At least half an octave below the input filter is big enough to avoid that voltage across the capacitor.
Let's say you set the input filter to 50ms. Then set the NFB filter to >75ms.
Let's say you set the input filter to 50ms. Then set the NFB filter to >75ms.
I set my input filter to ~90ms. That requires a combination of a 9k1 resistor and a 10uF capacitor.
That input filter makes sure that all frequencies significantly below 1.7Hz do not develop any significant voltage across the feedback capacitor.
At frequencies around 1.7Hz the feedback capacitor will develop a little voltage. But if the NFB capacitor is set to a value that ensures that frequencies around 1.7Hz do not develop a significant voltage across the capacitor, then that capacitor will not add any distortion to the signal being processed. By adjusting the impedance of the capacitor one can also adjust or eliminate the voltage across the cap. To achieve this desirable set of circumstances all one has to do is select a capacitor whose impedance at 1.7Hz is low in comparison to the NFB resistances (that set the amp gain).
You do this by selecting a RC time constant for the NFB capacitor that is at least half an octave below the input filter. i.e. Rc>140ms. This would require the NFB resistor to be 1k0 and the capacitor to be 150uF (150ms > 140ms) or any other RC combination that gives >140ms.
That input filter makes sure that all frequencies significantly below 1.7Hz do not develop any significant voltage across the feedback capacitor.
At frequencies around 1.7Hz the feedback capacitor will develop a little voltage. But if the NFB capacitor is set to a value that ensures that frequencies around 1.7Hz do not develop a significant voltage across the capacitor, then that capacitor will not add any distortion to the signal being processed. By adjusting the impedance of the capacitor one can also adjust or eliminate the voltage across the cap. To achieve this desirable set of circumstances all one has to do is select a capacitor whose impedance at 1.7Hz is low in comparison to the NFB resistances (that set the amp gain).
You do this by selecting a RC time constant for the NFB capacitor that is at least half an octave below the input filter. i.e. Rc>140ms. This would require the NFB resistor to be 1k0 and the capacitor to be 150uF (150ms > 140ms) or any other RC combination that gives >140ms.
Midrange, Bass, Midbass, Treble. . . hidden 4 band eq in there
Nichicon ES, Panasonic FC and other high efficiency caps could be on the shortlist of things to try. You can do a search for favorites to see what other people use. The idea of optimal differs per each application, so I'd try 5 favorites to see what you like. Price does not indicate performance, but datasheets can be helpful. Polar caps work fine in split rail amplifiers.
Midrange, forward, recessed or level: For Non-Inverting LM3886 amplifiers, it might be good to choose amplifier board power caps with large enough capacitance to help support a level midrange frequency response prior to attempting the selection of any signal caps. The Panasonic FC 1500uF that AudioSector uses for power caps of the amplifier board, might be a good, first place to start. Good power must come first.
Bass harmonic balance and clarity: The size of the NFB cap versus the size of the Input cap affects the H1/H2 bass harmonic balance whereby too small Nfb cap makes warm/boomy/muddy bass, too large Nfb cap makes rumble bass monotony and somewhere in-between would have a balance with neither problem. The muddy bass problem is most commonplace and that muddy problem happens when the input cap passes more bass than the NFB cap. So, the simple solution for clearer bass would be to have the Input cap smaller and/or the NFB cap larger.
AndrewT probably explained this better than I could, so long as your calculator is up for it.
Treble Bypass cap: For leveling out the treble, those inexpensive little green polyester dip caps from Tracon/Xicon are very easy to use in parallel to an electrolytic cap. Applicable size range are from approximately 1nF to 150nF and it takes some experimentation to see which is most attractive in combination with your electrolytic cap for getting the clarity that comes from leveling the treble response.
Nichicon ES, Panasonic FC and other high efficiency caps could be on the shortlist of things to try. You can do a search for favorites to see what other people use. The idea of optimal differs per each application, so I'd try 5 favorites to see what you like. Price does not indicate performance, but datasheets can be helpful. Polar caps work fine in split rail amplifiers.
Midrange, forward, recessed or level: For Non-Inverting LM3886 amplifiers, it might be good to choose amplifier board power caps with large enough capacitance to help support a level midrange frequency response prior to attempting the selection of any signal caps. The Panasonic FC 1500uF that AudioSector uses for power caps of the amplifier board, might be a good, first place to start. Good power must come first.
Bass harmonic balance and clarity: The size of the NFB cap versus the size of the Input cap affects the H1/H2 bass harmonic balance whereby too small Nfb cap makes warm/boomy/muddy bass, too large Nfb cap makes rumble bass monotony and somewhere in-between would have a balance with neither problem. The muddy bass problem is most commonplace and that muddy problem happens when the input cap passes more bass than the NFB cap. So, the simple solution for clearer bass would be to have the Input cap smaller and/or the NFB cap larger.
AndrewT probably explained this better than I could, so long as your calculator is up for it.
Treble Bypass cap: For leveling out the treble, those inexpensive little green polyester dip caps from Tracon/Xicon are very easy to use in parallel to an electrolytic cap. Applicable size range are from approximately 1nF to 150nF and it takes some experimentation to see which is most attractive in combination with your electrolytic cap for getting the clarity that comes from leveling the treble response.
Last edited:
Thank you all for your helpful & detailed explanation.
In order to summarize this, I have to do the following modifications.
One more question, why the schematic indicates non-polar caps at NFB?
In order to summarize this, I have to do the following modifications.
- I have 96.000uf of capacitance at PSU +/-35V ----- I guess no problem with midrange.
- Changing the input filter from 1k & 1.5n (Fc=106hz, 1.500msec) to 9k1 & 10uf,(Fc=1.7hz, 91msec) -----will this affect the input sensitivity of the amp?
- Replace the feedback resistor of 1k with a 687 ohm, in order to increase the gain from 20,6 to 29,5 approx.
- Replace the NFB capacitor with a 220uF = 151msec. ---- Is it Ok to try some non-polar MKT caps I already have in hand?
- What about the tweeter amps, do I have to follow the above example or it is ok to just increase the NFB capacitance from 10 to 22uf.
- Adding bypass caps on tweeter channel.
One more question, why the schematic indicates non-polar caps at NFB?
if the NFB cap has significant impedance at audio frequencies then it has an effect on the audio signal.
If you use a capacitor in the NFB that sees an actual audio signal then you must use a good capacitor. Some form of plastic film cap.
BUT THIS IS NOT the way to operate a Power Amplifier. It is not an active filter.
The NFB cap should not have a significant impedance in any part of the audio band.
If you make the NFB cap too small then you are asking the amp to operate as a filter. DON'T.
Use a cap that is big enough such that the passive filters at the input ensure that a significant voltage can NEVER build up across the NFB cap. If there is no voltage across the NFB cap, then there can never be any distortion due to the capacitor.
If you use a capacitor in the NFB that sees an actual audio signal then you must use a good capacitor. Some form of plastic film cap.
BUT THIS IS NOT the way to operate a Power Amplifier. It is not an active filter.
The NFB cap should not have a significant impedance in any part of the audio band.
If you make the NFB cap too small then you are asking the amp to operate as a filter. DON'T.
Use a cap that is big enough such that the passive filters at the input ensure that a significant voltage can NEVER build up across the NFB cap. If there is no voltage across the NFB cap, then there can never be any distortion due to the capacitor.
Very quickly...
1. 96000uf capacitance. 🙂 You could reduce that by a factor of... oooh lets not go there... 30 😀
2. 1K and 1n5 to 9K1 and 10uf will kill most of the audio. Also the 9K1 will reduce input sensitivity as it forms a divider with the 19.6k. It's reducing the level by nearly half.
3. Feedback resistor change is fine to alter the basic gain.
4. If you have non polar caps of that value then its fine to try. Electroylitics aren't all bad though 🙂
5. As I think I mentioned earlier increasing the 10uf feedback caps to 22uf just preserves the same low frequency roll off point due to the new value of fedback resistor.
6. Adding bypass caps.... not sure quite what and where you mean.
Why non polarised shown. Maybe just the way the "artist" drew it. I do that sometimes. If he really meant to use a non polarised its because he believes it may be sonically better. That's up to you to decide and evaluate. I would have no issue with electros when used correctly and correctly specified.
I wasn't talking about power board, but that is more than sufficient.
However. . .
I was actually talking about the amplifier board, such as your schematic shows.
The photo is from Linkwitz Lab. Do you see what's missing? The amp schematic shows the total capacitance on the amplifier board is a pair of 100nF??? There's a problem.
Let's fix.
An externally hosted image should be here but it was not working when we last tested it.
The photo is from Audiosector. See the two big blue cylinders on the amplifier board? You want them for your amplifier board.
In document AN1192, National Semiconductor promotes 470uF caps for the Power Caps Of The Amplifier Board; however, larger caps (like 1500uF) will have a more laid back presentation that is usually more useful.
Ah. That might be confusing. When I showed the AudioSector photo with the stereo version of the LM3886 chip, it does show exactly the Panasonic caps that you want. That was the good news.
Unfortunately, that board shows a bus ground that does not give easy stability. I don't wish to promote a bus ground scheme for LM3886. Please ignore the bus ground portion of the photograph. Thanks.
Unfortunately, that board shows a bus ground that does not give easy stability. I don't wish to promote a bus ground scheme for LM3886. Please ignore the bus ground portion of the photograph. Thanks.
This is very good. It works very well in practice. Thank you sir.
I have enjoyed years of good clean high resolution audio thanks to you.
Good question.
Amplifier design based on omission increases caveats.
Shorting/omitting the NFB cap will decrease the timeframe of amplifier fine tuning by not having to select the correct cap, decrease speaker safety, decrease chip reliability and decrease dynamics. And you would probably want to purchase a speaker protector circuit kit since typical chip amp failure mode is to emit one full rail directly into the speaker, potentially setting it ablaze.
I think that the negative consequences aren't worth the convenience of avoiding the cap selection.
P.S.
The majority of kits that promote NFB cap omission have listed a 47uF NFB cap value which is incorrect, a daft bass blocker, you hear the remaining signal as louder mids and thus omission would sound better than wrong component values. Please don't be misled. Better fix is to use correct component values.
Last edited:
The cap reduces the gain of the LM3886 to unity at DC. Without it and the circuit will "amplify" its own DC offset errors and cause much higher offsets.
I wonder if this is deliberate "salesmanship".
i.e. by allowing the purchaser to compare a "bad design" using the wrong capacitor value to a "good design" using no capacitor.
This way the salesman appears to be justifying the designer's choice of compromise: To DC block or not.
It is an honest error evidently caused by a bus ground scheme that interferes with frequency response purity.
To fix it, we'd have to add a signal star ground via groundlift resistor, conveniently elevating small signal groundpoint up off of the troubled little board.
To fix it, we'd have to add a signal star ground via groundlift resistor, conveniently elevating small signal groundpoint up off of the troubled little board.
Consider the case of LM1875 datasheet (split supply diagram), We don't have any low pass input filter here (offcourse, we can create one by inserting a capacitor || to R1(1M)). But we have a high pass filter(C1(2.2uf) and R2(22k)).
1.Do we consider/compare the RC time constant of this high pass filter capacitor(2.2uf) with RC time constant of NFB capacitor, in case we don't have a low pass filter at the input, to calculate the value of the NFB capacitor?
2.And, What if we have both high/low pass filters at the input stage?
3.What if we have none of them?
Last edited:
No. This thread is about LM3886. Therefore, please consider the case of the LM3886 datasheet.
Define the audio band at the input of the amplifier.
And you did it right.
Yes, we need both DC filter cap (input cap) and RF filter cap.
Last edited:
So with LM1875, we do not find the value of NFB capacitor as we do it with LM3886 by comparing RC time constants? The above rule (as explained by Mr.AndrewT) to find the NFB capacitor value, does not holds true for LM1875 or any other chip amp except for LM3886? RC time constant comparison between input filter capacitor and NFB capacitor is invalid for LM1875?
If it is so then how do we find the suitable value of NFB Capacitor for LM1875, for a particular configuration of the chip components?
Thanks.
Last edited:
- Status
- Not open for further replies.