Are all D/A conversions equal?

[Nyal Mellor]

Recently I have been wondering if the first D/A conversion e.g. from an optical disc or hard drive to digital is different in some ways than a real time A/D and then D/A conversion, for example that used in an inline DSP for speaker or room correction.

I have heard quite a few speakers (in particular the Lotus Granada) that have had inline pro level DSP that sounded spectacular. Yet the source D/A used for the first conversion and the quality of it was readily audible.

What's going on here?

 

[opus111]

I'd like to be able to suggest some answers but I can't understand your question. The conversion from analog to digital is just a comparator inside a disc drive. Also a DSP unit working from a digital source (a computer or CD player) doesn't use an A/D, the signal remains digital until the final D/A conversion.

 

[Nyal Mellor]

I'd like to be able to suggest some answers but I can't understand your question. The conversion from analog to digital is just a comparator inside a disc drive. Also a DSP unit working from a digital source (a computer or CD player) doesn't use an A/D, the signal remains digital until the final D/A conversion.

Ha ha! That's what you get for trying to formulate a question post lunch!

If you have a digital source, let's say a CD player, feeding a DAC. That goes into a pre-amplifier and then after the pre-amp the signal goes into a digital signal processor. The digital signal processor takes the analog signal, converts to digital, does it's stuff (EQ, crossover, whatever) then converts back to analog. The signal then goes into the power amps and from there the speakers.

My question is whether the initial digital to analog conversion is in any way different to the following analog to digital and digital to analog conversion and whether there are different factors at play in terms of what engineering affects sound quality. Is a real-time A/D, D/A conversion different than a D/A conversion on a CD?

Does that make any sense?

 

[opus111]

Ha ha! That's what you get for trying to formulate a question post lunch!

Ha, I know that feeling - too much blood diverted to the gut, not enough oxygen for brain cells... I've just had a light lunch so let's see how we fare.

If you have a digital source, let's say a CD player, feeding a DAC. That goes into a pre-amplifier and then after the pre-amp the signal goes into a digital signal processor. The digital signal processor takes the analog signal, converts to digital, does it's stuff (EQ, crossover, whatever) then converts back to analog. The signal then goes into the power amps and from there the speakers.

OK, I got ya So I'm curious now - why put the digital processor after the DAC and not (where in my estimation it should be) directly after the CD player? Is it because you need preamp functions to handle other analog sources? Having D/A then A/D afterwards is a recipe for SQ issues as no converter (in either direction) that I'm aware of is anywhere near the utter perfection of a digital link.

My question is whether the initial digital to analog conversion is in any way different to the following analog to digital and digital to analog conversion and whether there are different factors at play in terms of what engineering affects sound quality.

No I doubt that where the conversion is in the signal chain makes any difference. Think of A/Ds and D/As as somewhat grimy panes of glass - each one you have diminishes the transparency, no matter where it is in the stack. A digital link though is no glass at all.

Is a real-time A/D, D/A conversion different than a D/A conversion on a CD?

Perhaps the lunch induced oxygen starvation hasn't totally worn off - there is only ever real-time A/D and D/A because whilst digital doesn't have to be real-time, analog always does. A D/A inside a CD player will be essentially the same as a D/A inside a DSP box. There are HUGE (meaning of the order of night-and-day) differences between DACs though, which is something I've been gradually discovering for myself over the past two or three years, and the differences are not reflected in the traditional measurements we have. This is why two DACs whose measurements differ by less than a smidgen can be as different as chalk and cheese.

Does that make any sense?

Some - indeed a lot more sense but we're not totally there yet

 

[DonH50]

All conversions contribute to the signal at the speaker. Since data conversion (A/D or D/A) is a nonlinear process, you must multiply in time (convolve in frequency -- unless I got that backwords) everything to generate the end result. In the RF world we do it with chain matrices (at least some do; I use a computer to simulate or do the math for me, being a lazy ignorant engineer and not a theoretical mathematician). The data converters essentially perform mixing operations, with the ADCs aliasing their entire bandwidth into baseband, and DACs producing images extending beyond the signal baseband. And of course all the filters (ADC aliasing, DAC imaging) come into play along the way...

In general you will be limited by the least-common denominator for distortion, but both distortion and noise depend upon the gain trades. Where the gain is in the system (how much effective gain in each component) impacts the SINAD at the end of the chain. So, where components are placed does matter, at least theoretically. In an RF system you generally put the greatest gain at the input so noise contributions from the following stages contribute less. This leads to a trade in distortion, of course... And as opus111 said there are differences in implementation of all the analog buffers, filters, etc. in the chain. And differences in the digital filters, too.

Makes my head hurt...

 

[Nyal Mellor]

The devices I am referring to are often used in the high end home theaters after a pre-pro and before the power amps. They are also used in some really good sounding speaker systems such as the Lotus Group Grenada and Legacy Audio Whisper XD as well as others. Obviously the designers of these systems think that the additional conversions must be worth the benefit afforded in terms of removal of passive crossover components, control over the crossover, EQ, etc.

Interestingly for the standalone boxes that take an analog signal, do a digital conversion, then DSP, then a conversion back to analog, we can test their audio transparency by putting them in the signal chain with the DSP bypassed and compare that with the sound when they are not in the system.

 

[amirm]

Clearly this is the battle of large and small distortions. An outboard EQ will make a dramatic change to sound quality. Its contribution to low frequencies is essential. The trade off is potential addition of small distortions. I have not seen anyone quantify this trade off. For movie watching, I think it is no contest. The corrections for room response across multiple seats is critical to get right and there is no sense of "correctness" to the movie soundtrack as there is for music. So I am all out there. For pure 2-channel music listening as I said, more work needs to be there to quantify it.

One crazy thought I have had is to go to Steve's house and insert such an EQ and see if he likes it better or throws me out of the room . Let's see when he gets his new listening room done and we can visit that if he is willing.

 

[microstrip]

(...) One crazy thought I have had is to go to Steve's house and insert such an EQ and see if he likes it better or throws me out of the room . Let's see when he gets his new listening room done and we can visit that if he is willing.

Would you use the EQ only on subwoofers or in the whole system?

Since this thread is about EQ I will ask you a question about the SDEC-4500. As far as I know, one of its strongest points is that is can automatically optimize a multisub system using its own Sound Field Management algorithms and measuring systems. However, once you have carried the bass optimization, is not it possible to measure the responses of each channel and replace it with a much cheaper unit, e. g. Berhinger type units? In other words, can not this unit be used just to study the optimization of bass?

Please understand that my crazy question is purely academic in the good WBF tradition - I do not have access to such system and I am not considering EQ!

 

[amirm]

Would you use the EQ only on subwoofers or in the whole system?

I was going to subject him to the whole deal! But yes, one way out of the worry of A/D and D/A is to only apply it to bass.

Since this thread is about EQ I will ask you a question about the SDEC-4500. As far as I know, one of its strongest points is that is can automatically optimize a multisub system using its own Sound Field Management algorithms and measuring systems. However, once you have carried the bass optimization, is not it possible to measure the responses of each channel and replace it with a much cheaper unit, e. g. Berhinger type units? In other words, can not this unit be used just to study the optimization of bass?

If you can copy all the parameters one for one, yes. Here is the catch though. The calibration is a service for customers who buy JBL Synthesis. So you can't just get this portion as a practical matter.

 

[opus111]

Clearly this is the battle of large and small distortions. An outboard EQ will make a dramatic change to sound quality. Its contribution to low frequencies is essential. The trade off is potential addition of small distortions. I have not seen anyone quantify this trade off. For movie watching, I think it is no contest. The corrections for room response across multiple seats is critical to get right and there is no sense of "correctness" to the movie soundtrack as there is for music. So I am all out there. For pure 2-channel music listening as I said, more work needs to be there to quantify it.

I take it you were using 'distortions' in a kind of generic way - in my experience the primary 'distortion' introduced by D/A and A/D is noise modulation. They generally measure just fine for THD, what suffers is dynamics in a subjective sense. I have no experience of these systems myself so no comments to offer vis-a-vis the trade-offs.

@Don - yes you did get it backwards, convolve in time, multiply in frequency

 

[Nyal Mellor]

I take it you were using 'distortions' in a kind of generic way - in my experience the primary 'distortion' introduced by D/A and A/D is noise modulation. They generally measure just fine for THD, what suffers is dynamics in a subjective sense. I have no experience of these systems myself so no comments to offer vis-a-vis the trade-offs.

@Don - yes you did get it backwards, convolve in time, multiply in frequency

Hi Opus111 What do you mean by 'noise modulation'?

 

[opus111]

By noise modulation I mean dynamic changes to the noise floor which are correlated with the wanted signal. S-D DACs tend to have higher noise floor with larger signals - you can see that effect here: http://www.whatsbestforum.com/showthread.php?7103-Frugal-DACs&p=163843&viewfull=1#post163843. The Weiss Medea THD+N vs signal level plot also shows shifts in the noise floor occurring at particular levels of the test tone - between -30dB and -40dBFS if I recall correctly. This device uses an ESS Sabre32 DAC, also S-D.