LM3886: the effect of the compensation network Cc, Rf2 and Cf - diyAudio
Go Back   Home > Forums > Amplifiers > Chip Amps

Chip Amps Amplifiers based on integrated circuits

Please consider donating to help us continue to serve you.

Ads on/off / Custom Title / More PMs / More album space / Advanced printing & mass image saving
Reply
 
Thread Tools Search this Thread
Old 12th August 2012, 09:35 AM   #1
Marc Vi is offline Marc Vi  Netherlands
diyAudio Member
 
Join Date: Feb 2004
Location: The Netherlands
Default LM3886: the effect of the compensation network Cc, Rf2 and Cf

In the datasheet of the LM3886 you find:
- a schematic of a typical application (fig. 1) with no (extra) feedback compensation network and
- a test circuit # 2 and a typical application (fig. 2) with a feedback compensation network consisting of a Cc (220 pF), a resistor Rf2 (20 k) in serie with Cf (50 pF).

In many applications, especially the do-it-yourself kits, you find no feedback compensation network, but in some schematics you will find them, sometimes only the Cc of 220 pf and sometimes this Cc together with Rf2 and Cf.

I'am interested in the influence of these extra compensation components.

So I put a block signal of 20 kHz on the input of the amplifier (no input filter). This is what I saw on my scoop:

The first smooth image is without any compensation components. The second with the overshoot is with only the Cc of 220 pF. And the last image with the oscilations on the sides is with Cc, Rf2 and Cf.

I call for comments. So far I am sure: I will not use the (extra) compensation components, although National Semiconductor suggests with reserves to do so.

Marc.
Attached Images
File Type: jpg IMG_2055.jpg (50.8 KB, 418 views)
File Type: jpg IMG_2056.jpg (50.4 KB, 398 views)
File Type: jpg IMG_2057.jpg (50.8 KB, 393 views)
  Reply With Quote
Old 12th August 2012, 12:32 PM   #2
AndrewT is offline AndrewT  Scotland
diyAudio Member
 
Join Date: Jul 2004
Location: Scottish Borders
I recommend all the "optional" components. RF attenuation, Zobel R+C, Output inductor L//R,
Cs must be fitted very close to the power pins. The leads between the two sets of Cs must be very short.

I do not recommend that Cc (200pF) be directly connected to -IN. -IN is very susceptible to capacitance causing instability. Mount the two feedback resistors very close to -IN. Connect Cc to the other side of the lower leg NFB resistor. This is the ground end of Ri+Ci.

Try increasing the gain to see the effect on the overshoot. That middle plot showing significant overshoot looks typical of a PM<60 and probably much lower.

With the right combination of components and lead lengths and gain the overshoot should just about disappear before the square wave becomes too "rounded". This is a good clue to PM of 80

I would not use Cf + Rf2, until I had a lot more experience on stability compensation. I don't believe adjusting those two component values are a beginner task. However in discrete amps Cf~3pF to 10pF are often used to trim in the stability.
__________________
regards Andrew T.

Last edited by AndrewT; 12th August 2012 at 12:35 PM.
  Reply With Quote
Old 13th August 2012, 04:26 AM   #3
Marc Vi is offline Marc Vi  Netherlands
diyAudio Member
 
Join Date: Feb 2004
Location: The Netherlands
I already use a Zobel R+C, an output inductor L//R: see schematic.
I also use Cs's fitted very close to the power pins.

My question was if it is better or not to apply the feedback compensation network as suggested in the datasheet. So I did the experiment with and without those components and I found it is (much) better not to use them.

You confirmed my results. Indeed connecting the Cc of 220 pF to the -IN causes overshoot and makes the amplifier more sensitive for oscillation. You better connect it to the ground, making it an input filter. I apply an input filter of 1 Kohm + 1 nF, so the - 3 dB point will be app. 160 kHz.

You also confirmed not to use Cf + Rf2.

I think the LM3886 is a good product. I am suprised why they give in the datasheet the wrong advises.

Marc.
Attached Images
File Type: jpg LM3886.jpg (36.5 KB, 373 views)
  Reply With Quote
Old 13th August 2012, 09:54 AM   #4
AndrewT is offline AndrewT  Scotland
diyAudio Member
 
Join Date: Jul 2004
Location: Scottish Borders
Could you post the scope shots of the 3886 output with the components in the new locations.
This way others can see directly the before and after.

How long, in millimetres, are the lead lengths of the Cs capacitors? (8 leads for the 4 caps).
How long are the route lengths of the Cs capacitors? (4 routes from cap body all the way around the loop back to the cap body).
__________________
regards Andrew T.

Last edited by AndrewT; 13th August 2012 at 09:58 AM.
  Reply With Quote
Old 13th August 2012, 12:55 PM   #5
Marc Vi is offline Marc Vi  Netherlands
diyAudio Member
 
Join Date: Feb 2004
Location: The Netherlands
Hereby the schematics and the corresponding scoop shots of the output.

As you see the basis version gives a clear square wave, the version with the Cc of 220 pf gives overschoot and the version with the Cc, the Rf2 of 20 k and the Cf of 47 pF gives oscillations on the sides.

The value of R* is determined experimentely to cancel the offset voltage on the output. For one amplifier the value is 27k for the other 18 k. This gives an output offset of around 1 mV. This offset voltage hardly drifts (less than 1 mV) when the amplifier gets hot. The R*//C of 1 microF does not influence the scoop shots of the output.

The lead lengths of the Cs capacitors are 5 mm to the + or - and from there 20 mm over the PCB to the LM3886 connections. The lengths of the other sides to the ground on the PCB are 10 - 20 mm.

Maybe you know the answer why National Semiconductor suggests tot apply Cc, Rf2 and Cf?

Marc.
Attached Images
File Type: jpg LM3886 basis (1).jpg (45.2 KB, 358 views)
File Type: jpg IMG_2055.jpg (50.8 KB, 181 views)
File Type: jpg LM3886 with Cc (1).jpg (57.3 KB, 172 views)
File Type: jpg IMG_2056.jpg (50.4 KB, 166 views)
File Type: jpg LM3886 with Cc, Cf and Rf2 (1).jpg (59.4 KB, 603 views)
File Type: jpg IMG_2057.jpg (50.8 KB, 147 views)
  Reply With Quote
Old 13th August 2012, 01:43 PM   #6
AndrewT is offline AndrewT  Scotland
diyAudio Member
 
Join Date: Jul 2004
Location: Scottish Borders
Could you add the RF attenuator 220pF from Hot to Signal ground and show the corresponding scope shots? i.e. add to first and move on third.
__________________
regards Andrew T.
  Reply With Quote
Old 13th August 2012, 01:45 PM   #7
AndrewT is offline AndrewT  Scotland
diyAudio Member
 
Join Date: Jul 2004
Location: Scottish Borders
If that second scope shot is typical of what other builders get in their builds then I will surmise and state "no wonder they all hear changes in sound quality when they swap in and out other components.

First one must get the amplifier working properly.
Then one can look for "improvements".
__________________
regards Andrew T.
  Reply With Quote
Old 13th August 2012, 02:22 PM   #8
Marc Vi is offline Marc Vi  Netherlands
diyAudio Member
 
Join Date: Feb 2004
Location: The Netherlands
When I add a C of 220 pF from the + input to the ground in the basic version (the schematic most on the left), I see a very very minimal difference in the way that the bends rounds a little bit more on the tops. But the difference is too small to recognize it on a shot.

The difference is of course greater when you increase the value of the capacitor. Then you see more sooth bends with no oscillations or what so ever. What you is is a normal R-C filter.

I think I am going to use a capacitor from + to ground with a value of 1000 pf (- 3 dB point = 160 kHz) or 1200 pf ( - 3 dB point = 132 kHz). The phase shift at 20 kHZ is then less than 10 degrees. I sometimes hear that a phase shift of less than 10 degrees is not audible, but more dan 10 degres may be audible.

Marc.
  Reply With Quote
Old 13th August 2012, 02:32 PM   #9
AndrewT is offline AndrewT  Scotland
diyAudio Member
 
Join Date: Jul 2004
Location: Scottish Borders
Thanks for confirming that the "normal" location of RF attenuation Capacitor does not show a marked change in the stability response of the 3886.
Could you now show what happens in pic 3 with the RF moved to the normal location?

It seems that you are confirming what many have found:
The National datasheet was cobbled together without proper consideration for the skills of those who were about to read and implement.
Or more simply:
The 3886 datasheet is riddled with mistakes!

btw, those are my interpretations.
__________________
regards Andrew T.
  Reply With Quote
Old 13th August 2012, 07:32 PM   #10
Marc Vi is offline Marc Vi  Netherlands
diyAudio Member
 
Join Date: Feb 2004
Location: The Netherlands
Hallo Andrew,

Hereby the picture with Cc (220 pF) from + input to ground and with Cf (47 pF) in serie with Rf2 (22 k) from the output to the - input.

When I disconnect Cc, there is hardly no difference.

As you thought, the oscillations are caused by a Cc between the + and the - input!!

The conclusion must be: never use a RF attenuation capacitor between the + and - input but always - in the non-inverting situation - between the + input and ground.

Can you tell me what is best to do:

- Only a capacitor Cc between + input and ground to make a low pass filter or;
- A capacitor Cf in serie with a resistor Rf2 from the output to the - input or
- A combination of those two?

What will then be the best values?

Marc.
Attached Images
File Type: jpg IMG_2061.jpg (56.3 KB, 524 views)
  Reply With Quote

Reply


Hide this!Advertise here!
Thread Tools Search this Thread
Search this Thread:

Advanced Search

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off


Similar Threads
Thread Thread Starter Forum Replies Last Post
Miller compensation with a series RC network? Bricolo Parts 5 27th February 2010 01:19 PM
First Watt F1 - compensation network for fostex 126e recstar24 Pass Labs 6 16th October 2009 11:08 PM
Effect on freq. compensation by substituting transistors? h_a Solid State 12 16th April 2008 09:15 AM
Novice compression driver and compensation network Q... massappeal129 Multi-Way 1 22nd April 2004 09:59 PM
compensation network for tubeamplifier anodedick Multi-Way 3 16th July 2002 01:52 PM


New To Site? Need Help?

All times are GMT. The time now is 04:16 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