Hi,
In 2022, I've built an inverted LM3886 kit from aliexpress. It had stability issues and fried resistors in Zobel network before but I've resolved it (it was the missing feedback loop capacitor C3,C4). I noticed that absolute value of the gain/ feedback capacitor is a little high (220k/ 7.5k) but I left them as is.
I haven't use it much until recently I've moved to my new place and got a new pair of speaker (Wharfedale Diamond 12.1) and use the set as desktop setup. The amp is quiet and perform well, but occasionally, it distorts at tweeter of the speakers, it sounds like "scratchy" on some notes. It is louder on right channel than left channel. The problems goes away when I turn it off and turn it on again. This problem is so random, but if it exhibits the distortion on some part of the song, I can consistently try to play that part of the song to repeat the problem (I have a list of problematic songs).
Please suggest what procedure should I taken to diagnose this. I don't have oscilloscope but plan to get a handheld one soon. Here's the schematic of the amp.
Thank you,
AP
In 2022, I've built an inverted LM3886 kit from aliexpress. It had stability issues and fried resistors in Zobel network before but I've resolved it (it was the missing feedback loop capacitor C3,C4). I noticed that absolute value of the gain/ feedback capacitor is a little high (220k/ 7.5k) but I left them as is.
I haven't use it much until recently I've moved to my new place and got a new pair of speaker (Wharfedale Diamond 12.1) and use the set as desktop setup. The amp is quiet and perform well, but occasionally, it distorts at tweeter of the speakers, it sounds like "scratchy" on some notes. It is louder on right channel than left channel. The problems goes away when I turn it off and turn it on again. This problem is so random, but if it exhibits the distortion on some part of the song, I can consistently try to play that part of the song to repeat the problem (I have a list of problematic songs).
Please suggest what procedure should I taken to diagnose this. I don't have oscilloscope but plan to get a handheld one soon. Here's the schematic of the amp.
Thank you,
AP
If it burned up zobel before from hard oscillating
It might have been slightly resolved and more stable but certain passages
in music will trigger ringing oscillating.
assuming you have reviewed speakers with another amp and dont have
tweeter issues / distortion
unless of course crossover issue presenting very low impedance.
gain seems high and with inverting needs resistor on non inverting pin to balance DC
offset on output. Do you have basic meter to see if amplifier speaker outputs has high DC offset.
cannot read values on schematic well.
used google lens and found same pictures.
other wild guess is chips still work but damaged . drifted from original
heating / oscillating issue.
It might have been slightly resolved and more stable but certain passages
in music will trigger ringing oscillating.
assuming you have reviewed speakers with another amp and dont have
tweeter issues / distortion
unless of course crossover issue presenting very low impedance.
gain seems high and with inverting needs resistor on non inverting pin to balance DC
offset on output. Do you have basic meter to see if amplifier speaker outputs has high DC offset.
cannot read values on schematic well.
used google lens and found same pictures.
other wild guess is chips still work but damaged . drifted from original
heating / oscillating issue.
I had similar problem with fake LM1875 , replaced with another in the faulty ch. and the problem was gone.
High oscillations was also observed in my Genuine LM 3886/ 4870 build , which I got sorted with ceramic caps as shown in pic
Those "optional" 220 pF between inv & non inv. and 50 pF + 20K (220K in your case) in feedback are actually "must", read datasheets.
Higher value feedback resistors causes more Johnson noise but not crazy oscillations, but I stick to 1k/18K combo, else the vaue of Cf will change
Don't connect your Wharfedales, else your tweeters will get fried. Use some elcheapo speakers to check. U can also use series power resistors like 10 - 47R with speakers to minimise damage during initial tests.
https://www.diyaudio.com/community/...site-amplifier-achieving-0-0004-thd-n.262273/
https://www.circuitbasics.com/design-hi-fi-audio-amplifier-lm3886/
High oscillations was also observed in my Genuine LM 3886/ 4870 build , which I got sorted with ceramic caps as shown in pic
Those "optional" 220 pF between inv & non inv. and 50 pF + 20K (220K in your case) in feedback are actually "must", read datasheets.
Higher value feedback resistors causes more Johnson noise but not crazy oscillations, but I stick to 1k/18K combo, else the vaue of Cf will change
Don't connect your Wharfedales, else your tweeters will get fried. Use some elcheapo speakers to check. U can also use series power resistors like 10 - 47R with speakers to minimise damage during initial tests.
https://www.diyaudio.com/community/...site-amplifier-achieving-0-0004-thd-n.262273/
https://www.circuitbasics.com/design-hi-fi-audio-amplifier-lm3886/
Attachments
Last edited:
I don't know if you're aware of this but given its running an inverted configuration, don't disconnect the inputs with it powered up or it'll likely go unstable.
Thanks everyone!
1. Speakers work fine with other amps. With this amp, they work fine most of the time. Nominal impedance is 8 ohms but can dip down to 4 ohms (which should be fine with 3886)
2. Agree that the RC combo is necessary. I'll try to find some test power resistor.
3. I was thinking about the relay too because I only heard the distortion in R channel until I carefully test it again. I haven't inspect the relay yet.
4. I guess for ringing oscillation, I could test with 1kHz square wave signal. One channel handheld scope should be enough? Do I need any specific feature later (like FFT)?
5. When it had original oscillation problem, it oscillated both with and without inputs. I guess I can put 47k resistors across the input and ground to add more stability.
6. I was suggested to adjust the gain ratio/feedback resistors as well as move the 0.1uF decoupling C closer to the chip, like under the PCB, but haven't go ground to do that yet.
I'm thinking to check with scope to confirm the problem before mods. I'm trying to learn how to do this.
AP
1. Speakers work fine with other amps. With this amp, they work fine most of the time. Nominal impedance is 8 ohms but can dip down to 4 ohms (which should be fine with 3886)
2. Agree that the RC combo is necessary. I'll try to find some test power resistor.
3. I was thinking about the relay too because I only heard the distortion in R channel until I carefully test it again. I haven't inspect the relay yet.
4. I guess for ringing oscillation, I could test with 1kHz square wave signal. One channel handheld scope should be enough? Do I need any specific feature later (like FFT)?
5. When it had original oscillation problem, it oscillated both with and without inputs. I guess I can put 47k resistors across the input and ground to add more stability.
6. I was suggested to adjust the gain ratio/feedback resistors as well as move the 0.1uF decoupling C closer to the chip, like under the PCB, but haven't go ground to do that yet.
I'm thinking to check with scope to confirm the problem before mods. I'm trying to learn how to do this.
AP
In which case the nature of the source could affect oscillation - or even inductance of the cables from source to amp?
Thanks all. Sorry for late reply. I've just got an oscilloscope. I think the picture says it all.
Note that I'm not using function generator but an app connected through DAC's bluetooth. This is measured at connected speakers. Sine wave is fine.
Note that I'm not using function generator but an app connected through DAC's bluetooth. This is measured at connected speakers. Sine wave is fine.
You are right. I checked output and it show similar wave form. I changed the source to oscilloscope function generator, it doesn't ring.
(Yellow is input, magenta is output)
I guess I have to try to get it to distort again before next measurement. Or how should I test?
(Yellow is input, magenta is output)
I guess I have to try to get it to distort again before next measurement. Or how should I test?
I meant the it was ringing at input and I've changed the source so it doesn't ring any more. I'm not sure if I set the scope correctly.
1. I connected probe (ground and test pin) directly to binding posts and parallel to the test speakers.
2. Tried test signal at 1kHz Square.
Also, I tried lifting the earth prong and amp hum. Tried connected the scope but it is hard to display the hum. However, if I connected only test lead without ground, the scope pick up exaggerated hum signal (by induction?).
Any idea should I test this? I want to verify the problem before fixing it. I'm unable to reliably re create problem yet.
Thanks.
AP
1. I connected probe (ground and test pin) directly to binding posts and parallel to the test speakers.
2. Tried test signal at 1kHz Square.
Also, I tried lifting the earth prong and amp hum. Tried connected the scope but it is hard to display the hum. However, if I connected only test lead without ground, the scope pick up exaggerated hum signal (by induction?).
Any idea should I test this? I want to verify the problem before fixing it. I'm unable to reliably re create problem yet.
Thanks.
AP
I would suggest that both of the above are good measures to apply.
What a shame you've got an intermittent problem > they are the worst.
Thanks. I think I've caught it. It does look like mild oscillation at the peaks of sine wave. I'll try to connect the input resistors to see if the oscillation is gone. I only have a 100k and 22k around.
Also, I haven't yet found a reliable way to create the problem though. Last test was purely random. I guess I could try to drive the amp to clipping point as it may oscillate more, but I'm afraid it might damage the chip.
AP
Also, I haven't yet found a reliable way to create the problem though. Last test was purely random. I guess I could try to drive the amp to clipping point as it may oscillate more, but I'm afraid it might damage the chip.
AP
If the board is built according to the schematic, you really need to rework the decoupling network on the LM3886. You need a capacitor on the mute pin as well.
The chip is oscillating because it doesn't have the recommended decoupling capacitors at the power supply pins of the IC. See the LM3886 data sheet for TI's recommendations or below for mine.
You can read more here: https://neurochrome.com/pages/taming-the-lm3886-chip-amplifier
Tom
The chip is oscillating because it doesn't have the recommended decoupling capacitors at the power supply pins of the IC. See the LM3886 data sheet for TI's recommendations or below for mine.
You can read more here: https://neurochrome.com/pages/taming-the-lm3886-chip-amplifier
Tom
Finally, I have chance to do it. Adding 50k for each inputs and decoupling C at both rails. I have tested for 4-5 hrs and no problem so far.
Thanks you all for valuable recommendations.
AP
Thanks you all for valuable recommendations.
AP
Yup, those ceramic decouplers are required for RF decoupling, right on the pins like that is ideal.
Also the PCB should be cleaned of flux (immediately after soldering is best!) as old flux risks turning to carbon over time and starts to conduct, setting yourself up for a failure down the line. I find an old coarse toothbrush is a good tool for helping to remove old hardened flux. Most of those solder joints seem rather heavy in solder - solder surface should not be convex as that easily hides a dry-joint. However if its working leave it as is.
Also the PCB should be cleaned of flux (immediately after soldering is best!) as old flux risks turning to carbon over time and starts to conduct, setting yourself up for a failure down the line. I find an old coarse toothbrush is a good tool for helping to remove old hardened flux. Most of those solder joints seem rather heavy in solder - solder surface should not be convex as that easily hides a dry-joint. However if its working leave it as is.
I don't think flux forms carbon, but it does attract water, thereby becoming conductive over time. You can get some corrosive substances formed that way too so cleaning the flux off is definitely best.
The easiest way to clean the flux off is by using some flux remover. It's usually a mix of various solvents, including acetone and isopropyl alcohol. Work it in with a good stiff brush, rinse the board with more flux remover and then remove the residue with water (assuming the components on the board are washable). I then toss the board in an 80 ºC oven for 5-10 minutes to dry.
Tom
Just to report the result after weeks of use. I did get few occasion of the ringing oscillations. Much less than before (once every few hours before fix once every few weeks after fix). The oscillation is now doesn't need power cycle to fix but a large bass can also reset the amp's oscillation. I think I'll leave it as it is rather than goes down the rabbit hole.
For the flux residue, I did try to clean them by soaking with contact cleaner and toothbrush. Not much come off because it's harden, so I use flat head screw driver to carefully scrap them as much as I could without damaging the pcb coating. Cleaner but not that clean.
For the flux residue, I did try to clean them by soaking with contact cleaner and toothbrush. Not much come off because it's harden, so I use flat head screw driver to carefully scrap them as much as I could without damaging the pcb coating. Cleaner but not that clean.