[Circuitsoft]

So, I built a headphone amp using a TDA1517P that I rescued from a pair of cheap desktop speakers. Has lots of hiss, but adds so much impact and clarity to other sounds, that I figured I'd better keep using an amp. Anyway, I built it on a radioshack generic PCB, and it's been bare, on my desk, in-use for over 18 months.

The other day, I leaned back in my chair, pulled on the headphone wire, and shorted the 12Volt power supply directly into the left er6i. Needless to say, the earphone doesn't work anymore. The amp still appears to be fine, though I don't want to push it.

Deciding I should do it right this time, I figure I'll do SMT, with everything on one side so it doesn't short like that again.

Since the TDA1517 (which is made to drive 6W+6W into a 4ohm load) sounded so good, I figured I'd find an op-amp with high current output.

First look, OPA2677 - 500ma output, rail-to-rail, Burr Brown. Since I've seen no mention of it, whatsoever, I'm not sure what to think.
Next: AD8397 - Looking pretty good, I guess, but I wonder if it'll have the same impact as the TDA1517 did.

Can anyone weigh in here? I've even been tempted to do it out of discretes with a design modeled after my dad's Dynaco Stereo 80, but at much lower voltages.

 

[diditmyself]

Quote:

Originally Posted by Circuitsoft /img/forum/go_quote.gif So, I built a headphone amp using a TDA1517P that I rescued from a pair of cheap desktop speakers. Has lots of hiss, but adds so much impact and clarity to other sounds, that I figured I'd better keep using an amp. Anyway, I built it on a radioshack generic PCB, and it's been bare, on my desk, in-use for over 18 months. The other day, I leaned back in my chair, pulled on the headphone wire, and shorted the 12Volt power supply directly into the left er6i. Needless to say, the earphone doesn't work anymore. The amp still appears to be fine, though I don't want to push it. Deciding I should do it right this time, I figure I'll do SMT, with everything on one side so it doesn't short like that again. Since the TDA1517 (which is made to drive 6W+6W into a 4ohm load) sounded so good, I figured I'd find an op-amp with high current output. First look, OPA2677 - 500ma output, rail-to-rail, Burr Brown. Since I've seen no mention of it, whatsoever, I'm not sure what to think. Next: AD8397 - Looking pretty good, I guess, but I wonder if it'll have the same impact as the TDA1517 did. Can anyone weigh in here? I've even been tempted to do it out of discretes with a design modeled after my dad's Dynaco Stereo 80, but at much lower voltages.

AD8397 doesn't lack impact, quite the contrary. If you have low impedance phones and need a lot of current it can be a good choice.

 

[fortney]

Hi,

Before making a decision you might want to look at AMB's Mini 3. A complete kit can be purchased from Glass Jar Audio.

I built the lower current model; it is the best of my several small amps.

F

 

[Circuitsoft]

Quote:

Originally Posted by fortney /img/forum/go_quote.gif ... you might want to look at AMB's Mini 3.

I was hoping to build something a tad smaller. I'm an EE at a design services firm, so I'm plenty capable of building my own PCBs. I'm just unsure of either (A) how to charge multiple Lithium cells in series or (B) how to build a boost converter (3.7->12V) of reasonable efficiency.

 

[diditmyself]

Why not settle for a single cell. If size is such an issue I'd guess you're going to use this amp with a portable player and IEMs. In that case 3.7 V supply would be enough if rail-to-rail opamps are used like LME49721, AD8532, AD8616 or AD8656. If you go for a 3-ch topology you don't need large rails-to-ground caps to sink/source the return currents. This could make a really small, lightweight amp with very long runtime.

 

[Steve Eddy]

Quote:

Originally Posted by diditmyself /img/forum/go_quote.gif If you go for a 3-ch topology you don't need large rails-to-ground caps to sink/source the return currents.

The load currents return from whence they came, which are your rails-to-ground caps. A three channel amp doesn't change this. And if you want to keep ripple voltage the same, a three channel amp will need the same amount of rail capacitance as a two channel amp.

se

 

[Circuitsoft]

Quote:

Originally Posted by diditmyself /img/forum/go_quote.gif AD8397 doesn't lack impact, quite the contrary. If you have low impedance phones and need a lot of current it can be a good choice.

Planning on using low-impedance phones (12ohm nuForce or 16ohm Etymotic). What voltage does the 8397 need to produce such impact?

 

[Circuitsoft]

Quote:

Originally Posted by Koyaan I. Sqatsi /img/forum/go_quote.gif The load currents return from whence they came, which are your rails-to-ground caps. A three channel amp doesn't change this. And if you want to keep ripple voltage the same, a three channel amp will need the same amount of rail capacitance as a two channel amp. se

I think his point was that you don't need caps from the rails to the half-rail ground reference as the output doesn't load the ground.

 

[diditmyself]

Quote:

Originally Posted by Koyaan I. Sqatsi /img/forum/go_quote.gif The load currents return from whence they came, which are your rails-to-ground caps. A three channel amp doesn't change this. And if you want to keep ripple voltage the same, a three channel amp will need the same amount of rail capacitance as a two channel amp. se

Oh yes, it does change things. I'm not speaking of ripple. I think ripple mostly is a non-issue when opamps with extremely good rejection ratios are used.

I'm thinking about crosstalk. If you want acceptable crosstalk with a passive ground you need thousands of uFs when low impedance phones are used, and that means large caps, especially if you want "audio approved" ones.

 

[Steve Eddy]

Quote:

Originally Posted by Circuitsoft /img/forum/go_quote.gif I think his point was that you don't need caps from the rails to the half-rail ground reference as the output doesn't load the ground.

Wouldn't need that even for a two channel amp. So don't know why he said it in the context of a three channel amp.

se

 

[diditmyself]

Quote:

Originally Posted by Circuitsoft /img/forum/go_quote.gif Planning on using low-impedance phones (12ohm nuForce or 16ohm Etymotic). What voltage does the 8397 need to produce such impact?

I haven't noticed that the sound changes that much as the voltage drops. The only amp I'm using it in now is Mini3, and it starts to distort and shortly thereafter dies as the voltage drops. At about the same time the blue LED goes out so I guess it's somewhere in the regions of 3V. I don't know if it's OPA690 or AD8397 that goes out first.

 

[diditmyself]

Quote:

Originally Posted by Koyaan I. Sqatsi /img/forum/go_quote.gif Wouldn't need that even for a two channel amp. So don't know why he said it in the context of a three channel amp. se

I don't even know how to define a 3 ch amp. I really meant the use of a low impedance high quality active device as virtual ground vs resistors and capacitors. In a battery operated amp there aren't that many options. Battery center point is not a good alternative.

 

[jcx]

you should review my (jcx) posts with 3-channel, crosstalk keywords

a "active gnd" supply splitter that is not "3-channel": (signal and output gnd are the "star" point - no "3rd-channel" buffer in between)

Quote:

3/4 channel "active gnd" schemes are not going to beat a good implementation of star gnd - any buffer op amp will introduce noise from its input stage and distortion from the nonlinear operation of its AB output stage and frequency dependant output impedance from the buffer internal feedback properly layed out Cu trace and wiring to the output TRS connector as star gnd point doesn't introduce extra active device noise, can be orders of magnitude lower in Z at high audio frequencies and can avoid "gnd contamination" by virtue of separating the pwr_gnd branch of the star gnd so that load current doesn't flow in the sig_ref or fdbk_ref gnd branches pwr_gnd could be (preferably) the center of a symmetric battery stack or dual ps instead of the "active gnd" battery splitter shown but the star gnd scheme is still superior with the "active gnd" as imperfections in the supply splitter op amp oputput become common mode ps terms and the channel amplifier op amp's cm psrr is usually higher than its feedback loop gain and always at least equal to the internal loop gain

 

[Steve Eddy]

Quote:

Originally Posted by diditmyself /img/forum/go_quote.gif I'm thinking about crosstalk. If you want acceptable crosstalk with a passive ground you need thousands of uFs when low impedance phones are used, and that means large caps, especially if you want "audio approved" ones.

Crosstalk from what source exactly?

se

 

[diditmyself]

Koyaan I. Sqatsi, I'm not clever enough to understand your question. My point is that a low impedance high quality active device used for virtual ground sounds better, measures better, could be made smaller and cheaper compared to a virtual ground à la CMOY where capacitors do the job.

jcx, I've built my amps with star ground. My gut feeling told me it's the way to go. I regard your example as a 3 ch amp. From listening tests I've found no substantial difference from separating virtual ground from ouput ground as long as the same high quality active device is used. I think I agree to what your saying.