I intend to use a TDA1512 in an application where the closed loop will vary a lot, falling down to unity in some cases.
I know the usual tricks: increasing the compensation (it is external), increasing the noise gain, but although they work for small signals, the input seems unable to cope with even moderately large signals. As soon as the input signal exceeds a volt or two, all hell breaks loose.
Any idea what's causing this behaviour (the data available on the internal circuit seems scant), and the way to remedy it?
I know the usual tricks: increasing the compensation (it is external), increasing the noise gain, but although they work for small signals, the input seems unable to cope with even moderately large signals. As soon as the input signal exceeds a volt or two, all hell breaks loose.
Any idea what's causing this behaviour (the data available on the internal circuit seems scant), and the way to remedy it?
I've seen that with the LM3886 as well. I'm thinking it has to do with the slewing behaviour of the internal stages of the amp. Another possibility is that the open-loop gain of the amp changes as function of the output voltage. The LM3886's AVOL drops as the VAS approaches saturation, for example. So it's perfectly possible to have a circuit that's stable around 0 V output but unstable with, say, 25-30 V output.
You'll probably also find that you can get the circuit to work with sine waves but it'll lose its mind if you feed it a square wave.
The best remedy I've found is to be conservative in your compensation. So aim for 80-90 º phase margin in simulation instead of the usual 45-60 º.
Tom
You'll probably also find that you can get the circuit to work with sine waves but it'll lose its mind if you feed it a square wave.
The best remedy I've found is to be conservative in your compensation. So aim for 80-90 º phase margin in simulation instead of the usual 45-60 º.
Tom
It is not a compensation issue, I can overcompensate to death, it does not help. When the input becomes positive by more than 1~2V, a large non-linearity begins to appear, and if the voltage is further increased, the amplifier goes completely berserk and the power has to be removed to allow it to recover.
There is something in the internal structure of the input causing this behaviour, but I have no idea what, or how to circumvent it
There is something in the internal structure of the input causing this behaviour, but I have no idea what, or how to circumvent it
There is an internal schematic in the datasheet, but unfortunately it is only a simplified internal schematic. https://www.alldatasheet.com/datasheet-pdf/pdf/101832/ETC/TDA1512.html
As it only happens when the input voltage is too high, you could try using a lower input biasing voltage than the TDA1512 was meant for. I don't know if you can pull down pin 8 with an external resistor or need to make your own external voltage divider; if pin 8 should bias some sort of internal clamp circuit, pulling it down will lower the threshold of the clamp as much as the input bias voltage.
(Another option is first attenuate and then amplify, but that's so obvious I guess there is a good reason for not doing that.)
Last edited:
It looks to me like the network at pin 3 either compensates the local loop in the pulling side of the output stage or the according current limiter, but not the overall loop.
I've seen large-signal oscillations in amplifiers that were small-signal stable several times, always due to something grossly non-linear causing unintended couplings and/or bringing the circuit into operating conditions it wasn't meant for. For example, a saturating transistor in a current source in the output stage pulling down a base rail, reducing bias currents in the input stage of the amplifier. If parasitic diodes to substrate or to an N-well go into conduction, you can also get the weirdest couplings.
Last edited:
I have found the solution: the biasing network at pin 8 is not actually like it is drawn. Other internal circuits are tied to it, probably related to the input stage.
I connected the pin to Vcc to disable it, and used an external divider instead. I used a 330 ohm resistor instead of a dead short, because I don't know what's behind the network and I didn't want to take the risk of damaging something inside, but practically it takes the pin to ~Vcc, and it completely solved the problem.
I kept the original compensation scheme, but I added a noise-gain network between the inputs, 330R+2n2, and everything finally worked as it should
I connected the pin to Vcc to disable it, and used an external divider instead. I used a 330 ohm resistor instead of a dead short, because I don't know what's behind the network and I didn't want to take the risk of damaging something inside, but practically it takes the pin to ~Vcc, and it completely solved the problem.
I kept the original compensation scheme, but I added a noise-gain network between the inputs, 330R+2n2, and everything finally worked as it should
