Fascinating, I had no idea of this. Is it the case, therefore, that today's tube amplifiers are not suited to the task of driving a large proportion of the speakers currently available, or do they do something clever to avoid the problem?
I have built a couple of headphone amplifiers in the past, and wondered about the output impedance question. I measured a Sony CD player's headphone output at 47 Ohms, but in the end decided that there was no logical reason why it should be anything but 0R.
I think some of the preference for tubes today is about nostalgia/romance, some is about "euphonic distortion" that some like the sound of, and there are some really well engineered tube amps that have a low output impedance and impressive (for tubes) specs. So it really depends. But much has been written on the topic of what the higher output impedance of some does to the sound. It's almost exactly analogous to headphones running from a higher impedance source. And, just like with headphones, some might prefer the less accurate frequency response, loss of bass damping, etc. when using modern speakers on a higher output impedance tube amp.
Last edited:
I think "back emf" leads people to strange conclusions
while it is the mechanism for electro-mechanical coupling in dynamic drivers it is a heuristic level/modeling step below the level we use to describe amplifier and loads in amp design
a EE is happier with complex impedance for describing amplifier loads including dynamic drivers and cables
then damping factor is a good way to normalize to the specific load and amplifier- at least for the low frequency impedance bump of most dynamic single driver headphones that comes from mass-spring mechanical resonance of the voice coil/diaphragm and the surround/air spring load
if the peak and min damping ratio for a given amp/headphone combination are less than 100 then there is no audible difference in frequency response from "infinite" damping/zero amp output impedance
ABX Amplitude vs. Frequency Matching Criteria
in practice is it seldom likely that even damping ratio of greater than ~20 gives audible frequency response variations with single driver dynamic headphones
multidriver systems can have much bumpier load impedance due to crossover design - with the bumps/notches at more sensitive hearing frequencies
while it is the mechanism for electro-mechanical coupling in dynamic drivers it is a heuristic level/modeling step below the level we use to describe amplifier and loads in amp design
a EE is happier with complex impedance for describing amplifier loads including dynamic drivers and cables
then damping factor is a good way to normalize to the specific load and amplifier- at least for the low frequency impedance bump of most dynamic single driver headphones that comes from mass-spring mechanical resonance of the voice coil/diaphragm and the surround/air spring load
if the peak and min damping ratio for a given amp/headphone combination are less than 100 then there is no audible difference in frequency response from "infinite" damping/zero amp output impedance
ABX Amplitude vs. Frequency Matching Criteria
in practice is it seldom likely that even damping ratio of greater than ~20 gives audible frequency response variations with single driver dynamic headphones
multidriver systems can have much bumpier load impedance due to crossover design - with the bumps/notches at more sensitive hearing frequencies
@AUDIODH, generally when the output impedance starts to approach (or worse exceed) the impedance of the headphone/speaker it's an audible issue. I would say to be safe you want the output impedance to be less than 1/8th the load. Or:
Zout <= Zload/8
For 8 ohm speakers, you want an output impedance < 1 ohm. For 32 ohm headphones < 4 ohms. For 300 ohm headphones < 37 ohms, etc.
Cable resistance shouldn't be much of a factor except perhaps with really fine gauge cable on 16 ohm headphones.
There are two big issues. The first is the output impedance forms a voltage divider with the load. And because the impedance of the headphones (or speaker) change with frequency, that means the voltage is divided differently with frequency. So your flat amplifier is no longer flat. And, as I posted previously in this thread, and show in my blog impedance article, I've measured deviations of 14 dB (+/- 7 dB) due to output impedance which is hugely audible.
The second issue is the bass tuning of the driver. The mechanical mass of the driver interacts with the suspension/surround to form a tuned system that resonates at some frequency--just like a suspension bridge wants to bounce in the middle at a certain frequency. A low impedance source provides electrical damping (or "braking") to that resonance much like shock absorbers keep a car from bobbing up and down uncontrollably every time you hit a bump--this is related to what some describe as "back EMF" . As the source impedance approaches the driver impedance the electrical damping is significantly degraded so the driver is less controlled--like a car with worn out shocks.
The wild card here is some headphone designers still plan for a relatively high output impedance. So they make the driver suspension stiffer than it otherwise needs to be to provide more mechanical damping. Such a suspension is typically not as linear and will create more distortion--especially at higher drive levels. As the driver tries to move further the suspension resists more and the driver motion cannot accurately follow the input signal. It's like a cheap paper woofer with the old style according pleated paper surround versus a high-end long throw highly linear driver with a foam or rubber surround. But those free moving drivers need more damping from the source (and/or a properly tuned enclosure) to be well controlled.
Open backed headphones tend to be less colored by internal resonances and reflections. They're widely regarded as generally sounding better. But open backed headphones don't provide much resistance to the driver motion. And if a designer can't count on a low impedance source, they generally have to compromise the suspension of the driver element in ways that make it less linear.
Zout <= Zload/8
For 8 ohm speakers, you want an output impedance < 1 ohm. For 32 ohm headphones < 4 ohms. For 300 ohm headphones < 37 ohms, etc.
Cable resistance shouldn't be much of a factor except perhaps with really fine gauge cable on 16 ohm headphones.
There are two big issues. The first is the output impedance forms a voltage divider with the load. And because the impedance of the headphones (or speaker) change with frequency, that means the voltage is divided differently with frequency. So your flat amplifier is no longer flat. And, as I posted previously in this thread, and show in my blog impedance article, I've measured deviations of 14 dB (+/- 7 dB) due to output impedance which is hugely audible.
The second issue is the bass tuning of the driver. The mechanical mass of the driver interacts with the suspension/surround to form a tuned system that resonates at some frequency--just like a suspension bridge wants to bounce in the middle at a certain frequency. A low impedance source provides electrical damping (or "braking") to that resonance much like shock absorbers keep a car from bobbing up and down uncontrollably every time you hit a bump--this is related to what some describe as "back EMF" . As the source impedance approaches the driver impedance the electrical damping is significantly degraded so the driver is less controlled--like a car with worn out shocks.
The wild card here is some headphone designers still plan for a relatively high output impedance. So they make the driver suspension stiffer than it otherwise needs to be to provide more mechanical damping. Such a suspension is typically not as linear and will create more distortion--especially at higher drive levels. As the driver tries to move further the suspension resists more and the driver motion cannot accurately follow the input signal. It's like a cheap paper woofer with the old style according pleated paper surround versus a high-end long throw highly linear driver with a foam or rubber surround. But those free moving drivers need more damping from the source (and/or a properly tuned enclosure) to be well controlled.
Open backed headphones tend to be less colored by internal resonances and reflections. They're widely regarded as generally sounding better. But open backed headphones don't provide much resistance to the driver motion. And if a designer can't count on a low impedance source, they generally have to compromise the suspension of the driver element in ways that make it less linear.
@RocketScientist and jcx
Many thanks for some really interesting info. So the source impedance is crucial for many types of headphones (and speakers), but it's not something I've ever seen specified as a requirement.
I get the feeling that there's a whole world out there of randomly mismatched equipment with the owners thinking the sound is 'musical', and manufacturers who, rather than scare people off by specifying too many requirements, make compromises aimed to suit the lowest common denominator.
Many thanks for some really interesting info. So the source impedance is crucial for many types of headphones (and speakers), but it's not something I've ever seen specified as a requirement.
I get the feeling that there's a whole world out there of randomly mismatched equipment with the owners thinking the sound is 'musical', and manufacturers who, rather than scare people off by specifying too many requirements, make compromises aimed to suit the lowest common denominator.
Exactly! That's why I take this topic so seriously. Those who keep "promoting" or otherwise justifying higher output impedances are part of the problem, not part of the solution. The solution to this problem is black and white. It was done with amps and speakers 40 years ago and everyone has been happier since. A near zero output impedance is the only impedance that offers universal compatibility going forward and it also offers headphone designers maximum electrical damping so they can make lower distortion, more linear, headphone drivers.
It's also worth pointing out if a headphone designer, for some hard to imagine reason, wants less electrical damping they're free to put series resistors inside the headphones. It's trivial to passively raise output impedance but there's no good way to passively lower it.
A very small percentage of headphone nuts seem to like the fact there's an endless combination of "sounds" available by pairing various sources with various headphones. And, as I mention in my blog article, some manufactures aim for "different" to help distinguish their products. But, IMHO, raising the output impedance is a poor way to achieve "different" as it tends to degrade bass performance, bass quality, and the frequency response variations it creates will vary widely from headphone to headphone.
Good Morning gentlemen,
Thanks for the responses. JCX, and Rocket Man! Nice responses.
When I was designing and using high power four-way concert loudspeakers, I was hired to perform some acoustical test on an Army RF chamber..The basic system was 10KW of Crown power into boxes loaded with JBL components, 2-18", 4-10", two mid-hi 2445 drivers and two 2404 tweets - essentially a Clair S4 box.
I had to produce pretty high levels at 16 Hz and record it with a strip recorder and B&K analyzer. A HP function generator was used as the source. We experienced a problem when we reached the lower frequencies, because when the waveform was removed the system popped because the signal was not cutting off at zero crossing. I designed a circuit to solve the problem, but ended up doing some research in that area. While damping factor is a buzz work and many times over stated, I found that the feedback of the amplifiers has a lot to do with what ends up as a more potential problem than a damping issue. That being said, headphones are similar as you both pointed out and the cable resistance likely equals the amp output impedance. I read somewhere something like this: you can measure something that measures great but sounds bad, and something that sounds good measures bad...maybe you are measuring in the wrong place. In the end it's not about the numbers I've chased all my life, but rather the sound.
-Dave
Thanks for the responses. JCX, and Rocket Man! Nice responses.
When I was designing and using high power four-way concert loudspeakers, I was hired to perform some acoustical test on an Army RF chamber..The basic system was 10KW of Crown power into boxes loaded with JBL components, 2-18", 4-10", two mid-hi 2445 drivers and two 2404 tweets - essentially a Clair S4 box.
I had to produce pretty high levels at 16 Hz and record it with a strip recorder and B&K analyzer. A HP function generator was used as the source. We experienced a problem when we reached the lower frequencies, because when the waveform was removed the system popped because the signal was not cutting off at zero crossing. I designed a circuit to solve the problem, but ended up doing some research in that area. While damping factor is a buzz work and many times over stated, I found that the feedback of the amplifiers has a lot to do with what ends up as a more potential problem than a damping issue. That being said, headphones are similar as you both pointed out and the cable resistance likely equals the amp output impedance. I read somewhere something like this: you can measure something that measures great but sounds bad, and something that sounds good measures bad...maybe you are measuring in the wrong place. In the end it's not about the numbers I've chased all my life, but rather the sound.
-Dave
RocketScientist
I read the attached elsewhere. Care to comment on that statement ?
N.B. I have no personal position re that statement.
SandyK
"For headphones the damping factor is irrelevant as damping does not come from the amplifier but the used diaphragm materials and housing/pads not the output resistance."
I read the attached elsewhere. Care to comment on that statement ?
N.B. I have no personal position re that statement.
SandyK
"For headphones the damping factor is irrelevant as damping does not come from the amplifier but the used diaphragm materials and housing/pads not the output resistance."
@sandyK One possibility is what JCX already brought up. If they're comparing a damping factor of say 20 to a damping factor of 2000, the above quote is true. When some talk about "damping factor" they're assuming the output impedance is already reasonably close to zero. And you very much reach a point of diminishing returns rather quickly.
But when the damping factor starts to approach 1, I don't believe the quote is generally true--at least for dynamic headphones. To be clear, damping factor is the ratio of the load impedance to the output impedance:
DF = Zload/Zsource
For example 32 ohm headphones on a 2 ohm source yields a damping factor of 16, while 80 ohm headphones on a 120 ohm source yields a (horrible) damping factor of 0.6. Consistent with what I said earlier, any damping factor above 8 isn't likely to make an audible difference. But with headphones it's not uncommon to have damping factors around 1 or even less.
The guys who work for the headphone companies have told me they can design better headphones for zero ohm sources. I believe them. The voice coil of a dynamic headphone driver is very similar to the drivers used in speakers and can provide the same sorts of electromagnetic "braking".
The total system Q (an indication of how damped the driver is in the final design) is a combination of electrical damping (Qes) and mechanical damping (Qms)--both terms in the equation count equally. I don't see why headphone drivers would somehow magically be exempt from the same parameters and laws of physics. For more information on these parameters see Wikipedia Thiele Small Qualitative Descriptions.
I agree the "load" on the head side of the driver does likely provide some damping. But that varies a lot depending on the headphone design, the fit on different heads/ears, etc. Some headphones just rest lightly on the ears with a very poor seal. There might be specific headphones that seal exceptionally well, with certain types of drivers, where that's the dominant damping mechanism. But that doesn't dismiss all the other cases.
And, finally, it's not hard to hear the loss of damping and change in tuning with some headphones with a high vs low output impedance. The impedance plot also changes shape which is indicative of a different system Q as more objective proof.
It would be nice if an engineer from one of the headphone manufactures could comment on this. I'm very familiar with the engineering behind speaker design but headphones are different in some ways.
But when the damping factor starts to approach 1, I don't believe the quote is generally true--at least for dynamic headphones. To be clear, damping factor is the ratio of the load impedance to the output impedance:
DF = Zload/Zsource
For example 32 ohm headphones on a 2 ohm source yields a damping factor of 16, while 80 ohm headphones on a 120 ohm source yields a (horrible) damping factor of 0.6. Consistent with what I said earlier, any damping factor above 8 isn't likely to make an audible difference. But with headphones it's not uncommon to have damping factors around 1 or even less.
The guys who work for the headphone companies have told me they can design better headphones for zero ohm sources. I believe them. The voice coil of a dynamic headphone driver is very similar to the drivers used in speakers and can provide the same sorts of electromagnetic "braking".
The total system Q (an indication of how damped the driver is in the final design) is a combination of electrical damping (Qes) and mechanical damping (Qms)--both terms in the equation count equally. I don't see why headphone drivers would somehow magically be exempt from the same parameters and laws of physics. For more information on these parameters see Wikipedia Thiele Small Qualitative Descriptions.
I agree the "load" on the head side of the driver does likely provide some damping. But that varies a lot depending on the headphone design, the fit on different heads/ears, etc. Some headphones just rest lightly on the ears with a very poor seal. There might be specific headphones that seal exceptionally well, with certain types of drivers, where that's the dominant damping mechanism. But that doesn't dismiss all the other cases.
And, finally, it's not hard to hear the loss of damping and change in tuning with some headphones with a high vs low output impedance. The impedance plot also changes shape which is indicative of a different system Q as more objective proof.
It would be nice if an engineer from one of the headphone manufactures could comment on this. I'm very familiar with the engineering behind speaker design but headphones are different in some ways.
Things get more complicated with a tube amp. If you severely mismatch the transformer impedance to the headphones you may create other problems. I'm not a tube expert, but most tube designs have an output impedance high enough to cause significant frequency response variations if the load impedance varies much.
Some like those "variations" (which is often part of the "tube sound") and some don't. It's highly dependent on the impedance curve of the load.
With the HD650's, for example, the impedance approximately doubles at 90 hz. So you could end up with a several dB rise at 90 hz if you use them with a high output impedance. With most tube amps you'll get at least some rise peaking around 90 hz.
Personally any boost around 90 hz sounds muddy and "boomy" to me. A lot of small cheap speakers boost their response around 80 - 100 hz in an effort to give the illusion of more bass. And it's not a good thing to my ears.
Some like those "variations" (which is often part of the "tube sound") and some don't. It's highly dependent on the impedance curve of the load.
With the HD650's, for example, the impedance approximately doubles at 90 hz. So you could end up with a several dB rise at 90 hz if you use them with a high output impedance. With most tube amps you'll get at least some rise peaking around 90 hz.
Personally any boost around 90 hz sounds muddy and "boomy" to me. A lot of small cheap speakers boost their response around 80 - 100 hz in an effort to give the illusion of more bass. And it's not a good thing to my ears.
Tube HP Amp transformer
ET, I have some experience in Tube Amplifier design, primarily with guitar and bass tube amps.
I have always tried to match the transformer output Z to the load. The reflected load, as seen by the power tube(s), will achieve the best power transfer. This is some what dependent on the output configuration, etc., but still a good rule of thumb.
ET, I have some experience in Tube Amplifier design, primarily with guitar and bass tube amps.
I have always tried to match the transformer output Z to the load. The reflected load, as seen by the power tube(s), will achieve the best power transfer. This is some what dependent on the output configuration, etc., but still a good rule of thumb.
Hi Rocket Scientist,
I have most always seen feedback in tube power amplifiers. It is typically in the form a R and/or C negative feedback. Much the same as many solid state amplifiers. Depending on the amp design, that 90 Hz bump can be compensated in the cathode of the pre-amp section. A simple parallel RC or R/RC circuit. It can get you 3 dB and if designed properly can likely be set at or near the Fc point. EQ circuits can get more involved if needed, but I wouldn't get overly concerned about that up front....test and see how it works first.
I have most always seen feedback in tube power amplifiers. It is typically in the form a R and/or C negative feedback. Much the same as many solid state amplifiers. Depending on the amp design, that 90 Hz bump can be compensated in the cathode of the pre-amp section. A simple parallel RC or R/RC circuit. It can get you 3 dB and if designed properly can likely be set at or near the Fc point. EQ circuits can get more involved if needed, but I wouldn't get overly concerned about that up front....test and see how it works first.
Yes, but the feedback has to be on the load side of the output transformer to properly correct for impedance and losses in the transformer itself. Most of the tube circuits I've seen either are OTL (no transformer) or there's no feedback on the secondary like this "high end" SET design from the Gilmore site:
An externally hosted image should be here but it was not working when we last tested it.
Tube HP Amp transformer
good point. I'm use to a more typical class A/B higher power amps, which run cooler and are more efficient.
The feedback does indeed come from the transformer secondary side via AGND. While this does provide more power and stability, it also will introduce more distortion than a straight Class-A per the schematic you attached.
Stated circuits are of a push/pull configuration with the transformer Primary connected to the plates.
A point if interest, what do the amplifiers without transformers do to the HP's when the tube shorts internally to the plate supply - do the HP's smoke?
Hi Rocket Scientist,
good point. I'm use to a more typical class A/B higher power amps, which run cooler and are more efficient.
The feedback does indeed come from the transformer secondary side via AGND. While this does provide more power and stability, it also will introduce more distortion than a straight Class-A per the schematic you attached.
Stated circuits are of a push/pull configuration with the transformer Primary connected to the plates.
A point if interest, what do the amplifiers without transformers do to the HP's when the tube shorts internally to the plate supply - do the HP's smoke?
Attached is a Fender Twin Schematic...very typical of Guitar/Bass amp designs, regardless of manufacturer.
I would like to point out since the plate voltage is very high, the secondary is ground referenced for safety and acts as a feedback point as well.
If the event transformer shorts, the plate voltage would see ground and pop the supply fuse.
While I don't compare the two in performance, (distortion and power) the design you posted is not a high-power design....but looks appropriate for HP's.
Let's remember most electro-mechanical transducers are at or near 1% THD. So if a tube design introduces a little of it's own, doesn't necessarily mean it's bad. Just Sayin'
I would like to point out since the plate voltage is very high, the secondary is ground referenced for safety and acts as a feedback point as well.
If the event transformer shorts, the plate voltage would see ground and pop the supply fuse.
While I don't compare the two in performance, (distortion and power) the design you posted is not a high-power design....but looks appropriate for HP's.
Let's remember most electro-mechanical transducers are at or near 1% THD. So if a tube design introduces a little of it's own, doesn't necessarily mean it's bad. Just Sayin'
Yeah, if I were building a HP amp for expensive cans I'd be rather worried about that myself. Some OTL designs on the Gilmore Headwize site have direct coupled outputs. And even with a cap, the right failure might still damage the headphones from the high voltage pulse that would make it through.
Direct coupled solid state amps can also fail in ways that will damage headphones but it's statistically less likely due to the much lower voltages and higher reliability of transistors, FETs and ICs.
In either case, you can use a protection circuit. Although it might get a bit tricky to design one that's fast enough to prevent damage from high voltages. The solid state circuits generally just have to intervene before any thermal damage is done. But high voltages might cause nearly instant physical damage.
I was in an audio store once when an idling tube amp literally caught fire and filled the store with smoke--all from a grid failing in one of the output tubes. It had output transformers, and thankfully, the speakers were fine. But we all wondered what would have happened if it had been an OTL design.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.