diyAudio

diyAudio (http://www.diyaudio.com/forums/)
-   Chip Amps (http://www.diyaudio.com/forums/chip-amps/)
-   -   LM3886 without electrolytic feedback capacitor and (hardly) no DC offset voltage (http://www.diyaudio.com/forums/chip-amps/217712-lm3886-without-electrolytic-feedback-capacitor-hardly-no-dc-offset-voltage.html)

Marc Vi 10th August 2012 10:28 PM

LM3886 without electrolytic feedback capacitor and (hardly) no DC offset voltage
 
1 Attachment(s)
Hereby the schematic of the LM3886 without an electrolytric feedback capacitor. Such electrolytic capacitor may introduce distortion and therefor it is better to drop it.

But how to prevent a (multiplied) DC offset voltage at the output?

By inserting an extra resistor to the negative input, you can balance out the offset voltage. This extra resistor is short-circuited for AC signals by a capacitor, but this capacitor can be a MKT/MKP type of lets say 1 microF.

The value of the resistor is determined experimentely and I found 27 Kohm for one (see the schematic) and 18 Kohm for the other amplifier. The offset voltage is then only 0,9 milliVolt respectively 0,2 miliVolt !! This offsetvoltage doesn't run away through varied circumstances, like power, temperature etc.

I am very suprised that this method is not applied more often. Please give your comments.

Marc.

ddd 11th August 2012 07:51 AM

Nice idea with some demerit. You simply made better symmetry of input bias currents paths. But the thermal drift of input voltage and current offsets will be still multiplied by 20. Electrolytic C makes this factor to be 1.
In your circuit I wouldn't put the shunt C at all.

Marc Vi 11th August 2012 08:09 AM

Thanks ddd,

I do agree that the thermal drift of input voltage and current offsets will be still multiplied by 20. But how much will that be in reality? All the components are closely to each other on one chip.

I made the chip clipping and oscillating with a high voltage and high frequency input signal, so it got warm. Directly after that I measured the output offset voltage. The difference was in the range of less than 1 milliVolt.

You need the input capacitor because of
- the chance that there is a DC voltage on the input signal;
- if the input has a low DC resistance, then the symmetry will be broken.

I recommend the capacitor parallel to the extra resistor to the negative input because it decreases the chance of high frequency oscillations. This is an experience I have with building more amplifiers in the past.

Marc.

ddd 11th August 2012 08:55 AM

The data of the both offsets are given. And one can estimate easily the contribution of the both (20kOhm in account).
1mV drift of the offset is extremely good with k=20.
Very useful info about oscillations. I supposed C could causes them.

Marc Vi 11th August 2012 05:07 PM

ddd,

1. The C shunt (in my schematic the C parallel to the resistor of 27 K) doesn't cause oscillations: on the contrary!

Of course I measured without the C shunt and then you see some overshoot on the edges of blocksignals of 50 kHz and more. With the C shunt the blocksignals have smooth edges but it doesn't change the slew rate. I measured almost 15 V/microsec.

If you put a high signal (more than 3 V p-p) which a high frequency (more then 200 kHz) on the input (without an input filter), the amplifier will clip. If you make the input voltage higher and higher, you force the amplifier to oscillate. The C shunt makes that the amplifier will oscillate a little bit later.

2. Indeed the data of the offsets are given in the datasheet within broad limits. In practice all examples of the LM3886 have different input offset voltages and currents. What you do is adjusting the value of the resistor (in my schematic the resistor with the value 27K) in such way untill the output offset voltage is 0,00 V. For one amplifier I found a value of 27 K and for the other a value of 18 K. The output offset voltage is then only 1 - 2 millivolt!! And this output offset voltage hardly drifts when the amplifier gets hot.

So with only this extra resistor and by - experimentely - adjusting its value, you can tune out the effects of the (differences in) input offset voltages and currents.

Marc.

ddd 11th August 2012 08:47 PM

Very good results, Marc. I cannot find thermal drift data in the datasheet but the input offset voltage and current vary 10 times. Hope your chips have the typical low value.
I think your method deserves high mark and wide application.
I'll try it soon.

Marc Vi 11th August 2012 09:30 PM

ddd,

I appreciate that you will try it soon.

Indeed in the datasheet the input offset voltages and currents vary 10 times but that is between different examples. Within one example, the offset voltage and current will vary much less.

My experience is that the typical value applies to the most examples following the Gausse curve. See note 10 in the datasheet. The limit value is the guaranteed value for a single example. See note 11.

By adjusting the value of the resistor, you can cancel the difference between different examples.

I hope to hear your practical results.

Marc.

ddd 12th August 2012 05:42 AM

Agree. Furthermore even worst chips will get better symmetry.
The method could be applied in case with electrolytic C as well, improving further the offset.
The shunt C could be possibly by far less, taking account of its anti-oscillation role only.

PS. In case with electrolytic C the R(+in -> earth) will have intentionally a little bit higher value to provide adjusting headroom. Or your idea can be accomplished just changing this R without additional one.
Avoiding el. C is the basic value of your method anyway.

zakman35 12th September 2012 07:25 PM

Does this extra resistor effects the gain of the amp?

Marc Vi 12th September 2012 07:54 PM

Thank you for your question. I do not understand why there is so little interest in a solution to avoid the electrolytic feedback capacitor and still have almost no DC offset voltage.

No, this extra resistor doesn't effect the gain of the amplifier for two reasons. The first one is that the input resistance of the amplifier is so high that the extra resistor is very very small compared to the input resistance. The second reason is that the extra resistor is short-circuited by the capacitor of 1 microF, although the main reason for this capacitor is that it prevents the amplifier motorboating and oscillating. But even without this capacitor my amplifier didn't oscillating or motorboating so the capacitor of 1 microFarad is pure precaution.


All times are GMT. The time now is 08:46 AM.


vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2014 DragonByte Technologies Ltd.
Copyright 1999-2014 diyAudio


Content Relevant URLs by vBSEO 3.3.2