5
\$\begingroup\$

Why would an amplifier output be clipped asymmetrically?

I am building a small headphone amplifier using the Texas Instruments TPA6100A2. I have built up the circuit just as in the TPA61002 Evaluation Board User's Guide:

TPA6100A2 EVM

The opamp is designed for single supply, single ended use. The only difference is that I added 100uF electrolytic capacitors at ROUT and LOUT with 20 Kohm discharging resistors on the headphone side.

I am powering the 6100 from a 3.3v regulator and it works very well and looks like will meet my needs. I will add that the datasheet references driving 16 or 32 ohm headsets. However in my case I will be using it to drive an aviation headset with impedance around 300 ohms. The tests with said headset are very promising.

Below there are two screen captures of the amplifier output (yellow trace) being fed a 500Hz sine wave from a cell phone (blue trace).

I will note that at the level of output on the top picture with my headset the volume is too loud. Withing the listening level there is no clipping (at least for my use), meaning I don't expect to spend any time at all in this region of the performance domain.

Just to find out where the limits are I pushed it into full overdrive (bottom picture).

So it is not rail-to-rail (I knew that) but gets fairly close to the full 3.3v in the supply (not shown but the oscilloscope was displaying 3 volts as the amplitude). It is still clear that the wave is clipped in an asymmetrical manner.

Oscilloscope screen captures

Why would this be? I've double checked the wiring and setup, all that seems correct as per the schematic referenced above. Is it possible I have a bad opamp? Unfortunately I do not have another on hand to compare against.

Adding information: The datasheet provides this insight:

"For maximum signal swing and output power at low supply voltages like 1.6 V to 3.3 V, BYPASS is biased to VDD/4. However, to allow the output to be biased at VDD/2, a resistor, R, equal to RF must be placed from the negative input to ground."

Also, this setup is on a breadboard with jumper wires and 5% through-hole resistors and capacitors, not like the EVM with it's fancy SMD 1% components.

\$\endgroup\$
  • \$\begingroup\$ Yellow trace means probe tip at Rout and gnd clip at GND, right? Just curious: are all ground the same? I mean, power supply ground/negative, cell phone ground (was it connected to a charger or was it on battery?) and scope ground. \$\endgroup\$ – Sredni Vashtar Jun 4 '16 at 21:51
  • \$\begingroup\$ Can the amp swing equally far from midpoint in either direction at that voltage with that load? Not all amps can produce a rail to rail output under all rated conditions, some can't get all the way to the positive rail while some can't get all the way to the negative rail (although that's less common). What does the datasheet say for your operating conditions (or something close to it)? It may be a limitation on the input signal range, there are some amps with rail to rail outputs that can't handle rail to rail inputs. \$\endgroup\$ – Sam Jun 4 '16 at 22:35
  • \$\begingroup\$ @SredniVashtar: You are correct, I should have added those details. The cellphone was on battery. All grounds (regulator, headset, opamp scope probes, cellphone audio common) tied together. \$\endgroup\$ – codeZonkey Jun 4 '16 at 22:38
  • \$\begingroup\$ @Tom: I am not sure how to check that, the datasheet is surprisingly scant on performance details. \$\endgroup\$ – codeZonkey Jun 4 '16 at 22:40
  • \$\begingroup\$ The data sheet says things "the TPA6100A2D biases BYPASS to VDD/4" and that "to allow the output to be biased at VDD/2" you need a resistor = to Rf from the negative input to ground, but it looks like you've already got that. Otherwise, like Adam suggested, it sounds like a biasing issue (especially if you only get 1V bias rather than the 1.6V midpoint you should be getting at 3.3V) \$\endgroup\$ – Sam Jun 4 '16 at 23:09
2
\$\begingroup\$

First, thank you for the well-written question.

Based on your scope captures, I suspect the mid-rail biasing is not working correctly. You should be able to check the output bias voltage by shorting Vo1 and IN- together, then measuring the voltage at Vo1. (Turn your input sine wave off first, obviously.) You should see about 1.6V. If it's more like 0.8V, check R3 and R6. You should also verify that your electrolytic capacitors are connected the right way.

\$\endgroup\$
  • \$\begingroup\$ I shorted Vo1 and IN1- with a jumper but it made the TPA very angry. It began oscillating at a VERY high frequency. All I did was drop in the jumper, should I have disconnected anything else? In current configuration without any input the bias at Vo1 is ~400mV without R3 and ~1V with R3. I placed some additional information from the datasheet at the bottom of the question above in case this adds valuable insight. \$\endgroup\$ – codeZonkey Jun 4 '16 at 23:04
  • \$\begingroup\$ Hmm… maybe it's not unity-gain stable? Sounds like you were able to get it working. Odd that increasing the bypass capacitance would help; according to the datasheet they generate the mid-rail bias with a resistive divider. The cap should reduce noise, not change the voltage. \$\endgroup\$ – Adam Haun Jun 5 '16 at 5:39
  • \$\begingroup\$ I didn't think of messing with the cap initially because I thought the same thing. \$\endgroup\$ – codeZonkey Jun 5 '16 at 13:51
0
\$\begingroup\$

So I think Adam Haun was on the right track and the bias output voltage not being at VDD/2 was causing the bottom to shift up and clip the wave. I tested with other capacitor values for C3 and increasing the capacitance seemed to shift the bias towards the center. In the end a 10uF electrolytic at C3 produced a bias of 1.54 Volts. As a result the symmetry of the wave was restored to almost equal and the output power before overdrive was increased as the negative portion of the wave did not get clipped too early:

Better symmetry

At this point I think I could keep fine tuning C3 to shift the bias and increase the symmetry but it is close enough for my needs. Thanks everyone for their time and feedback!

\$\endgroup\$
  • \$\begingroup\$ If you also want to address the clipping at the maximum drive level try changing out R7 and R8 for smaller value resistors. \$\endgroup\$ – Captainj2001 Jun 5 '16 at 0:34

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.