The LM3886 data sheet specifies 28V for a 4 Ohm load, and 7 Amps max. Note that 4x7 =28. I'm surprised if your chip is not toast, or perhaps it is.
There has been serious problems with the LM3886 spice model, which was covered in the LTC spice thread here on DIYA. Tom has paralleled chips to boost the current. I suggest you have a look at his work. I would probably add external output transistors but there are issues with that as well. If you need more than 50W then I think you need to forget chip amps, or perhaps try class-D.
There has been serious problems with the LM3886 spice model, which was covered in the LTC spice thread here on DIYA. Tom has paralleled chips to boost the current. I suggest you have a look at his work. I would probably add external output transistors but there are issues with that as well. If you need more than 50W then I think you need to forget chip amps, or perhaps try class-D.
Well you still need to re-invert to get the correct phase (unless you are happy reversing roles of speaker terminals - and only possible if not planning a headphone jack connection).
So if you want inverted 3886 you'd be best adding an opamp inverting stage in front so you can drive it low-impedance and correct the phase error. Both stages then benefit from absence of common-mode distortion - job done.
If you do my suggested changes leave C3 at 33pF. These changes are to combat the occasional ringing oscillation you mentioned.
Have a look at the datasheet (Figure 19+20). You cannot squeeze more than ~70W@4Ohm RMS = 100W peak = 20V peak out of the LM3886. That's why it goes into protection. 31W@8Ohm seems reasonable, given your supply voltage will drop under load.
Have a look at the datasheet (Figure 19+20). You cannot squeeze more than ~70W@4Ohm RMS = 100W peak = 20V peak out of the LM3886. That's why it goes into protection. 31W@8Ohm seems reasonable, given your supply voltage will drop under load.
Thanks all. I have parts on order to try few things per recommendation in this thread now. I have two (imho) independent problems so I'll try to systematically solve them. Here's my plan;
Problem #1 Occasional ringing oscillation (I'm still not reliably recreate the problem)
-Nyx suggestion to replace Cc (C7) with larger value (470pf + R330). However, I believe Tom has pointed out worsen phase margin with Cc and suggests using an EMI filter at non-inverting input instead (https://neurochrome.com/pages/stability#StabilityComponents)
Problem #2 Stability at 4 ohms (SPIKe at lower load)
-Tom suggests changing Rf to 20k and Ri to 1k. This will significantly lower input impedance. I'll test this before working with buffer.
-Buffer will be added. Thinking of a regulated OPA1612/NE5532 preamp board with modified gain resistors. I'll have to think where can I mount the board and whether I can piggyback power supply because I have no room for additional transformer.
-Optimization for input coupling network if needed (RC input coupling network, etc).
Other possible solutions
-If I'm not successful on resolving those issues, I might try to modified the board to be non-inverting design. To be frankly, it seems easier than adding buffer board. I'd need to remove all feedback/ input network, remove Cc, add feedback resistor under PCB across the pin and few modified stability components, rewire input to use pin 10 instead pin 9. I'm thinking of reusing some components (Input coupling C = 2.2uF, Rf1 = 47k, Ri = 2.2k, Ci = 100uF, Rf2 = 47k, Cf2 = 33pF, Cc = removed, supply coupling capa = same, Zobel & Thiele networks = same).
I'll report back.
AP
Problem #1 Occasional ringing oscillation (I'm still not reliably recreate the problem)
-Nyx suggestion to replace Cc (C7) with larger value (470pf + R330). However, I believe Tom has pointed out worsen phase margin with Cc and suggests using an EMI filter at non-inverting input instead (https://neurochrome.com/pages/stability#StabilityComponents)
Problem #2 Stability at 4 ohms (SPIKe at lower load)
-Tom suggests changing Rf to 20k and Ri to 1k. This will significantly lower input impedance. I'll test this before working with buffer.
-Buffer will be added. Thinking of a regulated OPA1612/NE5532 preamp board with modified gain resistors. I'll have to think where can I mount the board and whether I can piggyback power supply because I have no room for additional transformer.
-Optimization for input coupling network if needed (RC input coupling network, etc).
Other possible solutions
-If I'm not successful on resolving those issues, I might try to modified the board to be non-inverting design. To be frankly, it seems easier than adding buffer board. I'd need to remove all feedback/ input network, remove Cc, add feedback resistor under PCB across the pin and few modified stability components, rewire input to use pin 10 instead pin 9. I'm thinking of reusing some components (Input coupling C = 2.2uF, Rf1 = 47k, Ri = 2.2k, Ci = 100uF, Rf2 = 47k, Cf2 = 33pF, Cc = removed, supply coupling capa = same, Zobel & Thiele networks = same).
I'll report back.
AP
I'm playing with TINA-TI and here result for all suggestion for problem #1
1. Original circuit (Cc = 220pf)
2, Cc = 470pf series with 330R
3. Cc = removed, but I add 220p from pin 9 to ground (try to limited high frequency to enter the chip).
What do u think?
Circuit of the last case
1. Original circuit (Cc = 220pf)
2, Cc = 470pf series with 330R
3. Cc = removed, but I add 220p from pin 9 to ground (try to limited high frequency to enter the chip).
What do u think?
Circuit of the last case
Yes, that is the best solution. If distortion remains at 4ohm then it is probably due to the LM3886 chip itself.
AFAIK, LM4780 spice model is not officially available from TI.com anymore because the chip is discontinued. Google search yield only old results. My speakers are 8 ohms and I use in nearfield environment. The 4 ohm test is just for stability stress test.
What do you think about these two schematics? I found info about adding LPF at the input of 3886 (at CircuitBasic) to limit HF response and experimenting with them.
1. Inverted design, I moved Cc 220pF to inverted and ground instead of between inverted and non-inverted inputs. This forms an RC filter with cutoff frequency at 96k (220pF and R7.5k). The phase response seem better (close to 180 in 20-20kHz range). This is an easy mod for me (this doesn't solve low input impedance problem though).
2. Non-inverted design with LPF (at 234kHz, value from CircuitBasic). This design doesn't suffer from phase shift issue.
What do you think about these two schematics? I found info about adding LPF at the input of 3886 (at CircuitBasic) to limit HF response and experimenting with them.
1. Inverted design, I moved Cc 220pF to inverted and ground instead of between inverted and non-inverted inputs. This forms an RC filter with cutoff frequency at 96k (220pF and R7.5k). The phase response seem better (close to 180 in 20-20kHz range). This is an easy mod for me (this doesn't solve low input impedance problem though).
2. Non-inverted design with LPF (at 234kHz, value from CircuitBasic). This design doesn't suffer from phase shift issue.
TI also has a PSPICE model for the component. There never was an LM4780 TINA model.
Just wanna post an update on how I resolve this.
I decided to un-inverted my circuit and retest. I got very similar performance as the inverted circuit. So, I decided to swap the chips and it solves the problem.
4 ohms load, got around 64.8W before clipping (it got some slight ringing oscillation).
I have tested at 15, 20,100, 1k, 10k, 15k, 20k, 25k and power are not varied much.
8 ohms load, same test, got around 42W, and I hit the limit of my sig gen voltage. I think the actual power can be higher.
Here's pics of replacement chips and modified circuit.
AP
I decided to un-inverted my circuit and retest. I got very similar performance as the inverted circuit. So, I decided to swap the chips and it solves the problem.
4 ohms load, got around 64.8W before clipping (it got some slight ringing oscillation).
I have tested at 15, 20,100, 1k, 10k, 15k, 20k, 25k and power are not varied much.
8 ohms load, same test, got around 42W, and I hit the limit of my sig gen voltage. I think the actual power can be higher.
Here's pics of replacement chips and modified circuit.
AP
Attachments
Hi!
So you changed the chips and problems are solved? Would be nice to post a picture of the chips that didn't work.
Thank you!
So you changed the chips and problems are solved? Would be nice to post a picture of the chips that didn't work.
Thank you!
Thanks for warning. 4-ohm load is for testing only. My speakers are 8 ohms.
You can see in post #1 what transformer the thread starter has recycled for his project.
So the chip has to life with the voltage that comes out of it, no discussion.
If the designer of this amp is lucky, the advanced protetion layers of the (real) LM3886 will precent the worst. If not, maybe the DC protection build into the board will protect the speaker.
So the chip has to life with the voltage that comes out of it, no discussion.
If the designer of this amp is lucky, the advanced protetion layers of the (real) LM3886 will precent the worst. If not, maybe the DC protection build into the board will protect the speaker.
The chips you show in this thread look like the ones I got from Mouser last year. The other one, here: https://www.diyaudio.com/community/...-and-genuine-chip-lm3886.422929/#post-7913357
just screams "fake". Even the white paint is ridiculous.
just screams "fake". Even the white paint is ridiculous.
My main voltage is suppose to be 220VAC but it's always measured 230+. However, voltage should be fine with 8 ohm load since the datasheet said it can do 50W at +-35V.
BTW, this is the final schematic I used. Just a reference for future reader. (I omitted the supply decoupling caps in schematic, they are 100uF electrolytic and 100nF/0.1uF MLCC as shown in post #32).
BTW, this is the final schematic I used. Just a reference for future reader. (I omitted the supply decoupling caps in schematic, they are 100uF electrolytic and 100nF/0.1uF MLCC as shown in post #32).
Last edited:
I got them during COVID when it was hard to find. The thing is, I paid almost TWICE the price of the authentic chips I acquired later on. Now I've just need to get more authentic chips to average out LOL 🤣...
