Speakers distortion

[Krunok]

.. and the 'speaker is raised off the floor by several metres.

S

My Castle has 22kg and 92cm height - I can't raise it off the floor by several meters!

 

[sergeauckland]

My Castle has 22kg and 92cm height - I can't raise it off the floor by several meters!

That's one of the problems doing loudspeaker measurements at home. The best you can do is to try and get the 'speaker as far from all boundaries (floor,ceiling and walls) as you can. Generally, that means that your measurement microphone will be around 1.2m, perhaps 1.5m if you have tall ceilings, from the nearest boundary. That limits your lowest frequency measurement pseudo anechoically to about 200Hz.

Below 200 Hz, you can do near-field measurements by putting the microphone very close to the driver, 1-2mm close. Then, you have to apply all sorts of correction factors to the LF measurement taking into account the distance, the size of the baffle and the size of the driver, and only then can you stitch the LF measurement to the MF/HF pseudo anechoic measurement.

Considering that below 200Hz, it's the room that mostly affects what you hear, I haven't bothered to do near-field measurements for my 'speakers, given also how much is an approximation, and just adjusted the LF by ear, checked by some in-room measurements at the listening position.

If you are a loudspeaker manufacturer, you can do a lot in-house using gated pseudo anechoic and near-field measurements, but ultimately, the only way of a true measure of what the loudspeaker can do independent of the room is to use a large and very expensive anechoic chamber that goes down to 20Hz, or outdoor measurements as KEF used to do. Pictures of a KEF 105 on a 10m pole on an empty airfield shows how difficult this is. It's also possible to use a large empty indoor space like an aircraft hanger, as that at least avoids wind noise.

Considering how difficult and expensive proper measurements of loudspeakers can be, I'm not surprised a lot of 'boutique' manufacturers dismiss the whole concept of measurements and design by ear, with somewhat predictable results.

S.

 

[Krunok]

That's one of the problems doing loudspeaker measurements at home. The best you can do is to try and get the 'speaker as far from all boundaries (floor,ceiling and walls) as you can. Generally, that means that your measurement microphone will be around 1.2m, perhaps 1.5m if you have tall ceilings, from the nearest boundary. That limits your lowest frequency measurement pseudo anechoically to about 200Hz.

Below 200 Hz, you can do near-field measurements by putting the microphone very close to the driver, 1-2mm close. Then, you have to apply all sorts of correction factors to the LF measurement taking into account the distance, the size of the baffle and the size of the driver, and only then can you stitch the LF measurement to the MF/HF pseudo anechoic measurement.

Considering that below 200Hz, it's the room that mostly affects what you hear, I haven't bothered to do near-field measurements for my 'speakers, given also how much is an approximation, and just adjusted the LF by ear, checked by some in-room measurements at the listening position.

If you are a loudspeaker manufacturer, you can do a lot in-house using gated pseudo anechoic and near-field measurements, but ultimately, the only way of a true measure of what the loudspeaker can do independent of the room is to use a large and very expensive anechoic chamber that goes down to 20Hz, or outdoor measurements as KEF used to do. Pictures of a KEF 105 on a 10m pole on an empty airfield shows how difficult this is. It's also possible to use a large empty indoor space like an aircraft hanger, as that at least avoids wind noise.

Considering how difficult and expensive proper measurements of loudspeakers can be, I'm not surprised a lot of 'boutique' manufacturers dismiss the whole concept of measurements and design by ear, with somewhat predictable results.

S.

Wow - I see your point, it definitely cannot be done at home!

So, when I'm measuring speakers distortion at my LP and they are positioned where they normally are, what am I really getting? How much of the "truth" is in this graph?

Would it be any better if I repeat this measurement with mic positioned 1m from the speakers when they are as far from the walls as possible?

 

[sergeauckland]

Wow - I see your point, it definitely cannot be done at home!

So, when I'm measuring speakers distortion at my LP and they are positioned where they normally are, what am I really getting? How much of the "truth" is in this graph?

Would it be any better if I repeat this measurement with mic positioned 1m from the speakers when they are as far from the walls as possible?

It rather depends on what you're trying to measure. If you want to know what your loudspeakers are doing, independent of the room, then measuring at 1m, as far from boundaries using some sort of gated measurement as you get in REW or ARTA, will tell you that. This is what you need to do if you're designing loudspeakers and they have to work in a variety of rooms.

However, if you want to know what the loudspeakers are doing in your room, and trying to optimise their position rather than change the design, then doing an in-room measurement from the listening position, and even better, over a listening window so perhaps the average of 5-6 measurements done with different mic positions, makes the most sense. REW can do both sorts.

As to your graph above, if I've read it correctly, distortion looks excessive, I would expect at least 10dB less, even 20dB at mid and high frequencies. What level was it done at? I suggest 85 or 90dB SPL. 90dB SPL is nowadays pretty standard, although some small loudspeakers struggle with that at LF. For setting SPL, a sound level meter App is probably good enough.

S.

 

[andreasmaaan]

My understanding is that for a near-field measurement, the microphone has to be within 1-2mm of the middle of the driver, as they're mostly used to make measurements on the bass driver below the critical frequency in-room, which for most UK rooms will be around 200Hz.

Far field measurements are done pseudo anechoically at 1m, perhaps 2m if the room is large (or outside) and the 'speaker is raised off the floor by several metres.

S

This is absolutely true for amplitude response measurements. However, for distortion measurements it can be better to do something in between near and far-field, giving on one hand an approximation of the anechoic amplitude response, and on the other hand drowning out (only to an extent ofc) the influence of reflections on the distortion measurement. Would certainly not recommend this measurement distance for amplitude response ofc

 

[Krunok]

It rather depends on what you're trying to measure. If you want to know what your loudspeakers are doing, independent of the room, then measuring at 1m, as far from boundaries using some sort of gated measurement as you get in REW or ARTA, will tell you that. This is what you need to do if you're designing loudspeakers and they have to work in a variety of rooms.

However, if you want to know what the loudspeakers are doing in your room, and trying to optimise their position rather than change the design, then doing an in-room measurement from the listening position, and even better, over a listening window so perhaps the average of 5-6 measurements done with different mic positions, makes the most sense. REW can do both sorts.

As to your graph above, if I've read it correctly, distortion looks excessive, I would expect at least 10dB less, even 20dB at mid and high frequencies. What level was it done at? I suggest 85 or 90dB SPL. 90dB SPL is nowadays pretty standard, although some small loudspeakers struggle with that at LF. For setting SPL, a sound level meter App is probably good enough.

S.

I don't know at what level, but, as it was done at 82% of the volume, my approximation would be it was done at least at 90dB.

I don't know how to average distortion measurement done with log sine sweep in REW.

When you say sound level meter App do you mean that kind of app for Android smartphone or something else?

 

[Krunok]

This is absolutely true for amplitude response measurements. However, for distortion measurements it can be better to do something in between near and far-field, giving on one hand an approximation of the anechoic amplitude response, and on the other hand drowning out (only to an extent ofc) the influence of reflections on the distortion measurement. Would certainly not recommend this measurement distance for amplitude response ofc

Would that be at the distance of 1-1.5m from the speakers or ..?

 

[andreasmaaan]

Would that be at the distance of 1-1.5m from the speakers or ..?

No this is the 20-25cm I mentioned earlier

For most small-ish speakers, 1-1.5m would be far-field (or close enough).

 

[Krunok]

No this is the 20-25cm I mentioned earlier

Ok, I see.

I'll post the results once I make the measurement.

@RayDunzl , do you have sweep distortion measurement of your JBL LSR 308 speakers?

 

[sergeauckland]

I don't know at what level, but, as it was done at 82% of the volume, my approximation would be it was done at least at 90dB.

I don't know how to average distortion measurement done with log sine sweep in REW.

When you say sound level meter App do you mean that kind of app for Android smartphone or something else?

I've never found it necessary to average distortion measurements as distortion doesn't vary with microphone position, whilst frequency response does, quite significantly. As to sound level meter app, yes, any of the apps for an Android smartphone will more or less work, at least well enough for you to get an idea of the sound level at which you're doing the distortion sweep.

S.

 

[Cosmik]

...distortion doesn't vary with microphone position...

If that is literally true, then distortion measurements should always be near field and will be very easy to do i.e. almost eliminating extraneous noise.

But is it literally true? If you place your mic 1mm off a driver, are there 'nonlinear effects' as a result of compression of the air space between driver and mic, or whatever..?

 

[Krunok]

If that is literally true, then distortion measurements should always be near field and will be very easy to do i.e. almost eliminating extraneous noise.

But is it literally true? If you place your mic 1mm off a driver, are there 'nonlinear effects' as a result of compression of the air space between driver and mic, or whatever..?

I would imagine there would ne no compression if that driver is a tweeter but if it is a bass driver delivering 90dB of SPL I would imagine there would be some air compression, probably significant..

 

[mitchco]

I was looking into measuring speaker distortion a while back and got some info from Earthworks, which is the only other measurement mic manufacturer that had a distortion spec for their mics. Here is what I received from one of their engineers:

"Attached is a plot of the worst case THD vs Frequency for an M30, on the hot side (35 mV/PA). This is probably the hottest omni we make. You can see that it hits 3% THD at 138 dBSPL. I mentioned that being the "worst case" as we make M30s that go from 25 mV/PA all the way up to 35 mV/PA. The hotter the Mic, the sooner the THD will kick in. So for an M30 with 25 mV per Pa, the THD will be 3% at 141 dBSPL."

The reason why I bring this up as measuring distortion in room has several issues, the two main ones being not having too much room influence by placing the mic as close to the speaker system as possible (<30cm). Howwever, that then gives rise to SPL to the mic and some electret mics distort ion goes up with SPL quickly easily. Which is why you don't see distortion specs for most measurement mics...

Finding the sweet spot based on distance from speaker with a signal loud enough not to have too much room influence but low enough not to overload the mic takes some effort. If reference is 83 dB SPL at the listening position, then what is the SPL at 30cm from the speaker? Hopefully not loud enough to overload the mic...

 

[DonH56]

^^^ That's the mic I use and I have a similar plot, someplace... I also asked them for a plot from 50 dB to 120 dB and blowing up the scale. I cannot for the life of me remember the numbers but it was in the 0.01% ~ 0.1% range.

For those who might be interested, I measured the frequency response of my system from 5 Hz to 25 kHz using my M30 and my CSL-calibrated UMIK-1. FR was essentially identical (plots overlie) though of course the M30 has much greater dynamic range with lower noise and distortion (not usually an issue when just measuring FR in the room).

A few years ago I tried to find out what the distortion was for various loudspeakers in the say $5k to $25k range and discovered that it was pretty good for all types. Not DAC good, but about 0.1% or less around 90 dB SPL at 1 m (standard measurement distance when I looked). I was a bit surprised, expected higher. My Maggies, and several ESLs, were not significantly lower than the really good conventional designs, and at LF my panels tanked (again do not recall exact, but it was like 10%~30% as you approached their LF spec). That was true for all designs but the panels seemed a little worse. Note many ESLs, such as most Martin Logan designs, are hybrids with a conventional cone woofer and aren't as bad in the LF region as pure panels. Hard to beat physics (though Marketing tries!)

 

[Krunok]

^^^ That's the mic I use and I have a similar plot, someplace... I also asked them for a plot from 50 dB to 120 dB and blowing up the scale. I cannot for the life of me remember the numbers but it was in the 0.01% ~ 0.1% range.

For those who might be interested, I measured the frequency response of my system from 5 Hz to 25 kHz using my M30 and my CSL-calibrated UMIK-1. FR was essentially identical (plots overlie) though of course the M30 has much greater dynamic range with lower noise and distortion (not usually an issue when just measuring FR in the room).

A few years ago I tried to find out what the distortion was for various loudspeakers in the say $5k to $25k range and discovered that it was pretty good for all types. Not DAC good, but around 0.1% or less around 90 dB SPL at 1 m (standard measurement distance when I looked). I was a bit surprised, expected higher. My Maggies, and several ESLs, were not significantly lower than the really good conventional designs, and at LF my panels tanked (again do not recall exact, but it was like 10%~30% as you approached their LF spec). That was true for all designs but the panels seemed a little worse. Note many ESLs, such as most Martin Logan designs, are hybrids with a conventional cone woofer and aren't as bad in the LF region as pure panels. Hard to beat physics (though Marketing tries!)

Interesting info!

How exactly did you measure distortion?

 

[sergeauckland]

If that is literally true, then distortion measurements should always be near field and will be very easy to do i.e. almost eliminating extraneous noise.

But is it literally true? If you place your mic 1mm off a driver, are there 'nonlinear effects' as a result of compression of the air space between driver and mic, or whatever..?

Yes that's possible in that air is not perfectly linear. One can compress air to any arbitrary amount, but can only rarefy it to 1 ATM. However, if you do the sums, (and I have seen the numbers but can't now remember them) you'll find that at any sensible SPL that won't kill you, air is very linear indeed.
When measuring distortion, I agree that getting close to the driver is a Good Thing as it improves the S/N ratio and so allows lower distortion to be less affected by ambient noise and room gain. In my own room, located in a quiet rural environment, the microphone 1m from the loudspeaker allowed both frequency response and distortion measurements to be made. I may repeat the distortion measurements one day with the mic much closer to the drivers and see how much difference there was with my earlier measurements.

S.

 

[DonH56]

Interesting info!

How exactly did you measure distortion?

I did not, Earthworks measured the mic, and I searched manufacturer's data and reviews for speakers. I measured mine using that M30 with their preamp and a spectrum analyzer (old HP unit IIRC).

 

[Cosmik]

Ah, so the near field thing may be a red herring - I wasn't thinking about high SPL affecting mics badly.

Distortion is presumably something you can measure with massive amounts of averaging. If you leave your measurement running at a fixed frequency and accumulating for minutes at a time, presumably you reduce the effects of extraneous noise a lot. The room doesn't matter - as long as its effects are linear. However, I do recall a time when I listened to a 1kHz sine wave playing over a speaker of mine and hearing distinct distortion that varied with level. After tearing my hair out over it, I eventually tracked it down to some antique glass object in the room touching another, and setting up a rattle that varied with level. Such an effect could be present at a lower level without being very audible, but it would show up in your measurements if the mic was a long way from the speaker.

 

[andreasmaaan]

If that is literally true, then distortion measurements should always be near field and will be very easy to do i.e. almost eliminating extraneous noise.

But is it literally true? If you place your mic 1mm off a driver, are there 'nonlinear effects' as a result of compression of the air space between driver and mic, or whatever..?

Yes that's possible in that air is not perfectly linear. One can compress air to any arbitrary amount, but can only rarefy it to 1 ATM. However, if you do the sums, (and I have seen the numbers but can't now remember them) you'll find that at any sensible SPL that won't kill you, air is very linear indeed.
When measuring distortion, I agree that getting close to the driver is a Good Thing as it improves the S/N ratio and so allows lower distortion to be less affected by ambient noise and room gain. In my own room, located in a quiet rural environment, the microphone 1m from the loudspeaker allowed both frequency response and distortion measurements to be made. I may repeat the distortion measurements one day with the mic much closer to the drivers and see how much difference there was with my earlier measurements.

S.

The main issues with a measurement at 1mm would not be air non-linearity (I would think), but rather the near-field measurement:

1. seeing the woofer as though on an infinite baffle and
2. giving the appearance of rolled-off harmonics above the upper frequency limit of the measurement (which is proportional to the diameter of the driver / width of the baffle)

EDIT: and yes of course, SPL limitations of the mic

EDIT 2: Have modelled the air non-linearity (or tried to), and it seems that this would not be an issue. A typical woofer producing 100dB @ 1m, for example, generates a particle velocity at the driver/air interface of under 3 m/s, at which velocity air behaves more or less linearly.

 

[sergeauckland]

The main issues with a measurement at 1mm would not be air non-linearity (I would think), but rather the near-field measurement:

1. seeing the woofer as though on an infinite baffle and
2. giving the appearance of rolled-off harmonics above the upper frequency limit of the measurement (which is proportional to the diameter of the driver / width of the baffle)

EDIT: and yes of course, SPL limitations of the mic

EDIT 2: Have modelled the air non-linearity (or tried to), and it seems that this would not be an issue. A typical woofer producing 100dB @ 1m, for example, generates a particle velocity at the driver/air interface of under 3 m/s, at which velocity air behaves more or less linearly.

A near-field upper frequency measurement is indeed limited by the size of the driver, as the near field condition only holds for driver dimensions much smaller than the wavelength. The larger the driver, the lower that frequency is, so it's quite possible, indeed likely, that if one's trying to measure the frequency response of a full-range loudspeaker with a large bass driver, in a small room, then there will be a gap between the upper range of the near-field measurement and the lower limit of the far-field pseudo anechoic gated measurement.

This does mean, as you suggest in 2. above, that it's possible for the distortion measurement to be inaccurate with a large driver when using a near-field measurement, and due to room gain effects if using a mid/far field measurement.

S.