Electronics > Projects, Designs, and Technical Stuff

Composite amplifier: LM3886 + LME49720

<< < (4/4)

T3sl4co1l:
Well hell, there you have it. :popcorn:

Tim

bson:
Great little project, can't wait to see how you like it.  :-+

I'd give it an amplitude sweep also to look for compression as it gets too close to rail on the output.  This can be used to set the peak output level, and from there determine the total amplifier gain.  Then move as much off this to the LME49710 up front, off of the LM3886.  The good news is the LM3886 specs 0.03% THD+N at +26dB, and presumably this improves correspondingly with a reduction in gain.  The 49710 will have THD+N specs orders of magnitude better, so get all the gain you can up front.

Do you really need a 100W amplifier?  The higher the power, the higher its total gain needs to be, and the more you need to attenuate the input and feed it ever weaker input signals - unless you actually need the power.  Maybe consider switchable gain on the LM3886 (with a dial or switches) so you can run it at lower power levels normally, but crank up the gain if you buy large speakers that really need that much power.  From a numbers perspective, less gain is better.

You may not need the DC servos; the offset will be Vos of the LME49710 times the total amplifier gain.  The LM3886 offset will be removed by the feedback.  If I recall the 49710 max Vos is on the order of 1mV, but realistically you get much less.  Does any real world loudspeaker care about a 20-50mV offset?

I'd also use the mute function to silence it at power-on, using something along the lines of the BJT below.  (It seems to require ≥0.5mA sunk out of the mute pin to unmute.)  With 100kΩ*120µF and Vdd=17V the base reaches 0.6V in about 0.4s, and then it will gradually open up until it sinks the full 0.5mA.  If Vdd ramps at power-on it will take longer to unmute.  Or something like it (just breadboard and see what works).


Just a few thoughts.  Cool project, and I may make one for myself to experiment with if you don't mind. :)

I like the idea of using an LME49724 and differential inputs, especially if the input signal is going to be very weak due to high amplifier gain! :-+

Edit: oh, and another reason to mute it during power-on is the DC servos need to prime or you can get a nasty DC spike on the output.

JeanLeMotan:

--- Quote from: bson on March 05, 2019, 01:42:20 am ---Great little project, can't wait to see how you like it.  :-+

I'd give it an amplitude sweep also to look for compression as it gets too close to rail on the output.  This can be used to set the peak output level, and from there determine the total amplifier gain.  Then move as much off this to the LME49710 up front, off of the LM3886.  The good news is the LM3886 specs 0.03% THD+N at +26dB, and presumably this improves correspondingly with a reduction in gain.  The 49710 will have THD+N specs orders of magnitude better, so get all the gain you can up front.

--- End quote ---

The gains are setup with these in mind:
- Use the minimal LM3886 stable gain, which according to the datasheet is ~10. I used 11 just to be sure
- Have the LM49710 clip first to get a much nicer clip behaviour without saturating it. Check the attachments for the 2 behaviors: LM3886 clipping or the LME49710 doing the clipping.
- Use as small resistors in the central divider R6/R25 in the new schematic here: https://easyeda.com/jeanleflambeur/amp-lm3886-composite2
- The central divider does 2 things: allows the 49710 to clip first, and helps with stability (the C1 cap across R5)


--- Quote from: bson on March 05, 2019, 01:42:20 am ---Do you really need a 100W amplifier?  The higher the power, the higher its total gain needs to be, and the more you need to attenuate the input and feed it ever weaker input signals - unless you actually need the power.  Maybe consider switchable gain on the LM3886 (with a dial or switches) so you can run it at lower power levels normally, but crank up the gain if you buy large speakers that really need that much power.  From a numbers perspective, less gain is better.

--- End quote ---

I'm listening to a lot of classical music at medium volume and the dynamic range is very high. Since I don't have the speakers yet (I want to build a set) to know their sensitivity, I wanted the head room.


--- Quote from: bson on March 05, 2019, 01:42:20 am ---You may not need the DC servos; the offset will be Vos of the LME49710 times the total amplifier gain.  The LM3886 offset will be removed by the feedback.  If I recall the 49710 max Vos is on the order of 1mV, but realistically you get much less.  Does any real world loudspeaker care about a 20-50mV offset?

--- End quote ---

How is the offset of the LM3886 removed in the feedback? I guess you mean the 49710 feedback, right? I imaging the 49710 replaces the LM3886 offset with its own, right?
In the new schematic I have on 49710 per paralleled group so I don't think I can remove the servos anymore...


--- Quote from: bson on March 05, 2019, 01:42:20 am ---I'd also use the mute function to silence it at power-on, using something along the lines of the BJT below.  (It seems to require ≥0.5mA sunk out of the mute pin to unmute.)  With 100kΩ*120µF and Vdd=17V the base reaches 0.6V in about 0.4s, and then it will gradually open up until it sinks the full 0.5mA.  If Vdd ramps at power-on it will take longer to unmute.  Or something like it (just breadboard and see what works).


--- End quote ---

I already have the datasheet mute circuit (cap + resistor). How is that circuit different?
Don't remember the time constant I calculated but it was several seconds to make sure the DC servos have time to do smth. Although in my simulations the servos need >15s to completely eliminate the offset... Muting for that long will not e user friendly as it might seem the amp is not powered on.


--- Quote from: bson on March 05, 2019, 01:42:20 am ---Just a few thoughts.  Cool project, and I may make one for myself to experiment with if you don't mind. :)

I like the idea of using an LME49724 and differential inputs, especially if the input signal is going to be very weak due to high amplifier gain! :-+

Edit: oh, and another reason to mute it during power-on is the DC servos need to prime or you can get a nasty DC spike on the output.

--- End quote ---

By all means - go ahead. The schematic and PCB are public domain, have fun with them. Just keep in mind that it comes with no guarantees as I'm not an expert (if that was not already obvious :) )

Check the second schematic and PCB first, I think it's an improvement over the first one.
https://easyeda.com/jeanleflambeur/amp-lm3886-composite2

Also, the corresponding (simple) PSU:
https://easyeda.com/jeanleflambeur/psu-lm3886-composite

bson:

--- Quote from: JeanLeMotan on March 05, 2019, 09:04:09 am ---How is the offset of the LM3886 removed in the feedback? I guess you mean the 49710 feedback, right? I imaging the 49710 replaces the LM3886 offset with its own, right?

--- End quote ---
Yeah, if the LM3886 adds an offset the 49710 will remove it.  The offset you get on the output is only the latter's input offset times the total gain.

Sorry, didn't see you didn't already have a mute circuit.

Navigation

[0] Message Index

[*] Previous page

There was an error while thanking
Thanking...
Go to full version
Powered by SMFPacks Advanced Attachments Uploader Mod