Evidence-based Speaker Designs

[ctrl]

Yes, as that is not a "perfect" cancellation. If you push it up by app 5dB with the filter with appropriate Q it will practically dissapear with 1/6 smoothing. The question here is would it sound better or not.

With appropriate smoothing, it would appear as if the cancellation by EQ has disappeared - it is quasi averaged out - but of course it is still present.
In this special case the EQ should be rather negative in sound.

 

[tuga]

You can add more stacked horns to combat the beaming, but which again leads to new serious issues IMO.

Could elaborate on those issues?

 

[QMuse]

You should keep in mind that the floor bounce is always position dependent (listening position height and distance to the loudspeaker) and the cancellation cannot, as already mentioned, be filled up with EQ.

IME it can - it is not a perfect cancellation as seen by the depth of the dip, so you simply increase SPL in that region and the dip will be filled, at least partially. That is not a narrow dip so IMHO it is hard to claim it will not be audible.

For stereo music listening EQ is always done at the LP that is at fixed distance from speakers so it doesn't really matter if that dip would move to higher freq as distance would increase.

 

[Bjorn]

Wouldn't a dipole interact with the room differently than a forward radiating box below Schroeder?

Yes, it would. But it doesn't trigger room modes any less. Just in a different way and require a different placement for an optimal (=flattest) result.

 

[QMuse]

I had similar dip at 400Hz and pulled it up easilly (red is uncorrected, blue is corrected, 1/12 smoothing). As a rule of the thumb, above 350Hz pretty much every dip can be pulled up.

 

[tuga]

IME it can - it is not a perfect cancellation as seen by the depth of the dip, so you simply increase SPL in that region and the dip will be filled, at least partially. That is not a narrow dip so IMHO it is hard to claim it will not be audible.

For stereo music listening EQ is always done at the LP that is at fixed distance from speakers so it doesn't really matter if that dip would move to higher freq as distance would increase.

If measured with MMM, would you apply a high-Q filter?

I had similar dip at 400Hz and pulled it up easilly (red is uncorrected, blue is corrected, 1/12 smoothing). As a rule of the thumb, above 350Hz pretty much every dip can be pulled up.

View attachment 64764

Is the blue curve the predicted or the actual in-room response?

 

[QMuse]

If measured with MMM, would you apply a high-Q filter?

Yes. If narrow dip or peak survived MMM averaging it certainly needs to be corrected. But I wouldn't apply more gain than 7-8dB for dips.

Is the blue curve the predicted or the actual in-room response?

Blue curve is actual in-room response. I'm doing corrections manually using rePhase. Both were measured with MMM with over 75 samples taken over the area 80cm*50cm*30cm (height).

 

[ctrl]

I had similar dip at 400Hz and pulled it up easilly (red is uncorrected, blue is corrected). As a rule of the thumb, above 350Hz pretty much every dip can be pulled up.

Of course I cannot judge whether the interference around 400Hz in your example is caused by the floor bounce, I will assume that it is and try to clarify the contradiction.

It is physically not possible to fill up the floor bounce (for a fixed listening position) caused by time delay difference (of a reflection) of a chassis with EQ.
In the example of the mid-bass horn, the delay time difference results in a phase shift cancellation - the reflected sound covers a longer distance than the direct sound. This cannot be eliminated by increasing the sound pressure level of the axis frequency response.

How is it possible that you managed to eliminate the floor bounce via EQ?

Fortunately there is not only this one floor reflection, but many more.
These other reflections always fill up the floor bounce a bit on average. And these other reflections allow the floor bounce to seemingly fill up by raising the axis frequency response in the affected area - the floor bounce itself is still there unchanged.

If these reflections were absent or very weak, it wouldn't work, because 5dB increase on axis also inevitably increases the floor reflection by 5dB and nothing would be gained.

This leads us directly to the price to pay to seemingly smooth the floor bounce with EQ.

If the sound pressure in the affected area has to be raised by 5dB to compensate, this naturally has consequences for the chassis: higher harmonic distortion, more IMD, ...

If the listener changes his listening position by only 0.2m towards the loudspeaker, this can already lead to a shift of the floor bounce at 400-500Hz, which means that the EQ is no longer optimal.

 

[QMuse]

If the sound pressure in the affected area has to be raised by 5dB to compensate, this naturally has consequences for the chassis: higher harmonic distortion, more IMD, ...

If the listener changes his listening position by only 0.2m towards the loudspeaker, this can already lead to a shift of the floor bounce at 400-500Hz, which means that the EQ is no longer optimal.

+5dB at 400-500Hz shouldn't be a stress for a good speaker in terms of THD/IMD.

And regarding what to correct and what not, as I said in the post above: if narrow dip or peak survived MMM averaging over the LP area done with +60 samples IMHO it needs to be corrected. Of course, without pushing dips with more than 7-8dB.

 

[QMuse]

How is it possible that you managed to eliminate the floor bounce via EQ?

If the listener changes his listening position by only 0.2m towards the loudspeaker, this can already lead to a shift of the floor bounce at 400-500Hz, which means that the EQ is no longer optimal.

I don't know. I will make new measurement within a circle with 0.2m diameter at the position where my head is and will post here.

 

[QMuse]

I don't know. I will make new measurement within a circle with 0.2m diameter at the position where my head is and will post here.

Here it is, "head area" MMM is green, LP area is blue. @ctrl , may it be that floor bounce dip of my speakers is at 600Hz as listening distance is close to 4 meters?

 

[ctrl]

Here it is, "head area" MMM is green, LP area is blue. @ctrl , may it be that floor bounce dip of my speakers is at 600Hz as listening distance is close to 4 meters?

Hard to say, because at 4m distance the floor bounce is only one of many, many reflections.
As already mentioned, this must be calculated or simulated for each listening position.

If you tell me the height of the driver (the driver that covers the frequency range around 600Hz) from the floor, height of the listening position and distance from the speaker I can quickly simulate it.

 

[QMuse]

If you tell me the height of the driver (the driver that covers the frequency range around 600Hz) from the floor, height of the listening position and distance from the speaker I can quickly simulate it.

So, meet my Castle Harlech S2. They are twin pipe quarter wave transmission line speakers. Both midwoofers are working together up to 1800Hz where they are crossed with tweeter by LR24 crossover. Lower midwoofer is at 86 cm and top is at 95cm. My ears are app at 90cm. Listening distance is app 4m.

Thank you in advance for your effort!

 

[ctrl]

So, meet my Castle Harlech S2. They are twin pipe quarter wave transmission line speakers. Both midwoofers are working together up to 1800Hz where they are crossed with tweeter by LR24 crossover. Lower midwoofer is at 86 cm and top is at 95cm. My ears are app at 90cm. Listening distance is app 4m.

Oh f....! That's tough.
We must consider two competing effects here. One is the time offset of the two chassis and the resulting floor bounce.
With the simple simulation tool used, this can no longer be reproduced.
But with a little trick it can be simulated very roughly. For a reliable simulation the effort (at least for me using ABEC/AkaBak3 for serious speaker simulations) would be very high.

At first we only look at the effect of the offset chassis. Since I don't know their distance/offset, I just assumed 0.28m. Since both chassis are operated up to the crossover frequency of 1.8kHz, this leads to cancellations due to the offset of the chassis and the resulting phase shift.
(red + blue the single chassis, in black the resulting frequency response)

Now we include the floor bounce at 4m listening distance.
(red + blue the single chassis with floor bounce, in black the resulting frequency response, again with phase shift of the chassis)

You can see that it has now become much more complex and the cancellations have shifted.
But both simulations say that the critical range at the current listening position is expected to be somewhere between 400-800Hz.

Whether one of the two effects is dominant at the listening position, or what the ratio is, cannot be determined reliably with such a simple simulation.

 

[QMuse]

LOL Yep, complex speaker. But it sounds lovely so I'm keeping it although it's 17 years old.

First of all let me thank you again for your effort!

I can offer 3 sets of measurements, all from LP.

Blue is left channel, green is right. Top 2 are MMM taken from "head area". 2 in the middle are MMM taken from seat area 80cm*50cm*30cm (height) and 2 at the bottom are single point sweeps from the center of LP at ears level. As you can see no major dips exist in 400-800Hz area. My correction was based on 2 responses in the middle as I like it to be as spatially robust as possible.

 

[andreasmaaan]

@QMuse I've always thought it strange that your speaker uses two midwoofers at such a distance from each other.

Have you considered taking nearfield measurements of each to confirm it's not in fact a 2.5-way design?

 

[tuga]

Always found those Castle very intriguing.

Do you know if the top woofer signal is delayed?

 

[QMuse]

@QMuse I've always thought it strange that your speaker uses two midwoofers at such a distance from each other.

Have you considered taking nearfield measurements of each to confirm it's not in fact a 2.5-way design?

Yes, I have measured both nearfield and indeed both midwoofers are working up to 1800Hz.

There was a larger model (Castle Howard) which had upper woofer working only up to 300Hz, but when listening both of them side by side mine sounded better.

 

[QMuse]

@QMuse I've always thought it strange that your speaker uses two midwoofers at such a distance from each other.

Have you considered taking nearfield measurements of each to confirm it's not in fact a 2.5-way design?

My speakers were designed by Steve Stanley Hewlett and IMHO they represent one of the best designs made in Britain those days. I would be very interested in asking him how he avoided that phase inverted bass coming from twin pipes TL at the bottom of the speaker also don't cause cancellation but unfortunately he passed away from cancer 2 years ago.

 

[andreasmaaan]

I would be very interested in asking him how he avoided that phase inverted bass coming from twin pipes TL at the bottom of the speaker also don't cause cancellation but unfortunately he passed away from cancer 2 years ago.

How far apart are the exits from the pipes?

I'd be more interested in asking him how he avoided cancellation between the direct radiation of the two woofers!