increasing DC offset

[ir]

I recently threw together an amp based on the 3875 national reference design. only key difference is an input cap (2.2uF NP) and changed feedback values (Rf =68k, Ri = 3.3k) so that it's sensitive to 1V for use as a guitar amp power stage.

when the amp is first powered on, i get an instant offset of 40mV - which is the highest offest i've ever seen on any of my amps (typically 2-3mV). it instantly starts decreasing at a rate of about 2mV per second and then heads into the negatives, i'm guessing it will eventually hit the rail voltage if i let it but it gets well up over a few volts

input is shorted to ground during testing, load connected (speakers, also tested with 10Ohm, 10W resistor). everything as per national reference design

one thing about the 3875 vs the 3886/4700 etc is that the 3875 has no ground pin - not sure if this makes a huge difference, but i for one feel more comfortable knowing the insides of that little IC have a proper ground reference. and all the other amps (many) i've ever made based on other chips have been fine. Also tried swapping out the IC for another 3875, no luck

Supply rails read +23.4, -23.4 and are dead stable (+/- 0.1V)

i'm all outta ideas. i'd rather sort this than rebuilt with a 3886. i've double checked it again and again. probably cocked something up no doubt

any ideas?

Cheers

[peranders]

You are spoiled 2-3 mV is extremely good and 40 mV is nothing to worry about. Worst case for the LM3875/86 is 10 mV times the gain. People here has got up to 300 mV or so when they have had DC gain 30.

[ir]

yeah i know 40mV would be fine, but the trouble is it starts at +40mV then goes down and down , to 0mV into the negs

highest (lowest?) i let it run was -7volts

obviously a major glitch somewhere there

[weissi]

Use a 68k//3k3 resistor in series before the +in terminal.... Maybe the feedback electrolytic is defective?? Try another one ( use at 220uF bipolar)

Markus

[AndrewT]

Hi,
inverting or non-inverting?
This sounds like a leaking/faulty cap.
Can you measure the voltage across the NFB and Input caps to see if they move with the changing offset.

Output offset should be measured and set with the output open circuit.
7volts across an output load is a very high dissipation in the chipamp and also liable to burn out a speaker.

[Juergen Knoop]

you didn't forgot the bias-resistor at the + input?

[ttan98]

A faulty chipamp possibly or most likely?

[ir]

yeah, a leaky cap sounds quite likely me thinks

will give your suggestions a try and see what happens

Cheers