[ekovalsky]

I now have in my possession a fully modded TacT S2150 amplifier.  Since I already have two stock S2150 amps, I finally have the opportunity to experiment with active multi-amping.

As before, the RCS 2.2X is used to divide signal between the mains and subs with 70hz, 48dB/octave high and low pass filters.  The sub output goes through the integrated DAC via a balanced cable into a Crown K2.  The main outputs (all digital) go through paralleled AES/EBU and S/PDIF links to S2150 #1 (modded amp, on tweeters right now) and S2150 #2 (stock amp, on midrange), respectively.  From amp #2 another S/PDIF link daisy chains to S2150 #3 (stock amp, on woofers).  I will be replacing the daisy chain link, which is using a borrowed S/PDIF cable, with another parallel direct link from the 2.2X using a glass Toslink cable which should arrive in the next few days.

Previously, two discrete sets of speaker cable or one biwire set was used to connect the amplifier(s) to the external passive crossover, which then had three sets of pigtail speaker cables that connected to the binding posts of the main channel that are in turn connected directly to the individual driver groups.  So it was extremely easy to bypass the passive crossover, I just disconnected it and connected each amps' outputs directly to the binding posts on the speakers.

I'm using wonderful new software for TacT called TACS.  It is developed by some gurus in France and is free.  It allows very flexible crossover design with semi-automatic parametric EQ, speaker correction, and level normalization.  Also it has a self-contained measurement module that is vastly superior to what comes with the RCS 2.2X.  It still uses the RCS to generate pulses and as mic preamp.

First step is to configure the TACS software to the system.



Next I measured the woofers and midranges with no crossover to check frequency response.  

Doing this with the tweeter would probably result in 18 barbequed Raven ribbons, so the tweeters were measured with a safety crossover, 60dB/octave @ 2khz high pass.  Using high frequency filters in the S2150 amps requires a reprogramming of firmware and system files; luckily the appropriate files are included with the TACS download so I did not have to buy them from TacT    

TACS fortunately has an awesome virtual remote control that lets you control the important functions of the each amp and the RCS within the software.  This includes firmware reprogramming!





After looking at the individual driver group responses (far nearfield) I designed basic crossovers.  They wound up identical to what the factory uses in the bass/mid (350hz) but I dropped the mid/tweeter a bit (3500hz to 3000hz).  Also I opted for higher slopes -- 24dB/octave Linkwitz-Riley.







Once these crossovers were programmed into the amps, I proceeded to repeat measurements to assess time delays and impulse response.







These pulses were obtained with the polarity reversed in the amps, to compensate for the mic which also has polarity reversed.   From these the delays of each driver group, related to the DSP filters, are evaluated.  Woofers have the most lag, followed by the midranges with the tweetesr the least lag.  So delays are added to the tweeter and midrange amp to align the pulses with the woofer.  This is further tweaked after additional measurements at the listening position to compensate for toe in (which alters relative distance to the mids and tweeters in the array) and vertical displacement of the woofers.

One thing to notice here is that there is measurable low frequency pre-ringing of the midrange and tweeters.  I believe this to be a "feature" of DSP crossovers.  Similar pattern can be seen in the measurements of the NHT xD (with XdA processor/amp -- based on DEQX technology) published in the review in the November issue of Stereophile.  

TACS lets each driver group be measured with bass, midrange, or full range pulses.  This helps greatly with time alignment and is a feature missing from the basic RCS software package.

Finally, a non-corrected measurement of the speaker is made at the listening position.



At this point, correction in the bass range is being implemented in the RCS as it was before the active multi-amping.  I was expecting to have switched correction into the amps by this point but have been distracted by listening sessions  

I will post more in the following days, but listening impressions have been very positive.  There is a cohesiveness of the sound that I would have only thought impossible with a large, multi-way speaker.  The treble through the modded amp is exceptionally clean and smooth, with no trace of harshness yet lots of detail and "air".  

Things should be even better when the amp is fed a cleaner digital signal from a fully modded RCS 2.2X which should be here in the near future.

Much more to follow...

[JoshK]

Nice setup and a very excellent tutorial!  Looks like you are learning a heck of a lot about speaker design.

[ctviggen]

Did you cut out the room response section of the time-domain signal?  Or does the TACT do that for you?  In other words, the time-domain signal from 10-14 ms should have undulations caused by reflections.  Yours doesn't, meaning that either they're cut off (they occur later) or you have excellent room treatments.

[ctviggen]

Oh yeah, about that preringing, try softer slopes.  From what I remember, steep slopes have the detriment of ringing.  See, for instance:

http://www.iar-80.com/page112.html

I don't think ringing is good, as it's intrinsic to the design of the system there's no way to remove it, and it would have to be deleterious to the sound.

[ekovalsky]

Did you cut out the room response section of the time-domain signal?  Or does the TACT do that for you?  In other words, the time-domain signal from 10-14 ms should have undulations caused by reflections.  Yours doesn't, meaning that either they're cut off (they occur later) or you have excellent room treatments.

You can set the x-axis limits (time in msec) in the TACS software.  I just excluded the reflections which are visible later.  The impulse measurements were obtained a few inches away from the drivers, so room effects don't show up until several msec later.

These measurements would be the basis for "speaker correction" which can be done with TACS, and is also available with the standard DEQX software.  Other than slight irregularity in the mids, though, there is nothing worth correcting at the driver level.  All the abberations that are visible in the farfield (i.e. listening position) measurements are from room effects or lobing/combing of the drivers.

If a speakers' individual crossiver/driver design produce fairly flat response there is probably little added value in driver level correction.  This would have come in handy with my previous speakers, though, which had a rather roller coaster response, even nearfield, from the planars.  This is probably to some degree a drawback of planar drivers (benefits being wide usable range and low mass) that is not as much an issue with cones and ribbons.

[ekovalsky]

Here is something interesting...

I was interested to see how different DSP crossovers affect the pre-ringing apparent in the measurements I made.  And of course I wanted to compare the impulse response of the active DSP crossover the passive Alon crossover.



This graph shows three separate tweeter impulses superimposed.  One was taken with all crossovers in the RCS and S2150 bypassed, with the Alon passive crossover in line.  The other two were obtained with the passive crossover removed, and 24dB/octave and 60dB/octave Linkwitz-Riley crossovers implemented in the S2150.  I adjusted delays to separate the impulses so each could be separately examined.

Anyone want to guess which is which ?

Red -- passive crossover only
Teal -- DSP crossover, 24dB/octave
Yellow -- DSP crossover, 60dB/octave

So the pre-ringing is not simply a modicum of the DSP crossover, nor is it particularly affected by the slope.  Presumably it is happening in the TacT amp (a digitally modulated class D design) itself.  Unfortunately the only other amp I have right now is the Crown K2, which is also a Class D design.  It would be interesting to remeasure with an analog solid state or tube amp.

I have done some more listening with the three S2150 (one modded, two stock) on the mains with DSP crossovers only, and with only the modded amp on the mains via biwire cable to the passive crossover.  Active DSP sounds better hands down -- mainly in the coherency of this large speaker speaker.

New toys arrive soon.... fully modded RCS 2.2X and glass Toslink optical link to obviate the need to daisy chain.  Both should help clean up the digital signal to the amps and further improvement performance, particularly at the higher frequencies.

[jermmd]

Eric,
I'm interested in the new TACS software. Is it only for use with the Tact amps? Can you provide a link? I have no desire for crossover functions as I'm sticking with my Stereo amp and the passive crossover that came with my speakers. Is the TACS superior to the Tact software for room measurement and correction? Is it downloaded into the TACT (or how does it interface with the TACT)? Is it supported by the TACT company? Finally, where did you hear about it?

Thanks,

[zybar]

Great stuff Eric!

George

[JohnR]

Here is something interesting...

I was interested to see how different DSP crossovers affect the pre-ringing apparent in the measurements I made.

Hi, does this software actually generate an impulse which you feed into your amps and speakers? It kinda looks like it's applied a lowpass filter to it. Can you convert the impulse responses into frequency responses (and display the full frequency range)?

[brj]

Thanks for posting all of this, Eric!

While not in a position to pursue it right now, I'm intrigued by the possibilities that digital crossovers and driver/room correction units offer.  I very much appreciate your efforts to write-up and post your experiences, and I look forward to further installments!

[ekovalsky]

Eric,
I'm interested in the new TACS software. Is it only for use with the Tact amps? Can you provide a link? I have no desire for crossover functions as I'm sticking with my Stereo amp and the passive crossover that came with my speakers. Is the TACS superior to the Tact software for room measurement and correction? Is it downloaded into the TACT (or how does it interface with the TACT)? Is it supported by the TACT company? Finally, where did you hear about it?

Thanks,

Joe,

TACS is freeware from France, written specifically for expanded capabilities of the TacT amplifiers when used in active multi-amp setups.  It is frequently updated, and the latest iteration allows full virtual operation of the connected amps as well as the RCS.  There are basically five modules:

1.  Configuration -- describe your system, speakers/drivers, amplifiers, etc.

2.  Remote control -- control all functions and settings of your amps and RCS

3.  Crossover filters -- crossover, EQ, correction, and output level

4.  Measurement -- uses the test signals and mic / mic preamp of the RCS.  Significantly more versatile than the measurement module in the RCS.

5.  Programming -- stores generated filters in the amplifiers

It does not generate RCS filters or program them, you will still need to use the TacT software for that.  So if you don't have TacT amps, it will not be of much use other than the more versatile measurement module.  It is downloaded and run from PC and communicates with the equipment via serial cable.  It requires the .net framework to run.  Here is download linK:

http://www.dirac.fr/tacs/

Even if you don't have TacT equipment, it may be worth downloading to check out its feature set.  It is free !

[ekovalsky]

Here is something interesting...

I was interested to see how different DSP crossovers affect the pre-ringing apparent in the measurements I made.

Hi, does this software actually generate an impulse which you feed into your amps and speakers? It kinda looks like it's applied a lowpass filter to it. Can you convert the impulse responses into frequency responses (and display the full frequency range)?

The software does not generate any pulses itself, rather it controls the RCS which generates the pulse.  It also uses the microphone and mic preamp built into the RCS.   TACS has a much better graphing module than the RCS, also it lets you choose which signal pulse (full range 2k- 2khz, mid range 400hz - 2khz, or sub 20-400hz) to send during any given test.  This lets you accurately meausure any individual driver.  The RCS software will only uses the sub pulse on the bass, which significantly restricts temporal resolution, and full range on the main output which makes direct measurement of woofers and midranges difficult -- the RCS will give "insufficient level" errors of the woofers and mids despite a loud pulse because it is trying to set its levels from the high frequencies.

You may be right about the lowpass filter.  I'm pretty sure the TacT digital amps use a low pass filter on the ouput to get rid of the carrier wave.  Maybe this is what shows up on the impulse response.  If so a lot of other class D amps may have similar behavior.

[csero]

Seeing these measurements I have serious concerns about how far fetched the results from reality.

 TACS measured an almost perfect symmetryc sinc function overlayed with an asymetric (without preringing) filter function for an analog filter. The sinc part simply should not be there. It can come ftom the amp reconstruction filter, but as the input is analog, the inverse of this should be in the analog aliasing filter and should not show in the measurement.

If you substract the sinc function from all 3 crossover measurement, then the analog would shows a correct analog filter step response and the digitals have significant pre ringing.

BTW have you seen the NTH measurements in the latest Stereophile?
The drivers measured individually have significant pre ringing, which more or less cancel out on axis. But it would be interesting to see off axis impulse measurements.
Let me quote myself from an old topic:

It can sometimes make sense to use a linear-phase (zero-phase) FIR filter, but most of the times the pre-ringing produced by such a filter is more disturbing that the frequency-dependant delay of a minimum-phase filter.

Most DSP studio-monitors that feature FIR filters to compensate the drivers and use them as crossovers have a recommended minimum-phase crossover setting, because the spot where the different pre-ringing artifacts played by the different drivers cancel each other out would be very small in linear-phase mode. ...

[JohnR]

Well, the sinc function is a (FIR) lowpass filter. Since it's an FIR I suspect it's being generated in the software. Judging from the plots, I'd guess it's about 25 kHz.

I expect this is easier on the equipment than trying to produce a pulse for a single sample (i.e. full bandwidth). If the plots only show up to 20 kHz, you won't see its effects in the frequency domain.

Eric, if you're willing to send me the data (in a text format) I'll plot the full frequency range.

[jermmd]

Eric,
Thanks for addressing my questions. I guess I won't need the new software for now.

[csero]

Well, the sinc function is a (FIR) lowpass filter. Since it's an FIR I suspect it's being generated in the software. Judging from the plots, I'd guess it's about 25 kHz.

I expect this is easier on the equipment than trying to produce a pulse for a single sample (i.e. full bandwidth). If the plots only show up to 20 kHz, you won't see its effects in the frequency domain.

Eric, if you're willing to send me the data (in a text format) I'll plot the full frequency range.

Yes, that is what I meant. The MLS sequence is not valid (too wide bandwidth) because it measure the measurement chain also, not only the DUT. This makes comparison of impulse responses questionable.

[ekovalsky]

Well, the sinc function is a (FIR) lowpass filter. Since it's an FIR I suspect it's being generated in the software. Judging from the plots, I'd guess it's about 25 kHz.

I expect this is easier on the equipment than trying to produce a pulse for a single sample (i.e. full bandwidth). If the plots only show up to 20 kHz, you won't see its effects in the frequency domain.

Eric, if you're willing to send me the data (in a text format) I'll plot the full frequency range.

JohnR, I'll try to figure out if there is a way to export the data into a spreadsheet or text format.  

csero, your understanding of this is far beyond mine so I cannot intelligently respond to your comments.

I greatly appreciate the input from both of you.  I think we are still in the early stages of this technology (limited to DEQX and TacT right now) but it will certainly proliferate in the future.

[csero]

The technology itself is getting close to maturity. The real question as always is what to measure, and what to consider from the measurements and what to ignore.

[JoshK]

The real question as always is what to measure, and what to consider from the measurements and what to ignore.

Ahh yes, the really tough question.   Seems like whenever someone like Dr. Geddes (or whoever else you wish to read) comes out with a study examining the link between measurements & our hearing, noboby wants to hear it.  

I did notice some pro equipment having suggested minimum phase setting even when linear phase was possible.  You explanation makes a lot of sense as to why.  I will have to consider this when making my FIR crossovers.

[csero]

Ahh yes, the really tough question.   Seems like whenever someone like Dr. Geddes (or whoever else you wish to read) comes out with a study examining the link between measurements & our hearing, noboby wants to hear it.

"The average audiophile is stuck that deeply in his trench that he'd rather be killed than accept advice based on reasoning other than an audiophile one. "