LM4562 based headphone amp

[zdr]

This is something I use everyday that sounds just right for me, but I am trying to make things a bit tidier so I redesigned the board from scratch. I plan to use LM4562 in a can which allegedly sounds better. Before it goes into homebrew pcb production, I would like to get the community's view of this design, in case I made some stupid mistakes. If someone finds it useful, I can also make eagle files available for everyone.

[sudhir kumar]

please post copper side in pdf for diy use.
thanks.

[agdr]

It looks like you are using +/-12Vdc power supply rails and a stage gain of 10x. What impedance of headphones or IEMs are you using the circuit with?

One suggestion for driving lower impedance phones would be to substitute a 70mA/channel NJM4556, like used in the O2 headphone amp, in place of the 20mA/channel LM4562, then drop the output resistors to 10 ohms or so.

http://www.google.com/url?sa=t&rct=j...xLoy5w&cad=rja (NJM4556 datasheet)

The pin-out on the two chips are the same so you could just swap back and forth. You could even socket those 100R output resistors with 0.1 headers.

Here is a TO-99 to DIP adapter so you could still use that metal can version of the LM4562 with the DIP socket:

http://cimarrontechnology.com/to-99t...n020601-1.aspx

The metal cans really are supposed to sound better. I saw an post or article somewhere along the way, a few years ago, from someone who worked at National saying the group was going to look into why that was, but about that time National disbanded the audio group.

[sgrossklass]

Originally Posted by agdr

The metal cans really are supposed to sound better. I saw an post or article somewhere along the way, a few years ago, from someone who worked at National saying the group was going to look into why that was, but about that time National disbanded the audio group.

Probably inductive coupling between signal and rails on the DIP8 package with standard pinout (remember, class B currents and stuff). Seems to be a significant problem when moving into RF territory, with special "low distortion" pinouts available as a consequence.

Looking at the circuit, 100 ohms of R_out is a little high for most anything save for 600 ohm cans. And yes, LM4562s aren't super high current, though with a NE5532 commonly doing a fairly nice job "barefoot", the '4562 certainly shouldn't do too badly either. (I'd still prefer a buffered arrangement, but space is obviously tight here.)

Board wise, I might look into making xmfr secondary side loop area a bit smaller (maybe move the xmfr to one side), but other than that it looks pretty good to me for a single-sided board. If you can find the space, a few small caps for RF suppression at the rectifier might not be that bad an idea either.

[Mooly]

Just a thought... the opamp feedback network is DC coupled and of unequal impedance to the input network. Have you measured for real what the DC offset is ? The LM4562 has fairly low input bias currents, a 5532 would certainly have a significant offset many 10's or 100's of millivolts which isn't good. Any bjt opamp will have this issue used like this.

Why not at least include space for caps in the feedback return. You can always link them out if not needed. and perhaps look at equalising the networks. Noise isn't such an issue in reality.

[sgrossklass]

Well spotted, Mooly... that 100k pot is a little on the high side, too. I'd go down all the way to 10k if that's not an issue in terms of sources. Given the low Vnoise level of a 4562, that would drop noise by (up to) almost 10 dB... nothing to sneeze at for sure.

[JensH]

Apart from the suggestions above you should also increase the value of C7 and C8. Probably to at least 100uF. 220 or 470uF could be a good starting point.

[zdr]

Thank you for all the good comments! Yep, those filter caps have grown in the meantime to 220uF. Latest PCB version below.

[zdr]

Originally Posted by agdr

It looks like you are using +/-12Vdc power supply rails and a stage gain of 10x. What impedance of headphones or IEMs are you using the circuit with?

One suggestion for driving lower impedance phones would be to substitute a 70mA/channel NJM4556, like used in the O2 headphone amp, in place of the 20mA/channel LM4562, then drop the output resistors to 10 ohms or so.

http://www.google.com/url?sa=t&rct=j...xLoy5w&cad=rja (NJM4556 datasheet)

The pin-out on the two chips are the same so you could just swap back and forth. You could even socket those 100R output resistors with 0.1 headers.

Here is a TO-99 to DIP adapter so you could still use that metal can version of the LM4562 with the DIP socket:

TO-99 to 8-pin DIP Adapter (p/n 020601B)

NJM4556 - isn't it supposed to be inferior to LM4562 (LME49720)?

I will be using GS-1000 and FA-003, hence 100ohm at the output. Board will be fitted with dil8 socket and with little leg bending, it should take cans as well as dip8s.

[nereis]

LM4562 is good, in fact it's better than the venerable NE5532 for driving difficult 600 Ohm loads, but even with your 100 Ohms series build-out resistor, GS-1000 would look like 132 Ohms to your opamp, and very few small signal opamps tend to cope well with that kind of load.

Choosing another opamp designed for heavier loads like the OPA551 should work pretty well with the kind of load you are giving it, or if money is no subject, and you want the lowest distortion/noise from just LM4562, just parallel them up and sum the outputs with 1 ohm resistors. 3x LM4562 per channel should make them quite comfortable driving the headphones in series with the 100 ohm resistor.

NJM4556 is nice because it has moderately good distortion performance for cheap, and is rather good at driving heavy loads. This works out nicely in an output stage as you are not using it for gain.