. The exception to this was Mojo, where the fact that it will be used with brighter and cheaper HP implies a warmer balance. Indeed, the SQ of Mojo was set at the early design stage and I was pleased that the balance I intended was there without needing to adjust the sound - the prototypes sounded exactly as I expected.

So yes ensuring the right yin-yang balance is a crucial design aspect.

 

[Mython]

Smearing of transients might also influence some of the subjective aspects of music reproduction you mentioned, Dobrescu George.

Let's not forget - it's not all about the playback equipment - the ADC is a major, major, major bottleneck in terms of realism of the reproduced music in one's home...

 

[514077]

Good question

Sometimes, I wonder, I am used to live instruments and music, but the more I analyse some of the recordings I'm listening to, the more I realise that some of the mastering was actually done considerably smoother than the thing sounded live

I'm talking only about rock / metal music, with large dynamic compression and lots of guitars and things, lots of it sounds different when recorded. I fear that with metal music, often the masters erase some of the brightness that should be there, leading to music that sounds falsely smooth. Makes me wonder if revealing that recording alone isn't enough and I actually end up needing more and more EQ simply as a result of their fault

Which kinda makes me wonder what is the most recommended way to add some brightness without introducing noise. As in, what is the cleanest EQ supported by a software, and such?

I've played in quite a few bands, and tuned pianos in professional consert locations. I've been there when the musicians are trying to get sounds. One time, the drummer was playing his symbols, which had a particular angelic shine to them which I can't describe. Can't say I've ever heard that sound successfully transfered to a recording. It wasn't sizzling with piercing high frequency. Therefore, I've come to conclude that natural high-end content doesn't mean super-sizzling high-hats etc.

 

[Mython]

I've played in quite a few bands, and tuned pianos in professional consert locations. I've been there when the musicians are trying to get sounds. One time, the drummer was playing his symbols, which had a particular angelic shine to them which I can't describe. Can't say I've ever heard that sound successfully transfered to a recording. It wasn't sizzling with piercing high frequency. Therefore, I've come to conclude that natural high-end content doesn't mean super-sizzling high-hats etc.

I pity drive-unit designers - be it Balanced Armatures, micro Dynamic Drivers, or fullsize loudspeaker drive units, of various kinds.

These devices are expected not only to mimic real instruments, but often using MUCH smaller diaphragmatic area than real instruments utilise to achieve their sound.

I mean, just as one example, consider a loudspeaker tweeter - a 1 inch soft-dome, for example. That thing is expected to reproduce the sound of a cymbal that may be 15inches in diameter, and at the same SPL!!

Poor Rob goes to so much effort to reconstruct an analogue signal with breathtaking, eye-watering, accuracy. and yet has to suffer the indiginity of that signal being transduced by extremely compromised transducers!

 

[Dobrescu George]

Actually this is a serious issue, and it is something I spend a lot of time over - the yin-yang balance. So let's say you reduce noise floor modulation, and eliminate the skirt issue(actually another form of noise floor modulation - there are actually several), then you can end up with a sound that is way too soft and warm. And for sure it is much more accurate; but if it means that you end up with a presentation that is too dark and warm, then it could end up being overall unsatisfactory. My standard answer to this is to make the system brighter - such as speaker location, etc. But with product design, there is another way of improving the balance - by improving the innate transparency. And there are a number of ways of doing this.

I recall with the Dave development, at one point it was sounding very very dark and warm - much warmer than most people's impression of neutral. Then we have the issue of what actually neutral really is - and it is simply an average of bright or dark products. But nobody has heard a perfect DAC - as that does not exist - so what is truly neutral? Now when you are designing state of the art, and you are looking for the DAC that neither adds or subtracts to the sound, then I was happy to accept that Dave being too dark was the correct approach. But then I discovered another aspect, and this was definitely more accurate - it was the 256FS WTA filter - and suddenly the sound shifted from too dark to something more average (I won't use the term neutral).

Given that Dave is now pretty much accepted as state of the art - the most natural and transparent DAC available now - and it's certainly my reference (rather BluDave!) - then that makes it easier when designing more inexpensive products - in that the overall tonal balance should not drift too much from the BluDave reference. And that means that at every price point I have to ensure that the degradations in refinement and transparency is evenly matched - and I actually do do this at the design stage. I have to take cost out of a particular design, and so make sure that cost savings impact both refinement and transparency fairly evenly. The exception to this was Mojo, where the fact that it will be used with brighter and cheaper HP implies a warmer balance. Indeed, the SQ of Mojo was set at the early design stage and I was pleased that the balance I intended was there without needing to adjust the sound - the prototypes sounded exactly as I expected.

So yes ensuring the right yin-yang balance is a crucial design aspect.

Thank you for explainig.

I've played in quite a few bands, and tuned pianos in professional consert locations. I've been there when the musicians are trying to get sounds. One time, the drummer was playing his symbols, which had a particular angelic shine to them which I can't describe. Can't say I've ever heard that sound successfully transfered to a recording. It wasn't sizzling with piercing high frequency. Therefore, I've come to conclude that natural high-end content doesn't mean super-sizzling high-hats etc.

Yes, there's smothing about the timing of that, a certain shimmer of sorts that needs to happen with real life instruments, that I've never really heard in a recording. Really hard to put in words, not necessarily bright, as much as present, efortlessly shimmering, without being toned down.

I pity drive-unit designers - be it Balanced Armatures, micro Dynamic Drivers, or fullsize loudspeaker drive units, of various kinds.

These devices are expected not only to mimic real instruments, but often using MUCH smaller diaphragmatic area than real instruments utilise to achieve their sound.

I mean, just as one example, consider a loudspeaker tweeter - a 1 inch soft-dome, for example. That thing is expected to reproduce the sound of a cymbal that may be 15inches in diameter, and at the same SPL!!

Poor Rob goes to so much effort to reconstruct an analogue signal, and yet has to suffer the indiginity of that signal being transduced by extremely compromised transducers!

I agree with this, it must be a real pain to the driver designers to make something so complicated, a true engineering miracle if you look at things this way.

Smearing of transients might also influence some of the subjective aspects of music reproduction you mentioned, Dobrescu George.

Let's not forget - it's not all about the playback equipment - the ADC is a major, major, major bottleneck in terms of realism of the reproduced music in one's home...

This actually makes me wonder, a lot of the bands I really listen to haven't had the chance to be recorded in super expensive studios, so most of them used less than adequate ADC stages, which might further accentuate the issue that makes me yearn for more treble sometimes, and such.