Help with GainClone gain setting

[johan851]

Hey all. I'm building a power amp for the first time, and I think I've settled on the LM3886 stereo kit from chipamp.com. I'm planning on not using a volume control because I already have a preamp. I'm using the M3 headphone amp as a preamp <http://www.amb.org/audio/mmm/>. I set up the M3 to have a gain of 9. The M3 has 26V rails, which should produce a peak to peak max output of around 13V. I know that's a little high, but the M3 is sort of a hi-fi class A headphone amp.

Does anyone know what a good ballpark gain figure for my power amp should be? I think it should be around 2, but that seems oddly low.

To give you a better picture, I'm using 8ohm bookshelves that I built. They're Modula MT's, which are supposed to be very efficient. Thanks!

And as an aside, I think what I want to use is a transformer with 18V secondaries, because the advice I've seen floating around seems to be that 25V secondaries are too much. Thanks!

 

[ionomolo]

This chipamp is not stable under a gain of ten and it sounds really ugly near that setting, 20 is ok and 30 is exellent.

You can as well add a voltage divider at the input to get the gain you want while keeping the LM3886 at it's optimal operating point. This will also help in keeping preamp noise away.

Remember that:

· low value resistors (243R/6K81/470uF) are better for offset and noise performance.
· feedback loop must be short
· You should place small capacitors at each supply pin as near of the ic as you can, otherwise you will hear it motorboating.
· Voltage overrated caps sound beter in the signal path.
· Input signal ground, feedback ground and input decoupling ground must be connected to the same point and this point should be connected to the board main power ground where decoupling caps, ic ground, speaker return and power supply go by a single wire.
· 18V is OK and is said to sound better that 25V, but you can also run it at 25 (35.3 Vdc) and get more power if the opamp is properly heatsinked.

Enjoy!

 

[johan851]

Alright, thanks for the input. So looks like a gain of 30 is the best choice. I get the feeling, though, that the amp is going to be extremely loud when used with my preamp. I guess that's something that I'll just have to deal with?

Hopefully part selection won't be too much of a problem since I'm using a kit. I did have some questions about grounding. Since AC earth is supposed to be connected to the chassis, doesn't that mean that I'm going to want to use RCA input jacks and speaker wire posts that are isolated from the chassis?

 

[tangmonster]

ionomolo ,have given you great input.

My addition:

The standard National application circuit is good for full output with 1V input.

I do believe it is better to not ground your input rca's and output speaker wires to chassis it is very easy to create ground loops when they are grounded to chassis and you don't want that.

One good grounding point for everything is better than multiple ones.

That is what i found and a few other threads that where about humm mostly was due to input rca's grounded to the case

Tang

 

[AndrewT]

Hi,
your 13Vpp is the same as 4.6Vac [=13/2/sqrt(2)].
As a maximum signal level from the pre-amp this is approximately 6db overhead on a maximum 2.2Vac signal coming from a CD player. This is normal or slightly low.

The attenuator in the pre-amp will allow you to reduce the volume to normal listening levels. A gain of 9 (=+19db) is high. Do you have some very low level inputs that you need to amplify? 2times to 4times is more normal for pre-amps with gain. Some are unity gain.

 

[Nordic]

And for chipamps, the load limits the supply voltage which in turn limits the maximum gain available...

 

[ionomolo]

Originally posted by tangmonster:
I do believe it is better to not ground your input rca's and output speaker wires to chassis it is very easy to create ground loops when they are grounded to chassis and you don't want that.

Absolutely right

Originally posted by tangmonster:
One good grounding point for everything is better than multiple ones.

My experience tells that when power ground and signal ground are separated there are less chances to get hum, but this might as well happen for some other reasons.

 

[johan851]

Hi,
your 13Vpp is the same as 4.6Vac [=13/2/sqrt(2)].
As a maximum signal level from the pre-amp this is approximately 6db overhead on a maximum 2.2Vac signal coming from a CD player. This is normal or slightly low.

The attenuator in the pre-amp will allow you to reduce the volume to normal listening levels. A gain of 9 (=+19db) is high. Do you have some very low level inputs that you need to amplify? 2times to 4times is more normal for pre-amps with gain. Some are unity gain.

I'm amplifying the output from a DAC, specifically a Zhaolu 1.3 with some mods. So I guess a gain of 9 is pretty high. That's probably due to the fact that my preamp is essentially a headphone power amp. If I had a scope I would be able to take some measurements from my source, but unfortunately I don't have one.

I suppose I could just build the chip amp with a normal amount of gain and see if it works or not. Worst case is that I have to change two resistors in my preamp to make it work better, which isn't too terrible of a thing.

 

[ionomolo]

You coud build yourself a precision rectifier on breadboard using one standard ultra-cheap opamp and place a capacitor with a resistor ( 1M ) in parallel at its output. Connect this to the output of the rectifier using a 50R resistor or otherwise it may (will) oscilate. The voltage across the cap will be the peak voltage of the input signal:

http://sound.westhost.com/appnotes/an001.htm

Save it as this will be an useful thing when it comes to measure peak voltages if you don't have a scope. If your voltimeter has a low impedance then you should buffer the cap. A TL072 will do the job very well and is cheap.

Then set the overal gain as to get the clipping voltage of the LM3886.

 

[johan851]

Thanks for the advice. I'm going to go ahead and modify the preamp, though. I plan to install a variable gain control so that I can adjust my preamp gain anywhere from 2-15. That should give me plenty of options.

Does anyone know where I could find an ALPS blue velvet 10k pot? I know AMB.org has the 50k version, but a 10k would be perfect for this application.

 

[danielwritesbac]

I'm a little bit confused. Why doesn't a "fixed l-pad" structure work at the amp input if you want to prevent excess input?

 

[johan851]

I'm a little bit confused. Why doesn't a "fixed l-pad" structure work at the amp input if you want to prevent excess input?

I'm sorry, I'm not quite familiar with what a "fixed l-pad" is... Is there something I'm missing here? I'm going to end up putting some kind of attenuator in the signal path anyway, I figure I might as well stick it in the preamp and get a lot of gain control flexibility, which will be nice to have with my headphones as well.

 

[johan851]

Sorry for the double post...it's not letting me edit. But I just learned that a fixed l-pad is a big fancy name for "voltage divider." And you're right, I don't see any reason why that wouldn't work. That saves me a decent chunk of money as well, since I don't have to buy another ALPS, another knob, and all that jazz.

I think that I'll go ahead and lower my preamp gain while I'm at it, because it seems to be a bit too high anyway. I'll probably bring it down to about the 5-6 range. What kind of resistors and what kind of attenuation should I be looking for here? I see a lot of stuff floating around about carbon vs. metal film, and I'm not quite sure what to believe. Would I want something like a 0.5k and a 1k resistor, or would I need more attenuation than that? I know how to make a voltage divider, but I'm pretty inexperienced when it comes to the right types of components and the specific values that would be appropriate for audio. Thanks for the help.

If it makes any difference, I think I'm going to go with the audiosector stereo classic kit that uses LM3875's. I read around and people seem to prefer it a bit. It's also cheaper.

 

[danielwritesbac]

Instead of a potentiometer at the amplifier input?

The sum of two resistors would be between 20k and 30k. This represents the 25k potentiometer shown as an optional component.

There's no need to modify the preamp, and every reason not alter it, if you already like the way it sounds.

Well, I think that's right. Can someone double-check my figures?

EDIT: Alpa A20K Dual Gang isn't bad for a potentiometer. You could use it, play with it, and find your most frequently preferred setting. Then you could remove it, measure it, and replace it with resistors that represent the same setting.

 

[AndrewT]

Hi,
the disadvantage of using a low attenuation fixed resistor voltage divider is that either the input impedance is low and difficult to drive or the output impedance is high and becomes a high Rs that the following equipment may not tolerate well.
High attenuation factors avoid most of this compromising of Zin and Rs

eg. 10 +1k1 gives -20db, with Zin~=11k and Rs~=1k.
but 10k + 5k1 gives -9.4db, with Zin~=15k and Rs~=3k4

 

[johan851]

Instead of a potentiometer at the amplifier input?

The sum of two resistors would be between 20k and 30k. This represents the 25k potentiometer shown as an optional component.

There's no need to modify the preamp, and every reason not alter it, if you already like the way it sounds.

Well, I think that's right. Can someone double-check my figures?

EDIT: Alpa A20K Dual Gang isn't bad for a potentiometer. You could use it, play with it, and find your most frequently preferred setting. Then you could remove it, measure it, and replace it with resistors that represent the same setting.

No, I'll lower the gain of the preamp AND put something at the power amp input. The preamp already sounds great, but I'm changing it because 1) I like to tweak things and 2) the volume in my headphones is comfortable at about 1/6 of the volume pot's rotation. It'll just give me a larger range of useful knob turnage.

I was thinking I might use smaller resistors because I heard that less current means more resistor noise. Maybe that's only true in the Megaohm range? Having low resistor values isn't a problem in terms of driving current, since my preamp is a class A headphone amplifier capable of driving loads in the 16ohm range.

I'll check out the pot you mentioned, thanks.

AndrewT, could you explain why a resistor divider at the input would affect the output resistance?

 

[BWRX]

I was thinking I might use smaller resistors because I heard that less current means more resistor noise. Maybe that's only true in the Megaohm range?

Noise Analysis - Resistor Example

 

[johan851]

Great link. Thanks BWRX.

 

[AndrewT]

could you explain why a resistor divider at the input would affect the output resistance?

Hi,
use the diagram in that link that BWRX sent you.
Add the source resistance for the preamp to V1. let Rs=1r0
Rlp1=Zin=22k

With r1=0r0 and r2=1M the power amp sees at it's input Rs'= [R1+Rs]//R2
The preamp sees Zin of the power amp //R2 and before it gets there it also passes through R1 so Zin'=R1+[R2//Zin]

Now set you resistors to attenuate by that 9.4db mentioned earlier.
R1=10k, R2=5k1
Rs'=[R1+Rs]//R2=[10k+1r0]//5k1=3378.
Zin'=R1+[R2//Zin]=10k+[5k1//22k]=14k14
Adding the 9.4db attenuator has changed Rs' from 1r0 to 3k4 and changed Zin from 22k to 14k

If you adjust the attenuating resistors downwards then Rs' and Zin' both become lower.

 

[johan851]

Thanks for walking through that. Now that I figured out what your resistor notation means (I'm used to doing this on paper, were I would just say 14.14kohm instead of 14k14) I understand what you're saying.

So what you're saying is that the higher the impedance seen by the preamp is (Zin') the more noise I'm going to have. But I also need to make sure that Zin' doesn't go so low that my preamp can't drive it. Correct?

I'm saying that since my preamp is designed to drive loads as low as around 20r0, it makes sense to have lower resistor values in the attenuator so that I reduce the amount of noise produced by the attenuator. That doesn't mean I want extremely low values...maybe R1 = 5k1, R2 = 10k.

The point I'm still missing, I think, is why Rs' matters. Does the impedance at the amp input / preamp output as seen by the power amp actually matter?