|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
65.19.73.221
In Reply to: RE: Wonder what the tolerances of the resistors are, or more importantly, what they need to be? posted by Ivan303 on September 09, 2015 at 08:37:07
If the resistors are 0.1% accuracy then the DAC will be accurate to roughly 10 bits. Unless there are other mechanisms to calibrate the circuit the result is unlikely to be good. Also, if the resistors are physically separated ("discrete") small temperature differences will affect performance. The result of inaccurate values will be distortion, especially ugly cross-over distortion on small signals.
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
Follow Ups:
Well darn if Ivan didn't make a good point when he actually tried. Would a .5% tolerance be any better in your opinion?
-----------------------------------------------------------
Big speakers and little amps blew my mind!
No 0.5% would be five times worse, i.e. 8 bit quality.
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
.01%, .02% or .05% resistors.
Link below:
0.01% resistors would give you 12 bits.
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
Not sure the resistor tolerance matters much? If you use 5% resistors, you will just have to measure several thousand or more to get the values you are looking for. If you are using .1% tolerance resistors, you might only have to go through several hundred in order to get the matched sets needed.
I'm sorry but this would be a tough google for me. What is the tolerance of a laser etched resistor on a run of the mill R2R dac?
------------------------------------------------------------------
Big speakers and little amps blew my mind!
Read the spec sheet for what the chips claim. Percentage for 16 bits would be 1/(2^16) * 100% = 0.0015%. For 18 bits it would be 0.0004%.
Hopefully, anyone seriously considering building a D.I.Y. R2R DAC would understand the issues here and where these numbers come from.
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
Darn if wasn't right there in Wikipedia.
> "The required accuracy doubles with each additional bit—for 8 bits, the accuracy required will be better than 1/256 (0.4%).
> On a printed circuit board, using discrete components, resistors of 1% accuracy would suffice for a 5 bit circuit, however with bit counts beyond this the cost of ever increasing precision resistors becomes prohibitive. For a 10 bit converter, even using 0.1% precision resistors would not guarantee monotonicity of output."
----------------------------------------------------------------
Big speakers and little amps blew my mind!
"For a 10 bit converter, even using 0.1% precision resistors would not guarantee monotonicity of output."
OK... And that could be a lock-up nightmare in a digital feedback system. But just listening? Not so much.
The thing I'm not clear on (and geanted I've prolly not read every post) is what the problem is that using discrete networks is thought to address. I must admit that the notion rather appeals to me because, well, it's more tweakable. Things that would be absolute disasters in volume production can be opportunities for fine-tuning for individual users...
Rick
> " what the problem is that using discrete networks is thought to address."
I think it's sole purpose is an attempt at increased sound quality. Looks like it causes problems.
------------------------------------------------------------
Big speakers and little amps blew my mind!
Two reasons why discrete will have problems:
1. components not matched to sufficient tolerance
2. components drift from previously matched settings.
Temperature changes will cause drift, so it there is uneven heating of components there will be problems. Temperature differences can be minimized by putting all of the resistors on a single substrate, but then this won't be a discrete approach.
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
Thanks Tony. I think you and Abe are in agreement on this one. Now I'm thinking of the copper foil on those metal foil resistors. Wonder if when they oxidize it will change the resistance?
What about my question regarding the AD5791? What do you make of the quote from the spec sheet?
Thanks.
-------------------------------------------------------------------
Big speakers and little amps blew my mind!
I wouldn't take too much stock in theoretical type answers.
I'd get real information from people who actually "Modify" gear in real life.
There are plenty of forums that have people who like to mod gear and have tried many types of resistors in real-life scenarios.
Some folks over at Headfi, for example, did mod work to the Havana dac and other products and post detailed results and how-to's etc. For example bypassing the internal opamp of the PCM56P dac chip. I myself modded one of my Havana Dacs, replacing some 54 parts including Shinkoh Tantalum resistors, Schottky Diodes and much more.
Theory is great, but real-world in better imo.
Dynobots Audio
Music is the Bridge between Heaven and Earth - 音楽は天国と地球のかけ橋
> "Theory is great, but real-world in better imo."
I agree. The simple fact is these discreet R2R dacs are rated in the A+ category by Stereophile, and I am a big sceptic with journalists having been one, but you have to take their opinion into account.
In reading the reviews I came across one statement that stuck with me. The following is by Michael Lavorgna in a review of the Totaldac.
> "I compared the Totaldac to the Auralic Vega (see review). The Vega strikes me as another DAC that offers up superb detail yet in comparison to the Totaldac with PCM playback it sounded a bit processed and not as fully voiced in terms of tonal complexity. The Totaldac edges out the Vega playing back PCM data with a more natural and fluid sound and it even digs deeper into the recording. And that's an impressive feat I've only heard a small number of other DACs accomplish including the MSB The Analog DAC (see review) which also happens to be a R2R ladder DAC and the Meitner MA-2 (see review). And that's some mighty fine company indeed. With the Auralic Vega playing back DSD, the gap between it and the Totaldac closes somewhat and here's the interesting thing—the Totaldac in my experience closes the gap between PCM and DSD the tightest of the DACs I've heard here. Or to look at it another way, in my experience less costly DACs offer up some impressive performance when playing back DSD, equalling more expensive DACS playing back PCM. The other DACs I just rattled off also play back DSD and if I have one wish list item for the Totaldac it would be a checkmark next to DSD."
The important bit being, "in my experience less costly DACs offer up some impressive performance when playing back DSD, equalling more expensive DACS playing back PCM."
He is saying all this in relation to a $10K dollar discreet R2R dac. Sometimes reading reviews is reading dreary ad copy, and sometimes they let a little of what they think onto the printed page. This is one of those times.
-------------------------------------------------------------
Big speakers and little amps blew my mind!
Yes sometimes writers do let their true feeling slip into an article, other times you kinda have to read between the lines.
I've got a few Dacs, been trying to get some of what I 'think' is the best offerings in different Dac chips/designs. As far as expensive vs less expensive goes I can attest that some less expensive Dacs, namely R2R do indeed deliver the goods. I have a $2500 PCM1704 Dac next to my $600 PCM56P Dac, while they sound different from each other, the $600 Dac has a certain...as the Japanese would say 'Umami', a yummy savory deliciousness.
FWIW, most of the gear I own I bought sight unseen and unheard. It was from reading reviews [between the lines] and what was said, cross referencing reviews and looking at the design etc. After about 6-8 months of this I can be pretty confident in my purchase. For the most part I have gotten what I expected out of components, sometimes gear pass my expectations others 'few' its goes right back on the selling block quick.
Dynobots Audio
Music is the Bridge between Heaven and Earth - 音楽は天国と地球のかけ橋
That's a very large difference, relatively speaking.
Then it sounds like 20 bits is the ceiling for R2R, unless production gets really advanced, like nanotechnology advanced.
Thanks for all the information Tony. You brought up some issues I haven't heard before.
One last question. You saw the measurements of the Totaldac. You said it was probably over 20 bits. You didn't say anything negative about distortion, or crossover distortion. Linked is the technical page. Do you see any unacceptable effects from the low tolerance resistors? Are there enough measurements there to tell if it should perform as a decent dac, as opposed to a junk dac.
Thanks again. I'm probably going to have to turn in. Have a good night.
----------------------------------------------------------------
Big speakers and little amps blew my mind!
BB made 24 bit R2R dacs and so does MSB.
Thanks Morricab. It made me go back and look at the PCM1704. The highest sampling frequency is 96kHz, so there is still advancement possible. The MSB I just learned about.
-----------------------------------------------------------
Big speakers and little amps blew my mind!
Edits: 09/10/15
It's a NOS DAC. I would never say anything good about a NOS DAC that runs at 44 kHz. These are a defective way of playing back digital audio and do not output the waveform that was originally captured in the digital signal.
The photos show a "perfect" square wave, which is not a correct analog signal that can be encoded in a digital format. This shows the presence of imaging distortion in playback. If you had a waveform for a sine wave at high frequency shown over multiple cycles you would see a wavy envelop caused by beating between the true signal and the image.
The "noise" floor is probably the analog noise floor of the DAC with a silent digital signal. There are no plots of sine waves at varies levels, as would show harmonic distortion, no jitter plots, etc... In addition, there are no plots that show whether (how much) the noise floor varies as a function of the input signals being played.
A proper spec sheet or measurement review would have considerably more information than the two plots on the link you provided. Take a look at any of the Stereophile measurements in DAC reviews for what is missing. It looks like the plots shown here were "cherry picked" to be the best possible ones.
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
Yet my Audio-gd R2R dac playing cds sounds extremely good. This again shows to me that measurements seldom relate to how a component sounds to an individual listener
Alan
> " I would never say anything good about a NOS DAC that runs at 44 kHz."
Agree with the 44kHz sample rate, but I need more experience with NOS and ladder dacs. Your opinion is duly noted.
> " The photos show a "perfect" square wave, which is not a correct analog signal that can be encoded in a digital format."
That's something I've never heard anyone say before, but I've thought something along those lines for a long time. That testing a dac's or amplifiers performance using odd order harmonics probably is not the best way to do it.
> " If you had a waveform for a sine wave at high frequency shown over multiple cycles you would see a wavy envelop caused by beating between the true signal and the image."
This sounds like a better way to test dac's and amplifiers. If I ever posted any of this on the SET forum there would be an awful blood (mine) bath.
Today I looked for the resistor tolerance specs in the AD 5791 datasheet, but I couldn't find it. I did find the following;
> "The six MSBs of the 20-bit data-word are decoded to drive 63 switches, E0 to E62. Each of these switches connects one of 63 matched resistors to either the VREFP or VREFN voltage. The remaining 14 bits of the data-word drive the S0 to S13 switched of a 14-bit voltage mode R-2R ladder network."
That sounds to me like it's only 14 bits of R2R not the 20 bits advertised. Am I wrong on this, or is that the standard way?
Thanks.
---------------------------------------------------------------
Big speakers and little amps blew my mind!
:-)
Link below:
This guy says that a 12 bit PCM file will better a Nagra 15 ips reel to reel. Would that were true, but I'm afraid it's not. At least not as I have ever heard. He's got it all backed up by numbers, but they might not be the right numbers.
In this article once again everything is reduced to dynamic range capability and highest frequency sampled. I never see anything in print about the effect of an increase in samples per second. And I may have this all wrong.
----------------------------------------------------------------
Big speakers and little amps blew my mind!
Post a Followup:
FAQ |
Post a Message! |
Forgot Password? |
|
||||||||||||||
|
This post is made possible by the generous support of people like you and our sponsors: