Planning a TPA6120A2 amp. Need advice

[Lucien]

Hi All,

After dragging my feet for a few years (I've been busy) I've finally decided to build myself a headphone amp. I'm planning on a TPA6120A2 amp, and I've already looked at what Per-Anders and others have done, but I have some questions about the design:

Everyone seems to use an inverting input, and Per-Anders wrote that this is done to reduce distortion by a factor of 10. I've read the datasheet, but I don't understand how he comes to this conclusion. What have I missed? I do guess that it would be easier to build it as inverting as there'd be no need for a bias balance resistor between +in and GND.

Looking at the graphs for SNR, would it make the most sense to run the TPA6120A2 at a gain of 2.5-3 to maximise SNR? Some other designs seem to run it at a gain of 1 though, is there a good reason for this?

Finally, when putting in a top ground plane, what's the minimum distance I should have between the edge of the plane and signal traces?

TIA for the help...

[nelsonvandal]

I use it non-inverted with a TLE2426 and OPA690 as ground channel. To cure the DC-offset I use a servo circuit. It sounds sweet, but it's not a miracle drug.

Why use TPA6120 with input buffers? You'll get two degrading stages. An OPA2134, AD8620 or AD8066 or many of the other common opamps used for audio, drives most headphones with no problems. If you just want to buffer an opamp, you'll get a better sonic result with a common discrete "diamond buffer" in closed loop with your favourite opamp.

[jcx]

read Walt Jung's Op amp Applications chapter on line driver/composite op amps for a differnt view - high quality input op amps with high speed high current cfa op amps providing added gain and buffering inside one feedback loop improve the character of Both amps

[BWRX]

Originally posted by nelsonvandal
Why use TPA6120 with input buffers?

Because the TP6120 is a CFA (current feedback amp) the inverting input has a rather low impedance that can be difficult for most sources to drive, hence the use of input buffers.

[peranders]

Originally posted by Lucien
Everyone seems to use an inverting input, and Per-Anders wrote that this is done to reduce distortion by a factor of 10.

This is a general rule and the distortion will be reduced, not always by 10 maybe but some. In this case hardly important since the distortion is ultralow when it's "high".

[peranders]

Originally posted by Lucien
Finally, when putting in a top ground plane, what's the minimum distance I should have between the edge of the plane and signal traces?

TIA for the help...

This is dependent of how you'll do the pcb. Boards like mine have 12 MILS (0.012") as clearance.

[Lucien]

Thanks for the comments guys. Yes, the low input impedence is why I'm putting a buffer ahead of the TPA6120. I'll probably start with the AD8610. This being my first headphone amp, there's lots of different things to try out.

Per-Anders,

I'm afraid I don't understand what you mean by"ultralow when it's "high"". But it seems I will be going for an inverting setup, so it doesn't matter anyway.

And for the PCB clearance, I'm using EAGLE and can configure how much clearance to give between trace and plane. There's going to be a tradeoff, and I understand ground planes are usually added for stability, but in this case the extra capacitance would degrade performance.

Curious, but has anyone tried to put the input buffer op-amp in an inverting configuration as well?

[head_spaz]

I'm afraid I don't understand what you mean by"ultralow when it's "high"

I think he means that even at the 6120's highest distortion level, it still measures "low" by comparison to others.
It was a relative statement.

[nelsonvandal]

Originally posted by jcx
read Walt Jung's Op amp Applications chapter on line driver/composite op amps for a differnt view - high quality input op amps with high speed high current cfa op amps providing added gain and buffering inside one feedback loop improve the character of Both amps

Mr Jung never mention the words "sound", "sonic", "timbre" etc. Is this a theoretical benefit with better measured performance, or is it supposed to sound better? I've tried to buffer opamps with other opamps in many different combinations. I've always found a single opamp to sound superior, unless there's too low output current and audible distortion. Opamps do color the sound. Two opamps color it more than one. But it's probably a matter of taste. I prefer an uncolored sound.

A diamond buffer in closed loop (Jung multiloop or simple common loop) sounds better to my ears, less colored, is cheap, easy to work with and can be made quite small. TPA6120 is not the amp that makes everything else obsolete even though it has a nice sound signature. It definately adds a coloration - silky and smooth.

[jcx]

multiloop operation can improve measurable characteristics - the input amp doesn't have load current issues driving the noninverting input of the output amp, maximizing the input op amp gain and linearity - the input op amp can have separately filtered supplies, the simple fact of higher loop gain also increases psrr
the output amp inside the feedback loop of the input amp is corrected for errors from noninverting drive and thermal modulation, both of which it cannot itself correct since they cause error in the input differencing operation

I think the majority of "opamp sound" discussions are suspect - there may be noticeable differences in noise, clipping behavior or psrr/interference rejection in differing circuit implementations but if the pro "Golden Ears" at Stereophile couldn't tell a tweaked Carver amp from a expensive tube amp driving speakers with large impedance variations:
http://www.diyaudio.com/forums/showt...392#post152392
I don't know why I should readily believe competently implemented op amp headphone amps really have different "sound" into very benign dynamic headphones loads attributable solely to op amp manufacturer rather than overall amplifier frequency response and output impedance - when used in their linear operating limits and subjectively compared under properly controlled conditions; blind and level and frequency response matched to 0.1 dB

I do however choose to design for lowest possible distortion and large output Voltage and Current headroom - not something many common single op amp only headphone amplifiers can achieve with low impedance phones - but which multiloop TPA6120 or other buffer in loop types can

High loop Gain Composite Op Amp Circuits