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 22nd January 2004, 12:02 PM   #21
Ken L is offline Ken L  United States
diyAudio Member
 
Join Date: Sep 2002
Location: deep south
Default Thanks Steve

For your reply to my question

and Thuneau for your comments on the Rane

Regards

Ken L
__________________
No longer powered by Linux - not enough apps and cross platform integration - but maybe one day
  Reply With Quote
Old 22nd January 2004, 03:47 PM   #22
Thunau is offline Thunau  United States
diyAudio Member
 
Thunau's Avatar
 
Join Date: Mar 2001
Location: PA USA
Quote:
Originally posted by Vadim
Jan,

Yes I can see that. It is applying the signal cut before the D/A becomes a problem How do we do that?
Yes, the Behringer software is not very flexible in that regard. I was hoping that the new revision would include some more gain elements. You are limited to 15dB of gain reduction. If you consider the origin/market of the unit then it is clear that it's enough for 99% of the users. Most will leave them at 0 anyway.
The unit was meant to replace a rack full of analog crossovers, EQs, delays and limiters that used to cost as much as a nice car. In that regard it accomplished a lot. But the fact is that it's not the best choice for a loudspeaker processor.

Quote:
I do not see that using the digital input solves the problem. Perhaps I am missing something here. I plugged in my trusty Denon CD player to DCX using digital input-output ports and I observed input level meters on the DCX are doing about -3 dB maximum no matter what program material I throw at it. Now, it is my understanding that I am sending an 16-bit encoded signal to the DCX.

Anyway, I see no input clipping, - good. Now, if I introduce a shelving low-pass filter with 10 dB of gain I clearly see the output meters show clipping, - no good. Now, to fix this I attenuate the output digitally by 10 dB, and the DCX allows you to do this. The clipping goes away, - great. However, now what I have done is I reduced the signal resolution by almost 2 bits.
Not really. Once the signal is processed internally it' precision becomes 24 bit. Notice that I said precision and not resolution because the resolution is limited by the source (16 bit). By processing with high precission, you preserve the resolution of the original. You reduced the signal digitally by 15 dB only to prevent clipping. Your D/A is maximized - it will reproduce audio over it's entire dynamic range. Due to high precission DSP your resolution is preserved.

Quote:
Considering that the DCX measures a little better then 16 bits in the real world environment this is a problem for me. I know that in the lab with A-weighting the DCX is specified at better then 18 bits, but in my lab I only see 16 bits and even that is rather amazing. Anyway, a loss of 2 bits is significant, don’t you think? Well, am I wrong about all this?
As long as you get a full 16 bit resolution out of your converters you are in good shape. We tend to concentrate on the numbers a bit too much. Consider the dynamic range of the actual music you listen to. If the producers and engineers are very careful you are lucky to get 40dB-50dB dynamic range of the music on the CD. That's really, really good. Most stuff clocks in at less than 20dB, and the top 40 is usually about 10dB. If you watch the meters and play some modern Rock you'll notice that they move in a 6 dB window 90% of the time. From my recording experience I can tell you that even a $100.000,00 analog mixing console with 40 channels open hums away at -60dB. Once the tape starts rolling it gets worse.

[/QUOTE] I understand that the internal processing of SHARKs has nearly infinite headroom, but how does it help me here. I think what might be happening is that the D/A is a delta-sigma type and as such it must collect all the charge pockets and convert then to analog voltage with an internal op-amp. This op-amp has defined saturation levels and fixed power rails and so the signal is hitting those levels due to its large amplitude. I do not see how we can help this.

With this in mind, although the DCX gives an ability to introduce at most 15 dB of gain, which is a significant amount, it is not advisable to use this facility liberally for the fear of possible clipping. The nature of dipole EQ requires the opposite. Perhaps this is why the DCX also provides limiters. I don’t know, perhaps I am wrong about this, but so far this is what I think.[/QUOTE]

There is more than one way to skin a cat. You can combine for example a 6dB boost with a 10 dB cut a bit higher up and achieve the same target only offset in level which will make the clipping problem much easier to deal with. It's all about gain staging. People who grew up with daisy chaining multiple analog processors understand this subconciously. I would think that Linkwitz offsets the gains in his Orion crossover, otherwise he runs a risk of clipping the LF amps. You can do similar things in digital domain. With a measurement setup on your computer and a bit of play with the parametric EQs you should be able to approximate the target very closely.

Quote:
Well, the all pass circuits are analog delay lines and the DCX does provide them. I am not sure what the cross-feeding circuits are.
All pass filters are not quite a pure delay. The delay varies with frequency. The crossfeeding circuit I was referring to is a subtractive crossover, where the transfer function of LP is subtracted from the signal to create a HP (with the help of an allpass filter). I just used a term that popped in my head because I couldn't think of the proper name at the time.



Another fairly easy solution to the gain problem would be to convert the 16 bit signal to 24bit and attenuate it before it hits the AES input of the DCX. You could achieve a decent headroom before loosing resolution - about 40dB. That in combination with the output attennuators would let you achieve huge boosts before running the risk of clipping or uder-utilizing the D/A converters. There should be an easy way to do this. I'll have to research it a bit.

On another note, I have access to the Behringer box below dealer cost, but decided to make an active crossover with opamps (OPA2134) and good passive parts. I got the circuit boards from ESP. It took a bit of work but with the help of lspCAD pro I had a good sounding design before buying the parts. The final product sounds quite a bit better than the crossover emulator through my Alesis AI-3 8 ch. converter, eventhough the transfer function is nearly identical (within the parts tolerance).
__________________
"Most people just say what they know, the wise ones know just what to say."
  Reply With Quote
Old 23rd January 2004, 02:13 AM   #23
Peter M is offline Peter M  United States
diyAudio Member
 
Join Date: Oct 2002
Location: Cambridge, MA
Quote:
Again the answer is kind of. The Orion as it stands using the Peerles XLS needs a boost in the bass that reaches 40dB below 10Hz. This what SL feels is necessary for flat response and low group delay. 15dB of this is due to the low Q of the XLS10....

IMHO, you would be better off using a different bass driver, or crossing to a monopole sub below 40Hz as SL now recommends.
Steve,

Based on this, it seems as if, to me, much of the problem of digital equalization can be solved simply by deciding not to use Linkwitz's driver choice for the Orions, and instead deciding to use more capable drivers with a higher displacement and a higher Qts values, even if it at this point stops being the Linkwitz Orion. I'm very ignorant when it comes to the actual theory, but how do you think the DCX 2496, or even the dbx and Rane units, would fare using drivers like the Stryke AV12 or other high-excursion 10/12" drivers with higher Qts values in place of the Peerless XLS 10"? Just two AV12s yield 30% more displacement at 30% lower cost than the XLS 10"s, but I don't know how much its .452 Qts changes the need for a 40 dB boost in the bass. Does this seem like a plausbile idea, rather than playing around with digitally EQing the original drivers?

Peter
  Reply With Quote
Old 23rd January 2004, 03:29 AM   #24
diyAudio Member
 
Join Date: Sep 2002
Location: Australia
That's what I have done. Personally I wouldn't use the Peerless drivers with any of the digital crossovers.

I use twin Adire DPL12s a side and get flat measured response down to 20Hz with a total boost of 18dB.

The AV12 would also be a good choice due to the very high Xmax.

One other point to note, depending on how you wire them up, is that these drivers will most likely be more efficient than the single mid driver of the Orion. This means you will need to lower their level in the crossover. This lower overall level helps avoid clipping the processor. In other words, if the bass level is 6dB lower than the mids and you boost it 18dB the gain is really only 12dB.

Incidentally, the AV12 has something like three times the displacement of the XLS 10 by my calculations.

Cheers

Steve
  Reply With Quote
Old 23rd January 2004, 07:43 PM   #25
Vadim is offline Vadim  Canada
diyAudio Member
 
Join Date: Jun 2003
Location: Toronto
Jan,
Quote:
Once the signal is processed internally it' precision becomes 24 bit. Notice that I said precision and not resolution because the resolution is limited by the source (16 bit). By processing with high precission, you preserve the resolution of the original. You reduced the signal digitally by 15 dB only to prevent clipping. Your D/A is maximized - it will reproduce audio over it's entire dynamic range. Due to high precission DSP your resolution is preserved.
This is interesting! If understood you, - you are making a clear distinction between the signal resolution, and processing precision, right? Now, the signal resolution is of course 16 bits if AES input is used and theoretically 24 bit if input A/D converter is used, right?

Let us assume for a moment that the analog input is being used. Let us also assume that I take care of the input signal amplitude by keeping it as high is possible around 8-10 Volts, so then the performance of A/D converters is maximized. Now am I to conclude from your post that after I used the EQ any subsequent digital reduction of the signal amplitude has no bearing on the signal resolution? Am I to think that the output of D/A converer will still deliver the same resolution that the signal was digitized with regardless of the attenuation?
Quote:
You can combine for example a 6dB boost with a 10 dB cut a bit higher up and achieve the same target only offset in level which will make the clipping problem much easier to deal with. It's all about gain staging. People who grew up with daisy chaining multiple analog processors understand this subconciously. I would think that Linkwitz offsets the gains in his Orion crossover, otherwise he runs a risk of clipping the LF amps. You can do similar things in digital domain. With a measurement setup on your computer and a bit of play with the parametric EQs you should be able to approximate the target very closely.
I must say that I am at a loss here, as I did not grow up with “daisy chaining multiple analog processors”. Let me try an example just to see if I understand what you are saying.

First, let’s assuming about 8-10 Volt analog input. Now, let us say I need to create a shelving low-pass filter that would span about 30 dB from 20 Hz to 300 Hz, that is it must be 30 dB at 20 Hz and 0 dB at 300 Hz. Naturally, this is possible to do with DCX2496 by using the entire +15 dB and -15 dB range.

I will end up with +15 dB at 20 Hz and -15 dB at 300 Hz, if I use 2 filters combined to achieve the approximate shelving low-pass curve I need. The first filter is defined as 6dB/oct low-pass at 20 Hz with +15 dB of gain and the second filter is defined as 6 dB/oct shelving hi-pass at 300 Hz with -15 dB of gain. The overall response now looks like one broad shelving low-pass filter from 20 Hz (+15dB) to 300 Hz (-15 dB).

Now I see that at around 20 Hz I will definitely clip the signal so I attenuate the output digitally by let us say 10 or 12 dB to prevent clipping. Would that be a correct gain staging procedure with no loss in resolution?

Vadim
  Reply With Quote
Old 24th January 2004, 12:46 AM   #26
Thunau is offline Thunau  United States
diyAudio Member
 
Thunau's Avatar
 
Join Date: Mar 2001
Location: PA USA
Quote:
Originally posted by Vadim
Jan,

This is interesting! If understood you, - you are making a clear distinction between the signal resolution, and processing precision, right? Now, the signal resolution is of course 16 bits if AES input is used and theoretically 24 bit if input A/D converter is used, right?
You're on the right track. The one exception is that the analog signal is still 16 bit (or less) if it originated at the analog outputs of a CD Player. If you plug in a pure analog source, like the outputs of a mixing console running a bunch of mics then you're getting a 24 bit representation of a signal. That signal might have a S/N ratio of only 60dB but you're sampling it with 24 bit resolution. Of course the A/D converter has it's own noise floor and harmonic distortion (sonic sinature) which adds to that of the input signal. If you plug your SACD outputs in a tube pre and that combination sounds great to you, then by any means you should use the analog inputs of the crossover. Just be aware of the fact that you're superimposing the sound of those cheap converters on your precious signal.

Quote:
Let us assume for a moment that the analog input is being used. Let us also assume that I take care of the input signal amplitude by keeping it as high is possible around 8-10 Volts, so then the performance of A/D converters is maximized. Now am I to conclude from your post that after I used the EQ any subsequent digital reduction of the signal amplitude has no bearing on the signal resolution? Am I to think that the output of D/A converer will still deliver the same resolution that the signal was digitized with regardless of the attenuation?
In case of a floating point processor, yes, as long as you utilize the full range of your D/A converter. You would have to attenuate digitally by over 20 dB to get back to 16 bit resolution though. We are starting with a theoretical 120dB dynamic range of the D/A converter here.

Quote:
I must say that I am at a loss here, as I did not grow up with “daisy chaining multiple analog processors”. Let me try an example just to see if I understand what you are saying.

First, let’s assuming about 8-10 Volt analog input. Now, let us say I need to create a shelving low-pass filter that would span about 30 dB from 20 Hz to 300 Hz, that is it must be 30 dB at 20 Hz and 0 dB at 300 Hz. Naturally, this is possible to do with DCX2496 by using the entire +15 dB and -15 dB range.

I will end up with +15 dB at 20 Hz and -15 dB at 300 Hz, if I use 2 filters combined to achieve the approximate shelving low-pass curve I need. The first filter is defined as 6dB/oct low-pass at 20 Hz with +15 dB of gain and the second filter is defined as 6 dB/oct shelving hi-pass at 300 Hz with -15 dB of gain. The overall response now looks like one broad shelving low-pass filter from 20 Hz (+15dB) to 300 Hz (-15 dB).

Now I see that at around 20 Hz I will definitely clip the signal so I attenuate the output digitally by let us say 10 or 12 dB to prevent clipping. Would that be a correct gain staging procedure with no loss in resolution?

Vadim
Yes, you pretty much got it. Except the 6dB/octave filters work out to:
+15 @ 20Hz
+9 @ 40Hz
+3 @ 80Hz
-3 @ 160Hz
-9 @ 320Hz
-15 @ 640Hz
You do get a nice graphical representation of the curve in the Behringer software. Just slide around until you get a nice smooth curve and you're done. Use the parametrics to smooth out room related gremlins. I don't think you will need full 30dB of boost. And since there is not much loud information in the very bottom octaves in most music, it's likely that you'll end up with much less attenuation then 15dB to prevent clipping. Movie soundtracks could prove different though. Play with it and have fun.
__________________
"Most people just say what they know, the wise ones know just what to say."
  Reply With Quote
Old 24th January 2004, 03:39 AM   #27
diyAudio Member
 
Join Date: Sep 2002
Location: Australia
In Vadim's example above, how much would you need to reducde the treble and mid levels to match this kind of boost and what would that do their resolution?

Cheers

Steve
  Reply With Quote
Old 24th January 2004, 12:49 PM   #28
Thunau is offline Thunau  United States
diyAudio Member
 
Thunau's Avatar
 
Join Date: Mar 2001
Location: PA USA
Quote:
Originally posted by sfdoddsy
In Vadim's example above, how much would you need to reducde the treble and mid levels to match this kind of boost and what would that do their resolution?

Cheers

Steve

That depends on the gain of amps used, individual driver's sensitivity and their EQ. Assuming that you have no other way of matching levels then the crossover, you might end up running out of available attenuation inside the DCX. If your crossover provides more range than that you should start around -15dB and tweak from there. Personally I would run the mids and tweeters wide open in the crossover with analog attenuators on the inputs of the amps. But even with the outs attenuated in digital domain as long as your input is nice and hot you should end up with good resolution given 24 bit D/A converters. What you are giving up is S/N ratio rather than resolution. Again, it's the whole analog gain staging issue. Your converters are humming away at -90dB. If you limit the maximum level you can get out of them to -20dB (bu using the digital attenuator) then your S/N becomes 70dB. But, if you utilize the whole D/A range (by not attenuating the signal digitally) and attenuate analog at the amp, you're pushing the D/A noise floor down together with the signal and preserve the S/N at the line level. The amp itself has it's own S/N and you can't do much about it other than using one that is properly sized for the application. In case of the Orion a lot of power goes toward LF EQ so the LF amp should be much more powerful than the mid and treble amp. If you're using a 200W amp for lows you will clip it before a 40W amp on mids or highs.
Remember that you need 10 times the power to gain 10dB. You are using a lot of power to correct the LF- about 12dB to 18dB at the prominent bass frequencies. So, use relatively small amps for mids and treble and preserve the dynamic range and resolution of your music.
__________________
"Most people just say what they know, the wise ones know just what to say."
  Reply With Quote
Old 24th January 2004, 02:36 PM   #29
Ken L is offline Ken L  United States
diyAudio Member
 
Join Date: Sep 2002
Location: deep south
Default _very_ interesting thread, guys

Quote:
Originally posted by Thunau

My money/hopes are on Multimedia PC though. A quiet Windows or Linux box with an RME interface running custom DSP program with all your CDs ripped to the hard drive and controlled wirelessly from a palm PC. It could live somewhere in the basement or a closet and be connected to the Converter by an optical cable
I think you're the _smart money_ on this one. I was told that Terry Cain of C&C speakers bought a similar unit for use in demonstrating his speakers at VSAC and CES. Used a laptop and was playing CD's that had been ripped to the hard drive. If I remember correctly it cost about $4500 US.

The vinyl guys that heard it said that it sounded like the best vinyl you have ever dreamed of - or words to that effect.

It's gonna have to come down a lot in price over time before it will be viable for me _big grin_.

The computer part should be easy to do and the DAC/ADC part does seem within the purvue of DIYer's with skills in that area.

I am sincerely appreciative of the contributors to this thread. It takes a certain amount of effort to make technical posts of this length. I am learning a lot reading it.

Best regards

Ken L
__________________
No longer powered by Linux - not enough apps and cross platform integration - but maybe one day
  Reply With Quote
Old 25th January 2004, 09:45 PM   #30
Vadim is offline Vadim  Canada
diyAudio Member
 
Join Date: Jun 2003
Location: Toronto
Jan,

I posted:
Quote:
“…Let us assume for a moment that the analog input is being used. Let us also assume that I take care of the input signal amplitude by keeping it as high is possible around 8-10 Volts, so then the performance of A/D converters is maximized. Now am I to conclude from your post that after I used the EQ any subsequent digital reduction of the signal amplitude has no bearing on the signal resolution? Am I to think that the output of D/A converer will still deliver the same resolution that the signal was digitized with regardless of the attenuation?”
you replied:
Quote:
“…In case of a floating point processor, yes, as long as you utilize the full range of your D/A converter. You would have to attenuate digitally by over 20 dB to get back to 16 bit resolution though. We are starting with a theoretical 120dB dynamic range of the D/A converter here…”
You know, I have been thinking about this and it does not sit well with me. I asked myself a question, - what happens to S/N and consequently the resolution of the signal when we digitally amplify it? Well, actually nothing! The signal is still characterized by same number of bits and consequently the same S/N ratio. All we have now is greater amplitude. Reason for this is that when we linearly amplify the signal, - we also amplify the noise encoded in the same digital word.

Now, the question is what happens when we want to attenuate the signal? I must say that I do not know how to accomplish such an operation without affecting the resolution of the signal. Attenuation must result in a bit reduction of the digital word. Because when we scale down the signal we also are attempting to scale down the noise, but the noise being stochastic (non-deterministic) in nature will resist that and its amlitude will (or perhaps should) remain the same.

So, I think a dithering technique is in order here if we were to attempt to accomplish amplitude reduction without any significant bit reduction. I somehow doubt that Behringer programmers did that.

Now, that although the A/D section is 24 bit capable, we naturally do not get the 24 bits if the signal originates at the analog outputs of a CD Player. Also, I am sure that there is no way for that 24 bit A/D converter to actually do 24 bits in a real world environment. Perhaps, if we were to place this converter inside a Faraday box (a completely electromagnetically screened metal enclosure), then maybe we can measure 24 bits, - and that is a much stretched ‘maybe’. Inside your equipment rack you will get 16 bits at best out of this 24 bit internal architecture.

With all this in mind I still think that we cannot afford to do any digital attenuation. I need to go back to my DSP books and look at exactly how to do digital attenuation. So I am not 100% sure that I am right about this, but …

Also you are correct that 6 dB/oct filter does not work as I described in my example. I should have simply called it a Shelving Low-Pass filter. Such 20 Hz to 300 Hz filter is possible to realize using a first order network. I should have been more accurate in my description.

Vadim
  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
Finally finished my orions srfranci Multi-Way 5 24th July 2009 11:29 AM
My week with the Orions, or 'why do we bother' cuibono Multi-Way 56 26th October 2008 12:51 AM
How hard to clone the Orions? Chaucer Multi-Way 46 8th July 2005 02:01 PM
Kls-9 Woofer Sustitute dinamic1 Multi-Way 0 13th March 2002 09:29 AM


New To Site? Need Help?

All times are GMT. The time now is 08:04 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