Sitting here musing about amplifier dampening factor and headphone driver resonance peaks.
Consider Sennheiser dynamic headphones; HD 600’s, HD650’s and HD 800’s plus others, they all have huge impedance / resonance peeks that need a huge amplifier dampening factor to tame.
Now I am looking at the impedance curve of the Audez LCD – 2 headphones. These planer magnetic drivers have a completely flat impedance curve without a hint of resonance. That is 70 pure resistance ohms of input impedance.
Typically when I think of vacuum tube headphone amplifiers I think of poor dampening factor. Now using these LCD – 2’s without resonance I am thinking that 30 ohms of amplifier output impedance does not matter nearly so much.
Any thoughts?
DT
Consider Sennheiser dynamic headphones; HD 600’s, HD650’s and HD 800’s plus others, they all have huge impedance / resonance peeks that need a huge amplifier dampening factor to tame.
Now I am looking at the impedance curve of the Audez LCD – 2 headphones. These planer magnetic drivers have a completely flat impedance curve without a hint of resonance. That is 70 pure resistance ohms of input impedance.
Typically when I think of vacuum tube headphone amplifiers I think of poor dampening factor. Now using these LCD – 2’s without resonance I am thinking that 30 ohms of amplifier output impedance does not matter nearly so much.
Any thoughts?
DT
I dunno about the industry, but when I see damping factor I consider the effects of driver back-EMF control, while with impedance peaks I consider effects on frequency response.
Many headphones have a resistive load. More true in the older days when outputs have high output impedance whether intentional or not - 47ohm resistor is common. Some even say some headphones were designed with that output impedance in mind.
Now with multi-driver ear/headphones, using them with high output Z is impossible.
Frequency response of high output Z and flat load Z will not be an issue - *IF* your output Z is constant across frequencies. May not be the case with tubes.
Many headphones have a resistive load. More true in the older days when outputs have high output impedance whether intentional or not - 47ohm resistor is common. Some even say some headphones were designed with that output impedance in mind.
Now with multi-driver ear/headphones, using them with high output Z is impossible.
Frequency response of high output Z and flat load Z will not be an issue - *IF* your output Z is constant across frequencies. May not be the case with tubes.
That's not necessarily correct.
300 ohm with 600 ohm peak and 30 ohm Zout changes FR by a measly .4 dB.
300 ohms is a factor of 10 larger than Zout. You see, with ~630 ohm the peak could be several MOhms .. and you'd still only see a .4 dB difference.
The 1/8th Zout rule of thumb gives you a max .5 dB difference if the peak has double the impedance which is a reasonable bad- or even worst-case assumption, but only for single-driver, dynamic headphones.
Last edited:
> I am thinking that 30 ohms of amplifier output impedance does not matter nearly so much.
You echo my observations. Several notorious 300 Ohm phones are audibly flatter with less than 120r in series, and ~~30r as a reasonable "low" value. OTOH many 32 Ohm phones are very flat impedance, and 30 ohms in series just drops level, does not change flatness.
Yes, damping, but the impedance at the bass resonance (may be mid-bass in some designs) IS the effect of damping. If we get the in-circuit rise at resonance down below 1dB then we have most of the electromechanical damping possible at the headphone plug.
Another point: ~~30 Ohms series from a 7V zero impedance source (or 0.25A with 30r shunt) gives roughly the "maximum" power into >90% of 'phones from 24r to 2K. The higher Z 'phones derive from more sensitive designs intended to monitor directly across studio transmission lines; the low-Z models are mostly cost/size constrained and can take more drive.
You echo my observations. Several notorious 300 Ohm phones are audibly flatter with less than 120r in series, and ~~30r as a reasonable "low" value. OTOH many 32 Ohm phones are very flat impedance, and 30 ohms in series just drops level, does not change flatness.
Yes, damping, but the impedance at the bass resonance (may be mid-bass in some designs) IS the effect of damping. If we get the in-circuit rise at resonance down below 1dB then we have most of the electromechanical damping possible at the headphone plug.
Another point: ~~30 Ohms series from a 7V zero impedance source (or 0.25A with 30r shunt) gives roughly the "maximum" power into >90% of 'phones from 24r to 2K. The higher Z 'phones derive from more sensitive designs intended to monitor directly across studio transmission lines; the low-Z models are mostly cost/size constrained and can take more drive.
with 600r (or even some 300r) phones at medium volume you can easily hit 50V level.
IMHO, the reason why hiZ phones with "+-15V amps" sound bad.
Ehh .. what?
HD600 needs about .25Vrms for 90 dB SPL. Aiming for 110 dB SPL (reasonable imo) we're at 2.5Vrms or 20mW.
HD650 needs the same or even less.
DT880-600 needs only about twice that.
For a comfortable 80 dB SPL average we're in the lower millivolt and microwatts range.
You may be confusing gain and power and the headphones "sounding bad" probably comes from the unnecessary fear of "not enough power" or simply not reaching the desired earsplitting levels.
HD600 needs about .25Vrms for 90 dB SPL. Aiming for 110 dB SPL (reasonable imo) we're at 2.5Vrms or 20mW.
HD650 needs the same or even less.
DT880-600 needs only about twice that.
For a comfortable 80 dB SPL average we're in the lower millivolt and microwatts range.
You may be confusing gain and power and the headphones "sounding bad" probably comes from the unnecessary fear of "not enough power" or simply not reaching the desired earsplitting levels.
Last edited:
Hi
I have a AudioTechnica ATH MSR7, 32 ohm impedance.
I made voltage measurements while listening to it.
At maximum level my ears can bear, the voltage of the crests of audio signals is 1.2 V peak to peak ! That is 0.42 V rms for the crests, 5.5 mW power in 32 ohm.
I use an amplifer capable of 20 Vptp with no concerns at all.
With a higher impedance, such as 600 ohm, the necessary voltage increases as impedance increases. 10 Vptp may be necessary.
Jacques
I have a AudioTechnica ATH MSR7, 32 ohm impedance.
I made voltage measurements while listening to it.
At maximum level my ears can bear, the voltage of the crests of audio signals is 1.2 V peak to peak ! That is 0.42 V rms for the crests, 5.5 mW power in 32 ohm.
I use an amplifer capable of 20 Vptp with no concerns at all.
With a higher impedance, such as 600 ohm, the necessary voltage increases as impedance increases. 10 Vptp may be necessary.
Jacques
That does sound about right. ATH-MSR7 was measured to need 23 mVrms for 90 dB at 1 kHz, but since it has a bit of a frequency response bump there, we can easily add 3 dB, for more like 32 mVrms. NwAvGuy would shoot for a conservative 110 dB SPL, with a corresponding signal level of 320 mVrms or 0.9 Vpp - close enough to 1.2 Vpp.
For a 600 ohm model of modest sensitivity like a DT880-600 (~0.5 Vrms @ 90 dB), we'd get 5Vrms or ~14 Vpp, which tends to be within reach of typical desktop headphone amplifiers powered from +/-12-15 V. Even 6.5Vrms would still be. Not having enough voltage gain may turn out to be an issue at this point, depending on source levels. But 50 V peak as claimed above? Not anywhere close unless you're semi-deaf and/or listening to raw unprocessed recordings of drum kits.
For a 600 ohm model of modest sensitivity like a DT880-600 (~0.5 Vrms @ 90 dB), we'd get 5Vrms or ~14 Vpp, which tends to be within reach of typical desktop headphone amplifiers powered from +/-12-15 V. Even 6.5Vrms would still be. Not having enough voltage gain may turn out to be an issue at this point, depending on source levels. But 50 V peak as claimed above? Not anywhere close unless you're semi-deaf and/or listening to raw unprocessed recordings of drum kits.
Last edited:
with 600r (or even some 300r) phones at medium volume you can easily hit 50V level.
IMHO, the reason why hiZ phones with "+-15V amps" sound bad.
I am more inclined to believe Xnor.
My very old Red Devils had a maximum Vpk recommended in the user manual of 5Vpk, even stating the measuring apparatus to be used to check they were not being overdriven.
I think the Pro4AA that came many years later had a similar maximum signal recommendation. At those peak levels the in ear SPL would be enormous
Based on those recommendations I would not design a Headphone Amplifier with rails higher than ±18Vdc and more probably ±12Vdc would be capable of over-driving most headphones.
Last edited:
resonance in the headphone drivers
It is not just frequency response or 0.4 db’s when we are speaking of a dampening factor of 5.
I feel what is more important is resonance and stored energy, at the impedance peek that smears the sound. Balanced outputs, balanced cables, and near zero output impedance are all expensive partial remedies for resonance in the headphone drivers.
DT
It is not just frequency response or 0.4 db’s when we are speaking of a dampening factor of 5.
I feel what is more important is resonance and stored energy, at the impedance peek that smears the sound. Balanced outputs, balanced cables, and near zero output impedance are all expensive partial remedies for resonance in the headphone drivers.
DT
300 Ohm / Zout=30 Ohm = 10
... but this number is meaningless.
The mass of the driver doesn't change, nor does the suspension. Do you understand why there is a resonance/impedance peak?
All that basically changes is the voltage applied to the headphones.
With Zout=0 Ohm 1V on the output will drop across the headphones, no matter if the load is 300 or 600 Ohm.
With Zout=30 Ohm 1V on the output will drop by -0.8 dB given a 300 Ohm load (e.g. the midrange) but will only drop by -0.4 dB given a 600 Ohm load (e.g. at the resonance peak).
So you effectively get a peaking EQ filter at the resonance frequency, +0.4 dB, with it's Q matching the impedance peak. This is minimal phase so can be easily equalized in both magnitude and phase. There's no smearing and I wouldn't call it that with the small phase shifts involved.
None of those can change the resonance/impedance peak. Btw, bridged amps will double the output impedance (but that's usually insignificant). Balanced audio is useful for stuff like microphones where the cables pick up noise. (Contrary to popular belief you don't need differential signaling for that - a headphone driver wouldn't know the difference anyway. What makes it balanced are equally large impedances.)
Imagine a simplified model where the cone mass is an inductor, the suspension is a capacitor and there's also a resistor for losses - all three in parallel.
Impedance of the inductor will rise with frequency ("moving the mass back and forth more quickly is harder"). With the capacitor it's the opposite. You'll get your impedance peak where these impedances cross.
Last edited:
Remember a pure resistive load has no capacitance or inductance and no resonance. Resonance is; capacitance, inductance and resistance in parallel.
Connect a resistance to a headphone amplifier and your favorite audio analyzer across the resistance you will measure distortion phase and whatever else you choose to measure. Record the results.
Now connect one of those 300 ohm headphone drivers of your choosing, you know one of those with a 600 ohm impedance peek. Now run the same measurements on the driver as you did to the resistor above. You will see easily measured differences. Add feedback to reduce output impedance, add balanced outputs and add balanced cables, you will see improved distortion, improvement in phase, improvement in whatever you measured.
To the point of this thread; planar magnetic headphones perform more like a pure resistance that the classic drivers with a huge impedance peek. I do not think that the rule of 8 need apply.
Go over to innerfidelity.com and see the test results for planar magnetic vs the other; distortion, phase and square wave.
DT
Connect a resistance to a headphone amplifier and your favorite audio analyzer across the resistance you will measure distortion phase and whatever else you choose to measure. Record the results.
Now connect one of those 300 ohm headphone drivers of your choosing, you know one of those with a 600 ohm impedance peek. Now run the same measurements on the driver as you did to the resistor above. You will see easily measured differences. Add feedback to reduce output impedance, add balanced outputs and add balanced cables, you will see improved distortion, improvement in phase, improvement in whatever you measured.
To the point of this thread; planar magnetic headphones perform more like a pure resistance that the classic drivers with a huge impedance peek. I do not think that the rule of 8 need apply.
Go over to innerfidelity.com and see the test results for planar magnetic vs the other; distortion, phase and square wave.
DT
What do you call distorsion phase ?
Adding balanced output and balanced cables will not improve distorsion or phase or whatever we measure. If so you have a big unidentified issue with your amp.
Jacques
One by one:
1) Zout=0:
If you plot the voltage across the headphones then you will get a flat electrical (not to be confused with anything acoustic!) magnitude and phase of the voltage.
If you plot the current through the headphones with the same source then you will get the inverse of the impedance plot - in both magnitude and phase.
V=I*Z, with a fixed V the current I needs to "cancel" the impedance Z.
Let's take a 300 Ohm (@ 1kHz) headphone that peaks at 600 Ohm (@80 Hz). So half the current will go through the peak for the same input voltage.
(Acoustically you will not result in a dip because the resonance is where the driver is more efficient, i.e. needs less current to produce the same SPL.)
2) Zout=30 Ohm
Now the voltage plot will not be flat anymore, instead it will follow the impedance plot both in magnitude and phase .. at least a bit (by the .4 dB mentioned before).
But now the current plot will be flatter proportionally.
Now if you EQ'd this signal by the aforementioned .4 dB with a min. phase filter it would again flatten both magnitude and phase of the voltage plot.
I've already explained the irrelevance of bridged/balanced stuff to this topic.
What distortion? What phase? What do you mean by improvement? What's an improvement e.g. in the voltage plot could be seen as a degradation in the current plot and vice-versa.
Of course. This is undisputed.
Again, the 1/8th rule of thumb takes a more or less arbitrary limit of max 0.5 dB if your headphone has a 2x impedance peak.
That is, if we take Zout = 300/8 = 37.5 Ohm with a headphone with a 600 Ohm peak, then we get +0.5 dB at the resonance peak.
And that's why I said that the damping factor is a meaningless number. We'd still have the same DF=8 with a 300 ohm purely resistive load.
Ferraris measure differently (in whatever random metric) from Lamborghinis. You seem to be saying that it's the driver .. but it's not. We can switch drivers and get the same measurements.
That is what I was trying to explain to you, but I was more asking if you understood the (simplified) mechanical aspects of the resonance peak.
There is no such thing called "distortion phase".
1) Zout=0:
If you plot the voltage across the headphones then you will get a flat electrical (not to be confused with anything acoustic!) magnitude and phase of the voltage.
If you plot the current through the headphones with the same source then you will get the inverse of the impedance plot - in both magnitude and phase.
V=I*Z, with a fixed V the current I needs to "cancel" the impedance Z.
Let's take a 300 Ohm (@ 1kHz) headphone that peaks at 600 Ohm (@80 Hz). So half the current will go through the peak for the same input voltage.
(Acoustically you will not result in a dip because the resonance is where the driver is more efficient, i.e. needs less current to produce the same SPL.)
2) Zout=30 Ohm
Now the voltage plot will not be flat anymore, instead it will follow the impedance plot both in magnitude and phase .. at least a bit (by the .4 dB mentioned before).
But now the current plot will be flatter proportionally.
Now if you EQ'd this signal by the aforementioned .4 dB with a min. phase filter it would again flatten both magnitude and phase of the voltage plot.
This is where you start to mix up things.
I've already explained the irrelevance of bridged/balanced stuff to this topic.
What distortion? What phase? What do you mean by improvement? What's an improvement e.g. in the voltage plot could be seen as a degradation in the current plot and vice-versa.
(*peak, not peek)
Of course. This is undisputed.
Again, the 1/8th rule of thumb takes a more or less arbitrary limit of max 0.5 dB if your headphone has a 2x impedance peak.
That is, if we take Zout = 300/8 = 37.5 Ohm with a headphone with a 600 Ohm peak, then we get +0.5 dB at the resonance peak.
And that's why I said that the damping factor is a meaningless number. We'd still have the same DF=8 with a 300 ohm purely resistive load.
You are mixing up electrical and acoustic measurements with completely different mechanical constructions - apples and oranges.
Ferraris measure differently (in whatever random metric) from Lamborghinis. You seem to be saying that it's the driver .. but it's not. We can switch drivers and get the same measurements.
Last edited:
real measurable differences
Take a look at the pdf test measurements linked below. Look at the low frequency distortion, phase shift and ringing on the square wave results. There are real measurable differences due to the impedance / resonance peek.
The mass of the diaphragm and coil resonating in the magnetic field generates a Back EMF. Add the Back EMF to the input of the headphone amplifier, this is feedback. (Increased amplifier output impedance reduces feedback.)
Feedback helps improve the negative effects of mechanical resonance in the headphone driver; however it is not a cure.
DT
http://www.innerfidelity.com/images/AudezeLCD2sn5423021Fazer.pdf
http://www.innerfidelity.com/images/SennheiserHD650.pdf
Take a look at the pdf test measurements linked below. Look at the low frequency distortion, phase shift and ringing on the square wave results. There are real measurable differences due to the impedance / resonance peek.
The mass of the diaphragm and coil resonating in the magnetic field generates a Back EMF. Add the Back EMF to the input of the headphone amplifier, this is feedback. (Increased amplifier output impedance reduces feedback.)
Feedback helps improve the negative effects of mechanical resonance in the headphone driver; however it is not a cure.
DT
http://www.innerfidelity.com/images/AudezeLCD2sn5423021Fazer.pdf
http://www.innerfidelity.com/images/SennheiserHD650.pdf
I can only repeat myself: you're completely mixing up things here.
That's an acoustic measurement. The distortion measured here is a matter of mechanical construction, i.e. size of the diaphragm, linear excursion (limits), ... and low frequency performance with headphones is also greatly influenced by (lack of) seal of the enclosed volume.
So completely irrelevant to the topic.
I have explained the effects of raising output impedance on this to you twice now.
Please read up on what a step and impulse response is.
I've already explained to you how an EQ could fix both magnitude and phase (and therefore the electrical step/impulse response) if you raised the output impedance to 30 Ohms.
So again: you're confusing electrical with mechanical things here.
And again: it's peak, not peek.
To summarize: There is no direct relationship between acoustic distortion, acoustic step/impulse response and impedance.
If you still don't get that then it should be trivial for you to find measurements on innerfidelity that directly contradict your conclusions.
What else do you think is going on? A moving, suspended mass on its own does not do anything electrically. We're always talking about a voice coil moving through a magnetic field here. That's why you get these impedance curves in the first place.
That's an acoustic measurement. The distortion measured here is a matter of mechanical construction, i.e. size of the diaphragm, linear excursion (limits), ... and low frequency performance with headphones is also greatly influenced by (lack of) seal of the enclosed volume.
So completely irrelevant to the topic.
That's an electrical measurement. It's part of the impedance.
I have explained the effects of raising output impedance on this to you twice now.
This again is an acoustic measurement.
Please read up on what a step and impulse response is.
I've already explained to you how an EQ could fix both magnitude and phase (and therefore the electrical step/impulse response) if you raised the output impedance to 30 Ohms.
So again: you're confusing electrical with mechanical things here.
No. Only one thing you've mentioned is part of the impedance and relevant here: phase.
And again: it's peak, not peek.
To summarize: There is no direct relationship between acoustic distortion, acoustic step/impulse response and impedance.
If you still don't get that then it should be trivial for you to find measurements on innerfidelity that directly contradict your conclusions.
This is what the impedance plot shows! I've also explained this. Inductor and capacitor, mass and suspension. One is the electrical equivalent of the (mechanical) other.
What else do you think is going on? A moving, suspended mass on its own does not do anything electrically. We're always talking about a voice coil moving through a magnetic field here. That's why you get these impedance curves in the first place.
No. At least not in the way you think it does.
Last edited:
Hi
I agree with xnor, DT you are mixing up everythings.
I thought you had measured something.
In fact no.
You are not reading papers the right way, sorry to tell you.
Jacques
I agree with xnor, DT you are mixing up everythings.
I thought you had measured something.
In fact no.
You are not reading papers the right way, sorry to tell you.
Jacques
I am not challenging the quality of you treasured headphones.
I am not challenging the quality of you treasured headphones.
I am just saying that the resistive, non-resonate; nature of planar magnetic headphone drivers makes them less sensitive to the output impedance of the headphone amplifier.
DT
I am not challenging the quality of you treasured headphones.
I am just saying that the resistive, non-resonate; nature of planar magnetic headphone drivers makes them less sensitive to the output impedance of the headphone amplifier.
DT
Well, that's true. It's essentially the flip side to them being quite inefficient, they're mostly a dissipative (resistive) load.
In some conventional dynamic driver closed headphones, by contrast, you can spot every acoustic resonance as wiggles on the impedance plot.
Classic hi-fi headphones tend to be quite heavily damped to tame the resonance peak, the most prominent examples being Beyer DT880-250 and the various AKGs.
- Status
- Not open for further replies.