Interesting topic: A/D converters and what makes them special?

[chessrock]

Second, I think it comes down to the quality of the D/A conversion. I think that for whatever reason it's a bigger trick to get it right going out as opposed to going in. I've had stuff recorded even on the lowest cost M-Audio card mix fairly well with improved D/A. Anyone else notice this D/A phenomenon???

That would certainly be a major, major factor for those of us who are routing back out to an analog mixer to do the summing . . . but not as much of an issue for those of us doing all of our summing "in the box," so to speak.

A certain eastern company by the name of Asahi Kasei Microsystems (AKM to you and I) have done a marvelous job of making affordable, quality A/D more readily available -- and I do believe a lot of the credit should go their way for the much improved performance of our budget sound cards . . . and even the not-so-budget ones like the Lynx, who use their technology.

Those looking for quality D/A may actually be able to find an affordable alternative in the form of a very familiar company with a familiar-looking box. And there are apparently some fairly reasonably-priced mods that can be done to further soup up the performance:

http://www.boldercables.com/Store.asp?m=TheBolderCableCompany&n=10&k=79313&s=ART+DI/O+Modifications

 

[ozraves]

That would certainly be a major, major factor for those of us who are routing back out to an analog mixer to do the summing . . . but not as much of an issue for those of us doing all of our summing "in the box," so to speak.

A certain eastern company by the name of Asahi Kasei Microsystems (AKM to you and I) have done a marvelous job of making affordable, quality A/D more readily available -- and I do believe a lot of the credit should go their way for the much improved performance of our budget sound cards . . . and even the not-so-budget ones like the Lynx, who use their technology.

Those looking for quality D/A may actually be able to find an affordable alternative in the form of a very familiar company with a familiar-looking box. And there are apparently some fairly reasonably-priced mods that can be done to further soup up the performance:

http://www.boldercables.com/Store.asp?m=TheBolderCableCompany&n=10&k=79313&s=ART+DI/O+Modifications

For those of us summing in the box, I do think that D/A matters a lot. After going from my M-Audio interface to an RME, I remixed four songs. The remixes sounded much better. If you can hear better, then you'll mix better. Part of what impressed me was how good the M-Audio A/D was for the price but I'd never gotten hear it due to the D/A conversion.

For those of us summing in the box, the master bus of the software is way important. Nuendo is better than most any software I've heard at summing. Does anyone know if Cubase SX has the same master bus?

I agree about the much improved performance of budget audio cards due to the improved converters.

Steve
www.mojopie.com

 

[mixsit]

Re: Re: Interesting topic: A/D converters and what makes them special?

>>snip...If you record at 16/44.1 the quietest sine wave that you can, the A/D would put out 1 LSB of information. This, obviously would translate into a square wave! The worst possible distortion we want. 24 bits, same test, would look significantly more like a sine wave. The distortion "floor" would be waaaay down there.

CD and 44.1Kz is the REAL problem. You must dither the bits to make the conversion. Dithering is "adding noise" on purpose to avoid the low volume "square wave" distortion. This is why CDs are always compressed. To avoid all the distortion, the CD mastering labs make sure that the dynamic range is limited to the top of the spectrum....

I'm in a bit over my head here., but this migh be untrue. As I understand, the first bit should produce the first 6db of dynamic range correctly, but the noise floor would be close behind. As you move up into higher the higher bits the noise as a % of the signal continues to diminish to where it is swampped by room noise and the analog limitations.

That was a good exorcize for me, and hopefully not TO far off...

Wayne

 

[Rev E]

I'll also add that if you're summing "in the box," it's probably a good idea to hook your monitors (or amp is using passive monitors) directly into the soundcard output. I keep everything plugged into the mixer, but for the final mix, I usually go directly into my active monitors. That way I cut out any effect the mixer has. Just remember to watch your output levels on the soundcard and turn off all system sounds so that if the computer started up in this configuration, the Windows startup sound won't blow your monitors to smitherenes.

Benchmark has a D/A converter that is quite the rave these days for the "direct connect approach" without the worries that I mentioned http://www.benchmarkmedia.com/digital/dac1/default.asp. A lot of respected folks have been raving about its sound quality as well (Bob Katz comes to mind). Pretty reasonable at ~$850. A/D to follow soon.

 

[Middleman]

Trying to catch up to you guys here. Let me summarize what I think is being said.

Most lo end cards do a great job at A/D, in fact as good as most devices up to say, $1000. It's the D/A that is making the difference with better clocks and jitter control in the higher end models.

Additionally, there are devices such as the one Chessrock points to, that can improve the D/A.

Several questions though.

Where would the D/A device go in the chain? Would this be off the existing lo end card, in my case an Audiophile 24/96? Would this come off the SPDIF port? If I come off the analog ports then I am already coming through the D/A of the soundcard, this does not seem right. Does the D/A get set up as an outward bound PCI card meaning; I would use my exisiting card for in bound signals and a more significant D/A device for outward bound?

 

[Rev E]

...Let me summarize what I think is being said.

Most lo end cards do a great job at A/D, in fact as good as most devices up to say, $1000. It's the D/A that is making the difference with better clocks and jitter control in the higher end models.

I'm not sure that's an accurate version of things... that D/A is all that matters. It's also critical to have good A/D. Hopefully others are saying that the problems of jittery input could be somewhat aleviated if one had a great D/A that enabled hearing the signal (albeit jittery) as unmolested as possible.

Where would the D/A device go in the chain? Would this be off the existing lo end card, in my case an Audiophile 24/96? Would this come off the SPDIF port?

For lower-end soundcards (or cards without AES/EBU) it would come out of the spdif out.

 

[Middleman]

Yeah, that was a little overstated. I am on a dentist high here and not thinking all that clearly. Thanks for the response though.

Chessrock opened a whole new thought process for me here. I was going to upgrade my soundcard but may go down another path which is an external DA device.

I wonder what the difference between the Lynx, RME and Delta products are on the A/D side of things? Would there be that dramatic of a difference in their interpretation of the audio data pushed to 1s and 0s? It would be interesting to compare the bits to see the differences. If they are not that great then a solid D/A makes alot of sense.

 

[Rev E]

I wonder what the difference between the Lynx, RME and Delta products are on the A/D side of things? Would there be that dramatic of a difference ... ? ... If they are not that great then a solid D/A makes alot of sense.

I can't say personally. But from the word on the street the Lynx card is in a class above the RME and Delta stuff. The Lynx cards are usually compared to high-end standalone converters like Apogee, Benchmark, and others in that realm. I can't say that the same is said of the RME Multiface or the Delta cards.

 

[ozraves]

The RME stuff is fairly good. Here's the thing: You'll see people upgrading from an M-Audio Delta 1010 to an RME DSP Multiface or a Lynx. But, it's doubtful you'll see RME DSP Multiface owners going to Lynx and vice versa.

It seems like lately there is a whispering campaign against RME. It's good stuff. The best sounding CD of 2002 to my mind was the Dixie Chicks' "Home" CD. It was recorded through RME converters (approx. $1,800 for eight channels) and an RME card (approx. $500) into Nuendo. I'd be very happy to get that sonic quality on any of my recordings.

Steve
www.mojopie.com

 

[Rev E]

The RME stuff is fairly good. Here's the thing: You'll see people upgrading from an M-Audio Delta 1010 to an RME DSP Multiface or a Lynx. But, it's doubtful you'll see RME DSP Multiface owners going to Lynx and vice versa.

...The best sounding CD of 2002 to my mind was the Dixie Chicks' "Home" CD. It was recorded through RME converters (approx. $1,800 for eight channels) and an RME card (approx. $500) into Nuendo...
Steve
www.mojopie.com

Steve,
The converter used by the Chicks was the ADI-8 PRO. I would agree with you that the word is and has been that the converters on the ADI-8 PRO are very good, certainly comparable to Apogee, the Lynx cards, and some high-end stand alone converters. However, my point was about Multiface and not the ADI-8 PRO.

The Multiface, while being a good set of converters for the price and likely use by home recordists, is NOT the same converters as the ADI-8 PRO. The $500 card that you talked about was just a Hammerfall ADAT card. There was no conversion going on in the card. All it does is accept a digital lightpipe signal from the ADI-8 PRO and send it to Nuendo (in this case). So the chain used by the Dixie Chicks was a bit more than your everyday Multiface chain. However, I do agree with you that the CD sounded good... a bit dry (spatial processing ... a.k.a. reverb), but certainly good. Plus the Chicks are known for not liking verb and have sworn it off since their first album.


Not that I'm looking at it, I can see how you thought I was referring to all of RME's goods when I said "the RME stuff." I meant the multiface.

 

[ozraves]

Rev E--

I agree with pretty much everything you say. I think we're talking past each other.

I'm going to try to get a LynxTwo for review.

--Steve
www.mojopie.com

 

[Teacher]

Yeah, that was a little overstated. I am on a dentist high here and not thinking all that clearly. Thanks for the response though.

Chessrock opened a whole new thought process for me here. I was going to upgrade my soundcard but may go down another path which is an external DA device.

I wonder what the difference between the Lynx, RME and Delta products are on the A/D side of things? Would there be that dramatic of a difference in their interpretation of the audio data pushed to 1s and 0s? It would be interesting to compare the bits to see the differences. If they are not that great then a solid D/A makes alot of sense.

from an ex-delta user to a current delta user....m-audio doesn't hold a CANDLE to the lynxtwo...thats my experience and i do even recall sjoko2 saying that the lynxtwo sounded better then the 1010 EVEN w/an external word clock...there was an article that said the AD on the lynxtwo compared to the Cranesong HEDD(approx. 3000 for 2chan AD/DA wit tape sim)i think its stretch personally....i think its safe to say the lo end cards have shit ad/ and da when comparing them to hi-end cards anyway...

 

[Teacher]

The RME stuff is fairly good. Here's the thing: You'll see people upgrading from an M-Audio Delta 1010 to an RME DSP Multiface or a Lynx. But, it's doubtful you'll see RME DSP Multiface owners going to Lynx and vice versa.

It seems like lately there is a whispering campaign against RME. It's good stuff. The best sounding CD of 2002 to my mind was the Dixie Chicks' "Home" CD. It was recorded through RME converters (approx. $1,800 for eight channels) and an RME card (approx. $500) into Nuendo. I'd be very happy to get that sonic quality on any of my recordings.

Steve
www.mojopie.com

i'm guessing the dixie chick LP sound had more to do with the 1 million dollar console...and 1 million dollars worth of mics, pres, compressors etc....not to mention the time they put in as well as the talent the 3 girls possess, rather then the converters sounding soooooo great

 

[ozraves]

i'm guessing the dixie chick LP sound had more to do with the 1 million dollar console...and 1 million dollars worth of mics, pres, compressors etc....not to mention the time they put in as well as the talent the 3 girls possess, rather then the converters sounding soooooo great

yeah, but the converters put the sound across rather well. here's the gear list of the studio where the cd was tracked: http://www.cedarcreekrecording.com/CEDAR.equip.html

it was mixed at a different on a euphonix board at a different studio.

 

[Teacher]

yeah, but the converters put the sound across rather well. here's the gear list of the studio where the cd was tracked: http://www.cedarcreekrecording.com/CEDAR.equip.html

it was mixed at a different on a euphonix board at a different studio.

the neve is enough to make you say...."ooooooo thats y....."

 

[chessrock]

word on the street the Lynx card is in a class above the RME and Delta stuff. The Lynx cards are usually compared to high-end standalone converters like Apogee, Benchmark, and others in that realm. I can't say that the same is said of the RME Multiface or the Delta cards.

I've heard much of the same thing. The magazine reviews seem to confirm the same thing . . . and I do remember Sonusman remarking on the improvement from the Delta 1010 to the Lynx, even when using an external clock with the 1010.

However, and this is the intersting part: In most of the reviews, these guys are doing side-by-side comparisons, and I don't recall anyone using the same D/A converter as a control in their tests in order to compare just the A/D sections.

What I'm wondering is if these people are just assuming the A/D is better because what's coming out of their monitors just sounds better . . . when in fact the reason it sounds better might have more to do with the better D/A of the Lynx (? ? )

Another interesting thing I remember Sjoko saying is that it's basically against the laws of physics to be able to make a high-end converter in the form of a PCI card-- due to the electric current needs of a converter versus a clock and so on.

Interesting.

 

[Teacher]

Hey nathan...fletcher....getta lynxtwo and compare to benchmark, mytek, lucid, cransong etc. ....let us know your opinions

...but the DA thing is an interesting thought but i mix 'in the box', so the higher quality i'm hearing may be due to my better decision making because of the better DA i have ...but for some reason i just can't buy that its only because i'm making better decision's in the mix, is y my tracks are starting to sound more hi fi...i'm guessing trackin with RNP pres instead of mackie's got something to do with it as well....

 

[Light]

Re: Re: Interesting topic: A/D converters and what makes them special?

Well, 24 bit is inherintley better than 16 bit as there is more dynamic range available. 44.1Khz has a theoretical top end of 22.05Khz. 96Khz has a 44.1Khz theoretical top end.

96k has a max reproducible frequency of 48 kHz. The bigger problem with 44.1 is not the max frequency, though, it is the anti-aliasing filter required to strip all information above 22.05k while at the same time leaving everything alone below 20k. That is a very serious piece of (analog) electronics, which produces some severe phase discrepancies. This leads to (among other things) a blurring of transients. It is this transient response which, in theory, can be improved with sample rates above 96k. With a 44.1 signal, your anti-aliasing filter needs to go from flat to -60dB or so in a little more than 1/10th of an octave. So what you are looking at is basically a low pass filter with a slope of 600dB per octave. Now, think about what EQ does to a signal at 12 dB per octave, this should demonstrate one of the major problems of 44.1. With 96k, you have a little more than an octave, which is a much less destructive piece of electronics (though still doing 60dB per octave. With 192k signal, you have almost 2 1/2 octaves. You are looking at a -20dB per octave filter. Much better. So yeah, I would say that the analog signal processing is a primary point of concern.

But you will also notice that I said earlier that this was the theoretical basis behind 192k being useful. Whether or not anyone has actually made a 192k converter which lives up to the theory, I do not know.

When I was in school, there was a very cool demonstration in one of the classes. The tech guy there had made a box which was just an ADC and a DAC, but with patch points after every stage of processing. You got to here exactly what each stage did. It was very educational.


Light

"Cowards can never be moral."
M.K. Gandhi

 

[Light]

What I've been hearing a lot of . . . and correct me if I've heard wrong . . . is that a lot of 24-bit converters aren't necessarily pumping out a full 24-bits worth of actual audio. So am I to assume that some of these bits are just "junk bits?" -- or as some people like to call them "marketing bits."

Data correction bits. CDs are not 16 bit either, becasue a few of those bits are dedicated to data corection. (at least that is what I remember from when I was studying this stuff, but that is closing in on ten years ago).

For instance, the last bit (or the first, I can't really remember) is a odd/even bit. If there are an odd number of "1s" in a , than the odd/even bit is a "1". If there are an even number of "1s" in the word, than the odd/even bit is a "0". If the odd/even bit and the rest of the word don't match up, then there is somthing wrong and the DAC dumps that word. If you have too many bad words in a row, you get skiping, errors, etc.

Like I said, I haven't studied this in quite a while, and I have not really had to think about it, so I could be remembering this wrong.


Light

"Cowards can never be moral."
M.K. Gandhi

 

[McQ]

Re: Re: Re: Interesting topic: A/D converters and what makes them special?

Originally posted by Light
96k has a max reproducible frequency of 48 kHz. The bigger problem with 44.1 is not the max frequency, though, it is the anti-aliasing filter required to strip all information above 22.05k while at the same time leaving everything alone below 20k. That is a very serious piece of (analog) electronics, which produces some severe phase discrepancies. This leads to (among other things) a blurring of transients.

This used to be true...before oversampling converters. Oversampling converters are the prevalent technology these days...

It is this transient response which, in theory, can be improved with sample rates above 96k. With a 44.1 signal, your anti-aliasing filter needs to go from flat to -60dB or so in a little more than 1/10th of an octave.

That's the correct idea, but ideally you don't want any aliased components to register at even the LSbit level.

So what you are looking at is basically a low pass filter with a slope of 600dB per octave. Now, think about what EQ does to a signal at 12 dB per octave, this should demonstrate one of the major problems of 44.1. With 96k, you have a little more than an octave, which is a much less destructive piece of electronics (though still doing 60dB per octave. With 192k signal, you have almost 2 1/2 octaves. You are looking at a -20dB per octave filter. Much better. So yeah, I would say that the analog signal processing is a primary point of concern.

That's one of the reasons why it's not done in the analog domain these days...

But you will also notice that I said earlier that this was the theoretical basis behind 192k being useful. Whether or not anyone has actually made a 192k converter which lives up to the theory, I do not know.

Sampling rate ain't the problem...resolution is (particularly with ever-diminishing IC gate geometries and the required operating voltage reduction...and a universally avoidable noise floor). Some (many?) of the communication systems operating these days are operating with ADCs in the tens-of-megahertz conversion rates (even at 16 bits).

When I was in school, there was a very cool demonstration in one of the classes. The tech guy there had made a box which was just an ADC and a DAC, but with patch points after every stage of processing. You got to here exactly what each stage did. It was very educational.

I wish more people could experience what you did...there'd be a lot less myth-masquerading-as-fact then!

Cheers!
McQ