Best approach - diyAudio
Go Back   Home > Forums > Source & Line > Digital Line Level

Digital Line Level DACs, Digital Crossovers, Equalizers, etc.

Please consider donating to help us continue to serve you.

Ads on/off / Custom Title / More PMs / More album space / Advanced printing & mass image saving
Thread Tools Search this Thread
Old 12th October 2013, 07:45 AM   #1
diyAudio Member
Join Date: Oct 2013
Location: Sweden
Default Best approach

I am new to the forum and am fascinated by all the great threads! I am planning to buld a pair of active 3 way speakers and am trying to decide on the best approach. The objective is to aim at the best accuracy in terms of amplitude and phase distortion, dynamic behaviour and noise (guess that's pretty obvious!). A couple of thoughts:

> It would seem smart to have a dsp crossover and power amp located close to each speaker and that the signal is received via a digital link (e.g. Toslink) in order to minimise any unwanted effects from long cables. Opinions?

> Another consideration is how to control the volume in a setup using dsp crossovers. There is a volume control on the MiniDSP for example but since most of the time the volume is say 10%, wouldn't this result in undue quantisation noise? Maybe a better idea (although awkward) would be to let the DSP run just below full range (i.e. no clipping) and then have an analogue volume control for each speaker channel. Opinions?
  Reply With Quote
Old 12th October 2013, 12:09 PM   #2
diyAudio Member
Avro Arrow's Avatar
Join Date: Sep 2010
Location: Toronto, eh
Something like a pga2320 can be daisy chained to give you as many channels as you need.
  Reply With Quote
Old 12th October 2013, 04:14 PM   #3
diyAudio Member
Join Date: Oct 2013
Location: Sweden
Looks like a good part. Low THD but only really at good signal levels (0,001% @ 1V RMS but this seems to grow by a factor of 10 for each factor 10 reduction in the input signal amplitude).
  Reply With Quote
Old 13th October 2013, 01:31 PM   #4
diyAudio Member
abraxalito's Avatar
Join Date: Sep 2007
Location: Hangzhou - Marco Polo's 'most beautiful city'. 700yrs is a long time though...
Blog Entries: 123
Send a message via MSN to abraxalito Send a message via Yahoo to abraxalito Send a message via Skype™ to abraxalito
If you're really after the lowest possible noise, then a passive line-level XO is the way to go, after a single DAC. A digital processor crossing over into separate DACs will give more noise because each operates in a wider (i.e. at least the full audio) bandwidth. The passive filter will bandwidth limit the input signal whilst adding minimal noise of its own.
When irony first makes itself known in a young man's life, it can be like his first experience of getting drunk - Robertson Davies
  Reply With Quote
Old 13th October 2013, 04:41 PM   #5
diyAudio Member
Join Date: Jun 2009
Originally Posted by Willafizz View Post
It would seem smart to have a dsp crossover and power amp located close to each speaker and that the signal is received via a digital link (e.g. Toslink) in order to minimise any unwanted effects from long cables. Opinions?
A balanced send from the DSP over to the power amps works too. Some care is wanted when co-locating the DSP+DAC and the amps; when a single secondary is shared the bounce between DAC and power ground can end up larger than when separate secondaries are used. Designs with sufficient CMRR for this tend to be fine with a DSP+DAC+triamp at one speaker and just a triamp at the other.

If you run the level budget for your playback volumes and power amp gains that'll tell you whether there'll be sufficient bit depth utilization in the DAC for digital volume to be attractive versus analog volume. We can comment more specifically if you share the budget but, generally speaking, if all the output bits are used for signal or volume things are usually fine. Sometimes you can squeeze in a couple bits for level management too, though it's preferable to get the power amp gain about right. In scenarios with high dynamic range requirements a level pad of 20dB or so between the power amp's balanced receive and output stage can be an attractive alternative to analog volume control.

You've already noticed the THD+N limit on the PGA2320's noise floor, a behaviour which applies to all circuits. The part's 10uV typ, 17.5uV max output noise is in the range typical of audio circuits. The easiest way to get larger SNR and hence a lower THD+N limit is to use higher signal levels. Instrumentation amp type topologies can be attractive as their buffering allows use of lower impedance feedback networks to reduce noise. However, managing temperature coefficients to avoid the distortion introduced by resistor mistracking rising above the sub-microvolt noise floor you're trying to create is harder than it looks. It's also nontrivial to obtain sufficient CMRR to hold ground bounce residuals below a microvolt. Backing off and designing in a world where tens of microvolts happen can make life significantly easier.
  Reply With Quote
Old 13th October 2013, 07:59 PM   #6
diyAudio Member
Join Date: Oct 2013
Location: Sweden
Thanks for all the responses! I spent some time reading a very interesting article from Standard stuff in a way but a crystal clear presentation and a lot of myth busting. The article shows that 16 bit resolution is more than adequate for the human ears. From this, I can imagine that a 'digital volume control' in a 24 bit system would be fine.
  Reply With Quote
Old 14th October 2013, 03:53 AM   #7
diyAudio Member
Join Date: Jun 2009
True. However, there's the minor problem there are no 24 bit systems. Delivering 24 accurate bits in DSP has a way of requiring 64 bit CPUs or DSPs with unusually wide data paths such as the NRND TAS3108. 32 bit platforms like the SHARC and M4 can mostly do it given careful coding. 28 bit parts like the SigmaDSPs MiniDSP uses, not as much. The DSP is relatively easy compared to the analog side, though. The highest dynamic range DAC is the 135dB ESS ES9012. That's a resolution of 22.4 bits and, if you have a look at ESS's eval board and related collateral, you'll see getting the power supply and output buffer noise floor low enough to yield that resolution is nontrivial. Maintaining that 167nV noise floor across a board to board or chassis to chassis connection is---surprise!---also difficult.

As an example, consider a link with a +-14V signal swing and 14uV of audio band RMS noise. This is about the largest swing you can get from op amps operating at the practical maximum of +-15V supplies without introducing clipping distortion. It's also about the best one can do with low noise op amps and optimized filters without resorting to thermally challenging implementations or parts which are outright expensive rather than merely pricey. A 14V signal over 14uV noise is 120dB DNR, which is a smidge under 20 bit resolution.

Careful design and implementation will yield a couple more bits in the link but, as most top end audio DACs are 20 or perhaps 21 bit the merits are debatable. There's also a tendency for the signal swing on the link to be a couple bits less than optimal, so a pragmatic approach is to figure you can get about 19 bits from the DAC to the power amp input. Usually one needs 20 to 25dB of volume control, which is four bits. Ergo, decent digital volume design offers 15 to 19 bits to the speaker. Call it 17 bits nominal.

If, as is often the case, the power amp gain is 25 or 30dB too high that consumes 4 or 5 more bits in level matching. So one ends up with something like 11.5 bits nominal. It's my experience this is easily noticed subjectively and easily measured objectively. Would 14 bits be enough? Possibly, depending on the material, the listener, and the equipment. It's my subjective experience 15 or 16 bits is preferable, though objectively I start hitting problems with ambient noise floors from sources the better part of a kilometer away even when getting up at 4 in the morning to measure on windless nights.

Hence the importance of implementing proper gain structure when using digital volume control. Analog volume does mitigate poor gain structure but it's still hard to get noise floors under 1uV.
  Reply With Quote


Hide this!Advertise here!
Thread Tools Search this Thread
Search this Thread:

Advanced Search

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off

Similar Threads
Thread Thread Starter Forum Replies Last Post
The Meridian Approach tim pattinson Digital Source 5 22nd November 2007 06:39 AM
rational approach traw Multi-Way 2 15th August 2006 07:39 AM
A different sub approach, will it work? Zero One Subwoofers 15 1st May 2006 03:06 AM
here's my approach hitsware Solid State 2 13th November 2004 04:54 AM
A less used approach SuppersReady Pass Labs 18 29th January 2003 11:16 PM

New To Site? Need Help?

All times are GMT. The time now is 03:10 AM.

vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2015 DragonByte Technologies Ltd.
Copyright ©1999-2015 diyAudio

Content Relevant URLs by vBSEO 3.3.2