It all depends on how much power you will actually use, in the worst case.
For class AB amps (all op amps, most buffers) power is cheap, so you can afford to design well over realistic expectation. For a class A design, the power budget unravels quickly, but everyone has their pain point. 5W? 10W? 40W? 100W? How high a circuit power consumption before the additional filtering, bigger power transformer, and larger heatsinks (and heat) become "not worth it" - that's up to you.
I can say however that 10 mW is sufficient from a physical clipping point of view. Of course both 16 ohm and 300 ohm loads must be considered, and the currents and voltages raised accordingly to cover both situations.
As the little diagram attached tries to show, 10 mW is a reasonable lower limit, while 100 mW is a pretty good upper one, based on experience diminishing returns kick in heavily after that (for class A). The diagram indicates the current and voltage required to deliver that power into the outer range of headphone loads.
For a class A output stage, that means 28x1.4=40 mA bias current and 3x1.4= +5, -5 V rails is "plenty enough", and 135 mA and +9 -9 V rails demarks the "wow that's a lot of power" territory.
350 mA and 15 V rails is still hugely over-specified, but getting within reach of a practical design. Make it 10 V and 200-250 mA and I think people will give it serious thought. The circuit has a feedback loop, it should run quite nicely in class AB. You could make a nice little amp that was almost always within class A for headphones, but happily moves into class AB for desktop speakers.
For class AB amps (all op amps, most buffers) power is cheap, so you can afford to design well over realistic expectation. For a class A design, the power budget unravels quickly, but everyone has their pain point. 5W? 10W? 40W? 100W? How high a circuit power consumption before the additional filtering, bigger power transformer, and larger heatsinks (and heat) become "not worth it" - that's up to you.
I can say however that 10 mW is sufficient from a physical clipping point of view. Of course both 16 ohm and 300 ohm loads must be considered, and the currents and voltages raised accordingly to cover both situations.
As the little diagram attached tries to show, 10 mW is a reasonable lower limit, while 100 mW is a pretty good upper one, based on experience diminishing returns kick in heavily after that (for class A). The diagram indicates the current and voltage required to deliver that power into the outer range of headphone loads.
For a class A output stage, that means 28x1.4=40 mA bias current and 3x1.4= +5, -5 V rails is "plenty enough", and 135 mA and +9 -9 V rails demarks the "wow that's a lot of power" territory.
350 mA and 15 V rails is still hugely over-specified, but getting within reach of a practical design. Make it 10 V and 200-250 mA and I think people will give it serious thought. The circuit has a feedback loop, it should run quite nicely in class AB. You could make a nice little amp that was almost always within class A for headphones, but happily moves into class AB for desktop speakers.
Attachments
rjm, stop talking watts. It is a meaningless measure. As loads for headphones range from 25 ohms to 600 ohms, the same 1 volt output can be 40mW to 2mW and an acoustical output of 122dB(SPL) (AH-D2000) to 101db(SPL) (K141M). Amplifiers are constant voltage devices and will supply whatever current is needed to reach that voltage. watts be damned.
Please talk about capabilities of the load instead such that an amplifier is transparent. If I had an amp that clipped at the point of 120dB(SPL) peaks, it would be a very painful experience. Painful not for the loudness, but the distortion caused by the amplifier reaching its limit.
Peaks are good, let them through. It is the distortion that hurts, not the loudness.
As for "overspecified", look at the Audio-Technica ATH-W5000. 40ohm@2W. To be transparent, you need 9Vrms (25.2 Vp-p / 315mAp) to drive it. Granted, that's 135dB(SPL) and I understand what you're inferring.
I don't think it is proper when you buy a Ferrari of headphone to speed limit it to 55MPH especially when that artificial limit will cause pain for the listener.
Please talk about capabilities of the load instead such that an amplifier is transparent. If I had an amp that clipped at the point of 120dB(SPL) peaks, it would be a very painful experience. Painful not for the loudness, but the distortion caused by the amplifier reaching its limit.
Peaks are good, let them through. It is the distortion that hurts, not the loudness.
As for "overspecified", look at the Audio-Technica ATH-W5000. 40ohm@2W. To be transparent, you need 9Vrms (25.2 Vp-p / 315mAp) to drive it. Granted, that's 135dB(SPL) and I understand what you're inferring.
I don't think it is proper when you buy a Ferrari of headphone to speed limit it to 55MPH especially when that artificial limit will cause pain for the listener.
The original AKG K141M data was wrong. Specs from AKG are really 200mW, not 80mW. That now takes my max voltage up to 11 Vrms. That's 15.3Vp, so rails now have to go to 18
Now I can settle on a final power supply. I'll get a couple of these. Being 30VA I can do my original 20 watts of class-A no problem. I got the fins, so why not?
Double stacked they will fit in my case. The next PCB will have to be shorter to allow space for the front panel items and include the rectifier and filter. It'll be tight, but workable.
Now I can settle on a final power supply. I'll get a couple of these. Being 30VA I can do my original 20 watts of class-A no problem. I got the fins, so why not?
Double stacked they will fit in my case. The next PCB will have to be shorter to allow space for the front panel items and include the rectifier and filter. It'll be tight, but workable.
For the purposes of discussion, a couple of points:
I guess the biggest one is "amplifiers are constant voltage devices". You know very well that the output current of an amplifier is limited, and the class A output current doubly so. We are discussing how much bias current is needed, and equivalently how much class A output power is a reasonable design point to choose for a push pull circuit that will typically transition to class B after that limit is reached, up to the limit of hard (voltage) clipping. That we are using voltage amplifier topologies is totally tangential to the conversation.
Second point: "max power" and "max SPL" manufacturer specifications are as meaningful as the "absolute maximum" rating for semiconductors. You make sure they are never exceeded, it is not necessary to design the driving amplifier to exceed them. The ATH-W5000 is a good case : 40 ohms and 102 dB(/mW?) sensitivity. The 2W max input is just a consequence of the voice coils being enormous by headphone standards and therefor able to absorb the extra heat safely. The typical power requirements are no different than a pair of SR-60's, i.e. a few mW is more than sufficient.
"To be transparent, you need 9Vrms (25.2 Vp-p / 315mAp) to drive it." should be
"To induce permanent hearing loss, you need 9Vrms (25.2 Vp-p / 315mAp) to drive it."
I'll pass. Thanks anyway.
" If I had an amp that clipped at the point of 120dB(SPL) peaks, it would be a very painful experience. "
It would be painful one way or another, for a number of reasons. First is that it would be very loud. Second because this is over the max SPL for most headphones, chances are the headphone chassis/driver would be distorting severely by that point. Third is, as you say, because most headphone amps would be clipping or distorting badly by this point.
It wouldn't in our case because we are talking about a class A push-pull design and discussing the bias current requirements in terms of maximum available output power. It would just transition to class B operation and it's unlikely you'd even notice. Your headphones would, however, most likely not be happy campers given the max SPL ratings. Your eardrums, who knows.
Anyway, more to the point, do you know that for sure? I mean, when listening to headphones, are you certain the peaks are that loud?
I know for a fact that I never (ever!) exceed 2V rms peak into my HD-600. That's 13 mW, 11 dB(mW), so 97(dB/mW) + 11 = 108 dB.
For class A push pull, and for the 300 ohm HD600's, I could have the bias current at 10 mA (about 10 mW class A) and be good to go, certain that 99% of the time I will be operating in class A.
For a general release design, you have to look at reasonable worst case situations: so lower sensitivity headphones, louder than normal output, and the variety of impedances.
And you conclude about 50 mW, or class A bias of about 100 mA and, allowing some extra class AB headroom, the voltage rails about 10 V.
There really is no point in exceeding these figures for a headphone amp, except for very specialized cases like AKG K1000, or as in your case dual-use. I'm not saying "don't", I'm just saying "don't pretend that there is any real world advantage in doing so".
I guess the biggest one is "amplifiers are constant voltage devices". You know very well that the output current of an amplifier is limited, and the class A output current doubly so. We are discussing how much bias current is needed, and equivalently how much class A output power is a reasonable design point to choose for a push pull circuit that will typically transition to class B after that limit is reached, up to the limit of hard (voltage) clipping. That we are using voltage amplifier topologies is totally tangential to the conversation.
Second point: "max power" and "max SPL" manufacturer specifications are as meaningful as the "absolute maximum" rating for semiconductors. You make sure they are never exceeded, it is not necessary to design the driving amplifier to exceed them. The ATH-W5000 is a good case : 40 ohms and 102 dB(/mW?) sensitivity. The 2W max input is just a consequence of the voice coils being enormous by headphone standards and therefor able to absorb the extra heat safely. The typical power requirements are no different than a pair of SR-60's, i.e. a few mW is more than sufficient.
"To be transparent, you need 9Vrms (25.2 Vp-p / 315mAp) to drive it." should be
"To induce permanent hearing loss, you need 9Vrms (25.2 Vp-p / 315mAp) to drive it."
I'll pass. Thanks anyway.
" If I had an amp that clipped at the point of 120dB(SPL) peaks, it would be a very painful experience. "
It would be painful one way or another, for a number of reasons. First is that it would be very loud. Second because this is over the max SPL for most headphones, chances are the headphone chassis/driver would be distorting severely by that point. Third is, as you say, because most headphone amps would be clipping or distorting badly by this point.
It wouldn't in our case because we are talking about a class A push-pull design and discussing the bias current requirements in terms of maximum available output power. It would just transition to class B operation and it's unlikely you'd even notice. Your headphones would, however, most likely not be happy campers given the max SPL ratings. Your eardrums, who knows.
Anyway, more to the point, do you know that for sure? I mean, when listening to headphones, are you certain the peaks are that loud?
I know for a fact that I never (ever!) exceed 2V rms peak into my HD-600. That's 13 mW, 11 dB(mW), so 97(dB/mW) + 11 = 108 dB.
For class A push pull, and for the 300 ohm HD600's, I could have the bias current at 10 mA (about 10 mW class A) and be good to go, certain that 99% of the time I will be operating in class A.
For a general release design, you have to look at reasonable worst case situations: so lower sensitivity headphones, louder than normal output, and the variety of impedances.
And you conclude about 50 mW, or class A bias of about 100 mA and, allowing some extra class AB headroom, the voltage rails about 10 V.
There really is no point in exceeding these figures for a headphone amp, except for very specialized cases like AKG K1000, or as in your case dual-use. I'm not saying "don't", I'm just saying "don't pretend that there is any real world advantage in doing so".
RJM, you are so far off the point I'm crushed. The user is in control of the volume setting, not you and your under-rated amplifiers. I've seen 5V peaks to my Denons. That equates to an instant SPL of 136 if you can believe the specs. My max average level is 1v, and that's 122. But that's the highest I go, I prefer 3 clicks down but I don't recall what that level is.
I won't hurt my hearing because it is clean and undistorted rather than your hardware which will cause hearing loss from the square edging.
The distortion that bothers me is doppler. I'm amazed by it as I've never heard it before yesterday.
I'll set my bias where ever I want. With my heatsink barely sweating 40W, even the worst case headphone loading, 350mA of idle works fine for me. TO-3s are barely 31C at that setting. More work well for speakers. 800mA each would never cross into B with 16V rails and that might be my final as of now.
And before you ask, measurable distortion does increase with lower idle current.
I won't hurt my hearing because it is clean and undistorted rather than your hardware which will cause hearing loss from the square edging.
The distortion that bothers me is doppler. I'm amazed by it as I've never heard it before yesterday.
I'll set my bias where ever I want. With my heatsink barely sweating 40W, even the worst case headphone loading, 350mA of idle works fine for me. TO-3s are barely 31C at that setting. More work well for speakers. 800mA each would never cross into B with 16V rails and that might be my final as of now.
And before you ask, measurable distortion does increase with lower idle current.
Last edited:
1) most does not mean all. Take the Sonys for example, the MDR-CD6 is rated for 136 max. If you don't allow the signal to it to be clean, your electronics are causing the pain to the user. If I follow your advice and under-rate for least common denominator so no headphones are ever damaged from being driven too hard, that would end up being very distorted for the ones that are capable. most headphone amps on the market think 100mW into 100 ohms is acceptable. And boy do they sound like crap from early distortion.
2) Pain is not loudness. Pain comes from the distortion.
Just for the record, when I do outside live sound I prefer 107 dB(SPL) c-weighted average with allowable peaks to 119.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.