kipman725

Here is the impedance compared with simulation:

So finally managed to find a place I can do outdoor testing! its a yard so not reflection free but a lot better than testing indoors.  You can see that the results are matching the in box measurements to around 100Hz.  What I don't get is why this sub is so efficient in the upper bass, the numbers seem impossible? >110dB/1W?

In the Hornresp sim I used a semi inductance model generated from impedance measurements of the driver and simulated as a bass reflex.  The box is a bit different in that the port goes into a cavity with the (isobaric) driver and that there is obviously quite a long acoustic path from one side the of the driver to the other.

Seems to work when the port is linear.  InBox is the compensated SPL inside a bass reflex cabinet which has dimensions that should result in valid data <80Hz.  HornResp is the simulation of this sub.  Room compensated was generated by measuring a known subwoofer in the same physical location as the DUT.  This known subwoofer is a sealed box and has previously been measured using close mic spliced with time gated measurements.   

You can see that the response shape for the in box measurement matches that of the compensated response.  This box differs a bit from the Hornresp simulation in that there exists a cavity into which the port and driver fire which may increase SPL outside the low bass.

Next stage I guess is to make a microphone that can cope with those savage in box pressures!

Has anyone experience with microphone in box measurements?

https://www.audioxpress.com/article/measuring-loudspeaker-low-frequency-response

I was thinking with a transducer that remained linear at high pressures such as the MPX2010 series:

https://docs-emea.rs-online.com/webdocs/0ef4/0900766b80ef40f2.pdf

The maximum SPL of capabilities of ported and sealed boxes could be measured without resorting to ground plane measurements.  I'm not sure though what port air velocity saturation would do to the relationship between internal and external pressure?

3 hours ago, lilmike said:

I know, right??

Please, someone buy the 9601s so I don't. 

would if I was in the US

On 10/2/2019 at 9:06 PM, Ricci said:

Also...Consider that you will still need to be wary of very high duty cycle signals over the short term. Things like air and water cooling will help with long term heat buildup in the motor and coil, but not so much with rapid heating of the coil over a short time period. Coils can be burned VERY quickly in some cases with a bad scenario. With that said I wouldn't assume that you will be able to increase the short term peak signal handling of the system at all and this is what has the possibility of reaching driver Xmax for the most part. The longer term signals that make most of the heat buildup aren't normally going to be causing excursion issues. In fact greater excursion helps cool the voice coil and motor. With that said I definitely wouldn't go smaller with the cabinet. 

If this method was implemented the short term power handling would also increase:

https://patents.google.com/patent/US4757547A/en

Active driver cooling check out "RC1 soundsystem" and "real horns sound system" real horns are using water cooling for the 8 drivers in the bass horn and also realtime excursion measurement while RC1 have an 'active cooling system' which I think is water cooling.  Tom Danley also has a patent on using a air blower through the pole vent.  The main benefit is long term power handling as the magnet can be kept cool and heat removed from the box.

2 hours ago, peniku8 said:

If your 4 tops are clustered that would make around 133db at 1m at 128W into each top (double the power equals ~3db gain, having twice the speakers makes ~6db gain if perfectly aligned). You can substract 5-6db per doubling of distance if the tops don't have an exceptionally narrow throw (like line arrays) so it should be about 8db less at 3m. Maybe the effect is less grave at close range thou since humidity doesn't come much into play at short distances, so it could be even louder.

For SPL measurement I use my iPhone, which was calibrated with a UMIK. Factory settings were (only?) like 2db off in the "Mobile Tools" app I'm using.

Only one top per side they are 3 ways with a passive mid high crossover and active bass/mid-high.  Coverage pattern is 90x40.  For stereo you are assuming that left and right side add coherently?  Good idea on the phone! I also have UMIK to check the accuracy.

I do some PA stuff on an amateur basis, my tops are ~100 dB/1W and I run them off 4 amp channels rated as 120W (180W peak) per channel.  On some drum and bass I have seen the clip lights flicker on the power amps would I be correct in estimating that would be about 110dB peaks and an average around 100dB at 3m from the speakers?   Also what kind of instrumentation should I get to monitor SPL? as sometimes I worry that its too loud as it can be hard to tell as distortion is low. 

Another data point.  I was stood next to the mixing desk at an Aphex Twin gig (electronic music) and could see the long term average SPL readout which was hovering around 120dB (no idea of weighting) for the majority and got to around 125dB towards the end.  I would say it felt about right and they seemed to have a lot of headroom, at the end a few very low notes were hit hard without any evidence of fading or limiting.  This was in a very large lossy indoor environment (warehouse effectively).

3 hours ago, Ricci said:

Sensitivity will measure low on a dual opposed cab when measured from 1 location. I've got an article on this subject on the main website.

Can you post your model inputs and the physical characteristics of the cabinet?

It should be possible to load the pyle.dat above into Hornresp. The previous simulation was with 250L box volume but I will endeavor to be more accurate... (very little change in response from reduced volume)

The box is 0.396x0.440 (drivers mounted on this face) x1.495 m with the drivers opposed at the bottom of the box.  The box has an internal frame of 34 mm wood square planed wood at all vertices and with additional reinforcement around the drivers.  There are also cross braces above the drivers which are 15mm dowels but these take up negligible volume these also hold the stuffing in place.  Each box has 5 kG of polyfill in it filling the volume immediately above the driver.

The internal volume is 239L - 11.0 L (Braces) - 12L (estimate for two drivers) =216L

Stuffing density = 23g/L

I have attached the .zma files I used to generate the complex inductance parameters using the spreadsheet; the added mass was 163.44 g of disc magnets around the edge of the cone.  I haven't included the stuffing as I don't know what values to put in as don't think it would radically change the response shape.

I don't think the whole problem can be directivityand opposed drivers as the sensitivity is 10dB lower than expected <60 Hz.

added_mass1_163_44.zma free_air1.zma

1 hour ago, SME said:

What orientation was it in when you measured?

laid on its side on a wooden pallet with one driver facing towards the mic and one facing away.  Microphone on the ground offset two meters from the central axis of the driver facing it.

6 hours ago, SME said:

Can you plot your measurement against the simulation so we can see how they differ?  There are definitely reasons why you're likely to see differences between the two.

If you look at the picture in the post above I have also added the hornresp response with the drive level adjusted so it should match the +3dB curve in lime green.

To match the voltage in hornresp should be -3dB from 2.83V to take into account the +3dB gain over the 2.83V and -6dB due to the measurements been done at 2m resulting in 2V.  The actual speaker displays lower than expected sensitivity and different high frequency rolloff character.  I will get an impedance plot of the box tonight for further comparison.

I also managed to measure the boxes outdoors (2m ground plane).  However the results don't match my hornresp simulations, unfortunately I was quite ill when I took the outdoor measurements and so may have made sensitivity errors.  Despite that not even the shape of the response matches simulation... I have measured a few of the drivers and they seem to have consistent T/S parameters and have used the semi-inductance model.  The attached data is for the dual driver subwoofers that have two drivers on opposed faces driven in parallel.

pyle.dat DualBox.mdat

So now I have two failed drivers I have had a chance to take one of them apart.  The top plate is 6mm thick and the voice coil is 20mm tall therefore the mathematical xmax is 7mm:

https://speakerwizard.co.uk/driver-ts-parameters-xmax/

This particular driver has failed during transport by the pole piece becoming unglued and then non centered trapping the voice coil.

My sub situation is pretty marginal as its only a hobby for me.  Normally I'm running 12*15"low cost  drivers in undersized sealed boxes of an NU6000 which I equalise flat.  This has just about been enough for indoors.  However with only 8 drivers at an outdoor event I just didn't have enough sub; however to good news is that on closer examination to have only lost one driver due to thermal failure during the event and the one that failed before hand in transit so should be able to get up and running again soon.  Longer term I'm looking into changing my boxes into eight isobaric bass reflex (weak motors) which will increase output about +10dB according to hornresp and should be fairly compatible with other better drivers as an upgrade path.

I seem to have burnt a load of drivers using the same amp this weekend.  4*15" per channel (only 2.5" VC though) @ 8 ohm total load per channel, sealed boxes.  Hitting the built in limiter a lot and for sustained periods of time.  I just didn't have enough output to match my tops and the required SPL after loosing a sub due to transit damage.  Noticed that one of the drivers seemed to be stuck in place at the end of the night... (totally melted) I presume there are more.

I'm pretty sure I could make a device that could plug into the inputs and outputs of most sub amps and measure the voice coil temperature while the drivers in use.  Sounds like there would would be some interest in the device.  I have a spare voice coil and have ordered some other bits to test the method so should be able to report back next month on my progress.  I don't quite understand why amplifiers don't implement voice coil resistance sensing though as it would be very easy to inject a small DC offset and measure using that, DC protect could just be set to a higher level than the small offset. 

Here is some MATLAB code in this case the exact frequency was known in relation to the sample rate and so the carrier is locally generated and didn't need to be recorded as the same sample clock is used for the ADCs as the the DAC that generated the carrier (thus only needing two ADC channels).  This was used in another impedance measurement device that used a similar method.  As you can see its quite a simple process and could be continuous unlike if you used Fourier methods, so doesn't need a lot of memory for long integration times.

function [ Z ] = CalcZIQ(V, I, Fs, F)
%CalcZIQ calulates impedance from the voltage and current results
%   by increasing the length of the voltage and current recording length
%   less energy is integrated into the result not from the test stimulous
%James Lawson 2018
%Inputs:
%V: voltage waveform at constant sample rate [V]
%I: current waveform at constant sample rate [A]
%FS: sample rate [Hz]
%F: frequency of interest [Hz]
%Outputs:
%Z: complex impedance [ohm]

%Trim off trailing zero values (caused by result storage method sometimes)
%do this by finding first non zero value
index = length(V);
while V(index) == 0
    index = index-1;
    if index == 1
        break;
    end;
end;
if index ~= 1
    V = V(1:index);
    I = I(1:index);
end;

%Generate sine waves at the sample rate to demodulate the data
TimePoints = (0:index-1)/Fs; %Sample times
demod = sqrt(2)*sin(2*pi*F*TimePoints); %Demodulation waveforms
demod90 = sqrt(2)*sin(2*pi*F*TimePoints+pi/2);
    
%IQ modulators
ReVmult = demod.*V';
ImVmult = demod90.*V';
ReImult = demod.*I';
ImImult = demod90.*I';

%Low pass filter to remove 2*f component 
PassbandFreq = Fs/length(V)*10;
lpFilt = designfilt('lowpassiir','FilterOrder',8, ...
         'PassbandFrequency',PassbandFreq,'PassbandRipple',3, ...
         'SampleRate',Fs);
%fvtool(lpFilt)
ReVmult_filt = filter(lpFilt,ReVmult);
ImVmult_filt = filter(lpFilt,ImVmult);
ReImult_filt = filter(lpFilt,ReImult);
ImImult_filt = filter(lpFilt,ImImult);

%integrate, there is not need to do division step as all parts equaly
%effected
ReVacum = sum(ReVmult_filt);
ImVacum = sum(ImVmult_filt);
ReIacum = sum(ReImult_filt); 
ImIacum = sum(ImImult_filt);

%Calculate final impedance
Z = (ReVacum+1i*ImVacum)./(ReIacum+1i*ImIacum);

end
 

yes this is how I understand the results of this calculation.  As long as the mouths are within 1/4 wavelength they should act as if they are stacked together.

You could inject a low amplitude sine below a the audible band (~10Hz) before the power amps (IE in your DSP).  Then use a sound card on a computer to measure the driver impedance at this frequency using the pilot tone.  It would work best to use 3 sound card channels so you can recover the carrier, current and voltage and use coherent IQ demodulation to recover the signal.  Given sufficient integration time you should be able to obtain good results up to almost the full dynamic range of the ADCs.  

For non horn loaded speakers each doubling the number of speakers with each speaker driven by the same power when they are within 1/4 wavelength should be +6dB.  You get +3dB for a doubling of the input power and you get +3dB from an increasing in effective radiating area increasing the efficiency.  For horn loaded speakers there can be the additional effect of increasing the effective mouth size which can give more low end extension and efficiency if the mouth was undersized.  Notice how the bottom end is boosted and ripple reduced in the predicted acoustic output of labhorns in multiples; 

One horn:

4 horns:

The Behringer NX shouldn't clip the signal as its limiter is always in place and has zero attack.  What happens if if you request a signal output beyond what can be produced at the output stage is that the gain is reduced.  Of course the signal can be clipped earlier in the signal chain whereupon what has already been said about low crest factor waveforms is true.  Its also true that it can output 2 kW per channel continuously and the limiter could be heavily compressing the input waveform still.  Considering the power of the amp and the rating of the driver running the amplifier on its limiter for an extended period of time would result in driver failure, without any signal clipping,  -6dB from 2kW is still 500W which would eventually kill the driver.  

If you recone the driver yourself you could add in a temperature sensor or two to the voice coil former, or perhaps even a non contact thermometer if you can see the voice coil?  There are also a lot of papers on modelling voice coil temperature but they would tend to be time consuming to implement.    

8 hours ago, radulescu_paul_mircea said:

Thank you @Ricci for the response.

I didn't get the phrasing right, because in fact I know the results are good and verifiable. What I wanted to know in fact is if we could make a correction curve to get the amp out of the way completely, or a variable compensation curve.

I am thinking of using a Rigol oscilloscope to record voltage and current over time in sweeps and then create a power compression file to add to the measured response so that the only result is the drivers PC.

 

Instead create a thermal model of the driver and then you can just compare thermal models and simulate power compression:

https://www.klippel.de/fileadmin/klippel/Files/Know_How/Application_Notes/AN_19_Nonlinear_Thermal_Parameters_(Convection Cooling).pdf

This can also be extended to combined thermal models of cabs and drivers to optimise cooling. 

They don't appear to have a UK distributor unfortunately.  Only difference in performance seems to be higher impedance compared to 21DS115-4.  This could actually make quite a big system cost difference though as it looks like you could get most of the performance of the B&C using something like an inuke6000 whereas the Lavoce requires a greater voltage swing amplifier to reach the driver limits.