Distortion in loudspeakers

[thewas]

Ha, ok, you're not bursting anything, but I can see you can find ways that Neumann can be better....

Well, its not just me as you have seen already and if you look around in other threads about the importance of IMD.
I myself wish it would be different and my very nice KH120, JBL etc. be also as great, but reality is always tough.
But the good thing is said that we can reach something like 80% of the qualities for 20% of the price, which is great, isnt it?

 

[Robbo99999]

Well, its not just me as you have seen already and if you look around in other threads about the importance of IMD.
I myself wish it would be different and my very nice KH120, JBL etc. be also as great, but reality is always tough.
But the good thing is said that we can reach something like 80% of the qualities for 20% of the price, which is great, isnt it?

Yeah, that's cool. That Neumann is 10 times the price of the JBL - £1500 vs £150 each so I don't know if you could say 80% of the qualities for 10% of the price! Ha, but fine, I'm happy with my JBL 308p Mkii, and I think they're showing low distortion at my listening volume based on my measurements.

 

[mitchco]

As @thewas mentioned there are additional factors to consider when measuring speaker distortion as the room can have an influence, especially below Schroeder, as does the unknown distortion levels of the UMIK-1. The UMIK-1 is spec'd at: max SPL for 1 % THD @ 1kHz 133 dB SPL @OdB gain setting. But tells you nothing what the distortion level is across the frequency band. This mic has a good distortion spec across the frequency range of interest: https://audioxpress.com/article/fresh-from-the-bench-earthworks-m23r-measurement-microphone

 

[Robbo99999]

As @thewas mentioned there are additional factors to consider when measuring speaker distortion as the room can have an influence, especially below Schroeder, as does the unknown distortion levels of the UMIK-1. The UMIK-1 is spec'd at: max SPL for 1 % THD @ 1kHz 133 dB SPL @OdB gain setting. But tells you nothing what the distortion level is across the frequency band. This mic has a good distortion spec across the frequency range of interest: https://audioxpress.com/article/fresh-from-the-bench-earthworks-m23r-measurement-microphone

Cool, well I'm working with the limitations of what REW / UMIK / and the room influence can deliver in terms of distortion measurement, and I'm absolutely fine with that, I'm satiated by my test results so far so I'm not gonna buy more equipment nor do any further testing. But good to know some of the limitations.....if you take it to the nth degree you measure it anechoically or with Klippel or something, ha!

 

[ernestcarl]

I remember reading about "stepped sine" in this thread, and I concluded I didn't want to do it for fear of damaging my speakers.

I did so as well, for a while. Since your monitors have limiters in place, you probably shouldn't be too worried with that as long as you do not exceed the max continuous rating, as well as use adequate rest time intervals -- I have used 5-10s depending on volume.

308p mkII specs:
SPL Continuous: 102 dB
Peak: 112 dB

Actually, there's really little need to go beyond 96dB max unless you are measuring a subwoofer.

It can take a hellishly long time, though -- depending on your settings. When doing a stepped-sine for my sub, I felt stressed as hell even though I was in the second floor, and the sub was in the basement. I could "feel" the bass all the way up there for well up to 30 minutes! I would have gone outside for a stroll -- but it's the middle of winter in Canada, so...

You will hear and "see" distortion way well below those specified volume limits for sure. Also you will hear adjacent objects and room related resonances being activated (if measuring indoors) if you stay in for the duration while bass is being measured. It helps if you have another speaker on hand that's uber clean like the KH310, or even the KH120 (above 300Hz) to compare. It doesn't sound god-awful, unless if it's really, really bad.

Quite frankly, I wouldn't worry too much about it. I consider it (the usual distortion sweeps) more as a diagnostic for examining what effects my EQ has on my speakers when they are pushed very hard. And wouldn't you know, lots of folks like loudspeakers that do not sound "clinically" clean anyway.

 

[DDF]

It might be of interest to those following this thread, but SINAD in a speaker is equally if not more important to measure than distortion, due to typically high contributions of port noise and other possible rubbing and buzzing. I suggested to @amirm that he might want to consider adding it to his test suite. If SINAD is important in electronics, it's doubly so in speakers because its audible there.

A bit more info here with an example test method (MLSSA Incoherency Distortion test)

 

[Robbo99999]

@thewas and any other interested people, I redid my distortion measurements of my 308p in light of recommendations to measure closer at 0.33 metres and in mono rather than stereo. Yes, so I redid the measurements at 0.33 metres (each speaker seperately) for an initial target of 96dB at 0.33m which would equate to Amir's graphs of 86dB at 1m (he measures at 0.33m as I have done & then converts)......I was initially targeting this, but fell short by one dB, my 1kHz measurements were at 94dB which translates to 85dB at 1 metre. I did 512K sweeps for greater accuracy than the standard 256K, but I didn't do 1M sweeps because I was holding the mic due to lack of an extendable tripod....(I compared to some results using a little tripod at a greater measuring distance and the distortion results were the same vs me holding the mic, so no influence there). Following are my results, essentially pretty much all of the response is under 2% distortion across the board, and most of it is under 1% distortion. There is one exception to this and that is in the left speaker at 60Hz where it spikes briefly to 10%, but if you look at my raw frequency response you can see that corresponds with a massive null in the frequency response, so that's due to somekind of room mode so I think that 10% peak can be discounted. It doesn't have the distortion peak at the crossover between 1-2kHz that Amir's sample had, and my 308p samples perform better in the bass re distortion, so I think there is some unit to unit variance to account for the difference with Amir's measured sample of the 308p. My distortion measurments are a little worse than the results I showed before when it was all under 1% distortion, but that was measured in a stereo environment at 2 metres, so these results should be more accurate. (Note the first graph in each set is just the frequency response so you can see what kind of dB level I'm running the tests at, and also to notice any anomalies which I've discussed).
Right Speaker:
View attachment 98987
View attachment 98988
Left Speaker:
View attachment 98989
View attachment 98990

 

[ernestcarl]

@thewas... I think there is some unit to unit variance to account for the difference with Amir's measured sample of the 308p.

The unit-to-unit variance thing is almost certainly always going to be a given here, esp. in these kinds of consumer massed produced monitors where tolerances are simply impossible to keep very high.

I see this variation between my LSR305s... and in two other sets of Mackie monitors...

Sadly, I even see differences in my Sceptres (more expensive!) in both distortion figures as well as frequency response after measuring them as precisely, and consistently as I could in the extreme nearfield. I partially attribute the latter to the coax drivers being considerably more complex to manufacturer -- and likely it would be too costly to keep the kind of extra consistency checks that is already done for their more expensive counterparts. The internals and overall construction quality of these Presonus monitors simply do not even come close to that of Fulcrum Acoustic's actual Reference Monitor line series. Same designer, but very differently run companies. One is made cheaply in China as feasibly possible, the other is made in the USA (or constructed, at least) where they pretty much test every unit meticulously. I know Dave Gunness has mentioned how they strive for driver to driver consistency. Last I checked, the cheapest in the line was between $4-5k which is similarly priced to Meyer Sound's own new Ultra series line of monitors. Those upgrade paths look just about as unlikely of me getting some pair of Genelec coaxes.

The Neumanns are the only monitors I have where I can confidently say that they are almost like mirror images of each other. Even the volume trims at the back of these monitors are the most consistent I've ever seen. The amp volume trims of my pair of LSR305 do not exactly match, wherein I have to make a -0.5dB vol. adjustment via software.

LOL! *I know I'm digressing here, but related to this is also my concern about using universal speaker equalization -- based on Amir's spinorama data -- for pairs of speakers that may not exactly be pair-matched to each other -- or even to Amir's own measured sample. Let's say it's likely going to be a 95% match at best. The only way to know for certain is you also have to measure each one of your own monitors in the nearfield as precisely as possible. Same way you cannot rely on headphone EQ soley based on measurements made by one specific rig by one specific person. Headphones, of course, are notoriously much more inconsistent in this regard.

 

[Robbo99999]

The unit-to-unit variance thing is almost certainly always going to be a given here, esp. in these kinds of consumer massed produced monitors where tolerances are simply impossible to keep very high.

I see this variation between my LSR305s... and in two other sets of Mackie monitors...

Sadly, I even see differences in my Sceptres (more expensive!) in both distortion figures as well as frequency response after measuring them as precisely, and consistently as I could in the extreme nearfield. I partially attribute the latter to the coax drivers being considerably more complex to manufacturer -- and likely it would be too costly to keep the kind of extra consistency checks that is already done for their more expensive counterparts. The internals and overall construction quality of these Presonus monitors simply do not even come close to that of Fulcrum Acoustic's actual Reference Monitor line series. Same designer, but very differently run companies. One is made cheaply in China as feasibly possible, the other is made in the USA (or constructed, at least) where they pretty much test every unit meticulously. I know Dave Gunness has mentioned how they strive for driver to driver consistency. Last I checked, the cheapest in the line was between $4-5k which is similarly priced to Meyer Sound's own new Ultra series line of monitors. Those upgrade paths look just about as unlikely of me getting some pair of Genelec coaxes.

The Neumanns are the only monitors I have where I can confidently say that they are almost like mirror images of each other. Even the volume trims at the back of these monitors are the most consistent I've ever seen. The amp volume trims of my pair of LSR305 do not exactly match, wherein I have to make a -0.5dB vol. adjustment via software.

LOL! *I know I'm digressing here, but related to this is also my concern about using universal speaker equalization -- based on Amir's spinorama data -- for pairs of speakers that may not exactly be pair-matched to each other -- or even to Amir's own measured sample. Let's say it's likely going to be a 95% match at best. The only way to know for certain is you also have to measure each one of your own monitors in the nearfield as precisely as possible. Same way you cannot rely on headphone EQ soley based on measurements made by one specific rig by one specific person. Headphones, of course, are notoriously much more inconsistent in this regard.

Yes, unit to unit variation, well I'm pleased I got a reasonable pair of 308p's when it comes to distortion then! Re Equalisation using Anechoic Measurement data from Amir....I'm quite pleased & confident with having used it because it allowed me to fix the crossover with confidence, and it also identified the peak in the treble at 16kHz. Also, his measurements combined with my own in room measurements to quantify the effects of the HF Trim Switch on the back of the speaker led me with the help of @Maiky76 to confirm that use of HF Trim Switch at -2dB was a positive move, backed up by my listening tests too....this little adventure was here (https://www.audiosciencereview.com/...-mkii-studio-monitor-review.17338/post-570282), and Maiky76 characterised it into a Preference Score along with some other analysis of my determined effect of the HF Trim Switch (https://www.audiosciencereview.com/...-mkii-studio-monitor-review.17338/post-570903) ......so yeah I think there's great value in EQ'ing around Amir's anechoic measurements, and the associated analysis you can do on these.

 

[Bear123]

Found this thread recently and read a good portion of it, but not every page, but I did some distortion measurements for my JBL 308p Mkii speakers using REW & UMIK and I wanted to get some conclusion on how valid the measurements are (in relation to anechoic measurements too).....also as to whether I have the various measurement options setup properly.

Basically, I'm measuring both speakers playing together at my listening position which is 2 metres distance from the speakers. Amir measures one speaker at 1 metre distance for his SPL / distortion measurements, so my understanding is that my 2 speakers at 2 metres is basically the same as measuring 1 speaker at 1 metre, in terms of mathematically the dB levels are directly comparable in terms of the speakers being under the same stress.....but correct me if I'm wrong about this.

I decided to measure my JBL 308p Mkii speakers for distortion because it was the most negative variable found when Amir reviewed this speaker, here's his review (https://www.audiosciencereview.com/...ds/jbl-308p-mkii-studio-monitor-review.17338/). So I wanted to see what kind of distortion levels I was getting at my normal listening levels as well as at my maximum listening levels....so I did measurements at 77dB and 86dB (in terms of that is what REW labelled for the fundamental level). Here are my measurements, do I have it set up right and what kind of conclusions can we draw from these? (THD is the highlighted line in each measurement) (Measurements done after applying elements of an Anechoic Listening Window EQ, as well as removing one room mode peak at 155Hz that was exhibited from the left speaker (so that EQ filter only applied to left speaker), room not symmetrical)

Normal Listening Level (77dB):
View attachment 95232

Absolute Max Listening Level (86dB):
View attachment 95233
and I also tried to display the measurement above as Percentage on the Y-axis (which is more comparable to how Amir graphs his measurements) - do I have the checkbox options setup in the most relevant manner (top right of the pic below)?:
View attachment 95234

How comparable is this to Amir's anechoic measurements?

Sometimes, I do legitimately wonder if something like a 120 dB SINAD DAC and a Benchmark AHB2 would "sound" better with my speakers at normal listening levels on normal content in my room, compared to my Denon X3300 AVR. Best I can estimate is that my Denon sends a signal to my speakers with .007% SINAD. Said speakers and room are generating maybe 0.5% distortion up to moderate listening levels.

In a double blind listening test on a variety of normal content i.e. music and movies, would anyone be able to reliably tell which electronics were being used? In other words, would people be able to reliably hear a difference between .507% and .5001%. Its hard for me to imagine the answer would be yes.

I understand the desire to chase perfection in this hobby, but I wonder if we kid ourselves on wether the above scenario makes an actual audible difference or not.

 

[Robbo99999]

Sometimes, I do legitimately wonder if something like a 120 dB SINAD DAC and a Benchmark AHB2 would "sound" better with my speakers at normal listening levels on normal content in my room, compared to my Denon X3300 AVR. Best I can estimate is that my Denon sends a signal to my speakers with .007% SINAD. Said speakers and room are generating maybe 0.5% distortion up to moderate listening levels.

In a double blind listening test on a variety of normal content i.e. music and movies, would anyone be able to reliably tell which electronics were being used? In other words, would people be able to reliably hear a difference between .507% and .5001%. Its hard for me to imagine the answer would be yes.

I understand the desire to chase perfection in this hobby, but I wonder if we kid ourselves on wether the above scenario makes an actual audible difference or not.

Really?

Of course not between 0.507% & 0.5001%, you must be trying to pull my leg....why are you responding to my post with that tangent anyway when it seems quite far removed from what I was posting & measuring.....I'm not getting your angle & point? (You also quoted my old post from weeks ago rather than my latest measurements)

 

[Bear123]

Really?

Of course not between 0.507% & 0.5001%, you must be trying to pull my leg....why are you responding to my post with that tangent anyway when it seems quite far removed from what I was posting & measuring.....I'm not getting your angle & point? (You also quoted my old post from weeks ago rather than my latest measurements)

Nothing directed at you personally, in any way. But seeing your in room distortion measurements for your speakers got me thinking about overall distortion, including that from our electronics. Many seem extremely concerned about whether they are sending a signal with .007% distortion(such as a Denon AVR), or paying many many many thousands more for electronics that send a .0001% distortion signal to their speakers. Yes, a bit of a tangent but it seemed sort of relevant on the basic topic of distortion.

The distortion from your speakers seems low enough to be a non issue, and is around 0.5%. This got me wondering about what distortion level from our components is no longer relevant to sound quality.

 

[Robbo99999]

Nothing directed at you personally, in any way. But seeing your in room distortion measurements for your speakers got me thinking about overall distortion, including that from our electronics. Many seem extremely concerned about whether they are sending a signal with .007% distortion(such as a Denon AVR), or paying many many many thousands more for electronics that send a .0001% distortion signal to their speakers. Yes, a bit of a tangent but it seemed sort of relevant on the basic topic of distortion.

The distortion from your speakers seems low enough to be a non issue, and is around 0.5%. This got me wondering about what distortion level from our components is no longer relevant to sound quality.

Ah, I see, yes, well you gotta put it into perspective. Surely no point worrying about those small fractions of percentage you cited re distortion in electronics.

 

[amirm]

(Note the first graph in each set is just the frequency response so you can see what kind of dB level I'm running the tests at, and also to notice any anomalies which I've discussed).
Right Speaker:
View attachment 98987
View attachment 98988

These graphs don't look revealing. Can you limit the number of harmonics used in REW to just 5? Going to higher order ones is not wise as they are likely all noise but then add up to smooth out the THD. Also, please set the scale from 0 to 5% for THD ratio. And what is the flat line in the distortion graph?

 

[Bear123]

These graphs don't look revealing. Can you limit the number of harmonics used in REW to just 5? Going to higher order ones is not wise as they are likely all noise but then add up to smooth out the THD. Also, please set the scale from 0 to 5% for THD ratio. And what is the flat line in the distortion graph?

Unless I'm just not aware of how to do it, its difficult(not possible???) to scale a graph in REW from 0-5%. REW ends up using half the graph with extremely low levels of distortion below 1%. I tried briefly before and couldn't get it to scale that way.

 

[March Audio]

As @thewas mentioned there are additional factors to consider when measuring speaker distortion as the room can have an influence, especially below Schroeder, as does the unknown distortion levels of the UMIK-1. The UMIK-1 is spec'd at: max SPL for 1 % THD @ 1kHz 133 dB SPL @OdB gain setting. But tells you nothing what the distortion level is across the frequency band. This mic has a good distortion spec across the frequency range of interest: https://audioxpress.com/article/fresh-from-the-bench-earthworks-m23r-measurement-microphone

I use an M23 (not sure of the difference to an M23R) and also have a Umik. I will try to get a chance to do a sweep so we can compare the distortion and FR.

 

[Robbo99999]

These graphs don't look revealing. Can you limit the number of harmonics used in REW to just 5? Going to higher order ones is not wise as they are likely all noise but then add up to smooth out the THD. Also, please set the scale from 0 to 5% for THD ratio. And what is the flat line in the distortion graph?

Hi Amir, yes, I've limited harmonics to 5 in the following post in another thread, but here's the pic (and changed scale to be as representative as possible):
https://www.audiosciencereview.com/...bl-stage-a130-review-speaker.18260/post-59963
Left Speaker:

Right Speaker:

Ah, you asked me about the "flat line in the graphs", I'm not sure what you're referring to? The only flat lines are the where I've positioned the mouse cursor to highlight the junction between 2% distortion and the frequency above which it is consistently below 2% distortion. The dark brown line which is jagged, but relatively flat throughout a large section of the graphs is the noise floor.

Unless I'm just not aware of how to do it, its difficult(not possible???) to scale a graph in REW from 0-5%. REW ends up using half the graph with extremely low levels of distortion below 1%. I tried briefly before and couldn't get it to scale that way.

Yep, I tried to scale it like Amir does in his graphs so it's easier to compare, but it's exactly as you describe it.