Loss of sense of dynamics for home speakers is most likely due to non-linearity of drivers that cause compression. If we just look at signals of music, the strong low frequency content sometimes drive the audio signal into clipping, and not only the low frequency content gets clipped, the higher frequency content with the low frequency content get's clipped as well, and thus gives sens of compression during listening. The same thing can also happen when drivers reach excursions into the highly non-linear regions where the peaks get clipped off. Just send a 50 sine wave through a woofer and see at what level visible distortion in the measured sine wave occurs. When other signals are mixed with this sine wave, they will become distorted as well, and usually percieved as compression.
When we get into higher listening levels, then thermal compression also becomes a factor which gedlee had mentioned. In this situation, complex passages without large amounts of low frequency contents can also sound compressed.
Loss of sense of dynamics can also occur when the playback polarity is not the same as the recorded polarity.
When we get into higher listening levels, then thermal compression also becomes a factor which gedlee had mentioned. In this situation, complex passages without large amounts of low frequency contents can also sound compressed.
Loss of sense of dynamics can also occur when the playback polarity is not the same as the recorded polarity.
ucla88 said:
Exactly my reason for the post.
It is well know that drivers can exhibit sub-harmonics and yet theoretically no time invariant nonlinearity can generate a sub-Harmonic, but a time variant nonlinearity can. A sub-harmonic is virtually unmasked and would be highly audible. This interests me.
soongsc said:
I believe that both of these statements are untrue.
All of my recent research says that nonlinearity in a loudspeakers is virtually inaudible and yet we know that there are audible problems. I suspect that thermal issues are involved.
No research has ever shown your last statement to be true. Please quote some if you know of any.
I have the Gander article right here in my hands (it is not only available as preprint, it was also published in the "Journal", September 1986).
He is talking of thermal time constants of less than a second for the voice-coil alone. The magnet/basket structure has a much longer time constant. It might be quite complex to model their interaction accurately IMO.
The paper is also dealing with nonlinear distortion and alingment changes with temperature rise.
Thermal compression was an issue for the Dynaudio engineers from the very beginning of the company. The first time I saw driver thermal power compression mentioned at all was in the 21W54 datasheet ! If you consult the data sheets of their ferrofluid cooled tweeters you will find frequency response measurements for bursts up to 1 kW !
Regards
Charles
He is talking of thermal time constants of less than a second for the voice-coil alone. The magnet/basket structure has a much longer time constant. It might be quite complex to model their interaction accurately IMO.
The paper is also dealing with nonlinear distortion and alingment changes with temperature rise.
Thermal compression was an issue for the Dynaudio engineers from the very beginning of the company. The first time I saw driver thermal power compression mentioned at all was in the 21W54 datasheet ! If you consult the data sheets of their ferrofluid cooled tweeters you will find frequency response measurements for bursts up to 1 kW !
Regards
Charles
Some numbers
Seems good to have in mind a few numbers when thinking about this very interesting topic.
Mass of the wire in a voice coil covers 3 orders of magnitude (I'd not really thought about that before) from ~100mg in a dome tweeter to ~100g in a pro woofer ("~" = of order so please don't tell me I'm wrong by a factor of 2).
Tempco of copper is around 0.004/K at room temperature, so it takes ~25K to make 1 dB (10%) compression.
Heat capacity of 1g of copper is around 0.4J/K so it takes (at least) 10J to cause 1dB compression in a 1g coil (or only ~1J in a dome tweeter, or a "huge" 1kJ in a pro woofer). All assuming instantaneous heating.
At few W levels (loud for me) the dome tweeter could experience a few mJ heat changes within its bandwidth (Fourier components within bandwidth), leading to quite small distortions within its bandwidth (so that does not seem to be a mechanism that will explain much).
OTOH heating due to ~100 ms bursts (from musical rhythms) could start to produce quite substantial modulation sidebands (~10%) on the in-band sound. TBH I don't know if that is audible or not, but it will produce messy spectra. Modulating music with its rhythm does not seem a very good idea.
Hope I got the numbers right.
Ken
or at least better than my spelling...
hmm - I wonder what the real thermal time constant of a tweeter voicecoil is?
Seems good to have in mind a few numbers when thinking about this very interesting topic.
Mass of the wire in a voice coil covers 3 orders of magnitude (I'd not really thought about that before) from ~100mg in a dome tweeter to ~100g in a pro woofer ("~" = of order so please don't tell me I'm wrong by a factor of 2).
Tempco of copper is around 0.004/K at room temperature, so it takes ~25K to make 1 dB (10%) compression.
Heat capacity of 1g of copper is around 0.4J/K so it takes (at least) 10J to cause 1dB compression in a 1g coil (or only ~1J in a dome tweeter, or a "huge" 1kJ in a pro woofer). All assuming instantaneous heating.
At few W levels (loud for me) the dome tweeter could experience a few mJ heat changes within its bandwidth (Fourier components within bandwidth), leading to quite small distortions within its bandwidth (so that does not seem to be a mechanism that will explain much).
OTOH heating due to ~100 ms bursts (from musical rhythms) could start to produce quite substantial modulation sidebands (~10%) on the in-band sound. TBH I don't know if that is audible or not, but it will produce messy spectra. Modulating music with its rhythm does not seem a very good idea.
Hope I got the numbers right.
Ken
or at least better than my spelling...
hmm - I wonder what the real thermal time constant of a tweeter voicecoil is?
I am quite interested what kind of signal you used during testing and what the amplitude was?gedlee said:
I personally ran signals through different 6" or so drivers and find that they normally start clipping or distorting visibly in the data starting from 20V p-p or so which is not many watts. Believe me, I think it's quite unbearable when distortion starts to take place.
I am sure that thermal is an issue also, specially when very complex music passages are played at high levels. But with a speaker that has good resolution calability listend at low levels, it may well not be the case.
I think the cause is too obvious when only conducting listing tests that it's hard to make a research project out of it. I have demonstrated it to too many people and to date everyone had been able to hear it.
soongsc said:
We used musical passages at normal listening levels.
I've heard this one before - "its too obviuos to be worth proving" - Boy you can hide a lot of falsehoods behind that one.
Re: Some numbers
Ken - thanks for this, most enlightening. Yes I did find that the tweeter was responsible for the vast majority of thermal issues. This being in the ears most sensitive region is why I think that this is an issue.
A small tweeter would have a VC time constant in the 10's of ms. Magnet etc. much much much longer.
I have thought about heating the VC to a certain temp and then just holding it there as a way to reduce the thermal modulation. But its far easier just to use different wire.
The issue that I am trying to get at is how audible all this is. To do that you have to find a way to quantify the effect. Then you can control or compare different levels and determine audibility. But you first have to have a scale by which to measure what it is you are trying to study.
kstrain said:
Ken - thanks for this, most enlightening. Yes I did find that the tweeter was responsible for the vast majority of thermal issues. This being in the ears most sensitive region is why I think that this is an issue.
A small tweeter would have a VC time constant in the 10's of ms. Magnet etc. much much much longer.
I have thought about heating the VC to a certain temp and then just holding it there as a way to reduce the thermal modulation. But its far easier just to use different wire.
The issue that I am trying to get at is how audible all this is. To do that you have to find a way to quantify the effect. Then you can control or compare different levels and determine audibility. But you first have to have a scale by which to measure what it is you are trying to study.
Depending on definition of listening levels and speaker sensitivity, the power range can vary significantly. Perhaps more specifically listening distance, speaker sensitivity, power amp capacity? Perhaps the speakers or and amp were never driven into the clipping region. Then there is even the amps capability to handle speaker back EMF. Really just too many aspects.gedlee said:
I'm sure you have. I'm also sure nobody wants to prove something using methods that are more than necessary. But if this kind of a topic might qualify for a PHD thesis, maybe it might be worth complicating the issue.😉
Not everyone is going to be correct all the time, but if people are willing to share their own experience in how to select drivers and design speakers that have realistic perceived dynamics, I think we should just be grateful.
One test we did on a speaker system changing between the power amp power supply. The voltages were adjusted to be the same, and we felt that both power supply should be able to deliver at least the same power. One power was a switching power supply, and the other was a linear power supply. We were expecting the switching power supply to give better dynamics because it was quite fast. In reality, at the same volume setting, the switching power supply provided less low frequency dynamics and kept tripping the protection circuit causing it to switch in and out, while the linear power supply just worked normally.
One other test we did was this same amp but using a different transformer topology. Both transformers were ordered from the same manufacturer with the same rating, one was an O core, another was an R core, very obviously the transient capabilities of the O core were very much better than the R core.
We did not spend the time to furhter identify what was causing this, but just the switch on current was so different that a fuse that blows using one transformer does not blow using the other transformer.
I point this out just to let people be aware there are other factors that may change the perceived dynamics of a speaker considerably. There are so many factors that can make us make a wrong judgement that we just have to sort these things out little by little.
One other test we did was this same amp but using a different transformer topology. Both transformers were ordered from the same manufacturer with the same rating, one was an O core, another was an R core, very obviously the transient capabilities of the O core were very much better than the R core.
We did not spend the time to furhter identify what was causing this, but just the switch on current was so different that a fuse that blows using one transformer does not blow using the other transformer.
I point this out just to let people be aware there are other factors that may change the perceived dynamics of a speaker considerably. There are so many factors that can make us make a wrong judgement that we just have to sort these things out little by little.
Back in probably the 80's, when I just started out with interest in the dynamics of speakers, I once asked a manufacture what the frequency response was at the maximum rated wattage of their driver, and I was told to "Do your own tests".
Now that test equipment is much more affordable, I just find that not one single test is the answer to determining perceived speaker dynamics, but rather, different tests are necessary to determine what might be done to improve the perceived speaker dynamics.
Now that test equipment is much more affordable, I just find that not one single test is the answer to determining perceived speaker dynamics, but rather, different tests are necessary to determine what might be done to improve the perceived speaker dynamics.
I like drums but don't know if practical to compare a recorded wav of snare/tom.kick drum and combo to recorded loudspeaker playing those wav.
Re: Re: Some numbers
I thought that Cu was best of the common materials in this respect: at least it seems to be better than Al. Lets take the magnet gap and VC length of a unit as fixed. Then, because of its higher resistivity, Al wire has to have fewer turns to get the same resistance in the same volume producing a weaker motor. The current needs to be a bit higher for the same sound output (setting aside the lower mass of the voil for the moment).
The specific heat capacity of Al is higher than Cu, but because of the lower density the heat capacity per unit volume is only about 2/3 that of Cu. The change in resistance /K is also a little larger than Cu so Al loses out all 3 ways.
(Of course Al is lighter which wins back some, and makes it attractive in pro woofers where the main thermal problem is the long term one, and the mass reduction most significant.)
... I'm very also curious to understand what processes lead to audible effects due to short term heating, particularly in tweeters, and will follow your thread with interest.
Ken
gedlee said:
I thought that Cu was best of the common materials in this respect: at least it seems to be better than Al. Lets take the magnet gap and VC length of a unit as fixed. Then, because of its higher resistivity, Al wire has to have fewer turns to get the same resistance in the same volume producing a weaker motor. The current needs to be a bit higher for the same sound output (setting aside the lower mass of the voil for the moment).
The specific heat capacity of Al is higher than Cu, but because of the lower density the heat capacity per unit volume is only about 2/3 that of Cu. The change in resistance /K is also a little larger than Cu so Al loses out all 3 ways.
(Of course Al is lighter which wins back some, and makes it attractive in pro woofers where the main thermal problem is the long term one, and the mass reduction most significant.)
... I'm very also curious to understand what processes lead to audible effects due to short term heating, particularly in tweeters, and will follow your thread with interest.
Ken
Re: Re: Re: Some numbers
Actually CU is one of the worst. If you add about 12% Nickel to the CU then the coefficient of resistance with temperature goes almost to zero. This is sometimes called Constantin because its resistance is almost completely independent of temperature. BUT, the downside is that the resistivity goes sky high. The sweet spot is about 5-7% Nickel because you get the best lowering of thermal change with the least resistivity rise.
I don't have any data on the audibility of thermal effects. Thats the point. I want a test to quantify the effect so that I can set up a test to scale it audibly.
kstrain said:
Actually CU is one of the worst. If you add about 12% Nickel to the CU then the coefficient of resistance with temperature goes almost to zero. This is sometimes called Constantin because its resistance is almost completely independent of temperature. BUT, the downside is that the resistivity goes sky high. The sweet spot is about 5-7% Nickel because you get the best lowering of thermal change with the least resistivity rise.
I don't have any data on the audibility of thermal effects. Thats the point. I want a test to quantify the effect so that I can set up a test to scale it audibly.
Many driver designs tend to increase the cross section of vc wire. The trade-off is higher mass thus limiting bandwidth.
Motor flux modulation probably is another thing to consider If one is into driver design.
Motor flux modulation probably is another thing to consider If one is into driver design.
soongsc said:
This is easily solved with a shorting ring and I would never use a driver that didn't have one.
Earle, in your first post, you qualified limiting responses to actual tests,
"Please, analytical tests only. I'm not really interested in subjective interpretations of a listening test."
IMO, it's actual listening to music that will highlight an issue. For example, I don't like dome tweeters for a number of reasons. Yet to date I have not seen specs to indicate why my dislike exists.
Perhaps this could reveal something. If you generate a pink noise source starting at a specified level, then decaying at a controlled rate, and observed the result on a CSD plot, you would see amplitude variances over the spectrum.
Reversing the envolope, ie starting at the noise floor, and increasing amplitude should produce a mirror image of the first - if heat is not an issue.
Geoff.
"Please, analytical tests only. I'm not really interested in subjective interpretations of a listening test."
IMO, it's actual listening to music that will highlight an issue. For example, I don't like dome tweeters for a number of reasons. Yet to date I have not seen specs to indicate why my dislike exists.
Perhaps this could reveal something. If you generate a pink noise source starting at a specified level, then decaying at a controlled rate, and observed the result on a CSD plot, you would see amplitude variances over the spectrum.
Reversing the envolope, ie starting at the noise floor, and increasing amplitude should produce a mirror image of the first - if heat is not an issue.
Geoff.
My experience with the shorting ring is that it induces a different pattern of energy storage that is evident when looking at the impulse. The inductance was indeed lowered which was evident in the SPL graph, but better response was possible altering the cones property. It could be that the ring was not properly designed. It was later requested to be taken out until more analysis was conducted.gedlee said:
Re: Re: Re: Re: Some numbers
Ken
(and searching this topic I came quite quickly to your gedlee.com/patents page)
Ah I see, thank you! I had not realised there was a useful alloy like that.gedlee said:
Ken
(and searching this topic I came quite quickly to your gedlee.com/patents page)
Geoff H said:
This is the point isn't it? If you can't quantify something that you think that you hear then you are just chasing your tail. The ear is simply not a reliable instrument.
While I do decry casual listening "tests", I must admit to going into our listening room virtually every day and listening. If there is something that I don't like after a very long time (weeks to months), I would certainly look into it. I have never found something that I could reliably hear that I could not find in the measurements.
If you don't have a nice tight closed loop correlating the subjective to objective tests then, for the most part, things will go astray. Most people start with the simplified subjective tests, but never close the loop. This is no way to find optimal designs of loudspeakers. I start with the objective test, show that it is reliable, repeatable and shows differences in systems and then I go on to scale the subjective effect using the objective test as the metric (ruler). Now I know the level of the effect subjectively and I can quantify a design or change on this subjective scale. NOW I can design the system to optimize it for the best subjective impression. Anything else is guesswork.
We currently do this test (the second one, but with white noise) and put it in our data sheets. But these are more steady state thermal compression issues. Its the very short 1-2 ms. variations that I am interested in with this post.
- Status
- Not open for further replies.