High DC Offset on Gainclone

[GM Creations]

Hello all.

After spending several months reading about Gainclones on various websites, I finally took the plunge and started to build my first amp.

I built the PSU first, and it seems to be working fine, giving a steady 35V.

I’ve just completed the amp for one channel, and decided to test it before starting the second one.

As advised, I connected a 10 ohm resister across the output and measured 400mV. This was with no input signal. I then connected an old mobile phone via the headphone output, and the DC rose to 800mV.

I’m no electronics expert, but I can read a circuit diagram and am pretty good at soldering.

I have attached the circuit diagrams of the PSU and amp. Any advice as to what the problem could be would be greatly appreciated.

 

[scottjoplin]

Hi, you've omitted the DC blocking capacitor to ground in the feedback network

 

[GM Creations]

Thanks for your quick response.

As I said, I am a complete novice at this game. I was under the impression that the DC blocking capacitor was the 4.7 uf cap (Cin) Could you advise me on the position and type of capacitor that I need.

 

[scottjoplin]

Figure 1 here http://www.ti.com/lit/ds/symlink/lm3886.pdf

 

[Mark Whitney]

Even with CI missing an offset of 400mV is too high. There could be more wrong. Perhaps you could post some photos.

 

[Mooly]

I think you will find 400mv is bang on the money given the component values used. It could have been a lot worse if the chip was at the spec limits.

Yes, you need to AC couple your 680 ohm with something like a 220uF. If you make this cap to small then it will become the dominant component in setting the LF response instead of the 4.7uF input cap.

 

[Mark Whitney]

I have never measured more than 45mV offset with these or similar component values without CI installed.

The datasheet claims a maximum offset of 10mV at the inputs, this multiplied by the gain would still be way less than 400mV.

 

[GOR3]

Was the input shorted when you measured the offset? It should have been. Also, with ±35 rails, I wouldn’t recommend driving a load of less than 8Ω, including a speaker “rated” at 8Ω, that dips below that.

 

[Mooly]

The input bias currents can be up to 1uA for the LM3886. That would develop 22mv at the non inverting input.

It would always be advisable to check for stability issues to. If a scope isn't available then you can always rig a simple filter at the output such as a 10k and 1uF and measure the voltage across the cap. In other words another similar network in parallel to the Zobel one already present, but with different values.

 

[GM Creations]

Thanks everyone for your advice.

It turned-out to be a stupid rookie error. The wire shorting the input had come loose. I replaced it and measured 75mA.

 

[scottjoplin]

You mean mV I presume, that's still high(ish), connect the DC blocking capacitor

 

[GM Creations]

Yes mV, of course! I will connect the blocking capacitor.

 

[Mooly]

Fit the capacitor polarity according to the voltage you measure across the cap. Although the voltage is small, you should always orientate the cap correctly, even if it means one channel has the cap fitted differently to the other.