DC servo high offset on startup?

[theAnonymous1]

I have two identical headphone amp boards based on the LME49600 reference design. One board uses LME49720 for feedback/servo and the other uses LME49990 for feedback and OP27 for servo. Other than that they are identical boards and components.

The board with LME49990/OP27 has a high DC offset of around 1V when it's powered up, but after a few seconds it settles to ~1mV. The board with the LME49720 doesn't do this.

Any clues as to what might be causing this?

 

[abraxalito]

Might be something to do with the servo capacitor getting some charge on it during power-up in the OP27 case. To investigate further check the output voltage of the servo amps themselves (at pin6 for the OP27, pin 1 or 7 in the LME case) when settled.

 

[theAnonymous1]

The output voltage at the servo for OP27 is 1mV on one channel and 2mV on the other; 49720 is 10mV and 13mV (you can see why 49720 isn't the best choice here).

Probably should have went with OPA132/4 for the servo, but OP27 has been used in many servo designs so I thought I would give it a try. Unfortunately the OP27 is soldered on the bottom side of a SOIC to DIP adapter and the adapter is soldered on the PCB. :/

I could just put a delay circuit with a relay on the output as it only takes about 6 seconds to settle, but I would really like to know why it's doing it in the first place.

 

[abraxalito]

If the OP27 is only putting out 1 or 2mV its rather unnecessary to have a servo.

 

[theAnonymous1]

Well, it's a DC coupled headphone amp. I think the servo is there for any DC that might show up on the inputs.

 

[abraxalito]

So is the OP27 output voltage at 1 or 2mV immediately after power on? Tracking down how the servo cap gets charged seems to me to be the way forward here, or you could consider changing the servo's filtering characteristics. The former looks to be a better solution than the latter as you say the problem's not there with the LME opamp.

 

[theAnonymous1]

So is the OP27 output voltage at 1 or 2mV immediately after power on?

No, it acts the same way the output of the amp does; spikes at close to 1V and then settles to ~1mV 6 seconds later.

Tracking down how the servo cap gets charged seems to me to be the way forward here, or you could consider changing the servo's filtering characteristics. The former looks to be a better solution than the latter as you say the problem's not there with the LME opamp.

Unfortunately I don't have parts on hand to play with the filter values.

 

[abraxalito]

Back to back diodes between 0V and the inverting input of the servo opamp might help.

 

[theAnonymous1]

Pardon my ignorance, but wouldn't that just limit the max voltage on the negative input to ~700mV?

 

[abraxalito]

My apologies - I was thinking the servo was a simple integrator where the -ve input is a virtual earth, but I see from AN-1768 it not a pure integrator, there are two caps. OP27 already has diodes between the inputs so that won't work as a solution.

I went looking for an internal schematic for the LME opamp (to compare with the OP27), didn't find it so I'm out of options for you here. Something related to how your power supplies come up is creating the initial charge on the servo cap(s) but I have no idea what.

Rather than go to the trouble of a relay, you can put JFETs (which will be low impedance on switch-on) directly across the servo caps then drive their gates strongly negative when the power supplies have settled.

 

[theAnonymous1]

The power supply is a BWR-15/330-D12A DC-DC converter (+/-15V outputs) with additional output filtering. I've used these for many projects and this would be the first time one ever created a problem if that's indeed the cause.

http://www.mouser.com/ds/2/281/bwr7-10wa-series-42689.pdf

 

[abraxalito]

Six seconds for a servo to settle isn't what I'd call a problem. Tomchr had a much longer servo settling time on his first generation Mod-86. Even on the second revision where the filter characteristics were specifically modified for improved settling time, he's only claiming settling within 10seconds.

 

[theAnonymous1]

Yeah, it's not that big of a deal. I was just more curious to learn why the two op-amps acted so differently. I think I might unsolder the DIP adapters and throw in an OPA132 to see how it acts.

I appreciate your help.

 

[abraxalito]

I'm curious too - to get any further it would be worth seeing if the voltage ramp up on both rails coming from your DC/DC converter is symmetrical, but it'll be a fairly tricky characterization to make, needing a digital scope showing both rails over a time period of a few mS.

I agree, much simpler just to swap out the OP27.