Do you mean that 'thermal compression' is long term thermal soak and 'power compression' is nearer instantaneous thermal fluctuations?
I suppose no matter what we do for cooling the driver, little other than improved efficiency is able to reduce thermal issues arising from heat that can not propagate out of the VC material fast enough to be dissipated elsewhere (heat propagation speed though copper with enamel coating for example).
FWIW I'm not looking to build a high output tweeter... I just wondered what the cause.
Cymbal, tubular bells, small bells, triangle, top octave Steinway or Bosendorfer piano. Those racks of tubes percussionists stroke that are not listed in wikipedia list of percussion instruments.
I think the reason vendors do not reproduce 15-20 khz is that 99% of males over age 12 cannot hear it. I can only hear to 14 khz.
One tweeter that is only down 5 db 15-20 khz is eminence N314X-8 .
indianjo,
Thanks!
I would really appreciate seeing a Spectrogram of:
Cymbal, tubular bells, small bells, triangle, top octave Steinway or Bosendorfer piano.
Do they show a power falloff rate, or the lack of a power falloff rate of the upper harmonics?
A square wave is a good example where the Total Power of all the Odd Harmonics is equal to the Total Power of the Fundamental.
10 Watt Harmonic Power = 10 Watt Fundamental Power; That gives 20Watts Total Power of the complete square wave.
These are all Odd Harmonics.
The odd harmonic voltages fall off at the following rate: 1/3 third H, 1/5 5th H, 1/7 7th harmonic, . . . etc.
Often forgotten:
If a square wave is modified to have un-equal up-time versus down-time (48% and 52%, for example), then Even Harmonics are also present in the spectrum.
All the Even Order Harmonics have the same amplitude level, there is no high frequency rolloff as the Even Harmonic order number goes up.
But at the same time, the Odd Harmonics are falling off at the Odd harmonic's higher frequencies.
That results at a frequency where the Even Harmonics have more voltage than their neighboring Odd Harmonics voltages.
Spectrograms are great learning tools, when used judiciously.
I suspect the JBL O75 Tweeter frequency response (Ice Pick to the Brain) does not roll off earlier than 20kHz.
Thanks!
I would really appreciate seeing a Spectrogram of:
Cymbal, tubular bells, small bells, triangle, top octave Steinway or Bosendorfer piano.
Do they show a power falloff rate, or the lack of a power falloff rate of the upper harmonics?
A square wave is a good example where the Total Power of all the Odd Harmonics is equal to the Total Power of the Fundamental.
10 Watt Harmonic Power = 10 Watt Fundamental Power; That gives 20Watts Total Power of the complete square wave.
These are all Odd Harmonics.
The odd harmonic voltages fall off at the following rate: 1/3 third H, 1/5 5th H, 1/7 7th harmonic, . . . etc.
Often forgotten:
If a square wave is modified to have un-equal up-time versus down-time (48% and 52%, for example), then Even Harmonics are also present in the spectrum.
All the Even Order Harmonics have the same amplitude level, there is no high frequency rolloff as the Even Harmonic order number goes up.
But at the same time, the Odd Harmonics are falling off at the Odd harmonic's higher frequencies.
That results at a frequency where the Even Harmonics have more voltage than their neighboring Odd Harmonics voltages.
Spectrograms are great learning tools, when used judiciously.
I suspect the JBL O75 Tweeter frequency response (Ice Pick to the Brain) does not roll off earlier than 20kHz.
It the JBL 075 does roll off below 20kHz , but if it's diaphragm is replaced with the 2405 (slot tweeter) the response rises with with frequency, around +6dB at 20kHz compared to 5kHz as I recall from the one sample I tested. In that configuration they were as ultrasonic transmitter/receivers in traffic light controllers for rural areas. Many of those drivers were stolen and used in PA systems..
Acoustic cymbals and bells can have large amplitude harmonics above 10kHz, I've seen examples on RTAs when cuing up live mix channels where their response looks "flat" from 5kHz to 20kHz (the limit of the RTA).
Some muted trumpets have surprised me too, harmonics acoustically removed by atmospheric absorption at relatively short distance show up in the close mic.
This is has a fairly comprehensive list of acoustic instruments with :
https://www.tnt-audio.com/casse/life_above_20khz.pdf
There’s Life Above 20 Kilohertz!
A Survey of Musical Instrument Spectra to 102.4 KHz
James Boyk
California Institute of Technology
"I. Introduction
Each musical instrument family — strings, winds, brass and percussion — has at least one member which produces energy to 40 kHz or above.
Some of the spectra reach this work's measurement limit of 102.4 kHz.
Harmonics of French horn can extend to above 90 kHz; trumpet, to above 80; violin and oboe, to above 40; and a cymbal crash shows no sign of running out of energy at 100 kHz. Also shown in this paper are samples from sibilant speech, claves, a drum rimshot, triangle, jangling keys, and piano.
The proportion of energy above 20 kilohertz is low for most instruments; but for one trumpet sample it is 2%; for another, 0.5%; for claves, 3.8%; for a speech sibilant, 1.7%; and for the cymbal crash, 40%. The cymbal's energy shows no sign of stopping at the measurement imit, so its percentage may be much higher."
the cause of
A passive filter to allow more power to compensate for the tweeters falling response above 12kHz could easily be doubling power at 20kHz relative to below.
Simple thermal compression from the voice coil heating and less excursion to transfer more heat generated could amount for the ~1.25dB compression.
Last edited:
There could be truth in this. Thermal compression seems less likely, as most tweeters consume less current at higher frequencies due to the impedance of the VC. The other explanation I thought of is nonlinearity of the dome suspension, at higher frequencies and relatively large excursions the behaviour could be less linear.
That would be a measurement issue then. But it could well be.
I am amazed!
One thing to think about is how effective the low pass filters are ahead of, and inside of microphone preamps, sound boards, and various recorders.
At many recorder's sampling rates, unfiltered harmonics are going to be mirrored down into the audio range (due to Nyquist number versus harmonic and sampling rate).
Most microphones are low pass filters (at some high frequency, depending on the microphone).
I have sat in the band, in front of French Horns and Trumpets, with very little high frequency attenuation due to the short distance.
With my then youthful ears, no wonder they sounded very hot / tizzy / intermodulated.
I do not hear above about 14kHz or 15kHz, but any high level harmonics in a recording system that are not filtered will be mirrored down into the audio band where we can hear them.
I think that may explain why certain recordings sound bad.
One thing to think about is how effective the low pass filters are ahead of, and inside of microphone preamps, sound boards, and various recorders.
At many recorder's sampling rates, unfiltered harmonics are going to be mirrored down into the audio range (due to Nyquist number versus harmonic and sampling rate).
Most microphones are low pass filters (at some high frequency, depending on the microphone).
I have sat in the band, in front of French Horns and Trumpets, with very little high frequency attenuation due to the short distance.
With my then youthful ears, no wonder they sounded very hot / tizzy / intermodulated.
I do not hear above about 14kHz or 15kHz, but any high level harmonics in a recording system that are not filtered will be mirrored down into the audio band where we can hear them.
I think that may explain why certain recordings sound bad.
You bet. To elaborate, ferrite comes in various grades, and for applications like transformer cores, some grades are designed for low loss at high switching speed, at the expense of maximum flux density. Without knowing the details of the speaker in question, we can only assume, but if it uses a ferrite motor, it seems reasonable that they chose a grade for a good balance of maximum field strength vs frequency response up to 20kHz.
Soft iron probably plays an equally important role, as it not only allows a much stronger field to be concentrated next to the voice coil, but it also 'buffers' the ferrite. Ideally, all of the flexing in the field lines happens only in the iron, and the ferrite is protected against being demagnetized.
At the power limit, what could be happening is the field lines in the portions of the ferrite closest to the glue line are just starting to move, creating a sort-of temporary air gap. Then, as the coil current drops to zero, 2x per cycle, the field strengthens again, but the ferrite needs time to 'click' back into place.
FEA simulations are showing deformation of the coil carrier so it could well be a mech limitation! (together with other side effects)
And I remember long time ago having read something like "self-induced heating" , where proximity/skin effect has a freq dependent skin depth(EM wavelength) , prohibiting the field of the coil reaching the iron pole . But the aluminium carrier is in very close position , short-cutting the VC field!
Thinking aluminium carrier will cool down the coil is a short thinking in case of a small tweeter . Every object has a natural heat capacity which depends on it's thermal mass . So it is a time related process and not a single value to consider ...
Fisix make your head go red 🙂
And I remember long time ago having read something like "self-induced heating" , where proximity/skin effect has a freq dependent skin depth(EM wavelength) , prohibiting the field of the coil reaching the iron pole . But the aluminium carrier is in very close position , short-cutting the VC field!
Thinking aluminium carrier will cool down the coil is a short thinking in case of a small tweeter . Every object has a natural heat capacity which depends on it's thermal mass . So it is a time related process and not a single value to consider ...
Fisix make your head go red 🙂
Yes, we might put equal voltage in, but it is not receiving equal power as the inductance tends to rise a lot.
The Wharfdale shown does not have a waveguide so I don't think that is the reason.
The highest frequencies on the PS Audio Aspirin have an interesting behaviour while the lower tweeter range remains under control. This is a push pull planar magnetic.
Corner mass should not be an issue here. However the cause could be totally different than in the dome tweeters.
From Erin's Audio Corner
@bikinpunk - Did you do the compression testing with a log sine sweep like you (and I) usually do? eg. ~2.7 seconds
Here I test a midrange with a log sine sweep with a sampling rate/sampling size such that the sweep takes a short duration- just 3 seconds.
First, with a voltage level that produces 92dB @1m.
Then with a voltage level that produces 102dB @1m.
Then with a voltage level that produces 112dB @1m.
Then we normalised to the baseline of 92dB and plot it on one graph:
The legend represents the cursor point @10KHz-
Testing with drive level that should produce 102dB, by the time sweep goes to 10KHz, we only get an extra 9.4dB instead of the whole 10dB.
(ie. max output 101dB)
Testing with drive level that should produce 112dB, by the time the sweep goes to 10KHz, we only get an extra 17.8 dB instead of the whole 20dB.
(ie. max output 109.8dB)
But why is there ANY compression at 102dB... when it can play at almost 110dB?
Well, remember that the highest frequencies are not excursion related. They are generally thermally related. It is due to the very test methodology ie. (Farina) log sine sweep, which goes from the lowest frequency to the highest frequencies and generally takes seconds, will result in thermal limiting of the voice coil/magnet system.
Reference:
Previously (dis)covered in looking for a tweeter that can reach >110dB@1m
If we do impulsive testing eg. Stepped sine taking less than <50ms, then the measured compression behaviour is different because the effect of the test on the system is different. Blimey! That pesky Observer effect getting in the way of our fun again!
So, again, the testing is the easy bit.
The interpretation is the challenging bit.
I personally don't think that music program material has that much energy in the top octave, for sustained periods, that would cause the kind of compression effect that is observed with a log sine sweep.
In general, in a log sine sweep, the top octave compression should be view with scientific scepticism.
On the other hand, any bottom octave compression reveals the displacement limits of the system, and middle-range compression is suggestive of dynamic range limitations.
Further reading:
Reference:
7th Klippel Live session- Amplitude Compression, slide 28.
Here I test a midrange with a log sine sweep with a sampling rate/sampling size such that the sweep takes a short duration- just 3 seconds.
First, with a voltage level that produces 92dB @1m.
Then with a voltage level that produces 102dB @1m.
Then with a voltage level that produces 112dB @1m.
Then we normalised to the baseline of 92dB and plot it on one graph:
The legend represents the cursor point @10KHz-
Testing with drive level that should produce 102dB, by the time sweep goes to 10KHz, we only get an extra 9.4dB instead of the whole 10dB.
(ie. max output 101dB)
Testing with drive level that should produce 112dB, by the time the sweep goes to 10KHz, we only get an extra 17.8 dB instead of the whole 20dB.
(ie. max output 109.8dB)
But why is there ANY compression at 102dB... when it can play at almost 110dB?
Well, remember that the highest frequencies are not excursion related. They are generally thermally related. It is due to the very test methodology ie. (Farina) log sine sweep, which goes from the lowest frequency to the highest frequencies and generally takes seconds, will result in thermal limiting of the voice coil/magnet system.
Reference:
Previously (dis)covered in looking for a tweeter that can reach >110dB@1m
If we do impulsive testing eg. Stepped sine taking less than <50ms, then the measured compression behaviour is different because the effect of the test on the system is different. Blimey! That pesky Observer effect getting in the way of our fun again!
So, again, the testing is the easy bit.
The interpretation is the challenging bit.
I personally don't think that music program material has that much energy in the top octave, for sustained periods, that would cause the kind of compression effect that is observed with a log sine sweep.
In general, in a log sine sweep, the top octave compression should be view with scientific scepticism.
On the other hand, any bottom octave compression reveals the displacement limits of the system, and middle-range compression is suggestive of dynamic range limitations.
Further reading:
Reference:
7th Klippel Live session- Amplitude Compression, slide 28.
Last edited:
It would be very interesting to reverse the sweep so highs come first and compare with the normal low to high sweep.
Some of the heat limiting the top octave may have come from the lower stimulus tones.
Some of the heat limiting the top octave may have come from the lower stimulus tones.
Could it be so simple, that when a small Sd tweeter must try to use high excursion to prodece high spl, it doesn't have time to do that? So it compresses because F is fixed but excursion/amplitude is not. Another reason is that suspension and membrane shape nonlinearity/resonances increase with high spl/excursion, causing decrease of fundamental but increasing harmonics (we can't measure harmonics of 15-25kHz with most measurement systems)
https://www.klippel.de/test-objects/high-frequency-drive-units.html
Video of cone resonance
FineCone demo https://loudsoft.com/finecone/
Early domes were textile and 19-20mm diameter. They may have high efficiency but limited spl capacity. Larger 27-29mm domes with more rigid membranes behave more linearly with high spl, but compression drivers make it still better with their large Sd and compression chambers. This is old and common knowledge, I assume...
https://www.klippel.de/test-objects/high-frequency-drive-units.html
Critical Issues
- Flat SPL response on-axis and off-axis in the far field
- Maximal peak displacement limited by suspension nonlinearities
- Heating of the coil and thermal power handling
- Rocking modes causing voice coil rubbing
- High local displacement on particular points on the diaphragm causing nonlinear distortion
Video of cone resonance
Early domes were textile and 19-20mm diameter. They may have high efficiency but limited spl capacity. Larger 27-29mm domes with more rigid membranes behave more linearly with high spl, but compression drivers make it still better with their large Sd and compression chambers. This is old and common knowledge, I assume...
Last edited:
Yes. My method for short and long term compression is listed on my site for every review. I haven't had a chance to get the TAD posted there yet but the method is the same as all speakers I've tested before.
Lmao - this forum 😆
This is a constant(-ish) directivity design. A CD speaker by definition loses 6 db SPL / oct on axis. To attain a flat frequency response on axis, the speaker has to be filtered flat, meaning receive 6 dB more amplifier power per octave. Hence, an increased power load with rising frequency, hence rising electrical-thermal compression with frequency. The end.
Every serious full spectrum power compression measurement is performed with a noise or multisine signal, to cause constant heating over both time and frequency. I hope Erin doesn't use a sine sweep, as that would, as someone here at least correctly remarked, distort the results, in that heat accumulated in the coil would rise with frequency - which would in this case then compound the compression in the VHF (very high frequencies).
This is a constant(-ish) directivity design. A CD speaker by definition loses 6 db SPL / oct on axis. To attain a flat frequency response on axis, the speaker has to be filtered flat, meaning receive 6 dB more amplifier power per octave. Hence, an increased power load with rising frequency, hence rising electrical-thermal compression with frequency. The end.
Every serious full spectrum power compression measurement is performed with a noise or multisine signal, to cause constant heating over both time and frequency. I hope Erin doesn't use a sine sweep, as that would, as someone here at least correctly remarked, distort the results, in that heat accumulated in the coil would rise with frequency - which would in this case then compound the compression in the VHF (very high frequencies).
