Joe
I had thought of the constant Re idea by using a DC current to keep the Re within some limits. Kind of complex processing however and this would seriously degrade the thermal capabilities of the driver.
But, to me, it comes down to that I don't think that I can hear thermal modulation on high power drivers like I use. Some people call this better "dynamics" - that's as good a description as any. What I want to know is if the thermal attributes are the reason for this apparent improvement in sound quality. If it is then I think that the solution is obvious - use higher power units with bigger voice coils. This IS NOT what is currently done in the industry as smaller is cheaper and the trend is to smaller and lighter all the time. I have always felt this was wrong, but I'd like some hard data to support this position.
Pan is convinced that small 1" dome tweeter have "great" thermal response and "dynamics". When I hear them I hear the exact opposite. Well, OK, lets get some real data and quite badgering about with meaningless subjective impressions.
Michael
I have a wave file now, I will post it on my web site and you can download it from there.
I had thought of the constant Re idea by using a DC current to keep the Re within some limits. Kind of complex processing however and this would seriously degrade the thermal capabilities of the driver.
But, to me, it comes down to that I don't think that I can hear thermal modulation on high power drivers like I use. Some people call this better "dynamics" - that's as good a description as any. What I want to know is if the thermal attributes are the reason for this apparent improvement in sound quality. If it is then I think that the solution is obvious - use higher power units with bigger voice coils. This IS NOT what is currently done in the industry as smaller is cheaper and the trend is to smaller and lighter all the time. I have always felt this was wrong, but I'd like some hard data to support this position.
Pan is convinced that small 1" dome tweeter have "great" thermal response and "dynamics". When I hear them I hear the exact opposite. Well, OK, lets get some real data and quite badgering about with meaningless subjective impressions.
Michael
I have a wave file now, I will post it on my web site and you can download it from there.
Joe Rasmussen said:
It's not as simple as that, Joe.
You are right in that the temperature itself isn't the problem – its the temperature change that's the cause of Thermal Distortion.
On the other hand all depends on the relationship of thermal time constant and music signals involved.
"Time constant" is a math figure – well known – rock solid – excellent to measure and to document Thermal Distortion - but not very intuitively in terms of audibility.
On top of that it dos *not* reflect how deep SPL melts down at a given power injection.
This is certainly of interest !
To show how SPL melt down occurs - again a simu covering power injection over 6 orders of magnitude – just to put things into perspective.
The upper plot shows – on a logarithmic scale – the voice coil resistance raise
The lower plot shows – on a linear scale – the melt down of SPL
Slightly above the pink trace there is usually the limit of continuous power injection, mind you.
BUT I'd recommend - we should not so much concentrate on the absolute figures - *if* we are after obvious audible impacts of Thermal Distortion.
As outlined there are basically three major sonic impacts from Thermal Distortion
- Melt down of SPL at a very large time scale – meaning after minutes and hours of power injection at the very limit of a driver (this is what is covered by the common "power compression" figures)
- Melt down of SPL at a time scale that is within the pianissimo and fortissimo course of music – meaning within 10-20sec (this is covered by - what I call- "thermal overshot")
- Melt down of SPL at the time scale of signal frequencies – directly affecting signal integrity (this is covered by wave form deformation due to TD)
Maybe now – after going through a lot of wave form deformation discussion – we should come back to investigate a little more in detail the sonic impacts of TD at the medium term.
Wave form deformation due to TD may be the juiciest part – but the most severe issue (for me) is TD at the medium term.
To explain why – lets recall that the "thermal overshot" is the highest – and most audible, when thermal time constant is in the range or less of the time span where pianissimo and fortissimo happen.
Now lets assume for a moment that the spectral content during pianissimo and fortissimo stays more or less the same.
If we now take a two way speaker design and look at what happens – It doesn't take too much imagination to realise that the *matching* of thermal time constants of each speaker plays an enormous role.
Case one : Time constant of the tweeter is much less than that of the woofer but both are equal in their SPL melt down after an infinite time
Result:
The tweeter drops in SPL much faster when fortissimo starts giving heavy tonal imbalance during that period
If you listen to heavily compressed music – at very different SPL levels – there will be *no* tonal imbalance at all, as in the long run tweeter and Woofer reach thermal equilibrium were both have equal SPL melt down
Case two: Time constant of the tweeter and that of the woofer are equal but SPL melt down of the tweeter is much steeper than that of the woofer
Result:
The tweeter drops in SPL heavier when fortissimo starts giving also heavy tonal imbalance during that period
If you listen to heavily compressed music – at very different SPL levels – there *will be* tonal imbalance as well, as in the long run tweeter and Woofer reach thermal equilibrium were both *do not* have equal SPL melt down
Now imagine to combine a tweeter with two dedicated time constants (like the Seas Millennium ferro fluid) with a single time constant woofer – and we clearly can predict – from Thermal Distortion *only* - that voicing of that two way speaker isn't valid for every kind of music at every listening level.
Not even when woofer and tweeter are operated within their linear range !
Michael
I'll post what I get here - yes.
Just take the posted wave file (www.gedlee.com) and play it such that the peak level is enough to seriously tax the driver DUT, but not burn it out. Record two channels of data first the voltage across the driver and second the current into the driver. Now if you use a sound card the way to do this is to get a small resistor of .1 ohm and put it in series with the speaker on the ground side. Then record the voltage across the speaker plus resistor (voltage) and across the voltage across the resistor only (current divided by ten). This way the ground side of the sound card input is common to both signals and the amp. Some amps don't allow this because one side of the amps output will be grounded through the sound card and this can blow up things if you are not careful so test for this. I use a large 100 uF cap on the ground leg of the amp and this works fine.
The voltage signal will likely be too high for the sound card so you can pad this out with a 10:1 resistor bridge and then the voltage and current are calibrated and their ratio, which is the impedance, is accurate. But to be sure of calibration you should record these same two signals using a resistor of know value and send this as then I can calibrate the entire chain. The actual voltage need not be high on the calibration run.
Let me know if there are any questions, or if anyone else wants to try this test.
You can also record the voltage and the SPL in the far field and I can gate out the room and give you frequency response changes correlated with the voice coil temp. Also, to actually calculate the temp I need to know the Voice coil material.
Good luck - don't blow up too many drivers!!
Just take the posted wave file (www.gedlee.com) and play it such that the peak level is enough to seriously tax the driver DUT, but not burn it out. Record two channels of data first the voltage across the driver and second the current into the driver. Now if you use a sound card the way to do this is to get a small resistor of .1 ohm and put it in series with the speaker on the ground side. Then record the voltage across the speaker plus resistor (voltage) and across the voltage across the resistor only (current divided by ten). This way the ground side of the sound card input is common to both signals and the amp. Some amps don't allow this because one side of the amps output will be grounded through the sound card and this can blow up things if you are not careful so test for this. I use a large 100 uF cap on the ground leg of the amp and this works fine.
The voltage signal will likely be too high for the sound card so you can pad this out with a 10:1 resistor bridge and then the voltage and current are calibrated and their ratio, which is the impedance, is accurate. But to be sure of calibration you should record these same two signals using a resistor of know value and send this as then I can calibrate the entire chain. The actual voltage need not be high on the calibration run.
Let me know if there are any questions, or if anyone else wants to try this test.
You can also record the voltage and the SPL in the far field and I can gate out the room and give you frequency response changes correlated with the voice coil temp. Also, to actually calculate the temp I need to know the Voice coil material.
Good luck - don't blow up too many drivers!!
No problem at all, just wanted to know if you wanted to process eletrical *and* acoustical measurements.
I do all my measurements on symmetrically inputs - anything else is calling for trouble.
So - usually - I don't include DUT *and* shunt for voltage measurements - unless you *want* me to do.
Why far filed acoustic measurement - this is of no benefit here, as far as I can see?
Michael
I do all my measurements on symmetrically inputs - anything else is calling for trouble.
So - usually - I don't include DUT *and* shunt for voltage measurements - unless you *want* me to do.
Why far filed acoustic measurement - this is of no benefit here, as far as I can see?
Michael
gedlee said:
Oh boy ! – have you done a quick calculation on that ?
To compensate for 10W only, you have do "pre load" your (woofer - for example) VC with the same 10W.
Meaning for a 8Ohm woofer you have to apply roughly 9V-dc
Have you ever seen your woofer membrane jumping out with a beefy 9V-dc applied?
Dual voice coils probably - but what a bunch of additional hassles...
Michael
gedlee said:
The only way this would work is if the heat generated by the DC currect far exceeded the hear generated by the current, I(t) associated the input signal. This is just going back to what I said pages ago: separate the temperature into a long time average and a fluctuation about the averave and look at what that tells you. This is just adding a DC component to the long time average. The thermal system still behaves as a thermal low pass filter. Tem variations about the mean will decrease in amplitude as the frequency rises.
If you want to keep Re constant make the VC out of high Tc superconducting wire and set Re by adding an external series resistance element that is externally cooled and/or made for wire with low thermal coefficient of resistivity.
gedlee said:
DUT-current measurement = voltage measurement *only* over the shunt
DUT-voltage measurement = voltage measurement *only* over the DUT
Of course acoustic measurement is of interest. But why not measuring in the *nearfield* ?
Much easier for testspeakers with no baffle at all - laying on the floor...
Also I was thinking if your file could reveal wave form deformation - what you are really interested in.
And this would require a different and more sophisticated measurement prodcedure
Michael
If your setup allows you to measure the voltage without including the shunt then fine, but this requires an issolated input on the sound card which most don't have - in other words, it requires a true voltage differential measurement and not a voltage to ground measurement. This is NOT very common. No sound card that I know of will do this. (Unless you consider ground to be between the shunt and the DUT, in which case you ahve to have a floating output on the amps since if one side is grounded then it will yield a flase measurement.) Including the shunt in the voltage measurement will allow any sound card and amp to work fine and the effect of the shunt can be removed, if desired, but its already negligable.
If you are only doing a single driver then nearfield is OK, but it won't show you crossover change effects from the heating.
You are correct that I cannot (easily) get waveform modification effects as the signal processing is all linear and will ignore these efects. But it will accurately determine the time constants of the different thermal parts of the thermal circuit which can then be used in a detailed model to simulate the waveform effects for doing a subjective test. You have to take things one step at a time.
If you are only doing a single driver then nearfield is OK, but it won't show you crossover change effects from the heating.
You are correct that I cannot (easily) get waveform modification effects as the signal processing is all linear and will ignore these efects. But it will accurately determine the time constants of the different thermal parts of the thermal circuit which can then be used in a detailed model to simulate the waveform effects for doing a subjective test. You have to take things one step at a time.
gedlee said:
Actually, pretty many of the pro-soundcards do have differential inputs and differential outputs. Symmetric in / out is essential - not only for measuring!
I take all measurements with the Mackie Onyx 400F which is sort of top notch.
Doing more subtle measurements – like done here for investigating "new" effects – a "standard" soundcard is of little help.
In fact – a decent pro soundcard (symmetric in / out 24/192) is the most sophisticated audio measurement tool available. A Neutric A1 isn't really any better (well sort of ...)
For Peter BTW:
Comparing a good soundcard to a good DSO – my DSO is 14bit resolution only and for sure there are better ones out there, but I haven't come across one with 24bit resolution yet. (did you receive my email?)
gedlee said:
Yes I agree.
But for TD measurement the quality and value of shunt plays a huge role – as we are measuring "at the border line" . This will all be outlined, but I found that I have to rearrange my thermal model first to be more user friendly before providing the promised "A How To for TD Measurement"
gedlee said:
Not sure what you mean – a SPICE simulation is also kind of "linear" analysis tool and it clearly shows "something" - as does the two tone measurement some postings back.
Matlab analysis you possibly use, shouldn't be less powerful AFAIK (though I don't use it –being 1. lazy, 2. no math-whizz ).
I send you the files for analysis as soon I get around to measure.
Michael
mige0 said:
I proposed a solution in a discussion at the Madisound board some years ago, John was joining that discussion.
Dual voicecoils biased by the outputs of the driving amplifier. IOW a class A solution where the VC acts as the bias load for the output stage.
There are many with balanced inputs but I wonder how common it is with a floating input, don't know. A ground connection (not floating) would make false readings with a grounded DUT. I learned the hard way..
Yes, you need them both. I have a 12bit Picoscope (oversamples to 16bit) for fast signals and use my pro gear for high resolution in the audioband.
I'll get back to you tomorrow.
/Peter
mige0 said:
I just use simple (cheap) soundcards for all my work. They work just fine when you get used to them and learn how to deal with the one-sided inputs.
When you do SPICE its a time step integration which can do the nonlinear calcs. When you use signal processing on real data, its all assumed to be linear.
A couple of additional comments that may be of value. If they have already been brought up ignore them. I haven't followed this thread closely and a lot has been done.
First, my brief look at the simplified physics tells me that the thermal system looks basically like a 1st order LP filter. I think you all realize this already as it has certainly been stated that temp variations can only follow the heat gen if the thermal time constant is much shorter that the period of oscillation for a sine input. The time constant is a pole in the transfer function for VC Temp.
Second thing is that for a sine input the heat generation term will be periodic at twice the input frequency. Thus any temperature variations will always happen at twice the frequency of the input signal in such a case.
Third, which regard to heat transfer across the gap I would suspect that if the thermal time constant is much greater that the period of oscillation of an input sine wave that the convective heart transfer coefficient would be sufficiently modeled by a mean or time averaged value to first order.
First, my brief look at the simplified physics tells me that the thermal system looks basically like a 1st order LP filter. I think you all realize this already as it has certainly been stated that temp variations can only follow the heat gen if the thermal time constant is much shorter that the period of oscillation for a sine input. The time constant is a pole in the transfer function for VC Temp.
Second thing is that for a sine input the heat generation term will be periodic at twice the input frequency. Thus any temperature variations will always happen at twice the frequency of the input signal in such a case.
Third, which regard to heat transfer across the gap I would suspect that if the thermal time constant is much greater that the period of oscillation of an input sine wave that the convective heart transfer coefficient would be sufficiently modeled by a mean or time averaged value to first order.
Pan said:I proposed a solution in a discussion at the Madisound board some years ago, John was joining that discussion.
Dual voicecoils biased by the outputs of the driving amplifier. IOW a class A solution where the VC acts as the bias load for the output stage.
Elegant thinking!
Pan said:
There are many with balanced inputs but I wonder how common it is with a floating input, don't know. A ground connection (not floating) would make false readings with a grounded DUT. I learned the hard way..
You don't need *true* floating inputs (transformers) – but you have to keep clear about (electric) mass referencing .
Best reading – in a very complete and comprehensive overview - I've seen so far from Bill Whitlock, Jensen Transformers, Inc.:
http://www.audiodesignline.com/showArticle.jhtml?articleID=196900873
Pan said:
If you use the soundcard also as your stimulus source you wouldn't need a scope at all.
Nothing to be measured beyond the capability of the soundcard
(assuming you have a stable (non oscillating) amp or anything)
Pan said:
Ok.
Michael
Thanks, John – its always good to throw some additional light from a different angel on the topic at hand – though most was already covered in detail.
The topic of heat isolation / transfer due to the gap and the thermal coupling mechanisms between VC and motor structure is a really interesting one – would like to know what have been already outlined in AES papers yet.
From own measurements so far, all can be broken down basically into a two time constant system.
Haven't reached the limit of this really simple aproach yet
Maybe at higher power injection levels the model breaks - we'll see....
Michael
The topic of heat isolation / transfer due to the gap and the thermal coupling mechanisms between VC and motor structure is a really interesting one – would like to know what have been already outlined in AES papers yet.
From own measurements so far, all can be broken down basically into a two time constant system.
Haven't reached the limit of this really simple aproach yet
Maybe at higher power injection levels the model breaks - we'll see....
Michael
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.