AudioNervosa

Specialists => Audiologists => Topic started by: HAL on August 09, 2017, 05:31:28 PM

Title: Digital speaker crossovers and room correction discussion
Post by: HAL on August 09, 2017, 05:31:28 PM
Is there any interest in digital crossovers for speakers and digital room correction discussions?

Been working with both for many years and will be happy to discuss ideas and equipment tried along the way.

 
Title: Re: Digital speaker crossovers and room correction discussion
Post by: Nick B on August 09, 2017, 09:19:24 PM
I'd welcome your insights and experience
Nick
Title: Re: Digital speaker crossovers and room correction discussion
Post by: HAL on August 10, 2017, 06:14:28 AM
So far the list of DSP based crossovers has been:

Behringer DCX2496 (own)
DEQX 2.6P (own) (and newer listening to demos)
MiniDSP (listening to demos)
Danville Signal Processing dspMusik LCD (own)
Perpetual Technologies P-1A for SOCS (own)

Room correction:

MathAudio RoomEQ (own)
Dirac (listening to demos)

Room measurements:

Room EQ Wizard (use)
CLIO (own)
CLIO Pocket (own)
MMK speaker measurement system (co-developed)

Presently the best sounding of the group to me with the Megalith speakers here and others tried is the dspMusik and RoomEQ.  Both have easy setup capabilities and very flexible.   

Only takes very simple measurements to get RoomEQ working and sounding very good in my room.

Not heard everything, but have heard other systems as well at shows.   

To me the most audible part of great DSP crossovers is the ability to capture the soundfield with very accurate channel to channel crossover matching that is extremely difficult with passive components due to their tolerances.  When accurate time delays are needed, it is very easy to add.

I use room correction as part of room tuning.  First are the acoustic treatments (absorption and diffusion) to correct for the normal room bounce problems and bass absorption.  Then use RoomEQ as the last step to integrate the speakers and room.  Done this way has given excellent results in my system.

Hope this starts a conversation.
Title: Re: Digital speaker crossovers and room correction discussion
Post by: rollo on August 10, 2017, 08:54:23 AM
Hey Hal glad to see you contributing. You are expert enough, thanks for the input.


charles
Title: Re: Digital speaker crossovers and room correction discussion
Post by: dBe on August 16, 2017, 11:34:49 PM
Great to see you posting. You have the Window applications wired!

Cool, let's see what drives this method.

You know that I tried for several years to get a Windows machine sounding the way I wanted.  Still have it.  Lead on Dartanian!!!
Title: Re: Digital speaker crossovers and room correction discussion
Post by: Nick B on August 17, 2017, 12:57:57 AM
So far the list of DSP based crossovers has been:

Behringer DCX2496 (own)
DEQX 2.6P (own) (and newer listening to demos)
MiniDSP (listening to demos)
Danville Signal Processing dspMusik LCD (own)
Perpetual Technologies P-1A for SOCS (own)

Room correction:

MathAudio RoomEQ (own)
Dirac (listening to demos)

Room measurements:

Room EQ Wizard (use)
CLIO (own)
CLIO Pocket (own)
MMK speaker measurement system (co-developed)

Presently the best sounding of the group to me with the Megalith speakers here and others tried is the dspMusik and RoomEQ.  Both have easy setup capabilities and very flexible.   

Only takes very simple measurements to get RoomEQ working and sounding very good in my room.

Not heard everything, but have heard other systems as well at shows.   

To me the most audible part of great DSP crossovers is the ability to capture the soundfield with very accurate channel to channel crossover matching that is extremely difficult with passive components due to their tolerances.  When accurate time delays are needed, it is very easy to add.

I use room correction as part of room tuning.  First are the acoustic treatments (absorption and diffusion) to correct for the normal room bounce problems and bass absorption.  Then use RoomEQ as the last step to integrate the speakers and room.  Done this way has given excellent results in my system.

Hope this starts a conversation.


Rich,
You mentioned using room eq as the last step. Wouldn't excellent software be able to correct for deficiencies without using acoustic treatments? Also, I was on the Amarra site tonight which is Windows compatible. It seems to be a Swiss army knife of sorts. Your thoughts on it?
Nick
Title: Re: Digital speaker crossovers and room correction discussion
Post by: richidoo on August 17, 2017, 03:24:10 AM
Room correction only fixes the spot where you measure. Treatments affect the whole room.
Title: Re: Digital speaker crossovers and room correction discussion
Post by: HAL on August 17, 2017, 06:46:13 AM
To me room correction with treatments helps to minimize the need to use software correction.   Have heard both methods with full room treatment only and systems like Tact/Lyngdorf without room treatment, but it sounds more natural to me with treatments and then correction.

The correction systems have a limit as to how long they can correct the response.  Any room artifact outside that will not be corrected.  This is another reason to minimize the room artifacts to begin the process.

I have not used Amarra, so cannot comment.  Others may have and would be good to see their comments.  I run Foobar2000 with RoomEQ as the player and room corrector.  Keeping the user interface as simple graphics is good from my experience. 
Title: Re: Digital speaker crossovers and room correction discussion
Post by: Nick B on August 17, 2017, 12:21:21 PM
Thanks, Rich(es),
As you are both of the room correction first method, that works for me. I'll order a mic and do the testing. 
Nick
Title: Re: Digital speaker crossovers and room correction discussion
Post by: steve on August 19, 2017, 07:24:35 PM
How fine of adjustment can be had with a digital crossover?

Cheers

Steve
Title: Re: Digital speaker crossovers and room correction discussion
Post by: HAL on August 19, 2017, 07:30:14 PM
Any of the crossover types can be adjusted for Fc in 1Hz increments.
Title: Re: Digital speaker crossovers and room correction discussion
Post by: steve on August 19, 2017, 09:00:10 PM
Any of the crossover types can be adjusted for Fc in 1Hz increments.

I would assume that amplitude adjustments are almost always continuous? Do they have something similar to a zobel network adjustment, but digital? If so, is there a digital correlation to capacitor and resistor adjustments?

Cheers

Steve
Title: Re: Digital speaker crossovers and room correction discussion
Post by: HAL on August 20, 2017, 05:57:01 AM
The equivalent is a peaking filter that has both gain and attenuation.  You set the gain(or attenuation in dB) center frequency and Q you want.  You can measure the original analog crossover and then adjust parameters in the software.

It does not simulate the crossover at the RCL level like SPICE. 

Amplitude adjustments in the scaling blocks are floating point math precision. 
Title: Re: Digital speaker crossovers and room correction discussion
Post by: HAL on August 20, 2017, 12:12:09 PM
If you are using a Zobel network to correct the impedance of a driver with a parallel network to work with the passive crossover, then it is not needed for the digital crossover.   If it is due to the driver impedance loading on the amp for feedback stability, then the Zobel network would still be required.
Title: Re: Digital speaker crossovers and room correction discussion
Post by: steve on August 20, 2017, 05:57:46 PM
If you are using a Zobel network to correct the impedance of a driver with a parallel network to work with the passive crossover, then it is not needed for the digital crossover.   If it is due to the driver impedance loading on the amp for feedback stability, then the Zobel network would still be required.

I may not have been clear with my question, which I apologize for. I have heard digital crossovers in systems (at shows etc) and still find the "sound" wanting. I am attempting to ascertain the reason.

My first comment is, I have never heard an amp directly connected to a driver have a flat response. The driver frequency response varies with frequency, is impedance related. A zobel attempts to flatten not only the impedance, but also the frequency response over a broad frequency range. This means extremely minute changes, not only in the analog realm, but in the digital realm as well.

For general information, the digital crossover would have to simulate a capacitor and resistor combination (simulate impedance change VS frequency) and be variable to extremely minute levels. I shall give you an example.

I am matching the amplitude level my full range driver to amplitude of the woofer in my test speakers, to 1 part in some 200,000. I am adjusting the full range driver zobel network resistance by less than 1 part in 20,000. Are the very best digital crossovers capable of such minute adjustments? Just asking.

Lastly, what about the need to AD as well as the DA when working with analog? Has those distortions been taken into account? (Distortion is anything that corrupts the music.)

Cheers

Steve


Title: Re: Digital speaker crossovers and room correction discussion
Post by: steve on August 20, 2017, 06:09:31 PM
The equivalent is a peaking filter that has both gain and attenuation.  You set the gain(or attenuation in dB) center frequency and Q you want.  You can measure the original analog crossover and then adjust parameters in the software.

It does not simulate the crossover at the RCL level like SPICE. 

Amplitude adjustments in the scaling blocks are floating point math precision.

Setting the Q and center frequency is a resonant type circuit. I don't see how it will mimick a zobel network.

Below shows the difference between a Zobel impedance curve and 2 resonant circuits impedance curve (2 different Qs). Please excuse the quick hand drawing.

Cheers
Steve
Title: Re: Digital speaker crossovers and room correction discussion
Post by: HAL on August 20, 2017, 06:54:43 PM
A driver driven by an amp directly without the digital crossover feeding the amp will not be flat response.  It will be the response of the driver. 

If the Zobel network is being used to give the driver flat response as a compensation for a driver impedance causing resonances, the digital peaking filter driving the amp can be used to compensate for the resonance.  That has amplitude change in 1dB steps for the peak/dip.  Center frequency is in 1Hz steps.  Q is in 0.1 steps.

Another option is using a single biquad digital filter with the parameters to simulate an impedance compensation network for the driver response.  That gives the finest control over the filter response.  Then you have 32bit floating point precision on filter input values.   Each channel is independent, so drivers can be worked individually.  This requires measurements of the individual drivers.

Another option is using a short convolver with specific driver measured data from both drivers to invert their responses for flattening. 

The last two methods are available, but measurement tools are not built into the unit.  I have tools that can be used to do them with correct measured data.

Both the A/D's and D/A's in the unit all run at 24bit/192KHz for DSP processing and without using linear phase filtering. The unit uses AKM AK5397 and AK4495S units for their analog performance for sound quality.  Much improved audio performance over the digital crossover units I mentioned earlier. 

I have used the system as a digital crossover and compensation for a Dayton Audio PS220-8 full range driver to 3x8" servo subs in an open U-baffle with very good results.  I used the peaking filter method above for compensation and also with a Butterworth filter for crossover to the subs.
 

Title: Re: Digital speaker crossovers and room correction discussion
Post by: HAL on August 20, 2017, 07:04:14 PM
The R and C of the Zobel network with the RLC of the driver builds a resonance circuit for compensation.  The Zobel Network RC is calculated to work with the driver impedance (RLC parameters) 

If it is a parallel circuit usually for impedance flattening for easier amp/passive crossover design and drive.
Title: Re: Digital speaker crossovers and room correction discussion
Post by: HAL on August 20, 2017, 07:11:18 PM
A schematic for the Zobel's implementation with the driver might be helpful.  Can be a series or parallel network with the driver feed.
Title: Re: Digital speaker crossovers and room correction discussion
Post by: steve on August 20, 2017, 07:15:40 PM
The R and C of the Zobel network with the RLC of the driver builds a resonance circuit for compensation.  The Zobel Network RC is calculated to work with the driver impedance (RLC parameters) 

If it is a parallel circuit usually for impedance flattening for easier amp/passive crossover design and drive.

I beg to differ, a zobel network is not a resonant circuit for compensation. Check the impedance curves I posted in my last post. A zobel network is simply to reduce "flatten" the impedance of a speaker, to flatten the frequency response. In fact, one can cause the frequency response to drop as the frequency rises.

Cheers

Steve
Title: Re: Digital speaker crossovers and room correction discussion
Post by: steve on August 20, 2017, 07:17:54 PM
A schematic for the Zobel's implementation with the driver might be helpful.  Can be a series or parallel network with the driver feed.

A zobel network is always across the driver itself. I think you are confusing a resonant circuit, often called a notch filter, which you described in an earlier earlier.

Cheers

Steve
Title: Re: Digital speaker crossovers and room correction discussion
Post by: HAL on August 20, 2017, 07:32:56 PM
Parallel Zobel network for flattening driver impedance is not possible.  Biquads in this system simulate series style filters.

That style of network affects the electrical impedance load of the driver which has no effect on a digital crossover.  It affects the driver load on the amplifier.   

Title: Re: Digital speaker crossovers and room correction discussion
Post by: steve on August 20, 2017, 07:48:07 PM
A driver driven by an amp directly without the digital crossover feeding the amp will not be flat response.  It will be the response of the driver.

The amplifier is virtually never accurate, and its output impedance also contributes to poor response. But you are right in that a driver's response is not a flat.

Quote
If the Zobel network is being used to give the driver flat response as a compensation for a driver impedance causing resonances, the digital peaking filter driving the amp can be used to compensate for the resonance.  That has amplitude change in 1dB steps for the peak/dip.  Center frequency is in 1Hz steps.  Q is in 0.1 steps.


What you are describing is often called a notch filter, which has a resonant and Q. It is in series with the driver. I believe you are confusing it with a zobel network, which is a CR and across the driver. I posted the two impedance curves for a notch filter and zobel network in one of my previous posts.

Quote
Another option is using a single biquad digital filter with the parameters to simulate an impedance compensation network for the driver response.  That gives the finest control over the filter response.  Then you have 32bit floating point precision on filter input values.   Each channel is independent, so drivers can be worked individually.  This requires measurements of the individual drivers.


That sounds promising if it truly is similar to a zobel network, and not a notch filter. However, the sensitivity of the adjustment is still my question. Can the adjustments to the frequency response be so small as to be around 105db down from the fundamental/midband? (For the general public, we are discussing frequency response, tonal balance, not SPL changes across the entire audio band like a voluime control adjustment.)

Quote
Another option is using a short convolver with specific driver measured data from both drivers to invert their responses for flattening. 

Not good enough for what I am doing.

Quote
The last two methods are available, but measurement tools are not built into the unit.  I have tools that can be used to do them with correct measured data.

It appears I am working with such minute adjustments that a digital won't be able to obtain the ultimate "blending" the drivers. But then I am really delving into "deepest crevices" for maximum accuracy/naturalness.

Quote
Both the A/D's and D/A's in the unit all run at 24bit/192KHz for DSP processing and without using linear phase filtering. The unit uses AKM AK5397 and AK4495S units for their analog performance for sound quality.  Much improved audio performance over the digital crossover units I mentioned earlier. 

I have used the system as a digital crossover and compensation for a Dayton Audio PS220-8 full range driver to 3x8" servo subs in an open U-baffle with very good results.  I used the peaking filter method above for compensation and also with a Butterworth filter for crossover to the subs.

I am glad to hear of your success Hal. Sounds like you have been doing some fine work. As I mentioned I am delving into that last bit of territory to discover what can be obtained in matching woofer/full range drivers, and sonic purity; maximizing drivers. I will probably have to test the digital crossovers myself to see the absolute maximum capabilities I can obtain.

Cheers and thanks for the information Hal. Keep up the good work.

Steve
Title: Re: Digital speaker crossovers and room correction discussion
Post by: steve on August 20, 2017, 07:58:02 PM
Parallel Zobel network for flattening driver impedance is not possible.

Are saying the digital xover cannot simulate a zobel network? If so ok.

If you are saying a zobel network cannot flatten the driver impedance, what the amplifier output sees, yes it can quite well, but not perfectly. I have been using it for a long time to "flatten" the response of the speaker. But one has to use great parts.

Quote
That style of network affects the electrical impedance load of the driver which has no effect on a digital crossover.  It affects the driver load on the amplifier.   

Correct, a zobel will not affect the digital xover. The zobel network affects the load that the amplifier sees.

Cheers
Steve
Title: Re: Digital speaker crossovers and room correction discussion
Post by: dBe on August 20, 2017, 10:15:52 PM
Parallel Zobel network for flattening driver impedance is not possible.

Are saying the digital xover cannot simulate a zobel network? If so ok.

If you are saying a zobel network cannot flatten the driver impedance, what the amplifier output sees, yes it can quite well, but not perfectly. I have been using it for a long time to "flatten" the response of the speaker. But one has to use great parts.

Quote
That style of network affects the electrical impedance load of the driver which has no effect on a digital crossover.  It affects the driver load on the amplifier.   

Correct, a zobel will not affect the digital xover. The zobel network affects the load that the amplifier sees.

Cheers
Steve
I always use zones on drivers with large peaks @ FS because I do not know of any amplifier that does not benefit from a flat impedance curves.  Especially tube amps in the contexts of tube life, punch and timbre.
Title: Re: Digital speaker crossovers and room correction discussion
Post by: steve on August 22, 2017, 11:48:04 AM
I have a NOS pair of System Audio 1270 speakers with a notch filter at ~6500hz. I call them NOS because I do not have 100 hours on them, maybe not 50 hours. Got them out this morning and I am so so impressed with them. Claim to fame is that the membrane is 1 gram. So one only has to worry about the mass of the voice coil. Quick, natural sounding, dynamics, that notch filter does wonders to tame the upper mids/lower highs that 5 1/4" drivers are known for.

Cheers

Steve
Title: Re: Digital speaker crossovers and room correction discussion
Post by: HAL on August 22, 2017, 11:58:20 AM
I like open baffle planar/ribbon line arrays using servo sub arrays.  No Zobel or Thiel networks needed for the solid state Class AB amps driving either the ribbons or the planars. 

Been very happy with the dspMusik as the DSP crossover, A/D's and D/A's with my PC Music Server for room correction.



Title: Re: Digital speaker crossovers and room correction discussion
Post by: steve on August 22, 2017, 02:28:24 PM
I like open baffle planar/ribbon line arrays using servo sub arrays.  No Zobel or Thiel networks needed for the solid state Class AB amps driving either the ribbons or the planars. 

Been very happy with the dspMusik as the DSP crossover, A/D's and D/A's with my PC Music Server for room correction.

Glad it is working out so well for you Hal. I did own some used maggies some years ago and loved them. I did have a break in a lead wire that I had to repair. Other than that, I really liked them.

I did have some problems finding a nice capacitor for my zobel, but now am very happy. These days, it is difficult to find a great, accurate capacitor due to all the mis-testing procedures used. I have not listening tested the expensive caps so I cannot say if they are good. Just too expensive to test them in the size needed.

Anyway, the walls just disappear, music is in space, and very dynamic and live sounding.

Cheers and continued good work Hal.

Steve
Title: Re: Digital speaker crossovers and room correction discussion
Post by: HAL on September 01, 2017, 11:53:12 AM
Doug,
I have a DEQX 2.6p and it is a good unit.  I did not like the sound of the ADC and DAC's or linear phase filters, so that was my problem with it. 

The dspMusik has better ADC's and DAC's to me using the AKM AK5397 and AK4495S units.  It has the capability for being used as a preamp, but can be set for 1:1 operation and not used. 

The dspMusik has 8 outputs that are independent, so can be used as a 4-way stereo crossover.  As long as 4 channels of crossover covers all the drivers per channel, it will work. All analog inputs and outputs are balanced, but adapters for single ended can be supplied.  For longer runs balanced really helps with noise reduction and ground loop elimination.   Crossover points for subs are variable in 1Hz increments.  Low pass filtering can be set independently of the high pass frequency if needed or can be the same.

The measurement system is not integrated like DEQX.  I prefer to have the customer make the measurements needed with my Audiomatica CLIO pocket system and PC I send and then I do the build of the crossover software.   Once completed, the dspMusik is programmed and sent for trials.  If you already have a crossover design and parameters from DEQX, those can be used and no speaker measurements needed. 

MathAudio RoomEQ only works with digital file replay, so room correction is not available as a standalone box.  A Music Server is needed to use with either Foobar2000 or JRiver or other player that uses plug-ins for processing.   Until the next generation ADSP21469 dspblock is available from Danville, not enough memory to do room correction. 

The dspMusik programmed for customer use is $2000.00 + shipping.  This covers the CLIO measurement system being sent and returned for data collection.  The dspMusik system can be reprogrammed at any later point if the speakers change for an additional cost. 

Hope that explains the system.

Rich Hollis
Title: Re: Digital speaker crossovers and room correction discussion
Post by: HAL on September 02, 2017, 05:14:04 AM
The dspMusik system will not be automated like DEQX.  That is why I got involved to be the measurements person to work with the system and customer.  After all the DEQX work, still just sounded ok.  Same with the Behringer DCX2496. 

DEQX 2.6 does both the speaker crossover and room correction.  The PC program does the measurements and calculations to program the unit. You just had to have the calibrated mic and program for the installer to do it.  I have it here for my unit.  Cost $500 for that for my system and then more for the Earthworks M30 calibrated mic as the ECM8000 mic measurements for corrections did not sound very good.

If you have multiple speakers that change, then you would need a measurement system to work with the system.  About the same price as DEQX for the 2.6.   Crossover diagrams have to be build for each speaker, which I do for the customers.

RoomEQ uses a calibrated Omni mic from Dayton Audio to make the room measurements with the speaker.  Very easy to run compared to DEQX.  Runs from inside Foobar2000.  Fully automated measurements and the user decides the correction type and level with a single slider for both channels that are measured independently by the program.   This runs on a Windows PC for file based music replay. I just setup the measurement mic at the center of my listening chair.  RoomEQ can also do multipoint room correction with up to 20 mic positions if needed. It does not run on the dspMusik.   The next generation dspblock version and room correction are not even started at this time until the hardware upgrade is available.

Sorry, no experience with pro gear other than the stock Behringer DCX2496. 

Danville Signal, DSP Concepts and HAL want a customized system delivered to the customer to make sure it sounds good.   That is why we are all involved.

Rich
Title: Re: Digital speaker crossovers and room correction discussion
Post by: rollo on September 02, 2017, 09:14:26 AM
   Mad Scientist and Capt. Nemo all in one. Full speed ahead.


charles
Title: Re: Digital speaker crossovers and room correction discussion
Post by: HAL on September 02, 2017, 01:48:12 PM
For the speaker measurements, there is no need to send the measurement system back to get the data.  You just send the measurement files and I create the program to load on the dspMusik that comes with the measurement system.  Once it is running then you just send back the measurement system if you are keeping the dspMusik after the demo. 

Since file based playback is becoming a big part of music systems, this just integrates the PC into the setup.  Maybe not for everyone, but that is why there is an A/D converter for analog playback in the dspMusik.  No other DAC's are needed with the dspMusik system, so digital is a cleaner path.

Never heard a CD/DVD-A/SACD transport that beats file based playback to date.  That is why I build Music Servers. 

Good luck!



Title: Re: Digital speaker crossovers and room correction discussion
Post by: HAL on September 02, 2017, 06:48:35 PM
Rim was using the dspMusik RevC for crossover from the mains to the subs with time delays to keep the subs from cancelling at his listening position.  Crossover to the subs was at 50Hz.

He still used his preamp, Marchand crossover, analog and digital rigs with the RM40 system.  It is the hybrid that he wanted. 

That is as of the last time I was at his place. Have not talked to him in awhile.
Title: Re: Digital speaker crossovers and room correction discussion
Post by: HAL on November 30, 2017, 04:08:19 AM
Will be trying out a new room EQ curve in work for the dspMusikLCD.  After some listening trials, looking to see if it can be done in Audio Weaver for use with analog sources as well.   

This one is in work at Danville Signal.   Should be interesting to test with The Megaliths.

Have an idea of how to improve room correction measurements for improved sound quality.  This idea was started by DEQX with offering a standard and upgrade microphone with the 2.6p system.  Taking it to the next stage with a better mic and mic preamp for measurements at 24bit/192KHz via the new ADC module. 

This maybe a very automated process if the latest versions of a few different programs work well.