Why is the TA7136 so bad? It's the datasheet.

If you're here, you're probably searching for info on the Toshiba TA7136 (aka the TA7136P or TA7136AP) op amp. You may be having trouble with a device that uses it. Let me guess: is it picking up FM radio? Does it sound a bit fuzzy? Is the noise floor strangely elevated?

You may be having stability issues. This was the case with the Onkyo A-5 I revamped.

Using the spice model of the TA7136 that I developed for the Onkyo, here's the simulated loopgain plot of the tone control reference circuit suggested by the TA7136 datasheet. It's really bad: simulated phase margin and gain margin are zero, as the loopgain phase crosses through -180 degrees right at the ULGF around 1.6MHz.

It's possible to design a circuit around the TA7136 with generous stability margins, and retain plenty of in-band feedback; Toshiba just didn't bother. It's yet another example of bad datasheet reference circuits informing bad mass-produced applications.
 

Attachments

  • ta7136.png
    ta7136.png
    28.7 KB · Views: 29
  • datasheet_ckt.png
    datasheet_ckt.png
    20.9 KB · Views: 29
For reference, here's the TA7136 model I used.

The resistor values are guesses, since the datasheet omits them. The critical "Rguess6" value of 270 ohms was verified experimentally (by drawing current from P6 and observing the change in voltage at P1.) That resistor's value affects the circuit's stability more than any other.
 
Revising the TA7136 model

Why is the reference circuit's stability so bad? Is the simulation in post #1 really right?

Maybe not: with a 100k resistor driving pin 5, the current sources at Q11 and Q3 only conduct tens of microamps. That can't be right, that's not enough.

Let's fix this. On the original datasheet, the two diodes are named Q1 and Q2. Not D1 and D2. That's a hint: they may be diode-connected transistors, with Vbe matched to Q3 and Q11. That would make sense.

Indeed, the model works better if we substitute diode-connected BJTs and revise the values of Rguess3 and Rguess5 a bit. Now, the current sources act much more like normal current mirrors -- Q3 and Q11 will supply current that's reliably within a factor of two of that supplied to pin 5, over a wide range. I bet that's what's on the chip.

Guess what: the reference circuit's stability is even worse with this change. It is a basket case.

Here's the revised TA7136 model. To prove that this opamp isn't 100% awful, here's a revised tone control "reference circuit" with generous stability margins and generous in-band loopgain. This is what the datasheet should have said.
 

Attachments

  • ta7136.png
    ta7136.png
    74.1 KB · Views: 15
  • revised_tone_reference.png
    revised_tone_reference.png
    74.2 KB · Views: 17
  • revised_tone_reference_stability.png
    revised_tone_reference_stability.png
    540.7 KB · Views: 16