what are the best types of DAC?
06-21-2005, 05:06 PM
#1
Registered User
Thread Starter
Join Date: Sep 2004
Posts: 271
Likes: 0
Received 0 Likes
on
0 Posts
what are the best types of DAC?
just got to wondering if anyone out there can toss together an order of DAC performance in the car audio world. i know there's 1 bit, regulated 1 bit, 24 bit, 24 bit sig delt, 24 bit burr brown and so on. what's the best, what's the worst? anyone really know, not just from what they've read in someone's product literature? i was always under the impression that the 24 bit burr brown was some of the best to be had.
06-21-2005, 05:55 PM
#3
Registered User
I'm probably one of the relative few who are qualified to answer this one... oh no, I came --> <-- this close to calling myself an "expert" 
(inside joke for the long-time EJs out there)
Burr Brown is a semiconductor manufacturer, not a specific type of DAC.
Funny enough, all of the DACs you listed fall into the same category of oversampling DACs. Essentially, you're sampling the same data multiple times, hence the term oversampling. By doing this, you can use a lower resolution DAC internally. It also allows the designers an opportunity to relax some of the specs for related components (such as the low-pass filters attached to the output).
Single-bit DACs are generally chosen due to their inherent linearity (which is a good thing when it comes to audio). Sigma-Delta DACs lower the average noise floor by adding in some dither to the signal.
If I'm left with a choice between a simple 1-bit and a Sigma-Delta, I'll go for the SD version everytime. However, the single-bit converter that's listed doesn't specify what sort of pre-/post-processing they're doing on the signal (which, in essence, is all the SD does), so their implementation could be better.
In the end, though, you're never going to hear any difference. You're talking the difference between tenths of a dB in a noise floor so low down already with such a wide dynamic range (CDs, remember?), there's no point in arguing who's better. It's all marketing speak. If I was designing a nanovolt-level DAC, I'd be reallly concerned about this stuff, but that's not what we're doing here...
(inside joke for the long-time EJs out there)Burr Brown is a semiconductor manufacturer, not a specific type of DAC.
Funny enough, all of the DACs you listed fall into the same category of oversampling DACs. Essentially, you're sampling the same data multiple times, hence the term oversampling. By doing this, you can use a lower resolution DAC internally. It also allows the designers an opportunity to relax some of the specs for related components (such as the low-pass filters attached to the output).
Single-bit DACs are generally chosen due to their inherent linearity (which is a good thing when it comes to audio). Sigma-Delta DACs lower the average noise floor by adding in some dither to the signal.
If I'm left with a choice between a simple 1-bit and a Sigma-Delta, I'll go for the SD version everytime. However, the single-bit converter that's listed doesn't specify what sort of pre-/post-processing they're doing on the signal (which, in essence, is all the SD does), so their implementation could be better.
In the end, though, you're never going to hear any difference. You're talking the difference between tenths of a dB in a noise floor so low down already with such a wide dynamic range (CDs, remember?), there's no point in arguing who's better. It's all marketing speak. If I was designing a nanovolt-level DAC, I'd be reallly concerned about this stuff, but that's not what we're doing here...
I get it Dan
06-21-2005, 06:28 PM
#5
Registered User
Thread Starter
Join Date: Sep 2004
Posts: 271
Likes: 0
Received 0 Likes
on
0 Posts
thanks dan, appreciate the clarification! so what other factors in the circuitry contribute to added or reduced noise in HU's? is there any single determining factor in what's going to provide the "cleanest" sound, assuming all other outside factors remain the same? for example...how come, to my ears, i can listen to a regulated single bit Alpine through my system and it doesnt sound nearly as clean as say my sony 910 with dual 24bit DAC (keeping in mind both have close to similar s/n ratios etc.).
06-21-2005, 07:00 PM
#6
Registered User
i'll just say this much... i have an Adcom GCD-700 here at the house that is my main CD player. it sounds absolutely fantastic. i've tried at least 10 different outboard DACs with it, but none sound as authentic and lifelike as the built in DAC.
it uses a "Dual 20 bit Burr Brown digital to analog converter employing a 352.8 kHz eight times oversampling filter." and it uses dual power transformers for the digital stage and the class A pre-amp stage.
i can't speak for other DAC's, but i love this one. 
(aside: i can recall many times stating that these were dual 24 bit DAC's, but having checked the manual, i see that i was wrong. oops.)
it uses a "Dual 20 bit Burr Brown digital to analog converter employing a 352.8 kHz eight times oversampling filter." and it uses dual power transformers for the digital stage and the class A pre-amp stage.
i can't speak for other DAC's, but i love this one.
(aside: i can recall many times stating that these were dual 24 bit DAC's, but having checked the manual, i see that i was wrong. oops.)
Trending Topics
06-22-2005, 02:34 PM
#8
Registered User
There's no single determining factor, but every little bit helps. Rather than view it as "what's best", view it more along the lines of "what does the least damage to the signal at every stage".
First, you want to make sure the signal coming in is as clean as possible. In the case of FM signals, that means a quality mixer and pre-amp section. If we're talking CDs, the bitstream can be degraded by multiple things... poor lens alignment, dusty/oily CD, flakey tracking electronics, etc. This problem is (partially) resolved by oversampling... read the same bit multiple times and take an average of the value. Note, however, this is not the oversampling often referred to in marketing pamphlets.
That oversampling has to do with the data once it has been read (and possibly corrected). The units create interpolated data points in between the actual data points. For example, let's say we read a data point every second... if we oversample by 2x, we interpolate data points in between the ones we read, which makes it appear as if we actually read data every 1/2-second. We didn't actually read the data that fast, we just took a "guess", if you will, at what the data would have looked like if we had sampled that fast.
As I mentioned earlier, this makes the specs for surrounding components more "loose" (which means less expensive). It also has the advantage of reducing phase shift at higher frequencies... phase shift can be a bad thing when it comes to audio. As a visual example, phase shift of a video signal's high-frequency components will cause color smearing across the screen... imagine being able to hear this "color smear" in your audio... not good.
Second, see if the unit dithers the signal any... some do, some don't, and they may or may not list it. You want dithering whenever possible, assuming they dither correctly, for the lower bit count DACs. This is pretty irrelevant for 24-bit DACs. As counterintuitive as it sounds, dithering is actually the process of adding noise to the signal. Although this will add in a very low level of hiss (your ear is probably not going to hear it), it has the advantage of removing heavy amounts of distortion, particularly with high dynamic range material with a lot of quiet passages (like classical music). In a nutshell, dithering allows us to hear more bits of resolution than the material was recorded with (I shit you not!)... since 24-bits is way beyond our hearing resolution capabilities, there's no point in using it with 24-bit DACs, as I mentioned earlier.
The manufacturer may or may not list it, but if you're using a lower bit-length DAC, select the steepest low-pass filter cutoff you can. See my previous comments on oversampling.
We could go into a long discussion about how components external to the DAC affect the sound quality. There is a lot of truth in the matter, but as components get better, it's beginning to matter less and less. The difference between components from the 40's and from the 90's are gigantic, and that doesn't even include the type of component, such as electrolytic cap versus polyester cap.
PCB layout can have a large effect, too. If not designed properly, digital noise from the various quartz crystals, ceramic resonators, and PLLs on board can bleed back into the audio spectrum. A poor ground leads to a poor ground reference, making the converted voltages less than exact.
Did I miss anything?
First, you want to make sure the signal coming in is as clean as possible. In the case of FM signals, that means a quality mixer and pre-amp section. If we're talking CDs, the bitstream can be degraded by multiple things... poor lens alignment, dusty/oily CD, flakey tracking electronics, etc. This problem is (partially) resolved by oversampling... read the same bit multiple times and take an average of the value. Note, however, this is not the oversampling often referred to in marketing pamphlets.
That oversampling has to do with the data once it has been read (and possibly corrected). The units create interpolated data points in between the actual data points. For example, let's say we read a data point every second... if we oversample by 2x, we interpolate data points in between the ones we read, which makes it appear as if we actually read data every 1/2-second. We didn't actually read the data that fast, we just took a "guess", if you will, at what the data would have looked like if we had sampled that fast.
As I mentioned earlier, this makes the specs for surrounding components more "loose" (which means less expensive). It also has the advantage of reducing phase shift at higher frequencies... phase shift can be a bad thing when it comes to audio. As a visual example, phase shift of a video signal's high-frequency components will cause color smearing across the screen... imagine being able to hear this "color smear" in your audio... not good.
Second, see if the unit dithers the signal any... some do, some don't, and they may or may not list it. You want dithering whenever possible, assuming they dither correctly, for the lower bit count DACs. This is pretty irrelevant for 24-bit DACs. As counterintuitive as it sounds, dithering is actually the process of adding noise to the signal. Although this will add in a very low level of hiss (your ear is probably not going to hear it), it has the advantage of removing heavy amounts of distortion, particularly with high dynamic range material with a lot of quiet passages (like classical music). In a nutshell, dithering allows us to hear more bits of resolution than the material was recorded with (I shit you not!)... since 24-bits is way beyond our hearing resolution capabilities, there's no point in using it with 24-bit DACs, as I mentioned earlier.
The manufacturer may or may not list it, but if you're using a lower bit-length DAC, select the steepest low-pass filter cutoff you can. See my previous comments on oversampling.
We could go into a long discussion about how components external to the DAC affect the sound quality. There is a lot of truth in the matter, but as components get better, it's beginning to matter less and less. The difference between components from the 40's and from the 90's are gigantic, and that doesn't even include the type of component, such as electrolytic cap versus polyester cap.
PCB layout can have a large effect, too. If not designed properly, digital noise from the various quartz crystals, ceramic resonators, and PLLs on board can bleed back into the audio spectrum. A poor ground leads to a poor ground reference, making the converted voltages less than exact.
Did I miss anything?
06-22-2005, 02:38 PM
#9
Registered User
Oh, and referring to your 24-bit Burr Brown DAC is akin to talking about your Hewlett-Packard printer or your Memorex hard drive. Burr-Brown is simply the manufacturer (actually, Texas Instruments purchased that division years ago), so mentioning it when talking about the DAC doesn't really tell you anything.
Saying it's a 24-bit DAC, or a Sigma-Delta DAC, now that tells you something useful.
Saying it's a 24-bit DAC, or a Sigma-Delta DAC, now that tells you something useful.