I've been out of the loop. Is there a consensus on the best headphone IC made today?
Parameters:
Properly drives all common headphones, from 32ohm to 600ohm, all sensitivities
Bipolar supply at least +/-8V (or more)
Ultra low noise (not obsessive about THD)
Minimal external parts preferred
Would prefer no external transistors
I know, loaded question. But all opinions appreciated.
Parameters:
Properly drives all common headphones, from 32ohm to 600ohm, all sensitivities
Bipolar supply at least +/-8V (or more)
Ultra low noise (not obsessive about THD)
Minimal external parts preferred
Would prefer no external transistors
I know, loaded question. But all opinions appreciated.
Last edited:
Properly drives all common headphones, from 32ohm to 600ohm, all sensitivities
BUF634A is just a buffer though, so no voltage gain out of this one. Plus, minimum output impedance is 5 ohms, much like it's been for previous similar parts. Mind you, two of them would make for a splendid composite circuit with your dual opamp of choice, The Wire style. There are some neat ones with very good input impedance nonlinearity despite their nominal FET frontend (e.g. OPA1652 if memory serves), though LM4562 / LME49720 would also do a good job.
What you are looking for is pretty much a niche application, and IC manufacturers are not exactly rushing out to provide suitable single-chip solutions.
OPA1622 and OPA1688 are pretty neat as far as "opamps with some more grunt" go (power dissipation on the VSON package parts even is surprisingly high with the integrated thermal pad, though soldering these is not a trivial exercise for the hobbyist), but these will merely provide normal output power as in 150 mW / 32 ohms max, not several watts as you may need for an HE-6.
The last higher-power part was the trusty TPA6120A2 @ up to 700 mA per channel, but it tends to appreciate 10 ohms of output series resistance for stability, which is a bit of a no-go for impedance-critical cans (mostly IEMs, but even for some of Sennheiser's 50 ohm fullsize models it's a bit above optimum). Presumably that's a result of its past as a DSL line driver. It also sort of requires a previous stage to drive it.
There definitely are multiple ways of skinning this proverbial cat. Even the medium-power chipamp LM1876 (a 2x 20 W / 8 ohm part) has been used with external compensation (draining open-loop gain) to get its gain down to 8 dB safely. (The entry-level Lake People amps are doing this.) Possibly not the last word in finesse but presumably near-indestructible. The O2 amp parallels two of the trusty NJM4556As (the earguably earliest headphone driver opamps, now 35 years old) for a good 0.5 W into 32 ohms.
Your list of requirements still needs some work.
How much output power do you consider enough to "properly drive" your loads?
How much noise is "ultra low", output referred? 5 µV? 2 µV? 1 µV or even less?
How much voltage gain is required, if any (or what input levels are expected)? Covering a wide range of headphones often necessitates a choice among several different options.
How much source impedance is expected, and how constant is it going to be?
What sort of IC packaging can you deal with?
What is your power budget?
I would very much suggest making a list of "black box" specifications first. The key to high performance is not necessarily using the latest, most fashionable parts, but rather getting the overall system design right and using appropriate parts. ICs can do a lot of the heavy lifting for you but performing magic is not generally a standard feature.
What you are looking for is pretty much a niche application, and IC manufacturers are not exactly rushing out to provide suitable single-chip solutions.
OPA1622 and OPA1688 are pretty neat as far as "opamps with some more grunt" go (power dissipation on the VSON package parts even is surprisingly high with the integrated thermal pad, though soldering these is not a trivial exercise for the hobbyist), but these will merely provide normal output power as in 150 mW / 32 ohms max, not several watts as you may need for an HE-6.
The last higher-power part was the trusty TPA6120A2 @ up to 700 mA per channel, but it tends to appreciate 10 ohms of output series resistance for stability, which is a bit of a no-go for impedance-critical cans (mostly IEMs, but even for some of Sennheiser's 50 ohm fullsize models it's a bit above optimum). Presumably that's a result of its past as a DSL line driver. It also sort of requires a previous stage to drive it.
There definitely are multiple ways of skinning this proverbial cat. Even the medium-power chipamp LM1876 (a 2x 20 W / 8 ohm part) has been used with external compensation (draining open-loop gain) to get its gain down to 8 dB safely. (The entry-level Lake People amps are doing this.) Possibly not the last word in finesse but presumably near-indestructible. The O2 amp parallels two of the trusty NJM4556As (the earguably earliest headphone driver opamps, now 35 years old) for a good 0.5 W into 32 ohms.
Your list of requirements still needs some work.
How much output power do you consider enough to "properly drive" your loads?
How much noise is "ultra low", output referred? 5 µV? 2 µV? 1 µV or even less?
How much voltage gain is required, if any (or what input levels are expected)? Covering a wide range of headphones often necessitates a choice among several different options.
How much source impedance is expected, and how constant is it going to be?
What sort of IC packaging can you deal with?
What is your power budget?
I would very much suggest making a list of "black box" specifications first. The key to high performance is not necessarily using the latest, most fashionable parts, but rather getting the overall system design right and using appropriate parts. ICs can do a lot of the heavy lifting for you but performing magic is not generally a standard feature.
Last edited:
I don't need gain, plenty of swing hitting the HP amp.
Really, if my solution drove 98% of common headphones properly, I would be happy.
Low self-noise for me is -110dBu broadband (20-20k), unweighted. Which I believe is roughly 2.5uVrms BB/UW.
THD doesn't really matter, as long as it's less than .01% or so into most loads and most levels.
Iq (quiescent power) isn't an issue.
QFN no problem -- just don't recommend a BGA. i ----HATE---- BGA.
Last edited:
the TPA6120 output Cload isolation can be an inductor, R||L or lossy ferrite bead core:
https://www.diyaudio.com/forums/hea...32-composite-headphone-amp-2.html#post2796117
CFA like the TPA6120 can be used unity gain but still require a feedback R to set bandwith for stability
https://www.diyaudio.com/forums/hea...32-composite-headphone-amp-2.html#post2796117
CFA like the TPA6120 can be used unity gain but still require a feedback R to set bandwith for stability
OPA1622 is very nice. I believe it is the state of the art in headphone amp ICs.
Here is a parallel 8x OPA1622:
Cheap as Chips OPA1688 Low-THD Muscle Amp
Here is a parallel 8x OPA1622:
Cheap as Chips OPA1688 Low-THD Muscle Amp
I've built several headphone amps with the TI TPA6120 and been very happy with it.
http://www.ti.com/document-viewer/TPA6120A2/datasheet
It's not really new, but it sure does work well. No problem driving some of my weird 8 or 16 ohm cans, as well as 32 ohm. Super low noise floor, clean and dynamic. With a decent +/- 15V supply my hunt for headphone amplification is over. Currently I use 2 of them for my collection of quadraphonic headphones.
http://www.ti.com/document-viewer/TPA6120A2/datasheet
It's not really new, but it sure does work well. No problem driving some of my weird 8 or 16 ohm cans, as well as 32 ohm. Super low noise floor, clean and dynamic. With a decent +/- 15V supply my hunt for headphone amplification is over. Currently I use 2 of them for my collection of quadraphonic headphones.
Thanks everyone. GREAT replies - I would buy you all a beer.
For my purposes (low total parts cost, simplicity, small PCB area, low-noise) I'm leaning towards the T/I parts 1688 (US$0.75) and 1622 (US$2.90). I've ordered T/I eval boards and will do some critical A/B listening tests. Will use the best sounding part! The 1622 is clearly the superior part on paper ==> noise, drive current, capacitive drive, open-loop output impedance, thermal metrics, etc.
For my purposes (low total parts cost, simplicity, small PCB area, low-noise) I'm leaning towards the T/I parts 1688 (US$0.75) and 1622 (US$2.90). I've ordered T/I eval boards and will do some critical A/B listening tests. Will use the best sounding part! The 1622 is clearly the superior part on paper ==> noise, drive current, capacitive drive, open-loop output impedance, thermal metrics, etc.
Last edited:
We're still talking about OPA1688 & OPA1622 ? Because at those prices you must be buying them 2000 at a time.
Yes, modest production qtys.
Was also just looking at the INA1620, but (comparing specs) it appears to be a 1622 with four laser-trimmed ultra-tight resistors added to the die. Beautiful part, but more expensive, and I don't need the fancy resistors.
Last edited:
Yeah, tiny QFN / VSON packages aren't very thermal efficient, though the 1622's Rja at 47.6 C/W is manageable for my application (actually surprisingly good for the tiny pkg size). The 1620 is even better at 33.7 C/W. With a 125C operating spec, my app will never get close to max junction temp.
8V/32ohm =250mA such current may handle just a few opamps, one of them AD8397. Relatively expensive, about $3 but if you'll get it no self-oscillating it could be near ideal HPA. I've got S/N -122dbA with ES9038Q2M and -123db harmonics level Review and Measurements of E1DA 9038S BAL Portable DAC & Amp | Audio Science Review (ASR) Forum
Nice! But I wonder about ear health listening to headphones with peaks over 100dB SPL. 100mW into 32 ohm cans roughly translates into >115dB SPL. That's freaking dangerous! Seems like 100mW is plenty for everything but perhaps 8 ohm IEM users?
I'm good with addressing 98% of the market. And the 1622 is a beautiful part on paper -- can't wait to hear it. It doesn't get much better than 700nVrms broadband (20-20k) noise or -121dBu. I think that's close to the effective noise limit of CMOS technology.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.