Revising my LM3875 project

[cumesoftware]

Hello!

I'm in the process of making a revised version of an amplifier that I already made. The amplifier is very good as it is, but I want it to be better, more stable and completely foolproof (especially against pre-amps that decide to output 8VDC instead of the signal).

This was based purely on the NS datasheet, but I'm planning to add a 0,7uH coil paralleled with a 10Ohm 1W resistor on the output. My real concern is with the input. I found out that Ci as suggested by the datasheet is not effective at blocking DC. That requires a capacitor at the input. Someone suggested that Ci could be eliminated. I really prefer that option since NP electrolytics are hard to get.

My questions are:
To avoid more offset voltage at the output (because I suppressed C13 and C14), should I use 33K resistors at the input (after the DC blocking capacitor) so the offset currents are equal?

Those resistors are to be place from the input to ground before the 1K resistors or after the 1K resistors?

The resistors should have 33K because R4 and R8 are 33K, right? Someone can explain the theory behind that?

If I use 33K resistors, is it correct to use 680nF polyester caps at the input for the DC decoupling? I want the -3dB low frequency cut-off to be a bit less than 10Hz.

Attached are the schematic and PCB artwork.

[cumesoftware]

Please, I really need help on this!

Components:
C1/2/7/8 – 10mF electrolytic capacitor (50V);
C3/5/9/11 – 100µF electrolytic capacitor (50V);
C4/6/10/12/15/18 – 100nF polyester capacitor (63V);
C13/16 – 220pF 63V ceramic capacitor;
C14/17 - 22µF non-polarized electrolytic capacitor (6,3V) [should be eliminated with the addition of the input DC blocking capacitor? - I rather prefer too];
D1-4 – 6A05 rectifier diode;
F1 – 2A slow-blow fuse;
F2/3 - 6,3A slow-blow fuse;
HS1/2 – 1,2°C/W passive heatsink;
IC1/2 – LM3875 audio power amplifier (LM3875TF);
J1 – Earthed AC power plug;
J2/3 – RCA socket;
JW – 26AWG stranded wire;
L1/2 - 16 turns 6mm dia. 23AWG coil;
LP – 230V~ resistored neon bulb;
R1 – 10KR double gang potenciometer (10% tol., 1/8W);
R2/8 – 4,7KR carbon resistor (5% tol., 1/8W);
R3/5/9/11 – 1KR carbon resistor (5% tol., 1/8W);
R4/10 – 33KR carbon resistor (5% tol., 1/8W);
R6/12 - 2,7R carbon resistor (5% tol., 1/2W) [should have higher wattage?];
R7/13 - R6/12 - 2,7R carbon resistor (5% tol., 1/2W) [should have higher wattage?];
S – DPST switch;
T – Toroidal transformer 230V~ pri. 2x18V~ sec. 225VA.

[AndrewT]

build an AC coupled amplifier !!!


The alternative of a DC coupled amplifier requires quite a number of other circuits to detect a fault condition and action preventative measures.

[cumesoftware]

Hi AndrewT,

What do you mean by AC coupled? Do you mean keeping Ci? Or using an input DC blocking capacitor? I did some calculations on the DC blocking capacitor and came to the conclusion that adding a 22K (Rin) resistor to ground and a 1uF or greater (Cin) polyester capacitor (while keeping the 22uF Ci ) would require a 2.2K linear pot and a 1K resistor in order to change the law. The problem with this is that the impedance starts to be a bit low (around 680Ohm to 2.2K). Is this a problem? Should I add a buffer before the pot? Or can I go around by choosing a different value for Rin? How high it can be?

I really need your advise.

P.S.: I decided to keep Ci as the main LF cut-off and add a Cin for DC blocking.

[AndrewT]

AC coupled means only AC signals are processed by the amplifier.
That requires DC to not be passed through.
An input blocking cap and DC gain reduced to <=1 and an output blocking cap and any interstage blocking caps must be considered. To ensure DC is not amplified.

[AndrewT]

Originally Posted by cumesoftware

Should I add a buffer before the pot?

No !
If the output impedance of the volume attenuator is a cause of a problem then that problem is that the attenuator cannot properly drive what comes after it.

If a Buffer is required it must come after the attenuator, not before it.
The Buffer now converts the output to make it suitable to drive whatever you decide it must be capable of driving. eg. 1m of cable and 22k/1nF, or 50m of cable and 100k//2n2F or whatever you think your attenuator needs to drive.

[mrfeedback]

You need a capacitor in series with the wiper terminal of the volume pot.
This keeps any external DC out of your amp, and ensures that the input pin of the IC does not see a variable impedance ground.
Keeping DC out of the wiper terminal of the volume pot also ensures that the volume pot does not go 'crackly' prematurely - I believe electrolysis is a cause of noisy volume controls.

Eric.

[cumesoftware]

The need for both the input blocking cap Cin and for the Ci cap I understand, because even with Cin any offset created by the amp is not mitigated without Ci. But output caps on a split supply design? Those need to be huge and non-polarized! If I'm thinking right, I might as well as forget any caps in series with the load. I hope you are not suggesting to use caps there too. Are you?

Oh, a lot more needs to be answered. Since I probably need a 22K resistor (to give an error of 10% or less considering that the input bias can go up to 1uA) to set my Rin, and a Cin of 1uF or greater, and thus I need to have 2.2K pot with a 1K resistor between the cursor and ground (again to guarantee that the error is less than 10% because Rin and Cin are there), is it a good idea to use a buffer to increase the impedance? My idea is to let Ci set the roll-off and Cin is just there to block any DC (thus I don't need to use a huge capacitor for Ci since I'm not using Cin as my LF cut-off). Can I increase my Rin (therefore reducing Cin and increasing the pot and resistor)?

Originally Posted by mrfeedback

You need a capacitor in series with the wiper terminal of the volume pot.
This keeps any external DC out of your amp, and ensures that the input pin of the IC does not see a variable impedance ground.
Keeping DC out of the wiper terminal of the volume pot also ensures that the volume pot does not go 'crackly' prematurely - I believe electrolysis is a cause of noisy volume controls.

Eric.

Good advise! I think Cin will solve that too.

[cumesoftware]

Originally Posted by AndrewT

No !
If the output impedance of the volume attenuator is a cause of a problem then that problem is that the attenuator cannot properly drive what comes after it.

If a Buffer is required it must come after the attenuator, not before it.
The Buffer now converts the output to make it suitable to drive whatever you decide it must be capable of driving. eg. 1m of cable and 22k/1nF, or 50m of cable and 100k//2n2F or whatever you think your attenuator needs to drive.

I'm not sure if I was understood. The problem is that, by my calculations and to do things properly, I end up requiring a 2.2K pot and a 1K resistor to change the law of the pot. The problem is that some pre-amps can't cope with that load. This is a standalone amplifier. I really don't get the need of the buffer after the pot and I really need some explanation. Unless the buffer allows the use of a bigger value attenuator right at the input? Is that what you're thinking? Seems to be a good idea and also copes with noise (I guess that using the buffer before would generate some noise because of the high impedance of the same, right?).

I'm eager to learn and I really want to know the basis of your thinking.

[cumesoftware]

So far, I have two alternatives. This is the orthodox one:
schematic_2a.pdf

This is the less orthodox one, but it allows a greater value pot, and greater input impedance. Unfortunately the impedance that Cin sees will vary (is that a problem?):
schematic_2b.pdf

Which is preferable?