it provides a load at very high frequency for the amplifier to operate into if the speaker becomes inductive at these high frequencies or if the speaker is disconnected.
The Zobel, or better still the Thiele Network, is put in there to help prevent the amplifier from blowing up.
The Zobel, or better still the Thiele Network, is put in there to help prevent the amplifier from blowing up.
The zobel in the National Semiconductor datasheet assumes a cheap high-esr type polyester, which is customary for such filters. The kit probably provides higher cost polypro. . . low esr polypro, thus causing erroneous resistance value in the output RC.
The capacitor that National Semiconductor assumes, can be found at the nearby Radio Shack. You can use the little green polyester cap of 0.22uF or the 0.1uF will have a slightly weaker, yet still functional, effect.
OK, when you put it that way, I think I'll leave it alone. I found posts by Peter Daniel suggesting that his kits sound better without the Zobel, but there was also a lot of feedback saying that it shouldn't matter.
Daniel... Glad to see your post. Can you comment on my post #59? I quoted a post you made in 2007 when another guy was complaining that his 3886 amp was emphasizing the high mids, including vocals. You thought that a bypass cap would address that problem by adding more highs. I think that's what I need to do, but I'm not sure how. Am I understanding that I would simply take a film cap and install it in parallel with the input cap, which is now 1uF (but I will probably change to something bigger, like 4.7uF). I think you mentioned a 4.7nF cap, and then trying larger ones. What would examples of reasonable larger ones to try be?
The kit includes little grey cube 0.1uF 63v BC polypropylene caps for the Zobels. Are you saying this cap is affecting my sound for the worse?
I feel that the amp sounds pretty good, but there is a lack of high end. It's as if there were a treble knob that someone turned down.
Daniel... Glad to see your post. Can you comment on my post #59? I quoted a post you made in 2007 when another guy was complaining that his 3886 amp was emphasizing the high mids, including vocals. You thought that a bypass cap would address that problem by adding more highs. I think that's what I need to do, but I'm not sure how. Am I understanding that I would simply take a film cap and install it in parallel with the input cap, which is now 1uF (but I will probably change to something bigger, like 4.7uF). I think you mentioned a 4.7nF cap, and then trying larger ones. What would examples of reasonable larger ones to try be?
The kit includes little grey cube 0.1uF 63v BC polypropylene caps for the Zobels. Are you saying this cap is affecting my sound for the worse?
I feel that the amp sounds pretty good, but there is a lack of high end. It's as if there were a treble knob that someone turned down.
Last edited:
I admit that I have lost track how your setup actually is by now. Is Ci installed and the output offset in the range of a few mA? If it is not that should be the first step.
100µF per rail as smoothing capacitance is definitely too small. Even the Gaincard uses ten times that and is already low in capacitance.
Increasing Cin by adding smaller caps in parallel won't solve your problem. It may fine tune the sound to a certain taste, but won't make a significant difference.
The 1k blocking resistor that Mick Feuerbacher chooses to skip has a protective function which is described in the datasheet.
A Zobel circuit needs low inductance. That makes X7R or better ceramic caps first choice and film caps second. The sonic difference is very small, if there is one at all.
100µF per rail as smoothing capacitance is definitely too small. Even the Gaincard uses ten times that and is already low in capacitance.
Increasing Cin by adding smaller caps in parallel won't solve your problem. It may fine tune the sound to a certain taste, but won't make a significant difference.
The 1k blocking resistor that Mick Feuerbacher chooses to skip has a protective function which is described in the datasheet.
A Zobel circuit needs low inductance. That makes X7R or better ceramic caps first choice and film caps second. The sonic difference is very small, if there is one at all.
Sounds like there is no reason to find a cermic cap if that will not make any difference in the sound.
I messed up the quoting function, so I put my replies in red...
Thank you for your thoughtful replies. Do YOU have a theory why the sonic deficit is there. Another way to describe it: If I had an old hissy recording, I don't think I would hear the hiss, but the detail is lost in that range.
Is this a signal path thing? A wire thing? A layout thing? Has anyone else built the chipamp kit and had good crisp highs?
Last edited:
Bigger smoothing capacitance makes for a stable power supply. Too small capacitance may lead to audibe hum and may affect the sound as the voltage drops under load and the amplifier cannot reproduce impulses. I would expect clipping or the protection system to kick in rather than a lack of highs, but it is worth a try to increase the capacitance.
Did you check the speaker polarities? If they are out of phase the spatial reproduction becomes diffuse and does appear undefined. The highs will then also appear differetn. Use a test disc with pink noise or download pink noise from the internet. If both speakers are in phase, the noise should appear right in front of you between the speakers. If they are out of phase, the noise will appear not to be in a defined spot.
Did you check the speaker polarities? If they are out of phase the spatial reproduction becomes diffuse and does appear undefined. The highs will then also appear differetn. Use a test disc with pink noise or download pink noise from the internet. If both speakers are in phase, the noise should appear right in front of you between the speakers. If they are out of phase, the noise will appear not to be in a defined spot.
The chipamp.com kit may be useful for the auctioneer or a fiercely loud center channel speaker. However, the good news is that the Audiosector LM4680 (internally, exactly, 2 LM3886's) design will fit for a more musical amp. Let's do that. I'm not going to re-invent the wheel today--maybe tomorrow or next week, but not today. 🙂
First, don't put a cap in between the pot and amp because that will cause DC offset problems. Locate the input coupling cap at the RCA jacks.
The size of the input cap is approximately 4.7uF and I suggest that you use Audiosector's recommend model of cap for this. You may substitute an Elna Cerafine 4.7uF if you also parallel it with a tiny value cap for extended bandwidth to get your crisp highs.
Second use either the National Semiconductor specified type of cap for speaker output RC (speaker zobel) or increase the resistor value. As indicated by the resistor value, the wrong type of cap is a low ESR ceramic or polypropylene. The right type of cap is the little green High ESR polyester 0.1uF located in the drawer of the nearby Radio Shack.
The power circuit contains no appropriate support for anything except for extra low bass, radio interference, and super loud vocals. I suggest that you change the 100uF caps from V+ and V- (leave the soft start cap in place), and replace the (four of) 100uF power smoothing caps with (four of) the tall skinny version of Panasonic FC 1500uF specified in the AudioSector support forum and used on the LM4780 kit.
Its fine to leave the 10,000uF caps on the power supply board.
However, Audiosector design doesn't use a NFB cap because its either a whopping huge 1000uF that won't fit the board, a DC servo that won't fit the board, or omission.
Omission of the NFB cap will better level the frequency response on LM3875, LM3886, and LM4780, but there are caveats.
You need to be able to measure 10k, 11k, or 12k (anything in that range) from Signal in to ground at each amplifier board or else DC offset will be too much. This is why we relocated your input coupling cap to the RCA jack.
You need to add a speaker protector kit.
After these steps, all of the data and knowlege and examples for the Audiosector kits will now work on your chipamp.com kit, and you should get the same results.
Input impedance:
For audiosector design, input impedance is at approximately line level spec.
Regardless, there are many sources that benefit from a. . . Jfet Unity Stable Op Amp employed as a Current Buffer.
Some inverting designs actually require a buffer.
The Jfet buffer will allow you to re-locate the volume control to the buffer, so that you may keep a steady impedance at the amplifier, and you can also use whatever impedance is required by the source device (just purchase that value of potentiometer). You can "please" both source and amp this way.
Output impedance:
For 3 way speakers, the resistor in series to midrange and tweeter sets this.
For 2 way speakers the resistor across the inductor at the BSC circuit and the resistor in series with the tweeter sets this.
For 1 way, full range speakers a resistor in series with the speaker sets this (resulting in tube amp type of drive for the bargain price of sixty cents).
I think that you have a combination problem:
1). Super loud mids because of the insufficient 100uF power caps. Those should be at least 470uF if not 1500uF.
2). Wrong resistance value for output RC has created a treble muffler (so swap out that box cap for a dip cap).
3). Its possible that the LM3886 Spike circuit is sounding off with a bit of extra distortion further increasing the upper midrange and likewise further throwing the frequency response; and the most likely culprit for that (and extra heat) is ground loops or improper grounding of some sort. A 220pF cap between the chip's in+ and in- can brute force some extra stability and help decrease occurrence of LM3886's inbuilt noisemaker sounding off.
Fixing the chipamp design was plan A.
Plan B is to disconnect the circuit boards, and use Decibel Dungeon's lovely sounding inverted designs (also flips the amp's frequency response for laid back upper mids), or a myref design or a current pump design, or a nested topology (which have level responses).
You also have the option to bi-amp with LM3886 on bass and LM1875+regs on treble. LM1875's don't have inbuilt noisemakers, and therefore don't have any treble distortion.
First, don't put a cap in between the pot and amp because that will cause DC offset problems. Locate the input coupling cap at the RCA jacks.
The size of the input cap is approximately 4.7uF and I suggest that you use Audiosector's recommend model of cap for this. You may substitute an Elna Cerafine 4.7uF if you also parallel it with a tiny value cap for extended bandwidth to get your crisp highs.
Second use either the National Semiconductor specified type of cap for speaker output RC (speaker zobel) or increase the resistor value. As indicated by the resistor value, the wrong type of cap is a low ESR ceramic or polypropylene. The right type of cap is the little green High ESR polyester 0.1uF located in the drawer of the nearby Radio Shack.
The power circuit contains no appropriate support for anything except for extra low bass, radio interference, and super loud vocals. I suggest that you change the 100uF caps from V+ and V- (leave the soft start cap in place), and replace the (four of) 100uF power smoothing caps with (four of) the tall skinny version of Panasonic FC 1500uF specified in the AudioSector support forum and used on the LM4780 kit.
Its fine to leave the 10,000uF caps on the power supply board.
However, Audiosector design doesn't use a NFB cap because its either a whopping huge 1000uF that won't fit the board, a DC servo that won't fit the board, or omission.
Omission of the NFB cap will better level the frequency response on LM3875, LM3886, and LM4780, but there are caveats.
You need to be able to measure 10k, 11k, or 12k (anything in that range) from Signal in to ground at each amplifier board or else DC offset will be too much. This is why we relocated your input coupling cap to the RCA jack.
You need to add a speaker protector kit.
After these steps, all of the data and knowlege and examples for the Audiosector kits will now work on your chipamp.com kit, and you should get the same results.
Input impedance:
For audiosector design, input impedance is at approximately line level spec.
Regardless, there are many sources that benefit from a. . . Jfet Unity Stable Op Amp employed as a Current Buffer.
Some inverting designs actually require a buffer.
The Jfet buffer will allow you to re-locate the volume control to the buffer, so that you may keep a steady impedance at the amplifier, and you can also use whatever impedance is required by the source device (just purchase that value of potentiometer). You can "please" both source and amp this way.
Output impedance:
For 3 way speakers, the resistor in series to midrange and tweeter sets this.
For 2 way speakers the resistor across the inductor at the BSC circuit and the resistor in series with the tweeter sets this.
For 1 way, full range speakers a resistor in series with the speaker sets this (resulting in tube amp type of drive for the bargain price of sixty cents).
I think that you have a combination problem:
1). Super loud mids because of the insufficient 100uF power caps. Those should be at least 470uF if not 1500uF.
2). Wrong resistance value for output RC has created a treble muffler (so swap out that box cap for a dip cap).
3). Its possible that the LM3886 Spike circuit is sounding off with a bit of extra distortion further increasing the upper midrange and likewise further throwing the frequency response; and the most likely culprit for that (and extra heat) is ground loops or improper grounding of some sort. A 220pF cap between the chip's in+ and in- can brute force some extra stability and help decrease occurrence of LM3886's inbuilt noisemaker sounding off.
Fixing the chipamp design was plan A.
Plan B is to disconnect the circuit boards, and use Decibel Dungeon's lovely sounding inverted designs (also flips the amp's frequency response for laid back upper mids), or a myref design or a current pump design, or a nested topology (which have level responses).
You also have the option to bi-amp with LM3886 on bass and LM1875+regs on treble. LM1875's don't have inbuilt noisemakers, and therefore don't have any treble distortion.
Basically, use the Audiosector design for a guideline. It contains "dodges" to get around the frequency response caveats of the noisy Spike Circuit built into LM3875, LM3886, LM4780.
The above suggestion has absolutely nothing to do with appropriate design practices, but everything to do with achieving an appropriate output without having to replace your circuit boards.
After getting the thing to output usefully, then you can attempt to install an NFB cap to see if you enjoy its dynamic enhancement, amplifier longevity enhancement, and speaker protection advantages.
The above suggestion has absolutely nothing to do with appropriate design practices, but everything to do with achieving an appropriate output without having to replace your circuit boards.
After getting the thing to output usefully, then you can attempt to install an NFB cap to see if you enjoy its dynamic enhancement, amplifier longevity enhancement, and speaker protection advantages.
The brightest brashest loudest bypass cap is made by AVX. Its their little yellow box caps with the T on the label. You'll definitely hear more treble. AVX is overkill, but if that's what you need, then go for it. As usual, pitch of the boost is according to the size of the cap, and also according to variety of/in the main coupling capacitor.
Plan B (and more sensible).
1) Go to the radio shack and purchase two of the little green "dip" (not box) high esr (cheap) polyester 0.1uF caps for speaker zobel. Its the easy and expected way to get the series resistance up where it should be.
2) Relocate your input coupling cap to the RCA jack.
3) Recycle your "little grey cube 0.1uF 63v BC polypropylene" as signal coupling bypass caps by placing them in parallel with your input coupling caps, and just see if that works (to boost the treble efficiency of the main signal coupling caps).
You can also try the 0.022uF from the Radio Shack for a tamer option.
Another possibility is 0.47uF Nichicon Muse ES // 4.7uF Elna Cerafine. The tiny Nichicon is bipolar, meaning that it will have more treble or less treble depending on orientation, so that's very useful. I use it to patch computer sound card output.
4) Increase the 100uF power cap values up much higher until midbass is in proportion to upper midrange. Since you need good clarity, use other people's well working examples rather than purchasing a random part.
5) Cause a short to radio frequencies by placing 100pF or 220pF ceramic caps in parallel to the contacts of the RCA jacks (as a load), because amplifying noise is unnecessary work for the amplifier and because that little capacitive load will subtract noise from input cables.
Question:
What value of NFB cap will provide flat response down to 20hz when the cap's series resistor is 680R, and how do you recommend fitting the correct value cap into the space provided on that tiny little board, or would you recommend an off-board DC Servo instead?
Question:
What value gain divider resistors will work well for flat response down to 20hz with the 47uF NFB cap provided in the kit, what value replacement input load resistor and replacement potentiometer are then necessary, and in this case would you recommend the addition of a current buffer to the amplifier input?
Question:
What value power caps would you put at the amplifier board, and would you make any other alterations/upgrades to the power circuit?
I'll bite the bait. But I will regret this.
For F-3dB @ 20Hz, apply the formula.
For DC servo see all the other threads that discuss the pros and cons.
Gain divider resistors? Where's the schematic?
Power caps @ the amplifier board? What?
Daniel- Thank you for all of your input. Now I'm trying to figure out what Andrew meant:
Seems harsh. I will be most interested in your responses, Andrew, if you have time to give them. I tried your 3 suggestions including the tiny 47 pF cap across the RCAs, the 330 pF polypropylene cap across R2, and a 1 uF (you suggested 3) film cap between the pot and R1. None of those seemed to fix the problem.
Thanks to all for reading and thinking about this. I'm learning a lot.
Seems harsh. I will be most interested in your responses, Andrew, if you have time to give them. I tried your 3 suggestions including the tiny 47 pF cap across the RCAs, the 330 pF polypropylene cap across R2, and a 1 uF (you suggested 3) film cap between the pot and R1. None of those seemed to fix the problem.
Thanks to all for reading and thinking about this. I'm learning a lot.
I guess I don't understand the thing about bypass capacitors. Doesn't adding another tiny capacitor in parallel with the AC coupling cap just decrease the cutoff frequency a little bit? Why would it enhance the highs?
Please repair treble muffler first.
The National Semiconductor Datasheet value for output zobel assumes that the customary part will be used, adding 3.5 ohms resistance--the internal resistance of THIS cap:
Its Radio Shack part # 272-1069 and similar caps are available world wide. The cap pictured is the normal and expected cap for use in output zobel. It is high ESR polyester (low efficiency).
Otherwise, an overly aggressive output zobel IS a treble muffler.
Capacitors have differing efficiency peaks, inductance, and internal resistance inside. Polypropylene (that came with your kit) will have high efficiency, low resistance and is NOT suitable for teaming up with a 2.7 ohm resistor for output zobel, because the resulting harsh load is about out of specs for the amplifier.
If using the datasheet values AND staying within suitable loads for the amplifier, then use the low efficiency cap pictured to repair your output zobel.
The National Semiconductor Datasheet value for output zobel assumes that the customary part will be used, adding 3.5 ohms resistance--the internal resistance of THIS cap:
An externally hosted image should be here but it was not working when we last tested it.
Its Radio Shack part # 272-1069 and similar caps are available world wide. The cap pictured is the normal and expected cap for use in output zobel. It is high ESR polyester (low efficiency).
Otherwise, an overly aggressive output zobel IS a treble muffler.
Capacitors have differing efficiency peaks, inductance, and internal resistance inside. Polypropylene (that came with your kit) will have high efficiency, low resistance and is NOT suitable for teaming up with a 2.7 ohm resistor for output zobel, because the resulting harsh load is about out of specs for the amplifier.
If using the datasheet values AND staying within suitable loads for the amplifier, then use the low efficiency cap pictured to repair your output zobel.
..but is a valid question. After reading half of those 65 lines I quit reading, because until that point, there was not a single thing correct.
The idea behind using smaller (high quality) caps in parallel with a high value (low to medium quality) cap is to achieve the good sound of the high quality cap and the high capacitance and relatively low price of the bigger cap at the same time. In your case that is however taking the second step before the first.
Schematic is given at chipamp.com.
http://www.chipamp.com/docs/lm3886-manual.pdf
Gain divider is 680R with 22k, so right size NFB cap doesn't fit on tiny board.
Input load is 20k pot directly parallel 22k resistor, unless broken by inserting a capacitor in-between, with the consequences of high DC offset.
Power smoothing caps at amplifier board are an insufficient 100uF, causing an unpleasant frequency response during amplifier output.
Amplifier output Zobel resistor value is 2.7 ohms, assuming the use of a high ESR cap that was NOT provided in the kit and so assembly results in wrong RC value that is a treble muffler.
It seems that you gave me a grade of "F-" earlier, but since the chipamp.com design is even worse, what grade do they get?
What is the first step? If the first step is installing a DC blocking cap, I have done that.
What about this idea of changing the capacitor in the Zobel/Thiele network? Do you, Pacificblue or Andrew, buy that the Zobel as configured is blocking my treble?
Thank you all for the consideration and debate.
Last edited:
Oscillation for LM3886 is reduced by a tiny value cap from inverting to non-inverting input of the chip, soldered directly upon the chip's pins (under the board), so see the National Semiconductor Datasheet. 330pF works very well, but is contra-indicated by the question in the first post. 220pF should be a good answer. However, 100pF will still yet work. Its sure better than omission. However. . .
Oscillation is one of many probable causes of an unlevel frequency response in an audio amplifier. And, common LM3886 kits are prone to oscillation.
Here's a practical blurb from a support forum concerning LM4780 (the dual LM3886 chip): http://www.diyaudio.com/forums/audi...kit-building-instructions-22.html#post2150004
The pinouts are different because LM4680 has two LM3886 inside. However, the capacitor values and preventing oscillation is applicable to your LM3886 kit.
Oscillation is one of many probable causes of an unlevel frequency response in an audio amplifier. And, common LM3886 kits are prone to oscillation.
Here's a practical blurb from a support forum concerning LM4780 (the dual LM3886 chip): http://www.diyaudio.com/forums/audi...kit-building-instructions-22.html#post2150004
The pinouts are different because LM4680 has two LM3886 inside. However, the capacitor values and preventing oscillation is applicable to your LM3886 kit.
Last edited:
- Status
- Not open for further replies.