diyAudio

diyAudio (http://www.diyaudio.com/forums/)
-   Chip Amps (http://www.diyaudio.com/forums/chip-amps/)
-   -   LM3886 component selection (http://www.diyaudio.com/forums/chip-amps/224914-lm3886-component-selection.html)

SirPlanALot 3rd December 2012 11:27 PM

LM3886 component selection
 
Hello everybody!

First, I have read through a lot of different threads, but haven't found exactly what I was looking for.
As I didn't want to hijack several threads, I decided to start a new one.

I already have successfully designed and built some microcontroller circuits, but I'm quite new to audio design.

Now I want to build a small amp do drive (diy) PC speakers, and I have chosen the popular LM3886.
I generally like my projects being reusable, so everything that could be of use later on should be included.

I hope this was enough background information, so let's start with my questions:
  • I want to follow the design notes written here: http://www.diyaudio.com/forums/chip-...ml#post3165612 and here: http://www.diyaudio.com/forums/chip-...ml#post3167216
    Ok, or not?
  • For supply bypassing, I have planned 470F per rail (Panasonic FR, 50V, low-ESR) and 100nF X7R ceramic in 0805 SMD case.
  • Cc (220pF) is connected between the inputs of the LM3886, but some said it would be better between non-inverted input and GND, so I will provide space for both options.
    I planned to use NPO ceramic in 0805 SMD case, is that ok or should it better be foil, mica or something else?
  • For Ci, the LM4780 datasheet specifies polarized electrolytic capacitor.
    Is that correct, or should a non-polarized capacitor type (bipolar electrolytic, foil) be used?
  • For the snubber (Rsn 2.7R and Csn 100nF), the LM4780 datasheet states 0.25W resistor and ceramic capacitor.
    Is the resistor rating ok? In many designs a 2W or 3W resistor is used...
    For capacitor, I planned to use X7R ceramic in 0805 case, is that ok or should it be a foil type?
  • For the Zobel network (10R || 0.7H in series to the load) there are no ratings and recommendations in the datasheet.
    What should be the power rating for the resistor, would 2W be sufficient?
    How should the inductor be made?
    Pure air core, or wrapped around the resistor? I have read somewhere it should be separate, but don't find it anymore...
Any help, hints and corrections will be appreciated. :)

PS: I'm not into that high-end-voodoo-thing :D, I just want to build an optimal, reasonable and stable design.

AndrewT 4th December 2012 01:24 PM

Ordinary polarised electrolytics are good enough for Ci. But, you must ensure they have virtually no AC across them. i.e. make them big enough (low AC impedance) and any distortion becomes unmeasurable.

If the amplifier never oscillates then the Rsn can never get hot.
If you intend "testing" the amplifier at high frequency, such that you end up "driving" a signal through Rsn, then you may want to make it big enough for however much power you intend to dissipate in it. That could be 60W of continuous 100kHz !

The Rsn + Csn = output Zobel.

Thiele proposed a different stability enhancing Network that had both the Zobel and an output inductor. I refer to this as the Thiele Network. Dr Cherry did more work in explaining the Thiele Network and the infinite variations that can exist.

The amplifier can only heat up the Thiele Network for similar abuse as explained earlier. Do you plan to amplify >100kHz signals?

R//L pure air core. Keep the air core away from the amplifier stage and if possible away from steel chassis. Run the R outside the coil, not through it. I have not seen data for the effect of running the resistor inside the coil, so I'm recommending "play safe".

ClaveFremen 4th December 2012 10:08 PM

Quote:

Originally Posted by SirPlanALot (Post 3269211)
  • For supply bypassing, I have planned 470F per rail (Panasonic FR, 50V, low-ESR) and 100nF X7R ceramic in 0805 SMD case.

Consider Elna RJH, excellent and good sounding in that role


Quote:

Originally Posted by SirPlanALot (Post 3269211)
  • Cc (220pF) is connected between the inputs of the LM3886, but some said it would be better between non-inverted input and GND, so I will provide space for both options.
    I planned to use NPO ceramic in 0805 SMD case, is that ok or should it better be foil, mica or something else?

Use Wima FKP2 or if you want to use SMD parts: CDE silver mica or Panasonic ECHU (PPS)



Quote:

Originally Posted by SirPlanALot (Post 3269211)
  • For Ci, the LM4780 datasheet specifies polarized electrolytic capacitor.
    Is that correct, or should a non-polarized capacitor type (bipolar electrolytic, foil) be used?

That cap, if dimensioned like Andrew suggest, has a neglegible distortion contribute but its signature, IMHO, is still audible, use an audio grade cap (Nichicon FG is fine)



Quote:

Originally Posted by SirPlanALot (Post 3269211)
  • For the snubber (Rsn 2.7R and Csn 100nF), (...) For capacitor, I planned to use X7R ceramic in 0805 case, is that ok or should it be a foil type?

Use a high quality polypropilene cap.

BTW I've never been convinced by that PS, read here.

SirPlanALot 5th December 2012 01:38 AM

Quote:

Originally Posted by AndrewT (Post 3270021)
Ordinary polarised electrolytics are good enough for Ci. But, you must ensure they have virtually no AC across them. i.e. make them big enough (low AC impedance) and any distortion becomes unmeasurable.

Ah ok.
I thought of using a bipolar or un-polarized cap because the output swings around GND, and so will the voltage the cap sees.

I planned to make Rin=22k and Cin=4.7F, which results in a corner frequency of ~1.5Hz.
Ri will be 1k and Ci=180F, which results in a corner frequency of ~0.9Hz.
I hope this will be big enough, or am I wrong?

Quote:

Originally Posted by AndrewT (Post 3270021)
If the amplifier never oscillates then the Rsn can never get hot.

Sounds logical.
But how can I prevent the amp from oscillating?
Is it just the component placement?
Or can I do anything more to achieve this, like the optional components Rf2 and Cf?
By the way, the formula for the corner frequency of that network is specified in the datasheet as follows:
fc=[Rf1 Rf2 (s + 1/Rf2Cf)]/[(Rf1 +Rf2)(s + 1/Cf(Rf1 +Rf2))]
What is "s"?

Quote:

Originally Posted by AndrewT (Post 3270021)
If you intend "testing" the amplifier at high frequency, such that you end up "driving" a signal through Rsn, then you may want to make it big enough for however much power you intend to dissipate in it. That could be 60W of continuous 100kHz !

Excessive "testing" beyond audio frequencies is not my intention.
The (worst case) 60W of dissipation, when the snubber becomes active, has lead to my question of a reasonable size.
But as I think about it, IF the amp oscillates, it fries the resistor whichever power rating it may have.
So it is quite important (at least for me) to prevent the amp from ever oscillating, as I want an audio amplifier and no RF source.

Quote:

Originally Posted by AndrewT (Post 3270021)
R//L pure air core. Keep the air core away from the amplifier stage and if possible away from steel chassis. Run the R outside the coil, not through it. I have not seen data for the effect of running the resistor inside the coil, so I'm recommending "play safe".

So the R|L should not be placed on the board?

@ClaveFremen:
Those Elna RJH seem not very well available...
I'll have a look at the other types you suggested, and at your thread.

Good night!

AndrewT 5th December 2012 04:03 PM

The NFB DC blocking cap has a very small DC voltage across it.
This DC voltage is due to the input offset voltage of the amplifier. This input offset voltage exists at BOTH the +IN and -IN inputs.
The AC voltage across the DC blocking cap should be very small and the peak voltage is likely to be less than the DC bias across it. This addition of two very small voltages is usually never a -ve bias across the polarised capacitor.
In addition some of the polarised capacitor manufacturers state that a reverse voltage of < ~1Vdc is OK for their capacitors.
Quote:

That cap, if dimensioned like Andrew suggest, has a neglegible distortion contribute but its signature, IMHO, is still audible, use an audio grade cap (Nichicon FG is fine)
I am very inclined to not agree this quality philosophy.
I can see that when the DC blocking capacitor is too small and thus develops a significant AC voltage across it that the added distortion of that too small capacitor will have an audible effect. I suspect that many have heard this and without realising the cause concluded that the NFB DC blocking capacitor has a "sound" effect.

When the capacitor is sized such that there is virtually no signal voltage across it then the distortion added by the capacitor is inaudible.

This is something I have found in practice. I cannot hear the "sound" changes alluded to by other builders.

SirPlanALot 5th December 2012 05:57 PM

2 Attachment(s)
Thank you for the explanation AndrewT, I really appreciate your help!
I have read a lot from you in other LM3886 threads.

Here is what I have done until now.
The schematics is still a bit messy, because I have changed it several times after studying several design "guidelines".

The color codes for the layout are:
Red: top copper layer
Green: bottom copper layer
Cyan: top silkscreen layer
Magenta: bottom silkscreen layer

The parts were selected from the available types at Conrad and Reichelt, two popular german parts suppliers.
They may still change, but here is what was used for creating the current layout:
Cs (C1, C4): Panasonic FR 470F 50V; (C3, C5): ceramic X7R 100nF
Ci (C2): Panasonic FC 180F 50V
Cin (C10): Wima MKS-2 4.7F
Cm (C6): Panasonic FC 100F 50V
Cc (C8, C11): ceramic NPO 220pF
Cf (C9): ceramic NPO 47pF
Csn (C7): ceramic X7R 100nF

Rsn (R6): Yageo RC1206 2.7R 0.25W
Output resistor+coil (R9): will be removed from PCB
other resistors: metal film 0805

What do you think of my attempt?

Regards

ClaveFremen 5th December 2012 07:58 PM

1 Attachment(s)
Quote:

Originally Posted by SirPlanALot (Post 3270929)
@ClaveFremen:
Those Elna RJH seem not very well available...
I'll have a look at the other types you suggested, and at your thread.

You can buy them in Germany from Distrelec

Quote:

Originally Posted by SirPlanALot (Post 3271781)
What do you think of my attempt?

I would trasnfer PGND to the other side or to a groundplane so that you can route differently several traces like the ones related to mute pin and V+.

Something like that:

http://www.diyaudio.com/forums/attac...1&d=1354740529

As you can see V+ doesn't overlap OUT ;)

SirPlanALot 5th December 2012 08:22 PM

Quote:

Originally Posted by ClaveFremen (Post 3271949)
You can buy them in Germany from Distrelec

Thank you for the hint!
Do they stock the other types you mentioned as well?
The site is down at the moment...

Quote:

Originally Posted by ClaveFremen (Post 3271949)
I would trasnfer PGND to the other side or to a groundplane so that you can route differently several traces like the ones related to mute pin and V+.

I have left them intentionally on the same side, because of the SMD bypass caps I use at the moment.
But I could trim the ground polygon under the IC, and place the V+ track under the IC (where the ground trace is at the moment).
How would that be?

I don't like the crossing of V+ and Output either.
But sometimes I need to get a hint from someone else to see a solution.

ClaveFremen 5th December 2012 08:54 PM

Quote:

Originally Posted by SirPlanALot (Post 3271978)
Thank you for the hint!
Do they stock the other types you mentioned as well?
The site is down at the moment...

You're welcome. :)

They stock Wima FKP2 (use 2.5% tolerance -> copper leads, better if 100V rated)

Quote:

Originally Posted by SirPlanALot (Post 3271978)
I have left them intentionally on the same side, because of the SMD bypass caps I use at the moment.

You can use vias... or use LM3886 pads as vias...or simply bypass rail to rail like I did...

Quote:

Originally Posted by SirPlanALot (Post 3271978)
But I could trim the ground polygon under the IC, and place the V+ track under the IC (where the ground trace is at the moment).
How would that be?

Post a picture...

Quote:

Originally Posted by SirPlanALot (Post 3271978)
I don't like the crossing of V+ and Output either.
But sometimes I need to get a hint from someone else to see a solution.

;)

SirPlanALot 5th December 2012 09:49 PM

Quote:

Originally Posted by ClaveFremen (Post 3272015)
You can use vias... or use LM3886 pads as vias...or simply bypass rail to rail like I did...
Post a picture...

The V+ trace would just look like yours, and Gnd trace would end between the SMD decoupling capacitors.

Has the rail to rail bypassing any advantages over bypassing each rail to ground?

Quote:

Originally Posted by ClaveFremen (Post 3272015)
They stock Wima FKP2 (use 2.5% tolerance -> copper leads, better if 100V rated)

Ah, I've found them.
But sadly they don't have a 47pF or 56pF type, as is needed for Cf...
And SMD types like Panasonic EHCU are also not available, but Reichelt has them only in 330pF and 470pF, and also 100nF.
Will the 100nF EHCU be ok for power bypassing and Csn?

Are Wima MKS-2 ok for NF input?

And are the Elna RJH also ok for Ci?
Because Distrelec doesn't have any Nichion caps.


All times are GMT. The time now is 10:29 PM.


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