The 'Ci' DC blocking cap in the LM3886 Feedback Loop

speakerguy79 · 6th February 2009, 04:34 AM

Why does everyone use a ~highish voltage electrolytic here? It should never see but 1/20th the voltage of the rails at most. A lot of folks even use polar caps here, which would pop if anything more negative than a volt or so went across it.

Can I use a teeny little 6.3V rated cap here?

sangram · 6th February 2009, 07:38 AM

I have a 6.3V 47uF Black Gate NX Hi-Q in Ci position and I haven't had a problem.

Of course in failure mode that node sees full rail voltage (sometime negative) so the safest bet is probably a bipolar rated to one rail voltage.

pacificblue · 6th February 2009, 08:05 AM

The lower the frequency is, the higher is the resistance of that capacitor. The higher the resistance is, the more voltage will drop across it. Next to 0 Hz nearly all the voltage drops across Ci. Ci should therefore withstand the rail voltage. It does not need a safety margin like e. g. the PSU capacitors, because hardly any current flows though it, and that is also, why it does not matter, if the voltage across it is reverse.
At audio frequencies the voltage drop across it should be next to zero. At DC no current flows at all. In between those two states there are little, if any signals present. You can get away with a too small voltage rating. Is it worth the risk, when you look at the price of such an electrolytic? Once you had to cope with the disgusting smell of a blown capacitor you won't think about it again.

Rule of thumb: the next voltage rating below the PSU capacitors.

crt · 10th February 2009, 02:17 PM

i'm going to make LM3886 too, but i will not use 10K (pot.).
and i will use 3,3uF DC blocking caps.
the amp will play direct to my PC soundcard, so i don't need volume control.

is there need modification for the original schematic (datasheet) ?
if yes, can you draw for me (skecth) please.

Thank's

Iliya · 27th June 2009, 05:28 PM

Is there also an issue of lower quality of low voltage electrolytic capacitors? As I am aware 6,3V capacitors usually have twice higher losses than 25V capacitors from the same series.

AndrewT · 28th June 2009, 10:16 AM

Quote:

Originally posted by crt
i'm going to make LM3886 too, but i will not use 10K (pot.).
and i will use 3,3uF DC blocking caps.
the amp will play direct to my PC soundcard, so i don't need volume control.

this arrangement could well show up the soundcard and the chipamp in it's worst light.

sangram · 28th June 2009, 12:22 PM

Any specific reason why, Andrew?

I have a very heavily modded E-mu 1212m going directly into a P2P LM3886, and it's pin-drop silent, even with my ear inside the tweeter, literally. Even a Buffalo DAC used this way sounds simply great, with absolutely no noise at the speakers. The amp is set up for 26dB gain, and directly coupled to the card, with not even a series cap. I also don't use bandwidth control, or push the volume too high even though my speakers are only 85dB efficient.

With poor construction and power supplies some noise may leak through and I've experienced this myself when I struggled to fix a ground loop issue with my Delta66, but if the construction is properly handled I see no issues. Actually I love the sound this way and would never again use a volume pot, as even a reasonably high quality stepped attenuator I used degraded the sound from this arrangement. Pots simply sound bad, IMHO.

I do know about bit resolution loss at lower volume and noise performance of soundcards being another reason, but with a sufficiently high quality card this should not be a problem. I have checked the outputs from a X-Fi feeding a LM4766 (again, set up with 26dB gain) and the results are better than using a pot.

I have seen the Alien DAC create a bit of hiss when used this way, so I prefer a pot on that source, but I think that's relatively low-spec so the pot doesn't harm it too much anyway.

But I'm sure you have your reasons - I'm keen to know what they are.

AndrewT · 28th June 2009, 01:28 PM

Quote:

Originally posted by sangram
Any specific reason why
...................would never again use a volume pot, as even a reasonably high quality stepped attenuator I used degraded the sound from this arrangement. Pots simply sound bad, .............
..............I do know about bit resolution loss at..........volume ....................but with a sufficiently high quality card this should not be a problem. I have checked the outputs from a X-Fi feeding a LM4766 (again, set up with 26dB gain) and the results are better than using a pot.

I find your conclusions that the digital attenuation being better than analogue attenuation, whether stepped or pot, to be surprising.
That would have been one of my reasons.

The lack of HF attenuation would be another.

The lack of DC blocking at the input would be a third.

sangram · 28th June 2009, 02:24 PM

I can understand, I never expected the same results myself.

I believe the mixer on the soundcard also has an effect, with the regular windows kmixer obviously being a pretty grave offender in this regard.

I think the professional cards mix internally at 32-bits (mine does) so this may have an effect on the precision of the amplitude calculation. I would not know about this right now, but I'll do my research before I commit to an answer.

Most soundcards have DC blocking of some sort. The E-mu has them right before the output opamps, the m-Audio has them right at the output itself (after opamps), which is the position I've almost always seen them in. The Buffalo DAC, of course, has none but it has no output offset either, so not sure they're always needed.

I agree with your observation about HF attenuation. I use none myself and no zobel on the output so it is a bit hairy in my head (I wish I had some more outside but meh) at times, but the few times I have checked, with reasonable loads and some attention to layout this is not usually a problem. Again, I don't use a PCB so the only stray inductance I have to deal with is the lead inductance of a few millimeters of each component. I've never had a single case of oscillation since I started laying the grounds out correctly. On stereo chips this is a nightmare!

I think HF attenuation is pretty much compulsory in all cases though, and I do not advocate running without , but it seems to be pretty easy to leave those bits in when connecting directly. A simple RC filter should do, cornered at about 100KHz or so.

wazzy · 29th June 2009, 06:03 PM

i used the lm3886 without this cap initialy running it from my own xfi sound card and the sound was pretty good, then i changed it and am now using a 10uf 50v elna cap as the ci and seem to have a cleaner sound, not hitting the spike protection like ti used to... interesting. also dc coupled no input caps and no volume pots. xfi sound cards are pretty good!

6th February 2009, 04:34 AM	#1
speakerguy79 diyAudio Member Join Date: May 2006	The 'Ci' DC blocking cap in the LM3886 Feedback Loop Why does everyone use a ~highish voltage electrolytic here? It should never see but 1/20th the voltage of the rails at most. A lot of folks even use polar caps here, which would pop if anything more negative than a volt or so went across it. Can I use a teeny little 6.3V rated cap here?

6th February 2009, 08:05 AM	#3
pacificblue diyAudio Member Join Date: May 2008	The lower the frequency is, the higher is the resistance of that capacitor. The higher the resistance is, the more voltage will drop across it. Next to 0 Hz nearly all the voltage drops across Ci. Ci should therefore withstand the rail voltage. It does not need a safety margin like e. g. the PSU capacitors, because hardly any current flows though it, and that is also, why it does not matter, if the voltage across it is reverse. At audio frequencies the voltage drop across it should be next to zero. At DC no current flows at all. In between those two states there are little, if any signals present. You can get away with a too small voltage rating. Is it worth the risk, when you look at the price of such an electrolytic? Once you had to cope with the disgusting smell of a blown capacitor you won't think about it again. Rule of thumb: the next voltage rating below the PSU capacitors. __________________ If you've always done it like that, then it's probably wrong. (Henry Ford)

Thread Tools	Search this Thread
Show Printable Version Email this Page	Search this Thread: Advanced Search

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Adcom GFA-555 Feedback DC-Blocking Cap?	EchoWars	Pass Labs	14	14th April 2010 07:12 PM
DC blocking cap in feedback path or DC servo?	roender	Solid State	27	19th September 2007 09:40 AM
feedback loop	rmenger	Instruments and Amps	4	21st October 2006 07:04 PM
DC blocking feedback	borges	Analog Line Level	1	31st March 2006 07:48 AM
feedback loop	rmenger	Parts	0	29th July 2004 01:24 PM

New To Site?	Need Help?
Register to Participate Search Privacy Statement Contact Us	Frequently Asked Questions Did you forget your password? Mark Forums Read


	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

6th February 2009, 07:38 AM	#2
sangram diyAudio Member Join Date: Sep 2002 Location: India	I have a 6.3V 47uF Black Gate NX Hi-Q in Ci position and I haven't had a problem. Of course in failure mode that node sees full rail voltage (sometime negative) so the safest bet is probably a bipolar rated to one rail voltage.

10th February 2009, 02:17 PM	#4
crt diyAudio Member Join Date: Apr 2007	i'm going to make LM3886 too, but i will not use 10K (pot.). and i will use 3,3uF DC blocking caps. the amp will play direct to my PC soundcard, so i don't need volume control. is there need modification for the original schematic (datasheet) ? if yes, can you draw for me (skecth) please. Thank's

27th June 2009, 05:28 PM	#5
Iliya diyAudio Member Join Date: Oct 2005	Is there also an issue of lower quality of low voltage electrolytic capacitors? As I am aware 6,3V capacitors usually have twice higher losses than 25V capacitors from the same series.

28th June 2009, 12:22 PM	#7
sangram diyAudio Member Join Date: Sep 2002 Location: India	Any specific reason why, Andrew? I have a very heavily modded E-mu 1212m going directly into a P2P LM3886, and it's pin-drop silent, even with my ear inside the tweeter, literally. Even a Buffalo DAC used this way sounds simply great, with absolutely no noise at the speakers. The amp is set up for 26dB gain, and directly coupled to the card, with not even a series cap. I also don't use bandwidth control, or push the volume too high even though my speakers are only 85dB efficient. With poor construction and power supplies some noise may leak through and I've experienced this myself when I struggled to fix a ground loop issue with my Delta66, but if the construction is properly handled I see no issues. Actually I love the sound this way and would never again use a volume pot, as even a reasonably high quality stepped attenuator I used degraded the sound from this arrangement. Pots simply sound bad, IMHO. I do know about bit resolution loss at lower volume and noise performance of soundcards being another reason, but with a sufficiently high quality card this should not be a problem. I have checked the outputs from a X-Fi feeding a LM4766 (again, set up with 26dB gain) and the results are better than using a pot. I have seen the Alien DAC create a bit of hiss when used this way, so I prefer a pot on that source, but I think that's relatively low-spec so the pot doesn't harm it too much anyway. But I'm sure you have your reasons - I'm keen to know what they are.

28th June 2009, 02:24 PM	#9
sangram diyAudio Member Join Date: Sep 2002 Location: India	I can understand, I never expected the same results myself. I believe the mixer on the soundcard also has an effect, with the regular windows kmixer obviously being a pretty grave offender in this regard. I think the professional cards mix internally at 32-bits (mine does) so this may have an effect on the precision of the amplitude calculation. I would not know about this right now, but I'll do my research before I commit to an answer. Most soundcards have DC blocking of some sort. The E-mu has them right before the output opamps, the m-Audio has them right at the output itself (after opamps), which is the position I've almost always seen them in. The Buffalo DAC, of course, has none but it has no output offset either, so not sure they're always needed. I agree with your observation about HF attenuation. I use none myself and no zobel on the output so it is a bit hairy in my head (I wish I had some more outside but meh) at times, but the few times I have checked, with reasonable loads and some attention to layout this is not usually a problem. Again, I don't use a PCB so the only stray inductance I have to deal with is the lead inductance of a few millimeters of each component. I've never had a single case of oscillation since I started laying the grounds out correctly. On stereo chips this is a nightmare! I think HF attenuation is pretty much compulsory in all cases though, and I do not advocate running without , but it seems to be pretty easy to leave those bits in when connecting directly. A simple RC filter should do, cornered at about 100KHz or so.

29th June 2009, 06:03 PM	#10
wazzy diyAudio Member Join Date: Sep 2007	i used the lm3886 without this cap initialy running it from my own xfi sound card and the sound was pretty good, then i changed it and am now using a 10uf 50v elna cap as the ci and seem to have a cleaner sound, not hitting the spike protection like ti used to... interesting. also dc coupled no input caps and no volume pots. xfi sound cards are pretty good!