Tutorial: Bi-Wire, Passive Bi-Amp, Active Bi-Amp

[g29]

Not me. I do use a mini dsp to cross my C1's and woofers.

I started out with passive XOs.
Then analog active.
Then digital active.
Then sold the active XOs and went DAC-2-amp-2-driver as tech progressed controlling XO, DSP/DRC in a single location.

 

[MakeMineVinyl]

I started out with passive XOs.
Then analog active.
Then digital active.
Then sold the active XOs and went DAC-2-amp-2-driver as tech progressed controlling XO, DSP/DRC in a single location.

DSP is a powerful solution, however people should be aware that if optimum sound quality is the goal, the digital route either locks you into full-chain digital playback, or an analog-to-digital/digital-to-analog conversion cycle if your original source is analog, like in vinyl playback. If you only play digital such as streaming or CD, then DSP is fantastic - not so much with analog sources if the utmost purity of those analog sources is important.

Myself, I use analog active crossovers of my own design. In my case, although digital quality is important, it is not as important as vinyl playback to me.

 

[g29]

DSP is a powerful solution, however people should be aware that if optimum sound quality is the goal, the digital route either locks you into full-chain digital playback, or an analog-to-digital/digital-to-analog conversion cycle if your original source is analog, like in vinyl playback. If you only play digital such as streaming or CD, then DSP is fantastic - not so much with analog sources if the utmost purity of those analog sources is important.

Myself, I use analog active crossovers of my own design. In my case, although digital quality is important, it is not as important as vinyl playback to me.

My vinyl went away a long time ago after putting all my music on tap.

In contrast, here is the noise floor of a 64-bit XO.

 

[MakeMineVinyl]

There are important considerations beyond SINAD, regardless of what is the dogma at ASR.

I'm glad you have a solution which works for you - in your circumstances with an all-digital signal path, I'd do exactly what you are doing. But being an engineer myself who designs the stuff, I also realize that an additional A/D to D/A conversion cycle for no other reason than to accommodate DSP is detrimental in the context of an all-analog system, with disadvantages which outweigh the potential benefits of the finer control available with a DSP solution.

In my own system, I can make the in-room response as flat (or non-flat if that's what I want) as I desire with the all-analog crossover and line level passive voicing filters I have designed for my particular circumstances - no DSP required, and no losses from unnecessary A/D - D/A cycles. Another important consideration is that using DSP to 'correct' for a non-optimum room is not the best approach - a bad room needs acoustic treatment, or we need to use another, more suitable room. When the room is correct, then the need for elaborate filtering, DSP or otherwise, is greatly reduced.

 

[g29]

There are important considerations beyond SINAD, regardless of what is the dogma at ASR.

I'm glad you have a solution which works for you - in your circumstances with an all-digital signal path, I'd do exactly what you are doing. But being an engineer myself who designs the stuff, I also realize that an additional A/D to D/A conversion cycle for no other reason than to accommodate DSP is detrimental in the context of an all-analog system, with disadvantages which outweigh the potential benefits of the finer control available with a DSP solution.

In my own system, I can make the in-room response as flat (or non-flat if that's what I want) as I desire with the all-analog crossover and line level passive voicing filters I have designed for my particular circumstances - no DSP required, and no losses from unnecessary A/D - D/A cycles. Another important consideration is that using DSP to 'correct' for a non-optimum room is not the best approach - a bad room needs acoustic treatment, or we need to use another, more suitable room. When the room is correct, then the need for elaborate filtering, DSP or otherwise, is greatly reduced.

Whether or not you have an aversion towards ASR's measurements, they give you a ballpark understanding of what you are dealing with from one component category to the next. As an engineer, that should be of value to you or anyone in the hobby.

FWIW, I have no ADCs and have room treatments. Actually gone through multiple iterations of room treatments over the years including both commercial and DIY (after dissecting a few commercial units). Even made 6' tall, nested RPG Fractal clones. 1D QRDs inside of 1D QRDs. 4 football field lengths worth of TS milling.

My latest changes was replacing commercial OC 703 based absorbers @ 1st reflections with horizontally mounted polies that have 180 degree vertical dispersion patterns. I have dipole line arrays so the majority of the line array reflections now get redirected to the floor and ceilings in an evenly distributed fan pattern.

Line arrays have very little floor and ceiling bounce so the redirections are not adding insult to injury. The redirection keeps the energy in the room, while attenuating the intensity of the reflection and varying/lengthening the arrival of the reflection. The sound is not dry while correcting the sound-stage by minimizing the front, side and cross-side wall reflection contamination (room induced cross talk elimination).

In short, the polies are acoustic room stretchers. I originally modeled their behavior in a ripple tank simulator. Subsequent sweeps proved they attinuated the reflections more than the commercial absorbers.

Every correction that can be done by treatments are done by treatments (except a replacement venue) leaving DRC as the last resort as you have suggested. Even replaced outside wall and attic fiberglass insluation with much denser Rockwool to attenuate outside ambient noise.

 

[MakeMineVinyl]

Whether or not you have an aversion towards ASR's measurements, they give you a ballpark understanding of what you are dealing with from one component category to the next. As an engineer, that should be of value to you or anyone in the hobby.

FWIW, I have no ADCs and have room treatments. Actually gone through multiple iterations of room treatments over the years including both commercial and DIY (after dissecting a few commercial units). Even made 6' tall, nested RPG Fractal clones. 1D QRDs inside of 1D QRDs. 4 football field lengths worth of TS milling.

My latest changes was replacing commercial OC 703 based absorbers @ 1st reflections with horizontally mounted polies that have 180 degree vertical dispersion patterns. I have dipole line arrays so the majority of the line array reflections now get redirected to the floor and ceilings in an evenly distributed fan pattern.

Line arrays have very little floor and ceiling bounce so the redirections are not adding insult to injury. The redirection keeps the energy in the room, while attenuating the intensity of the reflection and varying/lengthening the arrival of the reflection. The sound is not dry while correcting the sound-stage by minimizing the front, side and cross-side wall reflection contamination (room induced cross talk elimination).

In short, the polies are acoustic room stretchers. I originally modeled their behavior in a ripple tank simulator. Subsequent sweeps proved they attinuated the reflections more than the commercial absorbers.

Every correction that can be done by treatments are done by treatments (except a replacement venue) leaving DRC as the last resort as you have suggested. Even replaced outside wall and attic fiberglass insluation with much denser Rockwool to attenuate outside ambient noise.

Sounds like you have it covered. Like I originally mentioned, with an all-digital signal flow, what you are doing is optimum, and using acoustic treatments is ideal - you're doing everything perfectly.

My AD / DA comment was mostly aimed at people who have either an all analog signal flow or a mixed digital/analog one. In that case the decision to use DSP boxes and accept the dual conversion is something which should be thought about carefully. I've experimented with DSP myself and found it totally unacceptable but for the unlikely reason of lack of headroom. My system has horns with sensitivities of 107dB/w and in order to keep noise low enough for my very quiet room/neighborhood, I have to use elevated signal levels up to the power amplifiers which have abnormally low gain and are basically just unity voltage gain 'current buffers' (again to reduce noise). All practical DSP boxes simply clipped with the signal levels I use, and to pad down the input to the DSP so they wouldn't clip, I would loose those high signal levels since the DSP boxes couldn't supply those high output levels. It was a no-win situation all around, and to be honest, the added capabilities of DSP filters went mostly unused since I don't need to EQ very much because I have an optimum room (it used to be a small motion picture dubbing stage for my music composition / recording / editing).

I used to poke around at ASR and found them extremely tiring - it was basically a cult. The thing is, I agree with 99% of the usefulness of measurements (and I have a well equipped lab here at home), but I also keep an open enough mind to realize that the human ear/brain is complex and sometimes we want 'distortion' and other 'bad things' - contrary to the ASR dogma. Reducing the entire world down to a single figure of merit - SINAD - is extremely myopic. I'm an engineer, but I'm also a musician, thank goodness.

 

[g29]

Sounds like you have it covered. Like I originally mentioned, with an all-digital signal flow, what you are doing is optimum, and using acoustic treatments is ideal - you're doing everything perfectly.

My AD / DA comment was mostly aimed at people who have either an all analog signal flow or a mixed digital/analog one. In that case the decision to use DSP boxes and accept the dual conversion is something which should be thought about carefully. I've experimented with DSP myself and found it totally unacceptable but for the unlikely reason of lack of headroom. My system has horns with sensitivities of 107dB/w and in order to keep noise low enough for my very quiet room/neighborhood, I have to use elevated signal levels up to the power amplifiers which have abnormally low gain and are basically just unity voltage gain 'current buffers' (again to reduce noise). All practical DSP boxes simply clipped with the signal levels I use, and to pad down the input to the DSP so they wouldn't clip, I would loose those high signal levels since the DSP boxes couldn't supply those high output levels. It was a no-win situation all around, and to be honest, the added capabilities of DSP filters went mostly unused since I don't need to EQ very much because I have an optimum room (it used to be a small motion picture dubbing stage for my music composition / recording / editing).

I used to poke around at ASR and found them extremely tiring - it was basically a cult. The thing is, I agree with 99% of the usefulness of measurements (and I have a well equipped lab here at home), but I also keep an open enough mind to realize that the human ear/brain is complex and sometimes we want 'distortion' and other 'bad things' - contrary to the ASR dogma. Reducing the entire world down to a single figure of merit - SINAD - is extremely myopic. I'm an engineer, but I'm also a musician, thank goodness.

I am an extremely talented artisan and musician. I genetically inherited from my father's side of the family. He always boasted he could draw flies and play the radio !!! My genetics mutated to include playing the stereo and TeeVee. Just to be clear, the artisan and muscisian part is total kofka.

With the BS being out of the way, I have had musicians in my house thinking I had things coming from outside in my back yard. They did not believe it was in the audio track until I replayed it for them. He was a guitarist, so it could be he was still tripping ? My neighbor was a concert pianist with 2 baby grands in her living room. Chopin was her favorite composer. She was a tiny little gal, surprising her small hands could cover the keyboard as they did.

I am an engineer, chief cook and bottle washer. I have worked in the satellite, CATV, broadcast TV and consumer electronics market. I have A/V software running in over 8 million North American installations. The part about companies tracking your usage is true ever since they had a back channel to upload the data !!!

As for audio, it has been a learning journey for me starting with piecing together car stereos and filters as a kid. My neighbor across the street was a ham radio operator with a whole room and car full of electronics, trying to push me in that direction. He took me to mobile emergency drills where rural civilian communications posts were setup, tested, logged and torn down. He was proud that he used his ham rig along with two other guys named "Frank" to organize disaster relief supply deliveries to an earthquake stricken area in Nicaragua in the early 70's. He proudly displayed his framed news article entitled, "The 3 Franks". He told me about this thing called a computer and that some day, everyone would have them in their homes. Too bad he passed before his prediction materialized.

I made the RPG Diffractal clones to only later discover they conflict with DSP. They scatter phase, which complicates/conflicts with phase based DSP. The polies do not. For me, the best location for the QRDs is in the back of the room.

I started out with a DEQX unit, but quickly discovered it was locked in time, software, processor power, taps, features and hardware so I started to venture out into DIY. Tried multiple things since then and am currently using my own customized version of DRC-FIR. I modified the 90's code to do 128-bit internal math and work at the driver, not speaker level. This allows you to apply DSP corrections to only the driver(s) needing the specific corrections and not placing those corrections onto other adjacent drivers that don't need it. This greatly cleans up XO regions along with digital XOs. I also modified it to use the speakers 64-bit FIR XO's as the correction target because the FIR XO contains copious amounts of phase and frequency data and is the ideal target for each. Both the amplitide and phase is properly corrected over XO regions at the individual driver level.

This can't be easily done with analog XOs because a speaker level "solution" is an average of good and bad drivers. The bad driver gets better and the better driver gets worse instead of improving both drivers according to their individual needs.

I then modified the code to coaleasece all of the corrections back into a copy of the original FIR XO so the resulting FIR does not change # of taps, timing or latencies. This allows the convolution engine to then just use a single FIR for XO, measurement sweeps and corrected playbacks.

Still learning things and interactions as I trudge along this hobby. If I had to do it over again, I would use ASR's measurements (and similar) to sort issues from worst offendors to least offenders and work the list in that order, starting with the worst offender, the venue.

 

[PaulyT]

Thanks gents, for the good discussion. And @g29 good to have you here.

I'll just selfishly interject that I'm a scientist - but my training is in chemistry so it doesn't help that much with understanding electronics. I'm still using a 3-way analog crossover designed by @MakeMineVinyl (and assembled by me) for the fronts in my HT. Used to have a lot more time for these hobbies, then had two daughters (one now in college and the other in HS) so they suck up all my time (AND MONEY) lately... sigh. I also play classical piano. So yeah, music is very important to me. My philosophy with audio gear is to optimize what I have as well as possible (I have some basic OC 703 DIY absorbers around my room) and then sit back and enjoy without obsessing over it too much - unless that's a source of enjoyment for you too!

So basically what I'm saying is that I'm enjoying this discussion, even if I don't have the background to understand all that you're saying.

Anyway, carry on.

 

[g29]

@PaulyT , FWIW, here is a simplified look at DSP via FFTs and iFFTs. Please forgive me if this is common knowledge and others feel free to correct my errors.

Sound/Music is a summation of an infinite amount of individual frequency bands.
Each frequency band has a phase and a frequency range with amplitude.
FFTs and iFFTs allow you to isolate, access, analyze and manipulate each band individually switching back and forth between frequency and time domains.
The more taps you have, the more bands you have. The more bands you have, the finer grain resolution you have (e.g. bass and treble controls vs a 24 band graphic equalizer, but with phase control as well).

Here is an animation depicting individual frequency bands summing in realtime to create a square wave at the bottom. To get a perfect square wave, you need to sum ALL frequencies from DC to infinity, not just frequencies from DC to 20kHz. FFTs and iFFTs allow you to go back and forth between the top 4 traces and the bottom summation (decompose a complex signal into individual discrete frequencies, optionally modify and then reconstruct the whole).

To implement DSP, you need a number of "taps". A tap represents a frequency band. Its width represents the resolution in hertz.

The more taps you have, the finer grain resolution in hertz. The finer the grain, the more precise adjustments can be.

Lower frequencies require more taps because a 5Hz width at 20Hz is different than a 5Hz width at 10kHz (e.g. [20-40]Hz being an octave compared to [10 - 20]kHz being an octave).

For a more tangible example, imagine a long foosball table with a lot of rods. Each rod represents a frequency band with phase.
The "with phase" is important. Graphical equalizers typically work on frequency amplitude, not touching phase.
This allows you to not only adjust frequency amplitude but also phase. This is important to integrate speakers in separate cabinets (e.g. sub/mains, sub, bass, high cabs) or non-time aligned drivers in a single cabinet.

If you have more of these long tables and line them up end to end, you achieve finer grain resolution.

Each rod can be slid in and out and twisted by a player (processing power).
The rod manipulation can be envisioned as changing both the amplitude and phase in each band independently of every other band represented by other rods.
The speed of the players (which some players are truly amazing, definitely not me) equates to the processing power of the DSP. The more processing power (better players), the more taps can be processed simultaneously (better game outcome).

Boxed DSP solutions tend to be tap and processor limited. If they take analog signals, they add the overhead of an ADC to the mix which has its limitations that @MakeMineVinyl has mentioned. Current PC hardware allows you to bypass and periodically improve on those limitations from one CPU generation to the next. Single thread speeds are still important.

Hope this simplified depiction is of some benefit to someone.

 

[PaulyT]

Yeah I get FFT to go from time to frequency domain, chemists use that for NMR.

 

[g29]

If I better understood the real world applications/potential of DSP in college, I would have majored in math and then some.

 

[PaulyT]

Yeah, I hear you, I mean, I’m happy with where I am, I work for NIH, but if I started all over, I probably would’ve been more in the computer science or algorithms side of things and yes, digital audio tech would’ve been very interesting to me.

 

[g29]

Yeah, I hear you, I mean, I’m happy with where I am, I work for NIH, but if I started all over, I probably would’ve been more in the computer science or algorithms side of things and yes, digital audio tech would’ve been very interesting to me.

When I went through school, my HS didn't have a computer. Home PCs were not yet a ubiquitous thing. College computer courses were geared towards accounting/business (2,000 card pickup and JCL queues), languages and manufacturing/machine control. Not much in the way of mathematical modeling or DSP at that time, though I did write a realtime 60Hz filter in assembly with no math library. That was the extent of DSP back then so the curriculum had not matured into today's marketplace. The mainframes of the time could not sustain the realtime loads and were a tad bit too expensive and bulky for home entertainment use.

The passage of time sure has opened up possibilities.

 

[MakeMineVinyl]

I'm just a bum who stumbled onto this forum a couple years ago. I don't know nuth'in, and I don't want to know nuth'in. I'm retired now. And get off my lawn!