Head-Fi.org › Forums › Misc.-Category Forums › DIY (Do-It-Yourself) Discussions › What I did next with Texas headamp chip, a digital control desktop / portable.
New Posts  All Forums:Forum Nav:

What I did next with Texas headamp chip, a digital control desktop / portable.

post #1 of 21
Thread Starter 

 

A while ago I posted details of my first experiment with the TPA6120, a high-performance headphone amp IC from Texas Instruments.

 

More recently I have shown a design using multiple NE5532s in parallel, combined with a PGA2310 stereo volume control, a chip which again has an impressive performance on paper.

 

For some time now I have been working on a rechargeable headphone amp to fit in a quality, easily-available enclosure.

 

This design has gone through many iterations, I have drawn up versions with both analog and digital volume controls, some fully DC coupled with a DC servo, some employing the AD8397 or the LT1206 as drivers, sometimes in a composite arrangement with global feedback, some with dual DC wallwarts and some with a single AC wallwart as supply.

 

In this case I have departed from my usual practise of making the PCBs myself, doublesided home-made PCBs at this level of detail are very labour-intensive, although some prototyping was still done before committing to placing an order. The large majority of parts are SMT, the exceptions are the sockets, switches, trimmers and the handwound inductors.

 

This particular version has AC blocking at the input followed by a unity gain buffer to ensure that the PGA2320 (in this case) is driven from a source of <600 ohms impedance. Following the volume control chip is the output driver, a TPA6120, chosen, in this case, because of its output short-circuit protection. The main board fits into rails built into the enclosure.

 

6 CR123A protected lithium cells with a nominal capacity of 880mAH provide power when on the move, These are carried on a mezzanine PCB which also engages in the rails in the enclosure. Fully charged they provide a measured +/- 12.3V. The PCB has a dual- configurable layout, jumpers mean that it can be configured to run either from 2 unregulated DC wallwarts, or from a single AC wallwart. For this first build I have gone the single AC route. CRCRC filters precede the adjustable regulators which are configured to provide a constant-current / constant -voltage charge regime to the batteries or alternately power the amplifier directly. On-board ferrite beads following the socket are intended to keep PSU switching noise off the supply cable..

 

On/off switching is by pushbutton, one button for on, one button for off. The switching is performed by a relay, arranged to latch its own coil. The protected cells have an internal circuit which interrupts their output when their voltage falls below 3V. Consequently, if any cell interrupts its output momentarily the relay coil is no longer energised and the device switches off with no possibility of a restart unless the 'ON' button is pressed. This SMT relay was chosen for its very low coil power. The on and off buttons and the power socket are positioned at the rear of the unit.

 

A vertical daughterboard carries the uprocessor and display. The main board layout follows TI's recommendations for a split groundplane with analog circuits on one side of the split and digital circuits on the other. The PGA2320 straddles the split.

 

Three additional pushbuttons provide control inputs. The right-hand button causes the display to count up, or wakes the uproc from sleep. The left-hand button counts down. The centre button switches control to left, right or both channels controlled. Decimal points on the inverted displays indicate which channel is being controlled. When both channels are controlled the higher of the 2 volumes is displayed. The uprocessor clock shuts down after a few seconds to save power and to avoid introducing noise into the system. A single red LED provides an indication that the device is switched on.

 

The front panel is pierced in 5 places to accommodate the buttons and the input and output sockets, transparent acrylic has been used for the first build, but a smoked fascia will be cut in due course, this will permit the display and LED to be seen while hiding the rest of the interior.

 

When turned off and on the ampifier sets the volume to the same level as it was when it last went to sleep. The display wakes from sleep when the right button is pressed. The up/down count is measured at first but ticks over faster after the first five counts.

 

The assembly altogether weighs a substantial 400 grams, which detracts from its portability a little, it measures 112*90*48mm but the weight contributes to a reassuring impression of solidity and quality.

 

Components:-

 

real_neat_components.jpg

 

The main PCB:-

 

seed_built_top.jpg

 

seed_built_bot.jpg

 

How it goes together:-

 

TPA6120_2.jpg

 

Rear view:-

 

rear_view.jpg

 

Front:-

 

front_view.jpg

 

How does it sound?

 

It sounds exquisite. The amp is silent as far as I have been able to discover so far. Volume control is smooth and noiseless. Bass is powerful and distinct without any trace of pitch information being subsumed into boominess. Vocals are natural sounding without sibilance. Rough tones carry no additional edge. High pure tones are sweet and clear. There is an excess of volume available with all the phones I have available and about 100 steps of attenuation remain unused with the sources I have tried so far.

 

More trials and some measurements next.

 

w

post #2 of 21

Really, really cool project. I think your surface mount soldering might need a little practice though tongue.gif

 

What kind of battery life do you get? Looking forward to the measurements.


Edited by bcg27 - 12/18/11 at 7:17pm
post #3 of 21

Looks good.

 

I have a few questions.

 

What microcontroller are you using ?

 

What are your thoughts on the PGA volume controls ?

 

Have you experimented with rotary encoders and tried other displays like LCD/bargraph ?

 

Just curious, I want to experiment myself but need a microcontroller first, baby steps.

 

Coincidentally was reading through Error401's old TPA6120 thread just before... maybe it's a sign that I need to build one.

post #4 of 21
Thread Starter 

 

 

Quote:
Originally Posted by bcg27 View Post

Really, really cool project. I think your surface mount soldering might need a little practice though tongue.gif

 

What kind of battery life do you get? Looking forward to the measurements.


Thanks.

 

 

The board's seen some rework. I have a toaster oven which produces better looking results, but I need to make a solder stencil first, which I haven't got around to. I don't have time to waste on appearances, I'm only really concerned with functionality at this stage, and it works.
 
I'm anticipating 17 hours at best 13 at worst, but I have some variations to run before I get a final figure.

Quote:
Originally Posted by splaz View Post

Looks good.

 

I have a few questions.

 

What microcontroller are you using ?

 

What are your thoughts on the PGA volume controls ?

 

Have you experimented with rotary encoders and tried other displays like LCD/bargraph ?

 

Just curious, I want to experiment myself but need a microcontroller first, baby steps.

 

Coincidentally was reading through Error401's old TPA6120 thread just before... maybe it's a sign that I need to build one.

 


It's a PIC16F887. It runs off its own internal clock (no crystal or resonator). I originally drew in a 16F690 (20 pins), but the 44-pin TQFP is actually no bigger and I had used a 40-pin DIL version in a previous design, so I swapped to the '887 because I had a proven program and I could actually use the same pinout. The 16F690 layout had an awkward split of the functions between ports, and I didn't want any complications in the bit-bang control routines.
 
I like the PGA volume controls, I think the fidelity is good, in many ways I wouldn't thank you for a pot, moving parts are potential trouble. Now I've got push switches, not much to go wrong there. I've used a rotary encoder in the past, but they're more difficult to program for, they're much more expensive, bulkier and mechanically complicated. The PGA gives you left / right independent control.
 
I've used LCD and bargraph displays too. The bargraph is problematic when you have 256 levels to deal with. An LCD dot-matrix would be more versatile, but you have to work up a whole character set. The 3-digit 7-segment multiplexed uses 11 pins, I've got enough pins to go non-multiplexed (using the '887), but fitting in 24 tracks in the available space wasn't possible. The display I've used is a good minimalist solution.
 
I like the PIC micros, 5 pins + a cheap USB programmer, you can program the devices in-system, this means no problems with a ZIF socket programmer or worse, no SMT parts. You don't have to provide ROM or RAM, a clock unless you need precision, many have built-in A/Ds and will do touch-sense (no moving parts).
 
w
 

 

post #5 of 21

When you consider that most DAC's output a higher Vrms than a headphone can handle and a lot of them have a low enough input impedance to drive headphones directly.   Thats when you realize that we need headphone amps for volume contol/impedance matching in most cases.   So the volume control impementation is a majore factor in the headphone chain and I like that you are experiementing with a different technique than most headphone folks have tried.. 

 

I see you are using a PGA2310,  this has been around over a decade,  curious if you know of better chips out there now (lower SNR)?   

post #6 of 21
Thread Starter 

Yes, this chip would in fact work without problem driven straight off many DACs, the input is buffered only to ensure that it sees <600R regardless of source. I could have used the buffer amp (and the real estate it occupies) in a composite loop around the TPA6120, which would reduce the THD + N of that section, but the distortion overall is probably dominated by the volume chip with a paper performance of 0.0003% THD + N. This should still be below the threshold of audibility.

 

To my knowledge there are no better volume control chips available than the Texas PGA23xx series, they have better THD + N and a 4dB DR advantage over the Crystal (Cirrus) chip IIRC (CS3310).

 

There are at least 3 chips in the series the 2310, 2311 and 2320. The 2311 is for operation from +/- 5V supply. The 2310 is available in through-hole (DIP) and I used it in a (16*DIP8) parallel NE5532 amp which I have also written up here.

 

This amp in fact uses a 2320, there are slight differences in the spec. and of course this one is SMT, although the 2310 is also available in SMT. Easy availability on ebay drove the choice in this case.

 

If you should come across any better-performing or even close equivalent chips, please let me know. As you can see, this is an area of interest for me.

 

w

post #7 of 21
Thread Starter 

Some RMAA results for this amp, showing first the setup for testing the soundcard alone, then with the amp in circuit but with no additional loading, then with a 15 ohm load (in parallel with the soundcard input).

 

RMAA has been discredited to a degree, at least in comparison with industrial test sets, but in the absence of such expensive equipment it gives some indication of performance. The results are shown in the 4th pic.

 

setup_thru.jpg

 

setup_tpa6120.jpg

 

setup_tpa6120_load15.jpg

 

TPA6120_rmaa.jpg

 

I've also had the chance to evaluate the constant current/constant voltage charging for the protected lithium cells (CR123A). This works quite satisfactorily, limiting current early in the charge procedure. Even after 24hrs+ under charge the cells do not overheat and the charge voltage tops out @ ~+/-12.4V, the PSU being set to +/- 12.6V with the batteries out of the system.

 

I have yet to discover exactly how long the amp runs on a charge, certainly all day, and longer than I have listened in a single sitting.

 

Definitely a success.

 

w

post #8 of 21
Thread Starter 

I made up a black faceplate for the amp...

 

TPA6120_3.jpg

 

...you can see why I'm pleased with this build.

 

w

post #9 of 21
Thread Starter 

I redesigned and relaid the main board for this.

 

TPA_dc_servo.jpg

 

I built up a second one, one channel had 10mV of DC offset blink.gif

 

There's no reason other than an unfortunate combination of parts, but I re-did the circuit with a DC servo using an OPA2277, one of the better opamps for the purpose other than chopper-stabilized ones. I took the opportunity to build in switch-on -off pop elimination using a relay.

 

I'm hoping to show this off at the UK Head-Fi meet.

 

w

post #10 of 21
Thread Starter 

Got the boards for the new version yesterday...

 

TPA_dc_servo_brd.jpg

 

I'm still building the LME49600 amp, so this one will not be happening for a day or so.

 

w

post #11 of 21
Thread Starter 

Got somw nice shim brass yesterday, went to the library to make a laser print (all my printers are inkjets). The result, a new solder stencil for this board.

 

Here's the stencil and the first board after reflow and fitting TH parts:

 

TPA6120_stencil.jpg

 

This stencil is a lot thinner than the copper one I made for the LME49600 amp, about 5 thou. Although the other one worked, the results with this are better, there's less chance of one pad bleeding into another and small components pulling together. When you heat the board the solder paste sags and flows to a degree.

 

I'm missing some preset pots for this, and a SOT23 PNP transistor, and I also have to populate a daughterboard.

 

Daughterboard and batteryboard before separation. It's cheaper to buy them as a single board, I cut them up with a scroll saw. It's remarkable what an industrial level of finish you can produce at home now with a few power tools.

 

TPA6120_daughter.jpg

 

w

post #12 of 21
Thread Starter 

Got the daughterboard built and mounted...

 

TPA6120_dboard.jpg

 

...funny thing, the amp didn't work. Not a peep, and it just got hot frown.gif.

 

After days of work I found the errors in the PCB. Mine of course. Cost a few destroyed opamps. An upside-down opamp in the schematic didn't help.

 

The good news is that with a bit of white-wiring it's working now. The volume control now works without any clicking between steps, the output offset is 20uV one channel 120uV the other and switch-on / switch-off are completely noiseless.

 

Better order some modified boards I guess . After I extract some more test results from this onebiggrin.gif .

 

w

post #13 of 21

that is really something - troubleshooting something like that is the REAL testament to your skill, even moreso than the design :D 

post #14 of 21
Thread Starter 

It's nice of you to say so El_Doug, the truth is it was a complete PITA requiring the utmost determination and persistence on my part, but it was a brute-force solution consisting of disconnecting all the stages from the power and reconnecting them one-by-one starting from the input and following an injected signal through from the input using a scope.

 

You just have to believe that there ain't nothing you can't fix.

 

w

post #15 of 21

Wakibaki - is this a build you're planning to make available through a board run? I'm very interested in your PGA volume. I'm tired of hunting down pots that are out of balance, or sonically inadequate. 

New Posts  All Forums:Forum Nav:
  Return Home
Head-Fi.org › Forums › Misc.-Category Forums › DIY (Do-It-Yourself) Discussions › What I did next with Texas headamp chip, a digital control desktop / portable.