[purple_color]

So, in my choise A/V proseccors for 7.1. active (Dynaudio bm6a) setup, I see this information about this models:

http://www.kgjack.com/pre-pros.htm


Rotel 1066 ($1350) - DAC - AKM 4364 24bit/96KHz, Processor - Crystal 49326 24bit.

B&K ref. 30 ($2300) - AKM 4393 24bit/96KHz, Motorola 56366 24bit.

Anthem 20 ($3300) - AKM 4382 24bit/192KHz, Motorola 56366 24bit.


What we have? New processors have old 96/24 DAC, old 24 bit DSP - but brand new recievers have 32 bit DSP and 192/24 DAC on board - such as Pioneer 2011, Yamaha DSP AZ2, AZ1, etc.


If I'll chose optimum perfomance model for $1000-2000 price, what can I do? Is DSP bits so important? Is 192/24 DAC really needs? Separates components always better?


If chose best DAC, Yamaha Z1 is winner (10 x Burr-Brown PCM1704 24-bit/192kHz vs. lowest price DAC in Anthem. B&K, etc..)? What a confusion..

 

[goldear]

I've wanted to gain a little more knowledge on these specs as well. I know the SHARC 32bit "floating point" processor is better than the older (56366) 24bit "fixed point" DSPs. But I also read that the newer Motorola 56367 DSP is suppose to be better than the older SHARCS...depending on the application. Someone mentioned how it takes multiple SHARCS to do the job of a single 56367. Is this true?


I also notice how some newer AKM 4393 dacs only have 24/96 resolution...at 120db, while their older dacs, ala the AKM 4382, do 24/192. Can someone give an explanation for this disparity. I know implimentation is a key, but is that all that's to it. Thanks

 

[PlasmaSquid]

ADC/DAC specs can easily be found in the datasheets on the respective manufacturers websites. Just go there and do a part number search. But keep in mind that these are just the manufacturer's published specs, which are acheived under near-ideal conditions. The individual circuit design in each piece of equipment that uses the part will always cause variations in the final specs.


The particular DSP being used is somewhat irrelevent as far as audio quality goes, since we're just talking about number crunching. It's more applicable to compare the algorithms which are implemented. A 24-bit, fixed point 56xxx DSP can implement double-preciscion 48bit math, if the DSP software engineer wanted to, so the native internal memory-width of the processor can be extended with some extra coding. The Analog Devices SHARC is 32-bit, fixed and floating point, but if the algorithms are implemented poorly enough, then the extra precision doesn't matter. Also, depending on who you talk to, some say fixed pt. is better than floating pt. and vis versa. In the end, it really depends on the algorithm and the math you're trying to implement.


Something that is easy to compare between these DSPs, is the speed. A 200MHz processor will allow the software engineer more flexibility to write more complete algorithms or add processing functionality than a 50MHz DSP, for example.


Also, keep in mind that many chip vendors (and Dolby) offer libraries of pre-canned algorithms that can be used directly or built-upon.

 

[thomaspf]

DACs are not just all about implementation. As pointed out by a previous poster the specs given by the manufacturers in their data sheets are the best possible case. In an actual implementation the Op-amps to buffer the DAC outputs probably are the other major component on the way to the output socket but the quality will only go down whatever you do.


There are huge differences in the DAC chips of the models you listed which will almost inevitably be reflected in the sound. The true resolution of the processors listed is probably somewhere between 14 and 15 bits with the remaining lower bit overlayed with statistical noise.


Some of the newer DVD players uses the latest generation DAC ships and op-amps and put almost any processors to shame. I have no idea why that is so but assuming you are looking for the best audio quality in your price range I would at least try out one of those newer players.


Toshiba SD9500 which also has digital in!

Denon DVD2900 will be launched in May


On paper these measure much better. I have not had the chance to listen to either of them.


Cheers


Thomas

 

[purple_color]

Ok, let's see DSP speed:


- New Motorola chip 56367 - 150 MHz, but it's used only in "high end" level proseccors (Naim, Anthem) for $3,000.

- Analog Devices - 66 MHz 21065L SHARC, it's use in Denon, TagMaclaren, etc.

- Yamaha always use 32-Bit Original LSI (YSS-938) - (?) MHz

- Sony use original 32 bit x 2 - (?) MHz

- Rotel 1066 use old Crystal 24 bit 33 MHz, new Rotel 1098 processor use Crystal 60 MHz, but have $2,900 price.


Who's next?


DAC specification is more strange:


Anthem 20 ($3,100) use AKM 4382 24bit/192KHz ($2)

Yamaha AZ1 ($2,200) use Burr-Brown PCM1704 24-bit/192kHz ($12,5)

Pioneer 2011 ($1000) use Burr-Brown PCM1704 24-bit/192kHz ($12,5) !!!

Rotel use older AKM 4364 24bit/96KHz with poor S/N, but $7 price.

Tag use 192kHz / 24bit AK4395 DAC (?)

Arcam ($4,400) use Wolfson - and new Onkyo Integra ($1,700) use Wolfson.

B&K use new AKM DAC ($4)...


Op-amps for DAC outputs is important, but with poor DAC with 90 Db S/N is nothing to help this processors, I'm right?


Now, this is wery difficult choise in hardware: low price model have better DAC/DSP on board?!


Main questions is - were can we see full specification on this models range DAC/DSP & what models is winner in home cinema proseccing only (NOT amplification) in "up to $600", "up to $1000" & "up to $2000" retail price?

 

[PlasmaSquid]

As with everything else in Life, you can't isolate one particular item (a DAC, a processor, a power supply, cables, speakers, opamps, power amps, race, color, creed, ....), take it out of context and use that as a reflection on the sum of the parts. There are many variables.


In addition to the internal bus-width and the raw MIPs (or DSP/processor speed). There are many variables that make up a DSP -- things such as it's instruction set and register model, it's efficiency in moving audio samples in/out of the chip, whether or not it has DMA functionality to help move data, it's ability to perform parallel moves/memory accesses in conjuction with a MAC (multiply/accumulate) operation, use of a internal program and/or data cache, whether it is using internal RAM or if it has to go off-chip to fetch instructions and/or data, hardware facilities for logarithmic functions or other math operations, etc.... etc...... And then, of course, there's how the programmer utilizes all this functionality. As technology is progressing, more and more stuff is getting implemented in hardware and things are becoming more and more generic for high-volume products (compared to, say, 5 or so years ago when more stuff had to be designed and coded from scratch).


In the end, the DSP is just crunching numbers. The operations could be done in a Pentium or PIC. But the Pentium is overkill/overpriced and has no easy facility for moving audio samples. The PIC would be too slow. DSPs are designed for real-time audio applications. But in no way could you say that one DSP sounds better than another. It's the math that's being done inside.


I'm not sure where those converter prices you listed were found, but I can tell you that volume discounts, buying direct vs. using distributors, and many other factors play a HUGE role in what clients pay for electronic parts (especially items like audio converters). I would seriously doubt that ANY manufacturer is paying $12 for a stereo audio DAC in this day and age.


If you really want to examine converter performance, you should get the product in question and test it using an Audio Precision. In some cases, you will come close to the manufacturers published specs, but in many cases the results will be worse. This is because the testing procedures might be different, and because many manufacturers (converters and gear) exaggerate their specs.


Really, it's not worth worrying too much about specific components, as long as a product uses something that's within a certain range of "good". The circuit designs and DSP algorithms are much more important. Almost anyone can buy the absolute best/most expensive components and put them in a box and sell it. But only good companies and good engineers can make the stuff sound good and work well together.

 

[goldear]

Plasma,


I understand that DSPs don't have so much an effect on the sound, but it seems that using the correct DSP engine is important for functionallity and future upgradability. Alot of companies couldn't do DPLII and THX Ultra2 because of the older engines, so they had to install new ones...ala the AVM20 that went from the Motorola 56366 to the 367. While companies like Lexicon are using 4 SHARCS, which would seem to leave enough expansion, does the 56367 have enough room for future expansion? Is this something that a consumer should be aware of when purchasing a new SSP? Thanks

 

[hwc]

Guys:


This whole DSP issue is not that complicated. It's just like buying a computer. If you bought one last year, you got an Intel Pentium XXXX processor. If you buy one this year, you'll get a different Pentium processor because Intel has introduced a new processor at the old price.


The digital side of A/V processors is nothing but a dedicated computer. Just like the chips in your PC constantly change, the chips in the A/V receivers change with the seasons. As faster chips come along, the chip makers add more features to the software because the horsepower of the new chips will run more complex software. Just like with new versions of Microsoft Word, the vast majority of "new features" are things that you could live without and not suffer any great hardship. The fundamental stuff that you want a Dolby Digital receiver to do remains unchanged.


Two pieces of advice:


a) don't fall into the "spec" trap when evaluating an audio component.


b) There are reasons to buy an expensive A/V receiver, but what is happening in the digital processing is not usually one of them. Everybody is using the same chips and what chips they use is determined by when the product was designed.


For example, this year's big "new feature" is 24 bit/192kHz DACs. Why? Because this year's chips have 24bit/192kHz DACs. Now, I don't know where you are going to get a multi-channel 192kHz digital signal into the receiver to decode. But, those who peddle the newest greatest specs don't let trivial little details like that get in their way! They know that consumers will fall for the spec game, even when they don't even know what the specs mean -- after all, 192 somethings must be twice as good than 96 somethings.


PS: My volume knob goes to eleven.

 

[kaanage]

Beautifully put , hwc.

 

[thomaspf]

Well, one more reason to looks at those players I mentioned earlier up in the thread. There have been sightings of DVD-As with 192Khz stereo tracks.


Not that I pretend I could hear the difference between a 24/96 or 24/192 recording ....


One important detail that you are brushing over is the fact that a lot of progress has been made in engineering chips and systems that have a true resolution of more than 16bits.


Let's define true resolution as the lowest order bits that you can determine with 100% certainty from looking at the output voltage of your DAC. Dynamic range and signal to noise ratio are all nice numbers but THD+N is what is usually limiting true resolution.


Looking at the processors that purple_color introduced at the beginning of the thread.


Rotel 1066 : 0.05% ~ -66db THD+N
B&K ref. 30 : did not find the info

Anthem 20 : 0.008% ~ -81db THD+N

Compared to the latest generation of players this is not even close


Denon DVD-2900 : 0.0009% ~ -100.9 THD+N

or a good 2 channel stero DAC just for comparison


Benchmark Media DAC1 : 0.0005% ~ -106db

This improvement applies even to plain old CD-audio at 16/44.1 and is quite audible.


Thomas

 

[PlasmaSquid]

Quote:

Originally posted by goldear Plasma, I understand that DSPs don't have so much an effect on the sound, but it seems that using the correct DSP engine is important for functionallity and future upgradability. Alot of companies couldn't do DPLII and THX Ultra2 because of the older engines, so they had to install new ones...ala the AVM20 that went from the Motorola 56366 to the 367. While companies like Lexicon are using 4 SHARCS, which would seem to leave enough expansion, does the 56367 have enough room for future expansion? Is this something that a consumer should be aware of when purchasing a new SSP? Thanks

It's hard to predict the future of processing requirements. Who knows what new surround formats/encoding schemes will be the "new thing" a few years from now? Or how many surround channels Dolby will be saying we need? Like hwc said, the best processors and software of today will eventually be replaced by something newer and faster tomorrow. DSPs and audio formats don't change quite as quickly as computer OS's and applications. Note that the basic 24-bit Motorola 56xxx DSP has been around for what, about 10 years. They've done a good job in upgrading the complimentary hardware and speeding up the core to keep it competitive. There are 56k DSPs that are faster than the '367, although I don't know all the other differences without some research. The SHARC is a bit newer. There are different flavors of the SHARC, and I'm not sure what flavor is being used in the Lexicon. The architecture is different, and like the PowerPC vs. Pentium arguments, some people say you can do more with a slower SHARC than a faster 56k. Getting an upgradable pre/pro from a manufacturer that has a good track record of support and upgrade releases is something to consider if you're concerned. Anthem is one of these.


It's hard to avoid getting caught up in specs and numbers, but they are definately not what you should be the most concerned with.


Regarding Lexicon - they're different in that they place a huge emphasis on their signal processing chops. And yes, they do have a long and proven track record for producing great DSP boxes (480L, 960L, PCM80, etc..). They have a lot of proprietary signal processing intellectual property, and have put a lot more into their algorithms than most other companies. It's one of the reasons they can charge more for their products. The hardware components that they use only make up part of the price difference.


Thomas, I agree that THD+N is a revealing measurement and that very few pieces of equipment reach the 0.0005% number (specs again). But I don't know that this measurement applies when speaking of the lowest bit resolution that can be resolved (in practical terms). I say this because THD+N is measured by injecting a high-level sine wave at particular frequency, notching it out, and including the resulting harmonics (distortion) and noise in the measurement. When measuring or listening to the lowest level signal (resolution) a DAC can resolve, you don't have the harmonic distortion component along with the normal noise floor of the converter and/or device. IOW, when listening to a note on a piano die out or the fade of a reverb tail, there is no added harmonic distortion from some notched out sine wave. There's just the noise floor of the device and whatever's already in the source material. For that reason, I think S/N ratio is a better measure for this particular case. But that's me. But I definately agree that the analog world will never equal the theoretical S/N of converters 18-20bits is about as far as it's ever going to get. More bits in converters do give the advantage of keeping resolution/information during signal processing operations though.

 

[thomaspf]

I am not sure why resolution is only applicable to low level signals? My assumption so far was that with a THD+N of let's say -90db I can determine the state of all 14 significant bits in a 0db signal but the lower ones are masked in noise caused by distortion which happens to be of higher impact than the thermal noise floor of the chip.


Is that assumption correct?


If it is then the better resolution gives you a clearer signal much like locking into a radio station that is somewhat out of tune. You can usually still hear what is going on but it is overlayed with noise. The lower the noise the clearer the signal.


Some companies measure their distortion and noise over the whole operational frequency range. I am not sure I understand you train of thougth with regards to notching out the original signal from a piano note. A piano note is not a clean sinus wave but is a complex mix of frequencies. Isn't the likely result of this is that a DAC will have even more problems to faithfully reproduce the signal with more bits in the distortion noise.

Cheers


Thomas

 

[PlasmaSquid]

Sorry Thomas,


I wasn't saying that a notch filter is applied to the piano note. I was referring to the other case - in which THD+N is measured and you're injecting a sine wave at say 1KHz. A notch filter removes this fundamental, and then the remaining frequency spectrum is examined for resulting sidebands (harmonic distortion) and noise.


You're right that the extra bits & resolution matter no matter what the level of signal is that you're trying to reproduce -- whether it's a very low level piano note or a very loud guitar chord. But as the signal level approaches full-scale, the noise floor becomes less noticable (given a reasonably low noise floor). IOW, the signal masks the noise. So the greater the ratio, the better. But with a low level signal (such as a decaying piano note), the signal level is smaller and the S/N ratio also gets smaller. So the noise will become more and more apparent as the piano note (for example) dies out, as there's less signal energy to cover up the noise (to our ears anyway). Going back to the original questions, the limiting factor will either be the converter resolution or the analog noise floor. And I think we agree that in many cases, with modern converters, that the analog domain is the limitation.


Easy.

 

[arthurvino]

Now we can add Parasound Halo AKM 24/96 in there, with new 24/192 supposedly to be implemented by year end...

Quote:

Originally posted by purple_color Ok, let's see DSP speed: - New Motorola chip 56367 - 150 MHz, but it's used only in "high end" level proseccors (Naim, Anthem) for $3,000. - Analog Devices - 66 MHz 21065L SHARC, it's use in Denon, TagMaclaren, etc. - Yamaha always use 32-Bit Original LSI (YSS-938) - (?) MHz - Sony use original 32 bit x 2 - (?) MHz - Rotel 1066 use old Crystal 24 bit 33 MHz, new Rotel 1098 processor use Crystal 60 MHz, but have $2,900 price. Who's next? DAC specification is more strange: Anthem 20 ($3,100) use AKM 4382 24bit/192KHz ($2) Yamaha AZ1 ($2,200) use Burr-Brown PCM1704 24-bit/192kHz ($12,5) Pioneer 2011 ($1000) use Burr-Brown PCM1704 24-bit/192kHz ($12,5) !!! Rotel use older AKM 4364 24bit/96KHz with poor S/N, but $7 price. Tag use 192kHz / 24bit AK4395 DAC (?) Arcam ($4,400) use Wolfson - and new Onkyo Integra ($1,700) use Wolfson. B&K use new AKM DAC ($4)... Op-amps for DAC outputs is important, but with poor DAC with 90 Db S/N is nothing to help this processors, I'm right? Now, this is wery difficult choise in hardware: low price model have better DAC/DSP on board?! Main questions is - were can we see full specification on this models range DAC/DSP & what models is winner in home cinema proseccing only (NOT amplification) in "up to $600", "up to $1000" & "up to $2000" retail price?