Digital that sounds like analog

DonH50

Member Sponsor & WBF Technical Expert
Jun 22, 2010
3,947
306
1,670
Monument, CO
The droop at Nyquist (1/2 the sampling frequency, fs) in the output of a normal DAC comes straight from sampling theory, sinX/X. It is about 3.54 dB at fs/2 IIRC. Oversampling reducing the impact of the rolloff by moving fs higher, no magic there. Analog peaking filters or digital correction filters before the DAC are often used to compensate. More exotic compensation schemes shorten the output pulse width, making a narrrower output pulse instead of simple steps, then post-filtering. I am not aware of any audio DACs that do that.

I am pretty sure I posted the curve in one of my early threads on sampling or DACs, not real sure...
 

jkeny

Industry Expert, Member Sponsor
Feb 9, 2012
3,374
42
383
Ireland
..... More exotic compensation schemes shorten the output pulse width, making a narrrower output pulse instead of simple steps, then post-filtering. I am not aware of any audio DACs that do that....
Can you point to examples of other than audio DACs using this method?
 

LL21

Well-Known Member
Dec 26, 2010
14,411
2,509
1,448
Good question - if the filter I'm describing is in place, then we'd need huge numbers of DACs to overcome its attenuation at 20kHz. Looking at the graph its -24dB there which is about a factor of 16. So I'd need 16X as many DACs for the HF boost than for the wanted signal. That many extra DACs would increase the glitching by the same factor which is what I set out to minimize :)

If the market really wanted response to 20kHz then it would be best to modify my filter to give that - it would either have to be twice as complex (say 14th order rather than 7th order) or we'd have to give up something in other areas. Now if I had the software to calculate the DAC-filter coefficients then I could give up some passband flatness in the passive filter and compensate for that in the DAC-filter. Which would be really cool, but I doubt it would gain us very much in terms of reducing the components in the passive filter, maybe lose us at most 2 inductors? It might turn out that my filter is unnecessarily good in its stop-band rejection (in simulation about -53dB) and we could give up something there without loss of SQ. I don't though have a filter design program at the moment to experiment with different filter parameters. I hope to find one in the future though!

wow...the complexity is nearly geometric progression! Thanks for the answer.
 

Kindhornman

New Member
Dec 3, 2012
21
0
0
Los Angeles, California, USA
Richard,
Being a complete dweeb at all of this a couple of questions. I am following you here and was wondering while you are talking about 16bit DAC chips how are the higher sampling rate CDP's and other devices reaching 24bit. Is this a series of 16bit DAC's working in parallel or in series or are there in fact 24bit chips that you can use for your design? Going by your conversation about the Nyquist relationship and the 1/2 value that seems to be what is driving the selection of your analogue filters lower 17khz selection. If you were using a higher frequency such as 96khx instead of the 48Khz that I think you are referring to wouldn't that solve the upper frequency cutoff situation? I think that you have been basing all of your design based on the original Redbook 16/48Khz standard correct? Why lock yourself into that standard? Why not a higher bit rate and clock speed?
 

opus111

Banned
Feb 10, 2012
1,286
3
0
Hangzhou, China
Being a complete dweeb at all of this a couple of questions.

Keep them coming :) Have all your questions dried up now Lloyd? :)

I am following you here and was wondering while you are talking about 16bit DAC chips how are the higher sampling rate CDP's and other devices reaching 24bit.

I could write at length on this subject because there's a fair degree of sleight of hand going on with the manufacturers of the chips on this point. For me a 24bit DAC doesn't just accept 24bits of data, rather it resolves to 24bits, meaning that each output code can be individually recognised. No DAC chip that I've come across is very close to resolving 24bits because they're too noisy. Yet you do even get some manufacturers claiming '32 bits' for their chips. I call BS :D

24bits would mean a noise floor out of the DAC around -144dB to get the resolution of all the codes. The ESS Sabre comes closest I think around -133dB but this is just off the top of my head, I've not read the datasheet in a long time.

So to answer your question very briefly, no DAC reaches 24bits, but they do nowadays do a fair bit better than 16. They mostly do this with S-D modulators which shift the quantization noise of their low-bit DAC up to higher frequencies with noise shaping. So the high resolution does totally depend on filtering the output sufficiently to get a decent noise level in the audio band. If you measured the output in the full bandwidth you'd just see the resolution of the DAC part of the chip itself which nowadays is typically 6 bits.

Is this a series of 16bit DAC's working in parallel or in series or are there in fact 24bit chips that you can use for your design?

Given that the vast majority of audio products no longer use multibit DACs there's no advantage in paralleling DACs nowadays. That's because the noise between them isn't random, it would be correlated so no reduction when they're paralleled. That hasn't stopped at least one manufacturer I've heard of doing it though (can't recall who it was though, unmemorable over-priced product from what I vaguely remember):confused:

Going by your conversation about the Nyquist relationship and the 1/2 value that seems to be what is driving the selection of your analogue filters lower 17khz selection. If you were using a higher frequency such as 96khx instead of the 48Khz that I think you are referring to wouldn't that solve the upper frequency cutoff situation?

I'm not sure here if you're talking about changing the sample rate we record at, or whether you're suggestion upsampling the 44k1 data we get off the CD? I'm strictly designing this DAC for the existing music out there, so the first is ruled out. The second brings some of its own issues in relation to SQ. Those are in my estimation good enough reason to stick with 44k1 for now until I exhaust all the potential design space...

I think that you have been basing all of your design based on the original Redbook 16/48Khz standard correct? Why lock yourself into that standard? Why not a higher bit rate and clock speed?

I recall a while ago reading that Willie Sutton was once asked why he robbed banks and the answer came 'Because that's where the money is'. I have a similar answer to his to give you - 'Because that's where the music is'. :p
 

opus111

Banned
Feb 10, 2012
1,286
3
0
Hangzhou, China
Wazzup guys?:)

Here's another disc recommendation - for the soundtrack to a movie, Don Giovanni. In the past I've never been a fan of opera, never had any on LP before the age of CD, but bought a few boxed sets when I could get staff discount on Sony CDs. I never really played the discs, with the exception of La Traviata which I have on Decca. With my new filter-equipped DAC, I thought I'd give this one a whirl after all those intervening years. I'm beginning to work out now why I never much enjoyed my opera CDs - on this DAC this one is amazing in terms of the diction of the singers. Its really clear to me now why opera is sung in Italian - its the emotional content in all those fricative consonants, but all the digital systems I've owned before now turned those fine details to mush, sibilance. Only now have I fallen in love with opera :cool:

http://www.amazon.com/Mozart-Giovanni-Raimondi-Kanawa-Maazel/dp/B000GCFAHK
 

DonH50

Member Sponsor & WBF Technical Expert
Jun 22, 2010
3,947
306
1,670
Monument, CO
Can you point to examples of other than audio DACs using this method?

I have seen it used in RF DACs, particularly those using the 2nd or higher Nyquist bands. My work was mostly military so I am not sure I can point to commercial examples. Agilent used that approach in a signal generator at one time, IIRC, but I don't recall the model number off-hand. There is nothing particularly magical about it, but sense it descreases the energy SNR tends to be reduced, and depending on the implementation the output stage/switch can be a bear to design... I explored an approach using tunnel diodes but did not build it (could get the speed but not the linearity I wanted).
 

opus111

Banned
Feb 10, 2012
1,286
3
0
Hangzhou, China
I'm not clear on your question here. If you're designing a speaker with bandwidth out to 24kHz then I can't think why you'd be using that with redbook CD because that only goes up to 20kHz. Is your speaker designed to work with hires - (it would need to be 88k2 or above)?
 

opus111

Banned
Feb 10, 2012
1,286
3
0
Hangzhou, China
I'm still rather unclear as to what your question means here. If a speaker goes above 20kHz, wouldn't only dogs and bats appreciate it? Or is there some advantage within the audio band that's gained by going beyond 20kHz? Or am I out of date with modern research into what we're really able to hear?

Higher bit-rate systems are really higher bit-rate in that they have greater bandwidth. There are good technical reasons for going up to 88k2 - to allow less phase distortion primarily. It means the input (AAF) and output (AIF) don't need to be steep like I've been describing here, with bags of group delay close to the corner frequency. Going to 88k2 though does bring its own disadvantages in terms of DAC performance which few designers (other than Dan Lavry) seem to be aware of.
 

Kindhornman

New Member
Dec 3, 2012
21
0
0
Los Angeles, California, USA
Just as in a good power amp your bandwidth is much greater than 20 to 20 the same can be said for a speaker. The reason I want to get up to say 24Khz is that is where I want the first resonance to occur if I can do that. I want any resonance in the upper band out of band for your hearing. So I am looking at the mechanical system.
 

opus111

Banned
Feb 10, 2012
1,286
3
0
Hangzhou, China
I'm not so sure that a good amp has wider bandwidth, based on what I'm discovering with this filter I'd postulate the opposite. I do agree though with a speaker that the tweeter resonance is a major issue and moving it higher reduces the audible effects lower in the frequency band. An amp isn't a mechanical system though with resonances so these are definitely two different cases.
 

opus111

Banned
Feb 10, 2012
1,286
3
0
Hangzhou, China
I'm not quite at the proposing stage yet but based on this experiment in listening to the filter, I'm inclined to suggest that fairly aggressive passive LP filtering at the input to a power amp is a good thing, SQ-wise. But it may be impractical to implement because of the impedance conventions of amplifiers, which is that we drive a high impedance from a low. I think the active electronics part of the amplifier should be wider band than audio but I'm not sure by what factor. I suspect that wideband amps have been preferred SQ-wise because they tend to be more resistant to OOB hash on their inputs, not directly because of their wide bandwidth.
 

opus111

Banned
Feb 10, 2012
1,286
3
0
Hangzhou, China
Any time you have an ADC then theory requires you have severe band-limiting anyway. My hypothesis at the moment is that few people's band-limiting is severe enough in practice for best SQ and not just in cases just before the ADC. Anywhere that noise can get in - which specifically includes cables - we should strongly band-limit. Or just build it all in one box.
 

Kindhornman

New Member
Dec 3, 2012
21
0
0
Los Angeles, California, USA
I guess one of the things that I think is that the bandwidth should be wider than our so called hearing and not the other way around. I want to get everything out of band before it starts to roll off. I think that is part of the magic that is often missing is up high whether we think we can hear it or not. It is some of the harmonic structure that you are removing by doing that. Just my opinion, no proof here, just keep walking, nothing to see here..........
 

About us

  • What’s Best Forum is THE forum for high end audio, product reviews, advice and sharing experiences on the best of everything else. This is THE place where audiophiles and audio companies discuss vintage, contemporary and new audio products, music servers, music streamers, computer audio, digital-to-analog converters, turntables, phono stages, cartridges, reel-to-reel tape machines, speakers, headphones and tube and solid-state amplification. Founded in 2010 What’s Best Forum invites intelligent and courteous people of all interests and backgrounds to describe and discuss the best of everything. From beginners to life-long hobbyists to industry professionals, we enjoy learning about new things and meeting new people, and participating in spirited debates.

Quick Navigation

User Menu

Steve Williams
Site Founder | Site Owner | Administrator
Ron Resnick
Site Co-Owner | Administrator
Julian (The Fixer)
Website Build | Marketing Managersing