If I put my notes here, I might be able to find them again later!
Headphone amplifiers: thinking aloud
The sobering fact is that the built-in headphone jack on most modern consumer electronics provides pretty decent performance. Taking that output and routing it through an external headphone amplifier rarely improves things, and frequently has a negative impact owing to increased background noise.*
[* This is a simple consequence of adding a volume control which attenuates the signal, and a gain stage which amplifies it back up. Even if the gain stage has the same noise floor as the input signal, the S/N is reduced by the amount of attenuation.]
There are specific use cases, particularly with "outlier" headphone models that require unusually high voltages or currents to drive, but in the main, for generic 16 ohm IEHs and the generic headphone ICs used in consumer electronics, I've found that external headphone amplifiers aren't worth the trouble and expense.
Instead, I've taken (I realise now) an elitist approach to focus on a desktop headphone amplifier circuits that accepts a high-end line output and amplifies it to drive high-end headphones.
In other words, to add a headphone amplifier where there was none, rather bolt an external headphone amplifier to the back end of a built-in one. Also, I cut my losses and optimized for high impedance HD-600s, rather than trying for a circuit that can handle 16 ohm IEHs as well. The latter, though, was more lazyness on my part. There's no reason why a circuit can't do both and still be high end, it just complicates things.
I'd now like to go back to the drawing board, with a blank sheet, and start over. The goal is to build a better high-end, quasi-cost-no-object* headphone amp.
(* Cost and size are still considerations, but not overriding ones.)
There are two primary design vectors:
1. Low distortion.
2. Low noise.
Most people tend to get hung up on the former, but the fact is 100 mA though a complementary, matched transistor push-pull buffer is going to drive any headphone you care to name with low distortion, even open loop. For 300 ohm phones, 25 mA is more than sufficient.
The "low noise" part doesn't get as much attention, and I agree its the lesser of the two in importance. Noise scales with amplifier gain. The more gain, the more noise. First rule, then, is to set the gain to no more than you absolutely need. Next thing is to look at the input referred noise. Audio opamps run 3-10 nV/sqrtHz. It's unlikely that we can do better by going discrete. At the very least, though, we should make sure that the circuit is not adding any more noise than absolutely necessary: resistor noise, power supply noise, ground loops, and unwanted parasitic/EMI/RFI coupling must all be kept under control.
I keep coming back to the sapphire amp and the question "can we do better?" "How?"...
First thing would be to redesign the diamond buffer to run at 100 mA.
Second thing is to replace the op amp with a discrete voltage amplifier. No, I'm not sure it would sound better than the OPA134 I use presently, but it's a point of honor to give it a shot.
Third thing would be to move away from "generic" BD135,136 transistors to something more "audio grade", and, additionally, to match the output devices.
[* This is a simple consequence of adding a volume control which attenuates the signal, and a gain stage which amplifies it back up. Even if the gain stage has the same noise floor as the input signal, the S/N is reduced by the amount of attenuation.]
There are specific use cases, particularly with "outlier" headphone models that require unusually high voltages or currents to drive, but in the main, for generic 16 ohm IEHs and the generic headphone ICs used in consumer electronics, I've found that external headphone amplifiers aren't worth the trouble and expense.
Instead, I've taken (I realise now) an elitist approach to focus on a desktop headphone amplifier circuits that accepts a high-end line output and amplifies it to drive high-end headphones.
In other words, to add a headphone amplifier where there was none, rather bolt an external headphone amplifier to the back end of a built-in one. Also, I cut my losses and optimized for high impedance HD-600s, rather than trying for a circuit that can handle 16 ohm IEHs as well. The latter, though, was more lazyness on my part. There's no reason why a circuit can't do both and still be high end, it just complicates things.
I'd now like to go back to the drawing board, with a blank sheet, and start over. The goal is to build a better high-end, quasi-cost-no-object* headphone amp.
(* Cost and size are still considerations, but not overriding ones.)
There are two primary design vectors:
1. Low distortion.
2. Low noise.
Most people tend to get hung up on the former, but the fact is 100 mA though a complementary, matched transistor push-pull buffer is going to drive any headphone you care to name with low distortion, even open loop. For 300 ohm phones, 25 mA is more than sufficient.
The "low noise" part doesn't get as much attention, and I agree its the lesser of the two in importance. Noise scales with amplifier gain. The more gain, the more noise. First rule, then, is to set the gain to no more than you absolutely need. Next thing is to look at the input referred noise. Audio opamps run 3-10 nV/sqrtHz. It's unlikely that we can do better by going discrete. At the very least, though, we should make sure that the circuit is not adding any more noise than absolutely necessary: resistor noise, power supply noise, ground loops, and unwanted parasitic/EMI/RFI coupling must all be kept under control.
I keep coming back to the sapphire amp and the question "can we do better?" "How?"...
First thing would be to redesign the diamond buffer to run at 100 mA.
Second thing is to replace the op amp with a discrete voltage amplifier. No, I'm not sure it would sound better than the OPA134 I use presently, but it's a point of honor to give it a shot.
Third thing would be to move away from "generic" BD135,136 transistors to something more "audio grade", and, additionally, to match the output devices.
Total Comments 0
