High-resolution Speaker Frequency Response & Distortion Measurements and Why We Need Them (video)

[abdo123]

Classical needs both deep and loud,there are performances way more complex than any other kind of music.

Like?

 

[Sokel]

Like?

 

[Sokel]

Like?

All that is a hint,YT can't reproduce it right.
Hint:The A0 of the piano is at 27.5 Hz,the contrabassoon is a little higher.

 

[DWI]

Classical needs both deep and loud,there are performances way more complex than any other kind of music.

Personally, I find all loudspeaker systems incapable of making a worthwhile reproduction of large scale orchestral music. I go to a lot of opera and ballet, but never listen to it at home. We go to some live broadcasts in cinemas with massive sound systems, even those I don't find particularly worthwhile, my wife likes them more, but they have the advantage of being $20 rather than $200 for a ticket. For large orchestral works I will listen using headphones, now go less frequently mainly because audiences tend to behave better in smaller venues.

For smaller groups and vocal music, motets, that kind of thing, Quad ESL are astounding. Recoding quality is usually excellent, it can be more of an issue with some jazz recordings.

Like many people, for me home audio is only ever a limited and poor substitute for live performance. My wife was a performer, it is a big part of our lives. I know people who plan all their holidays around going to a gig, my kids sometimes do that, including Glastonbury every year. We are planning a summer holiday (driving, not flying with the current chaos) around a performance at La Scala, Milan.

So I don't need 120dB monitors that would be used for recording/engineering large classical music as I don't listen to it. It's part of the reason why consumer choices of speakers in particular are more subjective than for professional use.

 

[Sokel]

Personally, I find all loudspeaker systems incapable of making a worthwhile reproduction of large scale orchestral music. I go to a lot of opera and ballet, but never listen to it at home. We go to some live broadcasts in cinemas with massive sound systems, even those I don't find particularly worthwhile, my wife likes them more, but they have the advantage of being $20 rather than $200 for a ticket. For large orchestral works I will listen using headphones, now go less frequently mainly because audiences tend to behave better in smaller venues.

For smaller groups and vocal music, motets, that kind of thing, Quad ESL are astounding. Recoding quality is usually excellent, it can be more of an issue with some jazz recordings.

I will be too emotional to comment on the subject,I'm moved even by the poor YT performance through my little fostex's I have on the pc.
Memory is a b***c.

 

[DWI]

Like?

I don't think it is all about frequency range and going to 27Hz. There are pieces of music, for example Mahler Symphony No 1 and Shostakovich Symphony No 2 where the sound level and dynamic range of the music, especially in the opening sections, is so extreme that it s very difficult to engineer to make it effective for home listening. The engineering required given the limits of home listening rather ruin the live experience, less so using headphones.

In Shostakovich's wonderful opera, Lady Macbeth of Mtensk, at this point in this production there are groups of musicians playing from 4 different locations in the venue (on stage, the pit and in both wings). As they say on Chuck Norris videos, don't do this at home.

 

[DWI]

I will be too emotional to comment on the subject,I'm moved even by the poor YT performance through my little fostex's I have on the pc.
Memory is a b***c.

I also can watch hours of YouTube clips on my laptop of performances I've seen and get very emotional. I've completely forgotten some shows after a week, but others remain strong in the memory more than 40 years later. I was reminded recently that I went to Miles Davis' last concert in Europe, at La Vilette in Paris in July 1991. I'd met up with a girlfriend and got tickets on the day. I remember Chick Corea and John McLaughlin, as I knew some of their music, also Maria Pages, but no recollection of Miles Davis as I knew nothing about his music at the time, also the 2-hour walk back to Montparnasse after the show at 3am, but no memory at all of Miles Davis.

I listen to plenty of recordings of performers and performances I've seen and am convinced those memories has a big impact on perception of audio recordings.

 

[gnarly]

My question is, how do you connect a microphone to a Rigol scope? What's the signal path?

Been a while since I made those, but pretty sure i just hooked up the soundcard to the scope.
probably used a isemcon7150 mic with a UMC404 or RME babypro providing phantom.

 

[amirm]

Fan noise is an example of noise, it is not random room noise that can be averaged out to lower the noise floor. Fans have dominate noise bands that persist even with averaging. The typical dominate fan noise frequency is = fan RPM times the number of fan blades divided by 60 seconds / min, plus plus a family of other easily measurable frequencies. Fan noise frequencies can easily be within 40dB of the test frequencies at the test microphone location(s). The fan frequencies car be easily within 0.5% to 0.75% and included in the software harmonic distortion calculations.

That means that fan noise can show up as harmonic distortion in the plots.

I noted in the review that if you have an obviously noisy room, such measurements will be subject to error. There are good mitigations though. I use a very quiet PC for my Klippel NFS measurements. The measurement mic is placed far closer to the speaker than the computer (and other sources of noise). Noise falls with square of distance so physics is helping us in that regard. Finally, it is critical to set the measurement volume as to optimize SNR without causing distortion in the microphone.

 

[amirm]

Are you suggesting manufacturers don’t test their own products?

"Test?" Yes, they do "test." They are not running full suite of measurements as I do and with precision equipment. One major company I dealt with had to borrow the same analyzer I have to make measurements to compare with mine! Companies are not putting priority on making precision measurements because it hasn't mattered for decades. There has not been a watchdog like me as the press has been sleep at the helm with subjectivists reviews. And measurements as sidebars. A number of companies have now changed direction and with good results to show for it. The rest are using obsolete measurement gear to make rudimentary measurements. And yes, some are not even measuring and trusting chip specs, what their OEM tells them, etc.

 

[DWI]

"Test?" Yes, they do "test." They are not running full suite of measurements as I do and with precision equipment. One major company I dealt with had to borrow the same analyzer I have to make measurements to compare with mine! Companies are not putting priority on making precision measurements because it hasn't mattered for decades. There has not been a watchdog like me as the press has been sleep at the helm with subjectivists reviews. And measurements as sidebars. A number of companies have now changed direction and with good results to show for it. The rest are using obsolete measurement gear to make rudimentary measurements. And yes, some are not even measuring and trusting chip specs, what their OEM tells them, etc.

I think you hit the nail on the head. Precision measurements haven't mattered for decades because for many units THD + N has been inaudible for decades, which is why I referred to the 33/303, a landmark product that outperformed the standard test equipment of the time. I personally used 140w Quad 909 (factory refurbished) that were 25 years old and cost $700.

Whether a review is largely technical (the 1968 review of the 33/303 also provided a full analysis of the components and circuitry), a separate "Measurements" page like Stereophile or a sidebar like HiFi News, they are properly tested. The HFN tests are done independently by a company that has designed test equipment and carried out 3rd party testing for manufacturers for almost 40 years, because it is just more cost-effective to use a 3rd party with current quality equipment. How much testing is published is a function of the readership's interest, a decision by the publisher. I suspect the bigger issue are measured reviews that are rejected and that we never get to hear about.

Personally I would trust a manufacturer like NAD, Nord, Octavio or whoever using OEM chip specs from the likes of B&O, Hypex of Purifi because from the reviews I've read, including yours, it seems to take a lot of effort to significantly downgrade their performance and near impossible to be audible.

 

[PeteL]

As far as cost, look at two French companies. @davidminard 's Octavio, who make a €200 multi-channel streamer, financed by €10,000 of crowdfunding. Do they need, or could they afford a €30,000 test machine? I doubt it.

Naim Uniti Atom Review (Streamer & Amp)

https://www.octavio.fr/amp/ All in one with Tidal, Qobuz, Deezer, AppleMusic, tunelm and co... multiroom, UPnP, Airplay2, Blutooth, 24/192 (and a parametric correction is annouced)... I wrote to this french factory for Audio Science... For 700 euros... And also a simple Streamer...

audiosciencereview.com

Then you have Devialet, with over €100million of funding, and production and QC technology that makes the AP55 look like a toy. They can manufacture in France because they can afford manufacturing and QC technology that is almost fully automated.

I would expect any business considering significant capital investment in any equipment to ask three questions:
1. What additional sales/gross profit will it generate?
2. What costs savings will result?
3. Can we afford it?

This could equally apply to Octavio looking to spend €1,000 on an analyser, someone looking at €100,000 for Klippel or Devialet looking at €1million for a new board making system.

@davidminard makes the point that what his company does is add value to existing hardware with new software and a nice package, rather than do anything relevant to measurable performance of the OEM hardware.

Likewise, I would happily buy a NAD M33 (recommended there) simply because the Purifi Eigentact amplifier and ESS Sabre DAC are known quantities.

Another common and much more cost-effective option is to buy in testing when needed, common with more expensive facilities like anechoic chambers. Reminds me of a friend's daughter, a very successful fashion photographer, who has never bought her professional camera equipment. She always used to rent it and now Canon give it to her. The main benefit of renting is not owning redundant equipment. Better pay $200 per external test than $30,000 for the test equipment?

I have no doubt there are many different ways of measuring and testing audio equipment, often determined by financial constraints or use of OEM systems. 've been around long enough to appreciate there is more than one way to crack a nut, usually many equally effective ways.

I agree, most will find a way to assess performance quite precisely. For many purchasing a AP X555 is just way too much. When I was in this business I had an old well calibrated System One that has been modified to have a USB adapter for day to day along with a 8903B. An ADC with some software can also do many things as VintageFlanker demonstrate in his reviews. I would rent a 555 for a few days near production to be able to have more of a full report along with the IOs flexibility which in Canada was about 300$ per days but I already knew what to expect for the most parts. I highly doubt that any manufacturers would go and release equipment shooting in the dark. The fact that some audio gear don't get those 100 dB+ Sinad numbers don't mean they don't know about it... Just different design goals.

 

[Axo1989]

There has not been a watchdog like me as the press has been sleep at the helm with subjectivists reviews.

We all need an emotional support animal from time to time.

 

[milosz]

I think there is some value to high resolution measurements IF a measurement system is used that doesn't also measure reflections from walls, equipment, tables....

Using an anechoic chamber is a well known method for measuring speakers and not reflections. I believe that Klippel's contraption, along with it's mathematical manipulations, also is able to measure the speaker and ignore the reflections. Even DIY folks can use gating to eliminate getting reflections in your measurement, at least for the shorter wavelengths.

In general if there are a few really high-Q (i.e., narrow) peaks or valleys in the amplitude response at various frequencies listeners are not going to strongly perceive them because music is unlikely to match those exact frequencies very often, statistically speaking. Nonetheless if response is really ragged- LOTS of narrow peaks & valleys - listeners can probably perceive this and it's won't sound "right." So looking at detailed FR graphs does tell you something about the behavior of the speaker in terms of sound quality- as long as you have a way of making detailed measurements that is accurate, it does tell you something about the speaker's ability.

Distortion is an even more important thing to characterize than precise frequency response, in my opinion. Bass distortion is a fact of life with any but the largest drivers, and human hearing is not terribly sensitive to distortion down in the bottom octaves - nonetheless low frequency distortion ought to be kept as low as possible; occasionally designs are produced that have quite respectable LF THD using reasonable-sized drivers, and these systems ought to be recognized for their good design.

There is no excuse for a speaker that has levels of distortion above -40 dB in the midrange - the vocal range- our hearing is rather sensitive to distortion in this range, and if you see high THD here, there is something wrong with that speaker.

Harmonic dIstortion in the upper midrange is best avoided but in general most good drivers are pretty clean in this region - except ribbon drivers that are crossed over lower than they should be. Designers really have to watch the distortion from ribbon tweeters if they try to cross them at too low a frequency.

I can't recall ever seeing IM distortion measurements for speakers. I hope maybe Amirm can do some some day, I'd be curious to see them.

 

[Axo1989]

Watching the beginning if the video, it occurred to me that high-resolution was considered in two aspects: the measurements themselves and their presentation. If the target audience spans a broad range of prior knowledge levels, a quick visual comparison of different smoothings would help convey the presentation issue.

I didn't watch it all, so apologies if this was shown later or if I missed it in the intro. Definitely good to have it up front though.

 

[lashto]

Square wave is not an audio signal.

That is something I did not expect to hear from you. There is quite a bunch out there who keeps saying that "test sinewaves are not audio/music", do you want to spend your week talking to them ?!
And those 'chirps' aren't exactly audio/music either ...

P.S.
that does not mean I disagree with you. And many thanks for the very useful video ... an indepth look at step/impuse/timing measurements would be highly appreciated too. I googled a lot of stuff but just could not find a good-enough explanation.

 

[lashto]

What you see is how the driver swings around its .5v position. Couse it wont stop there imediatly. A ideal driver wont do this and the acustic output would be zero.With DC a driver cant produce acoustic output. For acustic output you need to move air. Thats why drivers cant produce a rectangle wave. Or in other words if you see there something like a rectangle its the bad behaviour of the driver. So thats why you cant replace Voltage with dB. If you would measure the movement of the driver with a laser than you could replace the Volt with mm. But when it has reached top level it stops moving, your db's are 0. You can not go from the electronic signal world 1 to 1 into the acustic world.

Thanks for clarifying the issue with square waves & acoustical driver measurements. I have no audio-related qualifications but that sounds simple enough (even) for me.

However, when you say "a rectangle its the bad behaviour of the driver" it implies that a measurable "good behavior of the driver" exists too. Identifying the good/bad components is the whole point of those reviews/measurements, so it seems that square waves might actually be an useful test.
Or am I missing something?!

 

[tomtoo]

Thanks for clarifying the issue with square waves & acoustical driver measurements. I have no audio-related qualifications but that sounds simple enough (even) for me.

However, when you say "a rectangle its the bad behaviour of the driver" it implies that a measurable "good behavior of the driver" exists too. Identifying the good/bad components is the whole point of those reviews/measurements, so it seems that square waves might actually be an useful test.
Or am I missing something?!

See it depends just on the time look at it.

A huge heavy driver with low magnetic force swings longer. A very low mass driver with high magnetic force stops faster. But heavy drivers used for low frequences and low mass for high frequences.How this swing in, swing out behavior relates realy to what we hear i dont know. I just wanted to make clear that a driver cant have acustic output with a dc signal. A commen measurement for a driver is the step response, What is just off, and DC on. What you exactly can read out off it, in relation to sound i dont know. You could say the faster and the less swing the better, but i dont think its that easy. Couse you have to see it in relation to the frequence the driver has to do. With a rectangle you do the same just more often in a given time. Dont know what you should read more out of it then with the step responce? With signals like that you have to dig into fourier transformations and stuff like that. My math ends at sin,cos,tan.

 

[dc655321]

See it depends just on the time look at it.

A huge heavy driver with low magnetic force swings longer. A very low mass driver with high magnetic force stops faster. But heavy drivers used for low frequences and low mass for high frequences.How this swing in, swing out behavior relates realy to what we hear i dont know. I just wanted to make clear that a driver cant have acustic output with a dc signal. A commen measurement for a driver is the step response, What is just off, and DC on. What you exactly can read out off it, in relation to sound i dont know. You could say the faster and the less swing the better, but i dont think its that easy. Couse you have to see it in relation to the frequence the driver has to do. With a rectangle you do the same just more often in a given time. Dont know what you should read more out of it then with the step responce?

A speaker system or driver is a high-pass filter. That is why they cannot reproduce DC.

A step response for a single driver contains much useful information. In fact the system transfer function can be estimated from it to produce a mathematical model.

If a speaker system is not time aligned, the utility of the step response is limited as there are then N responses to consider for an N-way system.

 

[tomtoo]

A speaker system or driver is a high-pass filter. That is why they cannot reproduce DC.

A step response for a single driver contains much useful information. In fact the system transfer function can be estimated from it to produce a mathematical model.

If a speaker system is not time aligned, the utility of the step response is limited as there are then N responses to consider for an N-way system.

"..A speaker system or driver is a high-pass filter. That is why they cannot reproduce DC..."

Yes but i think its much more easy to understand, that a speaker that not moves cant produce sound. In case of DC on a speaker.