Does DSP belong in State of the Art Systems?

[Brucemck2]

I am an outlier on this forum, but, in my experience DSP done right - and “done right“ is an enormous caveat - can materially improve the sonics in even great rooms with great equipment.

In my case I have a purpose built room (Dennis Erskine and Keith Yates) with extensive room treatment (primarily RPG bass absorbers at key pressure points and a wide array of diffusors at key point, augmented buy limited broad band absorption), with great electronics and speakers and DSP is the difference between “impressive” and mind-blowing great. In prior two channel rooms (one also custom designed) with serious mainly analog gear (Acoustic Research, Basis, Aesthetix, Koetsu, VTL, BAT, etc. etc.) various incarnations of DSP also provided meaningful improvements.

Dialing in DSP is not unlike dialing in a cartridge setup (including cartridge loading), or dialing in speaker and chair locations. It’s easy to get to “just ok” and takes many years of experience to really nail. Automated tools like Dirac or Audiolens help, but only get part way there unless you have years of experience and invest hundreds of hours dialing them in.

Note: original post edited to reflect Schlager’s comments in post #163

 

[SeagoatLeo]

I am an outlier on this forum, but, in my experience DSP done right - and “done right“ is an enormous caveat - can materially improve the sonics in even great rooms with great equipment.

In my case I have a purpose built room (Dennis Erskine and Keith Yates) with extensive room treatment (primarily RPG bass absorbers at key pressure points and a wide array of diffusors at key point, augmented buy limited broad band absorption), with great electronics and speakers and DSP is the difference between “impressive” and mind-blowing great. In prior two channel rooms (one also custom designed) with serious mainly analog gear (Acoustic Research, Basis, Aesthetix, VTL, BAT, etc. etc.) various incarnations of DSP also provided meaningful improvements.

Dialing in DSP is not unlike dialing in a cartridge setup (including cartridge loading), or dialing in speaker and chair locations. It’s easy to get to “just ok” and takes many years of experience to really nail. Automated tools like Dirac or Audiolens help, but only get part way there.

I've heard a few mind blowing great (and expensive) rooms that did not use DSP. It is possible. I have not heard the same in DSP rooms; unfortunately, as I haven't heard 1% as many DSP corrected rooms (out of 400-500 total rooms).

 

[schlager]

Automated tools like Dirac or Audiolens help, but only get part way there.

Audiolense is no more automated than Trinnov and it will get you all the way and even further. People will rarely hit it right, the first time they dial in DCR based EQ. It takes time, to listen through different settings and compare simulations, because the settings are room, speaker and listening position dependent. There is a learning curve to be climbed and can actually been seen as a craft, within it´s own rights. It took a couple of month for me to reach my final filter settings and it has been playing for more than a year now, without being changed.

 

[schlager]

To DSP or not to DSP....
Many are reluctant to DSP, because they feel/think/hear the sound is being degraded running through the AD-DA conversion, also know commonly known as digitalis.

A forum member of a Scandinavian hifi forum, did a test, using a Yamaha LS9 mixer. Unfortunately they removed the test files.
In the 1. test people could hear a small difference, running a piece of music 10 time through the, in this case, DA-AD loop. The guy doing the test, did a test sweep through Yamaha LS9 and found it was not 100 % linear.



After he did some EQ to the signal, coming out of the mixer, people could not tell the difference, between the original music piece and the 10 x loop.

Then he did a 100 times loop, with the conclusion that the noisefloor and distortion was raised. That is to be expected, when you run a copy of a copy of a copy etc. in the analog domain, then there will be a loss in the dynamic roof (S/N).

As seen the Yamaha LS9 has rather mediocre / poor specs.
How much is 0,05% THD times 100 ....

I just google translated it, but it should be readable.

"I made a small test for fun since I have a simple home studio. I took about 30 seconds of a well-known track and looped it through a Yamaha LS9 digital mixer 10 times like this:
Digital playback from PC -> analogue out on mixer -> XLR cables -> analogue in on mixer -> digital back to PC -> recording
This then becomes a DA - AD conversion.

The recording was then used in the next round, etc., so that the final result I share with you (along with the original) is a tenth-generation copy that has been DA and AD-converted 10 times. Or switched between analog and digital domain 20 times if you want.

To avoid clipping, the original was pulled down 3 dB in level. In order not to reduce the resolution, the original is actually a 24-bit copy with a 3 dB lower level. All recordings were made in 24-bit 44.1 kHz.

Now it must be said that the Yamaha LS9 is really a mixer for PA use. It's not exactly high-end. It is not level correct at 0 dB gain, and I found that it rolls off a bit in the bass and peaks in the top. I have compensated for this in the final result, so that you avoid getting a difference due to level, channel balance or frequency response. The latter I can't guarantee 100% as a bit of work was done on the feel here..."

 

[godofwealth]

I have extensive experience in using the Tact room correction system, owning various versions of their products (RCS 2.2XP, 2150 XDM etc.). In the beginning, of course, it’s a bit thrilling to see the before/after response curves. Sometimes, DSP can work miracles. Once I bought a used pair of Proac Response 3.5s. The buyer cheated me because one of the speakers had a defective crossover. For a lark, I used the Tact to fix the obviously broken FR curve. It was magical. Without Tact correction, that speaker was unlistenable. With Tact, it sounded like nothing was wrong with it and it sounded like the non-defective partnering speaker. Quite mind blowing.

With all that said, I don’t use Tact anymore. It corrects the FR at one spot — the center listening area — but not over a wide range. I have a Lyngdorf 2170 that tries to correct room imbalances by asking you to measure the room at not just the center location, but a bunch of random locations. It gives you a measure of how well it’s modeled your room. It works well, but you’re stuck with the pulse width modulated amplifier of the 2170 that can sound a bit dry and sterile with some speakers since its response depends on speaker impedance.

I now prefer to listen without room correction. Fortunately my Quads are well designed so they interact very little with the room. My Klipsch La Scalas have little low bass, almost nothing below 40 Hz. They sound fine to my ears with a small 2 watt SET and I position them along the back without toeing them in. To my ears they sound very nice indeed. I have run the Lyngdorf RoomPerfect correction on them. I can’t say it makes a huge difference.

We often forget that recording differences are so huge they simply dwarf any room imbalances. Many recordings made with bad mikes or in dry sounding studios will sound that way, with or without room correction. Great recordings will sound nice with or without room correction.

 

[godofwealth]

I see my microphone vs ears comment, was misunderstood quite a bit. My point was, that to study and analyse what a hifi setup does in a room, we should use a microphone (calibrated) and not our ears.

A frequency sweep gives us a graph to look at peaks and dips, so you can quickly see the problem. It is rather difficult to hear the difference between 2000 and 3000 hz by ear, so with a mic you can pick out the problem and be in control.
If something is off in the sound, it would be quite difficult to point out, by ear, where the problem is. This example only goes for the actually sound (frequency), if we put phase, decay, resonances etc. into the mix, it becomes impossible to sort out by ear.

A microphone then becomes a very strong tool, for the user, for further analysis.

Agreed that mikes can be extremely useful in quality control. When the Quad ESL 63 was first released in 1981, a bunch of journalists were invited to see how they were produced, and tested. They were slack jawed to see that the final QC step was to test each production unit with a very carefully measured control unit simply by feeding the production unit with a square wave, out of phase with the same signal fed to the control unit. A microphone was used to measure the null response as the two out of phase reproduced square waves would acoustically cancel at the microphone. If they did, the production model was packed, otherwise it was sent back for further tests.

The astonishing part was that the ESL 63 was so accurate in reproducing a square wave that it could be QC tested in this way. The overwhelming majority of loudspeakers made today are not phase true. They do not reproduce a square wave properly at all. A tiny few with first-order crossovers, like Thiels, get it right. But 99.9% of loudspeakers would fail at this test.

Of course, it’s not yet universally accepted whether the ear is phase accurate. People don’t seem to complain given that almost every loudspeaker they’ve ever heard destroys phase.

 

[schlager]

godofwealth, I would not qualify Tact or Lyngdorf as top notch DSP/DRC. For one they don´t use FIR filter and you don´t have much user control. Tact is an old product and the algorithms of DRC has moved on quite a bit.

Every speaker interacts with the room, how could it not. With constant directivity speakers you have an edge over the rest of the bunch, in that you control the dispersion evenly.

PWM amps have very low output impedance and have good control over the speakers (bass), with low distortion, that is what you heard and thought it sterile and dry. That is a good thing in my book. A clean unaltered signal is what high fidelity is about.

Not many speakers can reproduce a square wave, because of CTC driver distance. Some panels can. Vandersteen comes to mind, as well. Synergy horns can do it over many octaves and will do it over a wide listening area.

Regarding phase, studies show that we are phase sensitive in the area around 800-3000 hz. Also the ears most sensitive region (something with evolution, speech and baby cry). I find imaging, clarity and transients improves, when the phase (timing) is right.

 

[Bjorn]

It seems like people are discussing room correction. DSP isn't necessarily room correction. DSP is far more, thus this discussion is based on wrong premises.

If a passive crossover is replaced by an active with the exact same settings as the passive, it's DSP. So saying that DSP has nothing to do in a high-end system is based on either misunderstanding or
wrong use of the word.

 

[schlager]

Bjorn, TS asked "Has anyone gone down this road? Are there devices on the market that minimize the negative impact of this pursuit?"

So is he referring to the DSP process itself (digitalis) or EQ with MINIDSP (or any other EQ digital devise)? I guess both and I also wrote DSP/DRC, to kind of cover both topics.

 

[Hear Here]

With all that said, I don’t use Tact anymore. It corrects the FR at one spot — the center listening area — but not over a wide range. I have a Lyngdorf 2170 that tries to correct room imbalances by asking you to measure the room at not just the center location, but a bunch of random locations. It gives you a measure of how well it’s modeled your room. It works well, but you’re stuck with the pulse width modulated amplifier of the 2170 that can sound a bit dry and sterile with some speakers since its response depends on speaker impedance.

Agreed. I've used amps with RoomPerfect (Lyngdorf), Dirac Live (NAD) and MARS (Micromega) and none actually improves the listening experience despite flattening the response curve to some extent. Of these I found RP to be the worst, though the TDAI 3400 is a pretty good amp without engaging RP. I've concluded that there are so many better methods to get the best sound from a system than chucking room correction DSP at the problem. If rooms' acoustics are so poor that none of the primary improvement methods work, then split the bass from top end first and apply DSP only to the bass amp, leaving the top end unmolested by this extremely complex signal processing.

Speakers such as the newer Martin Logans and Avantgardes do just this by employing active bass within the speakers. Newer fully active designs such as Dutch & Dutch, a couple of Dynaudio models and a few others take this a stage further, but as long as the top end escapes the thrashing that DSP gives the signal, the sound should not lose its excitement factor. That's my opinion, and I've backed this up by listening to these processors in my own system.

 

[godofwealth]

godofwealth, I would not qualify Tact or Lyngdorf as top notch DSP/DRC. For one they don´t use FIR filter and you don´t have much user control. Tact is an old product and the algorithms of DRC has moved on quite a bit.

Every speaker interacts with the room, how could it not. With constant directivity speakers you have an edge over the rest of the bunch, in that you control the dispersion evenly.

PWM amps have very low output impedance and have good control over the speakers (bass), with low distortion, that is what you heard and thought it sterile and dry. That is a good thing in my book. A clean unaltered signal is what high fidelity is about.

Not many speakers can reproduce a square wave, because of CTC driver distance. Some panels can. Vandersteen comes to mind, as well. Synergy horns can do it over many octaves and will do it over a wide listening area.

Regarding phase, studies show that we are phase sensitive in the area around 800-3000 hz. Also the ears most sensitive region (something with evolution, speech and baby cry). I find imaging, clarity and transients improves, when the phase (timing) is right.

Not true: it’s well known that PWM amplifiers interact with amplifiers in ways more difficult to understand than conventional class A/B amplifiers. Robert Greene of TAS has written more about DSP correction since the 1990s than anyone I know, and he’s commented extensively on this topic.

Here‘s a review of the Tact Millenium, one of the most well known PWM amplifiers.

SoundStage! Equipment Review - TacT Millennium Digital Amplifier (6/1999)

Here’s what this reviewer says to explain why speaker impedance matching becomes critical in PWM amplification.

“The second "but" that applies is the fact that the treble response is dependent on the impedance of the speaker. Because the output is coupled via the second-order filter at 60kHz, the actual load of the speaker connected cannot help but become part of the filter function. If the speaker impedance is around 6 ohms, then the Millennium is ruler straight. If it is 8 ohms, the response is up by 0.3dB at 20kHz, at 16 ohms a tad under 1dB, but at 2 ohms will leave the output more than 3dB down. This is certainly no worse than a lot of single-ended tube amps, and TacT's point is that this is easily corrected with a bit of (dare I say digital) equalization.

The fact is that not every speaker sounds wonderful with the Millennium. My trusty old Rogers LS 3/5a sounded very thin, which I suppose (and it's my best shot since they are not defective) could be traced back to their 15-ohm impedance. Otherwise, most speakers I have tried showed pretty much their basic character with a slightly varying final result. A pair of B&W 803s played better than they ever had. So did my old ribbon-hybrids, which do tend to lack some sparkle up top. Another lovely combination was made with a pair of Martin-Logan Sequel 2s that have been retrofitted with SL3 panels.

The best speaker/amp combination in this house was achieved with the beautiful Dali Grand Diva, which I shamefully underestimated in my Frankfurt report. It turned out to be a much better speaker than I had expected, not just with Millennium. I hope to provide you with a separate report on this. Another speaker that I have just heard fleetingly is the B&W Nautilus 802, which showed tremendous promise together with the Millennium.”

Quads, like most electrostatics, have very low impedance in the treble (2 ohms) but very high impedance in the bass region (> 50 ohms). These swings can be very hard to handle, and typical PWM amplifiers can sound really odd with such speakers, as has been my experience. If a bookshelf speaker like the legendary Rogers LS 3/5a sounds “very thin“ with the Tact Millenium, as the above reviewer notes, you can be rest assured it’s not the speaker. The LS 3/5a has a very rich and warm sound and is an exceptionally truthful loudspeaker (check out its amazingly flat measurements in Stereophile — John Atkinson has had a pair for decades. It’s got a designed bump in the bass to make it sound rich — it’s definitely not a “thin“ sounding speaker). A mint pair of Rogers LS 3/5A can sell for over 6-7 grand on ebay. It’s a true legend designed by the BBC for accurate broadcast monitoring in a small environment.

PWM amplification has its advantages, not least of which is energy efficiency. But as the famous saying goes, “What the Lord giveth, the Lord taketh away”. There’s no free lunch in audio as elsewhere in life. It’s all a compromise in the end, and you have to decide what compromises you are willing to live with. Sterile dry sound is not a compromise I can live with for the sake of energy efficiency. Keep in mind that the PWM process, like DSD, generates enormous ultrasonic noise that has to be filtered. There’s no getting away from that.

 

[schlager]

Pointing to a +20 year old design, when PMW still was in an early stage, probably don´t hold for modern PMW designs. They are practically a wire with gain, with no FR deviation vs load.

Maybe we can agree, that what we hear as differences in amps, regarding topology, most likely is frequency and distortion related. Often related to impedance mismatch. If it is frequency related then EQ would fix that.

 

[Hear Here]

Pointing to a +20 year old design, when PMW still was in an early stage, probably don´t hold for modern PMW designs. They are practically a wire with gain, with no FR deviation vs load.

Au contraire! The more recent room correction systems just add hugely to the processing that is done to achieve even bigger changes in the original signal. Early systems just attempted to flatten frequency response. Over the 2 decades since that early TacT, lots more processing is undertaken - to adjust time alignment, etc. All this phenomenal processing may be good for getting the bass improved, but it does the top end no favours - and a full-range amp with DSP cannot avoid having the entire frequency range passing through the processor.

 

[godofwealth]

Ah, “a wire with gain”, how many times have I heard that phrase in the past 35+ years of owning high end audio. If you look at the advertising literature from the 1970s, a lot of Japanese manufacturers used that analogy to tout their products. I recall a Sony advert that claimed its preamplifier was better than a wire as it had lower distortion!

All the while, no one seems to acknowledge the elephant in the room. The world‘s best loudspeakers can barely manage to resolve 8 bits in the treble or bass as their distortion exceeds -50 dB (even the ultra pricey 30 grand JBL M2s with their active DSP equalized crossover with individual 1000 watt Crown amplifiers have several orders of magnitude higher levels of distortion compared to even 50 year old 16-bit PCM technology). Want to enjoy truly low distortion sound? Invent a better loudspeaker! The world’s lowest distortion loudspeaker, e.g., the Quad 63 or its newer variants, can barely manage - 70dB (12 bits of resolution). And only at moderate volumes. Above 90 dB, Quads overload quickly. My Klipsch La Scalas will work at far higher levels (e.g., 100 dB) with distortion around 0.1-0.2%, but are not phase coherent like the Quads.

If I wanted a low distortion amplifier, a Quad 303 from 1960 will do very nicely, thank you. It’s Alan Shaw’s reference amplifier that uses to design his Uber expensive Harbeth Monitor 40.3XD (which now sells for 25 grand — amazing considering the first pair of Monitor 40s I bought in early 2000 cost me 2.5K!). The Quad 303 has a fully regulated output stage, uses an innovative “triples” circuit that renders the amplifier immune to transistor aging or bias inaccuracy, and according to Mr. Shaw still offers textbook accuracy. Like in everything else he touched, Peter Walker’s designs had a true magic. His current dumping 405 or 606 or 306 still has lower distortion in the treble than any modern PWM amplifier I’ve seen tested. The output bridge circuit in the current dumper combines feedforward and feedback to keep distortion low and harness the strength of a low power class A amplifier with the power of a class B amplifier. Devialet copied the same idea, except they use a class D amplifier for high power.

All said and done, I agree with Mr. Shaw. Loudspeakers are the bottleneck. They remain the most colored element in the reproduction chain and the most distorting component. For a true revolution in music reproduction, we need to invent a better loudspeaker. After Paul Klipsch and Peter Walker, I can’t think of any real innovation in loudspeaker technology in the past 60 years. Just bigger boxes with more drivers in them (e.g., Magico’s M9 made here in the Bay Area will cost you a million dollars, but it’s just a moving coil loudspeaker at the end of the day whose physics was understood over 125 years ago).

 

[schlager]

If you look at the measurements of modern PMW amps (Ncore, Purifi, etc.) they practically are a wire with gain.
Loudspeakers are the bottleneck. Yep! Don´t forget the room

 

[Atmasphere]

'A straight wire with gain' is incapable of acting as a voltage source. Nearly all speakers are designed with a voltage source amplifier in mind.

The phrase is almost entirely fiction (except for those speakers which are meant to be driven by a power source as opposed to a voltage source...). Probably best if no-one ever uses that phrase again for any reason ever

 

[morricab]

If you look at the measurements of modern PMW amps (Ncore, Purifi, etc.) they practically are a wire with gain.
Loudspeakers are the bottleneck. Yep! Don´t forget the room

Please show me the papers that correlate those distortion profiles with perceived sound quality.

 

[microstrip]

The real bottleneck is the stereo system. Next the listener , and from there on all else depends on the listener...

 

[Kal Rubinson]

Please Guys: It's PWM for Pulse Width Modulation.

 

[schlager]

Kal, we all know that threads often goes OT.
Back to DSP.