High end all digital dsp crossover ? - Page 10 - 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
Reply
 
Thread Tools Search this Thread
Old 7th November 2012, 01:20 PM   #91
diyAudio Member
 
Join Date: Jun 2009
Quote:
Originally Posted by googlyone View Post
ADAU1442
Do you happen to know if a SigmaDSP biquad operating in double precision is 28 bit coefficients and feedback with a 56 bit accumulator, or 56 bit coefficients, feedback, and accumulate? The datasheet's ambiguous and I don't recall seeing anything that spelled it out in the SigmaStudio docs or over at the Analog EZ forums.

Quote:
Originally Posted by Barleywater View Post
Why do you insert a power amplifier, speaker, microphone, microphone preamplifier into signal path with DAC as device under test?
End to end system measurements seem the best way of verifying one is correctly composing individual components' characteristics when desiging to some overall performance target. Strangely, they're not often done in audio, so the result is often performance limits on the integration between components rather than the components themselves. I mentioned that particular case in this thread as I was looking at the interaction of digital volume control with power amplifier gain. The measurements are swept sine in HOLMImpulse with the DSP XO+EQ operating on 16 bit samples with double precision floating point internally and the DAC output going direct to the power amps.

That's more or less the standard/default/mainline configuration one gets with a PC crossover so it serves as a good example of how analog and digital components have to be designed to work together.
  Reply With Quote
Old 7th November 2012, 03:12 PM   #92
diyAudio Member
 
Join Date: Jan 2008
16bits? LSB spectrum is only -90dB from full scale. Why this limitation when working with electronics?

Andrew
  Reply With Quote
Old 7th November 2012, 07:32 PM   #93
diyAudio Member
 
Join Date: Jun 2009
The majority of audio data is in redbook format (or non-uprezzed compressed versions thereof) and the majority of DSP implementations do not increase the bits per sample on the fly---the primary exception I'm aware of is the recently discontinued Squeezebox line, though there are probably others. If the source is local and storage is unconstrained it's trivial enough to change up to 24 bit and avoid the problem that way. But that's often not an option when streaming internet radio or such. In a DIY DSP it's often trivial to output 24 (or 32) bit samples regardless of the input sample size in forward time processing (reverse time processing is more involved but not commonly used).

Upping the bit depth from 16 to 24 unlocks someplace between 0 and 5 additional bits of resolution depending on the DAC's analog performance and may or may yield benefit from the increased number of guard bits (depends on the filtering and gain structure and how they're implemented) but that wasn't what I happened to want to measure in that particular case. Few DACs have more than 19 bit resolution and even fewer implement 19 bit linearity so the sample size increase has a way of yielding somewhere around 15dB of improvement (depending on the DAC used and part to part variations in fab). So it's actually kind of hard to throw digital bits at the problem of excess power amp gain.
  Reply With Quote
Old 7th November 2012, 08:29 PM   #94
diyAudio Member
 
Join Date: Jan 2005
Location: Sweden
Is there any open source code for a dsp xo that would be useful if you actually
already had the hardware ?
  Reply With Quote
Old 7th November 2012, 09:13 PM   #95
diyAudio Member
 
Join Date: Nov 2006
Location: Adelaide, South Australia
Blog Entries: 20
Question: "Do you happen to know if a SigmaDSP biquad operating in double precision is 28 bit coefficients and feedback with a 56 bit accumulator, or 56 bit coefficients, feedback, and accumulate? The datasheet's ambiguous and I don't recall seeing anything that spelled it out in the SigmaStudio docs or over at the Analog EZ forums."

Yhea - the datasheets are enough to use the devices, but leave a fair bit of the internals either minimally explained or not at all. That said, I have found Analog Devices to be pretty helpful with queries.

You configure the SigmaDSP as either single or double precision in their "programming tool".
- Single precision is 28 bits - which all arithemitic operations within the biquads are done in 28 bit precision. I do not use this.
- Double precision uses 28 but data and coefficients and all arithmetic within the biquad calculations is done in 56 bit and truncated at the end of the calcs.

The 28 bit part is implemented to allow filters gain, so 24 bit input coefficients are represented as 0X0.xxxxxx (i.e. zero plus 24 bits).

If you put in 16 bit data the last 8 bits are zero - at least until you run the data through a filter!

While the idea of all data being in 56bit format might seem "better" the math of it points out the futility. 24 bit is 140 odd dB DNR... With this "double" bit depth I have not been able to see artefacts or results of coefficients in filters messing me around - though I have not gone to the extent of digidal domain analysis of the data.


Qusp: Not sure who it was that prompted me to this anymore - I do recall lending them my old AD1940 DSP development board through....
  Reply With Quote
Old 8th November 2012, 12:18 AM   #96
diyAudio Member
 
Join Date: Jun 2009
Quote:
Originally Posted by dahlberg View Post
Is there any open source code for a dsp xo that would be useful if you actually already had the hardware?
If you're targeting SigmaDSP parts SigmaStudio generates the necessary code. It's a free download. If you're targeting Cortex M4 look at ARM's CMSIS DSP library and examples. More generally, there are several open source implementations of the Bristow-Johnson cookbook.

Quote:
Originally Posted by googlyone View Post
That said, I have found Analog Devices to be pretty helpful with queries.
Yah, though I don't feel it's fair to take up Brett's time if I'm not planning to use the part. I've found low frequency biquads can require 16 or more guard bits in coefficients and feedback to maintain accurate Q0.23 output with the usual [ -0.25, 0.25 ) scaling---more here---so a 28 bit filter path with 56 bit accumulation is maybe not so attractive depending on the number of accurate bits one requires.
  Reply With Quote
Old 6th January 2013, 06:33 AM   #97
diyAudio Member
 
Join Date: Jan 2009
Send a message via MSN to Adrculda Send a message via Yahoo to Adrculda
I'm in the same boat..
was looking around and the most simple and "elegant" solution without having to code a ton of stuff was from AD. I've ordered up some samples and hopefully will have some free time to work up at least a rough schematics by mid week.
I'm going with ADAU1445 and ADAU1966

This Combo will open the door to a 16ch system, not that you will ever need to use all of them, but why not start with a solid stepping stone rather then with loose gravel.

it will be controlled with HiFiDuino through an Arduino Due. Will be using SPI as main protocol rather than I2C due to its speed.

Anyways...
I will most likely start a thread once i get the parts and have a schematic in hand
__________________
Regards,
Adrian
  Reply With Quote
Old 7th January 2013, 08:10 AM   #98
diyAudio Member
 
Join Date: Nov 2006
Location: Adelaide, South Australia
Blog Entries: 20
Adrculda,
if you want a kick start on this I am happy to help. If you look at my blog I have put all the CAD files and source code for the ADAU1442.

Happy to assist with schematic reviews and will happily try to asist with some of the oddities that I am sure you will bump into on the implementation for these chips.

That said, once running they have proven to be bullet proof and very stable.

You wont find that I2C / SPI speed is such a big deal - there is not a lot of code to be loaded, and unless you are planning to do some really funky things with real time changes to coefficients, the amount of data you send to configure filters etc is pretty minor. If you have either an I2C or SPI bus logger, then go for that protocol!

Have fun!
  Reply With Quote
Old 7th January 2013, 08:21 AM   #99
diyAudio Member
 
Join Date: Nov 2006
Location: Adelaide, South Australia
Blog Entries: 20
By the way...
Just this weekend gone I connect my ADAU1442 DSP to my computer using the optical interface (in and out).

The optical output from the computer into the DSP worked just fine, but buggered if I could get the audio from the DSP into the computer.

The weird thing was that I could hook two of the ADAU1442 DSP's together (optical out to optical input) and things worked just fine - but nogo in sending data to the PC!

After several hours of self doubt and head scratching I finally dug into the DSP configuration code, then into the datacheet, which led to an errata from Analog Devices which I would paraphrase as:

"Ooops. Some pillock has set the SPDIF 'V' bit high on ADAU1442 serial data out, and done it in hardware. Every sane piece of audio hardware will ignore any data we send out. Our bad, Sorry."

Which translates to the SPDIF data output from this IC having the Data Valid bit set to "INVALID" meaning that most devices ognore the data and go into mute. Bugger.

Ardculda - just check this with the ADAU1445...
  Reply With Quote
Old 7th January 2013, 01:58 PM   #100
qusp is offline qusp  Australia
diyAudio Member
 
qusp's Avatar
 
Join Date: Oct 2009
Location: Brisbane, Australia
Quote:
Originally Posted by googlyone View Post
Qusp: Not sure who it was that prompted me to this anymore - I do recall lending them my old AD1940 DSP development board through....
not me then, no I just remember sending you a PM maybe a year ago asking if the design could be bent to some ESS dacs.

but no I havent absconded with your eval board =)
  Reply With Quote

Reply


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
DSP StudioX: active/digital crossover simulator with features... Under construction. s3tup PC Based 11 22nd November 2012 06:03 PM
High End digital 50W amplifier board sunlight009 Vendor's Bazaar 0 19th March 2012 03:12 AM
High quality DSP crossover setup on a budget? Thunau Multi-Way 2 2nd November 2007 10:55 AM
A high-end digital amplifier for everyone DAXgroup Digital Source 82 30th June 2007 03:16 PM
High end digital cable MichaelJHuman Everything Else 1 30th August 2005 02:41 AM


New To Site? Need Help?

All times are GMT. The time now is 10:43 AM.


vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2014 DragonByte Technologies Ltd.
Copyright 1999-2014 diyAudio

Content Relevant URLs by vBSEO 3.3.2