Fixing the LM3886 stereo amplifier oscillation problem

[@listerine-pr5lt]

Thank you John for your excellent modifications.

[@martinda7446]

I think JohnAudioTech has given us some wonderful videos, and here is another. Excellent stuff. Re the single rail operation. Not sure if 'virtual ground' is the correct terminology, but anyhow, for any new guys all that is happening is the positive supply is connected as usual to the +ve pin and that supply is also connected to the collector of a bipolar transistor which is arranged as an emitter follower - the bias supply set as a voltage divider having half supply on the base (slightly unequal to account for the BE junction) this will in turn give a low impedance half supply to the chip ground. The -ve pin is taken to actual ground and that works fine. Quite often people use the basic layout as suggested by the manufacturer. It is so common to see exactly that 1000 times over. Usually it is fine that way too. More than likely if there are issues it is with board layout and not omission of parts, but there are no rules here.

[@cassvirgillo3395]

Hello John, I am impressed with your knowledge of chip amp circuit design, duh, that's why I'm a long time sub. I don't know of anyone better than you when it comes to chip amp builders and developers. You really know your electronics. Glad your doing well and I hope your Brother is doing well. Thanks for the great information about the LM3886. Happy last month of Summer, all the best from Oregon, C.

[@laurynaslauris9677]

chip amps are built from transistors too. So the knowledge is for all kind of amps.

[@quananginh9446]

It's the look that brings in the sales. Naturally, a board with two big caps would get the attention much more easily.

[@ahmedalshalchi]

I am highly respecting your usage of VFD calculator for your math work ....

[@MichaelBeeny]

It's so easy to make a PCB these days, the Chinese take manufactures data without knowing anything about layout or electronics at all. They don't realise this data is a building block for manufactures to expand on to make a quality amplifier. They often fit parts they have to hand and often fake parts. Not that this matters as none are inspected or even tested. They don't care if the product is safe or even likely to kill you. It's just all about money!! As for specifications...just think of a number and double it, and throw it on Ebay. If too many people complain, close the shop and start all over again with another name.

[@johnsonlam]

In most cases (not all), they copy the sample circuit from the data sheet and remove those extra components, (e.g. standby) then make the board without measuring or testing, they know a few component missing or value not correct usually didn't affect the operation, they definite need to send the boards to John (same cases for Siglent sent an oscilloscope to Dave Jones for test), a FREE ADVISER for them!

[@PeterMilanovski]

No one is going to give you the circuit that's already been perfected! If that were the case, you would only be able to get the datasheet accompanied with a license. Everything that you need to get this amplifier up and running is in the datasheet, if you read closely, you will also find the key areas where you need to pay special attention to make it work properly, but you are going to need to do some experiments and a lot of testing to get it to work correctly! Some Chinese person took a look at the datasheet and thought, oh wow they just gave us the complete amplifier! I just have to copy it and before you know it, it's on eBay! It takes a lot of work to get an amplifier circuit working correctly, that's why no one else was building them and palming them off on eBay, but the Chinese don't care if you destroyed your speakers! It's not like you can do anything about it! And that's why videos like this are extremely important! Supporting channels like this keeps the bastard's honest! They will keep making this junk until we stop buying it.

[@MichaelBeeny]

@@PeterMilanovski I agree with you 100% Peter, my channel is full of faulty amplifiers PCBs that claim the world and often offer little. But once in a while I find something that's really quite good, it can happen. Probably 1 in 5. The sad thing is with just a few minor changes can make all the difference. I guess we never learn, we keep buying. It is a very steep learning curve.

[@PeterMilanovski]

@@MichaelBeeny I just might have to give your channel a visit then... There's no such thing as a topic being repedative as long as the content provider can add at least one thing that someone else didn't, then that's good enough for me!

[@MichaelBeeny]

@@PeterMilanovski Thanks Peter, hope you find something of interest.

[@ugott1]

A comparison with the TDA7294 should be interesting.

[@lee26carl]

You're gifted, John...

[@mysock351C]

When I built one of these LM3886's I just used the standard non-inverting resistor network for the amplifier and the Thiele network on the output with attention to minimize capacitive coupling and track impedance. Since the closed loop bandwidth will extend beyond 100kHz with just resistors I added RF decoupling capacitors at the input for interference suppression. Im wondering if the issues with this primarily come from capacitive coupling in the board layout? That will definitely cause oscillation if there is coupling back to the input or feedback network. For yucks I just popped a function generator and a test load on the amp, and its stable with a square-wave up to many 100's of kHz, albeit with rather severe attenuation of the square wave at those frequencies.

[@mysock351C]

14:16 If the capacitor for DC compensation is included (100uF at the bottom of the feedback network on the schematic), its also a good idea to include an output relay with a delay of a second or two to prevent turn-on thump. While the input bias currents are charging that cap, the amplifier will be in saturation, so its a good idea to include a relay between the amplifier and the speakers to prevent turn-on thump and possible damage to the speakers. Alternatively if you wont ever be using a turn table with the amplifier, it (along with the relay) can be omitted. The resulting DC offset should only be around 25-50mV for typical run of the mill ICs, but worst-case DC offset calculations for your particular circuit would be a good idea.

[@JohnAudioTech]

They use a 100uf and 100k resistor in the mute circuit for delay. The time constant seems to be enough to eliminate the turn on pop.

[@mysock351C]

@@JohnAudioTech Ah! Missed that when I looked at it the first time.

[@SupremeRuleroftheWorld]

John, i have a hypex UCD400 V6 amp board. would you like to test that board? (i do want it back when you are done) it would be nice to see a really high end board getting the treatment.

[@pliedtka]

That would be awesome. Heard the older Uber version and I did like it, but I'm very interested in less expensive regular UCD400 or ICE amps for DIY multi channel amp.

[@mysock351C]

The 47p/22k compensation on the feedback will only cut the gain by about half at high frequencies since there is a resistor there, and that will probably be well after the internal compensation is already heading towards unity. Not a bad idea to have it, but its limited in its effectiveness since if you try to lower the series resistance, or omit the resistor altogether as you might do in an op-amp circuit, there will likely be enough stray inductance to cause it to oscillate at RF. I tried 1k to see if I could get a steeper roll-off of the gain, and the amplifier would oscillate at around 2-3MHz from stray inductance. Power supply decoupling, RF decoupling capacitors at the input and an inductor at the output seemed to give the best results, and also keep the AM out from one of CBS's AM broadcast stations that's directly behind my house.

[@Tegelane5]

I don't have a scope but I added those components to my point to point build as without those amps sounded grainy. Adding those helped, on TDA2030As I used only caps those muffled highs noticeably.

[@mysock351C]

@@Tegelane5 There's a trade-off between actually getting the gain down far enough to allow some immunity to oscillations and RF interference, and actually killing the gain in the audio band. If the impedance of the compensation network is low at audio frequencies, it will kill off the high end gain. You can use free spice simulators to see what the response of the amps will be, and help with selection of any external compensation. Another approach is to include RF decoupling capacitors at the input. These should be chosen to be about 10x the impedance of the input network to the amplifier at 20kHz. If your using something like a 20kOhm volume potentiometer, the impedance will be about 10kOhms. A suitable capacitor would be 47pF across the input to each channel of the amp IC. That will usually provide good RF rejection above about 200kHz, keeping noise out of the amplifier and combined with the 1uH inductor at the output will help prevent RF oscillations.

[@mysock351C]

@@Tegelane5 I used the simple RF decoupling caps at the input on my build, which also has a flyback switch mode power supply integral to the amplifier PCB I made, and that seems to do a good job of keeping the hash out of the audio when suitable star-point grounding is used.

[@lllllllllllllllllllll1lll1]

Those fakes are the norm these days with Aliexpress/eBay boards. One of the common, easily identifiable counterfeit slip-ups- improper voltage rating for the cap series. Seen many "Nichicon" power filter caps labeled 105C when the real ones are 85C. Whenever I look into prebuilt boards, I factor an immediate re-cap into the total cost. Just seen it so many times, and they're such vital parts to the overall function of the board.

[@muxallopeniot9194]

That's a shame about the terrible quality. At least you know why it rattles now John. Great video!

[@arnekarlsen528]

Most probably an old, used capasitor. I have found similar, even dryer ones... in new chinese gear...

[@joohop]

Well Done Buddy Bless Up

[@Helectronics]

Really good video, you are great at troubleshooting!

[@MrAletube]

that RL network on output was well known by 70s constructors,amazing that 50 years later there are people that ignore these tactics,chinese don't really know hot to copy something.

[@jeffmassey4860]

The SPIKe circuitry may have something to do with the notch on the negative sine cycle. Looking at the description,the protection circuitry seems to do a lot of various things...

[@linandy1]

Cool. I wonder how those can be configured to run two in bridge mode?

[@pliedtka]

Awesome video. John, they didn't bypass the preamps op-amps rails. Can you check what happens if you bypass them, 🤔 Maybe slapping some compensation in NFB of op-amps would help. We didn't even check what the signal looks like on the input of the 3886, maybe the damn thing oscillates even before. I would put larger coil, say 3uH at the output if it oscillates with 0.27uF. Too bad we don't have more time for playing with compensation components ☹️ to cure the ringing.

[@hummusdude]

Hey John... Thanks for awesome videos! I'm a recent EET grad and I find the level of discourse on your channel to be really helpful. I have one request... can you add prefixes to the cap values on your schematics? I'm sure it's obvious to you and that's probably why you don't bother. But to us noobs it can be challenging to infer if a cap is pico, nano, or microfarads without the prefix.

[@buddhistsympathizer1136]

A question, please. If I were to bridge two LM3886s together in order to produce a higher output into a higher impedance, would you suggest a theil network on both outputs . . . or just one. Thank you

[@JohnAudioTech]

It probably wouldn't matter, but I'd treat them as if they're not bridged.

[@buddhistsympathizer1136]

@@JohnAudioTech Thank you

[@e.michaelkessler6921]

Back in the mid-90's, you could buy all sorts of fake name brand clothes at street markets in Beijing. I never thought anyone would fake a name brand capacitor to fool an end user.

[@PeterMilanovski]

Oh how much times have not changed.

[@cuongnguyenquoc6356]

Great video, thank you!.

[@rogerwalter2500]

Are those feedback capacitor ceramic or polyster?

[@amitghosh6966]

Nice video again

[@maxhifi1232]

I think clipping behavior is characteristic of National Semiconductor's SPiKe protection, which limits current when the amp is overloaded. I bet if you used an 8 ohm load it would clip more normally. This is covered by a paper on the overture series of chips you can find online.

[@johnyang799]

It's not. They are not the same behaviour.

[@maxhifi1232]

@@johnyang799 I meant the clipping at the bottom of the waveform, after all oscillation and bad behavior is removed at the end of the video. I've seen other LM3886 clip the bottom of the waveform the same way, and assumed it must be the protection circuit activating

[@johnyang799]

@@maxhifi1232 I know what you mean. But that's not the behavior of SPiKe. It's power supply decoupling related. SPiKe behaves much more aggressively. It also activates at or after the voltage peak. It works by sensing the temperature of a internal resistor as a better equivalent of SOA protection.

[@martinda7446]

A quick observation. The ultrasonic nulling capacitor in the feedback circuit is rendered fairly useless by the 20k resistor. I think that is an error. The idea is to pretty much stop anything over 30-40khz. Many designs just stick the cap straight across the standard feedback resistor. It would certainly do no harm to do just that. (making sure the roll off was starting at over 25k) Edit though the spec sheet says I'm way off!!! Anyone going to do the math? I'm not gonna bother...

[@JohnAudioTech]

Bypassing the feedback resistor with the cap alone might make the amp oscillate at high frequencies as they are not compensated for low gain. I'm sure they had good reason for adding it. It cuts the gain in half at high frequencies, probably at the the cost of phase margin.

[@teamnery1747]

How do you bypass the opamps on that board?

[@tonyj7205]

Do you put these up on ebay after you correct them? There's a couple you made videos of that I would have liked to have.

[@dgjm7129]

possibility to test tda8954th amp boards?

[@DeeP_BosE]

Thanks for the ton of information. when i connect my PAM8610 to my PC audio n run it on 12v from the same PSU as my PC it goes into a fenzy of suicidal oscillations using another power souce for the Amp is just perfect music sans noise hum etc. what is causing this? is it that the same ripple in the input PS is present in the input AS, which makes the class d volt reference point oscillate ?

[@JohnAudioTech]

Probably a pretty bad ground loop.

[@mysock351C]

To elaborate a bit more, since the PSU is powering both the PC and the amplifier you will have severe ground loop issues as the sound card and the amplifier will share multiple ground paths thru both the audio cable and the power supply itself at both the amplifier input and power supply. The resulting offset in ground voltage will appear as an input, and at the very least will be amplified as though it were a signal. Combined with poor decoupling of the power supply this will cause it to oscillate wildly. Since the other power supply you tried it with is likely isolated, the input of the amplifier will be floating, so there is no DC path for ground loops at the input. If you just want it to work you can include an audio isolation transformer to break the coupling between the sound card and the amplifier input, but if its noise from the PC bus thats the issue it wont help much. A more involved solution is to use an on-card isolated DC-DC converter to create a local isolated supply for the amplifier, but thats somewhat involved. But from my experience since the PC generates so much noise its best to use an external amplifier with its own fully isolated supply, so there is no easy path back to ground for the noise and DC currents. Also watch out for using a large 50/60Hz transformer as the shunt capacitance is easily high enough to allow higher frequencies through and cause interference in the audio. Best bet is to either use a screened transformer, or a clean SMPS that will fully isolate the power.

[@bernardmarx6705]

Try 2R2 instead of 10R parallel to L.

[@JohnAudioTech]

No effect observed.

[@bernardmarx6705]

@@JohnAudioTech it's about sqare wave test with lower, real life loads, 4 ohms or less... Btw are you a member of diyaudio.com? This is not a platform for "chating" 😁 cheers.

[@newENIO11]

is bypassing the preamp opamps as easy as moving the jumpers? tia

[@JohnAudioTech]

Both channels of the dual NE5532 opamp chip are used. The jumper appears to bypass only one of them. To eliminate the preamp from the circuit you would have to remove the NE5532 chip and input the signal where the final preamp output was.

[@newENIO11]

@@JohnAudioTech thasnk you for the prompt reply! this particular board has some hissing on it. its barely noticeable in the listening chair. but i wanna mod it to see if i can lower it down still . so i was wondering if its the opamps gain is the source or there is a problem with the design of the power supply section . i already removed the fake main filter caps and swapped in 2 genuine 8.2 k uf nippon chemicons. and i also added the LC network at the output.

[@newENIO11]

if its not too much to ask. im a noob to this so excuse my ignorance. but i was wondering if you can show me where to solder the input underneath to bypass the preamp section completely?

[@martinda7446]

Ha, ha, great ending cat n' fluff, puss n' bungle, kitty n' blooper.

[@botabaixo]

Can you bridge the two channels?

[@pliedtka]

They can run in parallel with v small resistor at the output of each chip, no bridging option as far as I know. But I didn't try it. The notes should show the schematics.

[@smartups1]

Made a video on LA4440 sound ic .

[@jeffclark5206]

Looks like they put one real and one fake one, you can tell by the black covering across the top where vents are.

[@PeterMilanovski]

The only thing real was the new plastic wrapping on the capacitor.... Nah, that was fake too!

[@quananginh9446]

They are truly artists at faking capacitors. The fake one measures just like the real one =))

[@laxr5rs]

Cool.

[@brunomiguel6634]

Maybe the ic's aren't authentic also..

[@quananginh9446]

I've heard that in China, there are "domestic" version of these ICs.

[@wolidhassanbadol4045]

SONY DXA-C70 Transformer AC 27 0 27....yes/ no....LM 3886

[@jeffm2787]

Umm, technically ringing is a form of oscillation. Just not a stable form.

[@JohnAudioTech]

Ring is a form of damped oscillation.

[@1pcfred]

It's hydromatic.

[@1959Berre]

Systematic.

[@1pcfred]

@@1959Berre I figured no one would get the reference.

[@PeterMilanovski]

Why it destroys speakers like lightning ⚡. Oh look at what you have done now! Can't get the bloody song out of my head...

[@gamedemo6589]

Damn.. Too many prejudice. 😅

[@doctordorkmeister5330]

That IC is probably fake too

[@1pcfred]

Genuine CE (China Export) quality.

[@johnyang799]

As explained in the last video, ICs weren't fake.

[@1pcfred]

@@johnyang799 it was conjectured. They could just be very good fakes though.

[@johnyang799]

@@1pcfred hmmm

[@JohnAudioTech]

@@1pcfred I can't be 100% sure but I believe these are authentic. The Chinese use counterfeit die of smaller size for less cost. Not having the room for proper output transistors on the counterfeit die, I find the fakes struggle with heavy loads. Since these were delivering 90w at 4 Ohms, +/-32v supply, they were doing the business. I also see the same notch at clipping in authentic LM3886 ICs.