Hi,
I'm testing my inverted LM3886 and found significant distortion when testing a 4 ohms load. Here's my setup and schematic.
DC = +- 35.8V (a little high)
Green = components added for stability
P1, P2 = Probe location, using ground at input
Load = 4/8 ohms 100W dummy load
Here's the distortion at 4 ohms load
Here's peak power at 8 ohms
My question is;
1. At 4 ohms, I got much less power with significant distortion. Is this SpiKE protection? or what is it?
2. At 8 ohms, I got about 31W, which seems low given the supply voltage?
3. Possibility of fake chip? I got the chip 2 years ago when it was scarce as "old stock".
(Known issues of this board
-Occasional ringing oscillation, still can't fully catch it but it happens less after more stability components added
-If C3 is removed, it will oscillate and R9 in Zobel network will burn (happened twice!))
Any suggestion would be appreciated.
AP
I'm testing my inverted LM3886 and found significant distortion when testing a 4 ohms load. Here's my setup and schematic.
DC = +- 35.8V (a little high)
Green = components added for stability
P1, P2 = Probe location, using ground at input
Load = 4/8 ohms 100W dummy load
Here's the distortion at 4 ohms load
Here's peak power at 8 ohms
My question is;
1. At 4 ohms, I got much less power with significant distortion. Is this SpiKE protection? or what is it?
2. At 8 ohms, I got about 31W, which seems low given the supply voltage?
3. Possibility of fake chip? I got the chip 2 years ago when it was scarce as "old stock".
(Known issues of this board
-Occasional ringing oscillation, still can't fully catch it but it happens less after more stability components added
-If C3 is removed, it will oscillate and R9 in Zobel network will burn (happened twice!))
Any suggestion would be appreciated.
AP
I would run the stability analysis on that circuit. Especially the 220 kΩ feedback resistor worries me. That along with the input capacitance of the LM3886 (and likely also the 220 pF across its inputs) will set a pole that's pretty low in frequency. That could be trouble.
A better approach for an inverting LM3886 is to add a buffer in front so you can use a more reasonable feedback network, such as 20 kΩ + 1 kΩ. You could then entertain the possibility of putting some of the gain in the opamp/buffer so the LM3886 has more feedback, thus, lower distortion.
Also note that you'll want 470 uF (min) || 10-33 uF || 100 nF (min) for the decoupling network. Read more here: https://neurochrome.com/pages/supply-decoupling
Tom
A better approach for an inverting LM3886 is to add a buffer in front so you can use a more reasonable feedback network, such as 20 kΩ + 1 kΩ. You could then entertain the possibility of putting some of the gain in the opamp/buffer so the LM3886 has more feedback, thus, lower distortion.
Also note that you'll want 470 uF (min) || 10-33 uF || 100 nF (min) for the decoupling network. Read more here: https://neurochrome.com/pages/supply-decoupling
Tom
The feedback resistor too high a value (necessitated because of the input impedance in inverting configuration. This was a cause of instability problems in a lot of inverting LM3875 DIY designs in the early 2000's. The amplifier probably happier with a gain of 20 (26dB).
Tom is the LM3886 guru. The only thing I would note is to place the decoupling caps as close to the power pins as possible.
FWIW, there is a Texas Instruments LM3886 SPICE model for their proprietary TINA software. That will help you get a better fix on the compensation capacitor C3.
Tom is the LM3886 guru. The only thing I would note is to place the decoupling caps as close to the power pins as possible.
FWIW, there is a Texas Instruments LM3886 SPICE model for their proprietary TINA software. That will help you get a better fix on the compensation capacitor C3.
Thanks all. Is it possible to replace R3 with lower value, such as, 100k? Or do I need to replace the pair R3,R7? How to calculate input impedance? Is it just the value of R7?
As Tom suggested, adding buffer stage is proper approach but it is beyond my capability to design and do that. I'm running this off a USB dac which manufacturer claimed 100 ohms output impedance.
As for power supply decoupling capacitors, I have them at the pins under PCB, so it's as close as I can.
Thanks,
AP
As Tom suggested, adding buffer stage is proper approach but it is beyond my capability to design and do that. I'm running this off a USB dac which manufacturer claimed 100 ohms output impedance.
As for power supply decoupling capacitors, I have them at the pins under PCB, so it's as close as I can.
Thanks,
AP
Your feedback values are high.
It looks like the SPiKe overload current limiting built into that chip.
Your 8 ohm waveform being normal virtually guarantees overcurrent protection.
Buffer transistors on the output will solve this. The feedback would have to include the new transistors along with a 0.68 ohm resistor between bases and emitters.
It looks like the SPiKe overload current limiting built into that chip.
Your 8 ohm waveform being normal virtually guarantees overcurrent protection.
Buffer transistors on the output will solve this. The feedback would have to include the new transistors along with a 0.68 ohm resistor between bases and emitters.
I don't see any good reason to make an inverting amplifier. Instead of this torture, you just need to change the configuration to non inverting and everything is solved.
I think an inverted lm3886 design benefits from lack of common mode distortion. It's more popular in the past prior to availability of better non-inverting design like Tom's excellent Modulus series that use different technics to achieve even better results.
I read through older threads in this forum when inverting gainclone were more popular in the past. 220k and 10k value were commonly used for feedback/ gain settings.
I guess I could try experimenting with those resistors' value. Any precaution should I taken?
I read through older threads in this forum when inverting gainclone were more popular in the past. 220k and 10k value were commonly used for feedback/ gain settings.
I guess I could try experimenting with those resistors' value. Any precaution should I taken?
I don't know about that. I don't like the low input impedance and phase inverting. Those distortions from the first post do not seem to be related to configuration. Where did you buy the LM3886? They may not be genuine.
Last edited:
Also, the 2.2uF coupling capacitor is small for this input impedance. It needs a much bigger one.
I'm thinking of 200k and 10k. Also, the input 47k resistor is used to avoid floating input and it is before the 2.2uF coupling cap. I'm not sure why I need higher value? Please suggest.
I got the chip from online when it was out of stock in most places. I have no way to verify if it's genuine or not.
I got the chip from online when it was out of stock in most places. I have no way to verify if it's genuine or not.
The input impedance for the inverting configuration is simply the impedance from the input of the amp to the inverting input of the LM3886. That's because the negative feedback ensures a virtual ground at the inverting input. So with the schematic as drawn, the input impedance is 7.5 kΩ.
Tom
Tom
That won't solve the stability issue, though. That's caused by the 200 kΩ (and the input capacitance). Go with 20 kΩ and 1 kΩ, but then you have 1 kΩ input impedance, so you add a buffer.
Technically, the input impedance seen by the source will be 47k||(2.2uF+7.5k), so within the audio band, you're looking at about 6.5 kΩ. That's pretty low. Increasing the 47 kΩ to 470 kΩ increases the input impedance (mid-band) to 7.4 kΩ; still pretty low.
Too many people rely on memorization rather than understanding.
Tom
TINA is actually not proprietary. You can buy a license to it if you want to. TINA-TI is freely available from TI.
Good suggestion to simulate the stability though. That'll save a lot of F..... eh .... messing around in the lab.
Tom
If stability is the problem and you want to keep your resistor values, here's what you can try:
- Increase your noise gain capacitor C7. Double or triple the value.
- Add a series resistor to C7. About 1k. If you make C7 larger you can live with less, e.g. 470pF ~330R.
2.2uF with 7k5 input impedance is RC filter of -3dB at about 10Hz. The practice is to make it at 1-2Hz. So try 10uF or more.
Thanks everyone for the clarification. I'm trying to wrap my head around 🙂
Possible fixes;
1. Tom suggests change feedback/ gain to 20k/1k and add buffer. Something like this?
(OPA1612 or similar)
2. Nyx suggests increasing C7 (Cc in datasheet) and series R to reduce potential quarsi-saturation oscillation (470p + 330R) but I have to consider value of C3 (Cf?) as well?
3. Nixie62 suggests change input RC filter to more proper response at low frequency.
4. Jack pointed out that phase shift issue with current C3 value. Suggests lower value...
Not sure where to start. I'm trying to learn more about TINA-TI to analyze stability
Possible fixes;
1. Tom suggests change feedback/ gain to 20k/1k and add buffer. Something like this?
2. Nyx suggests increasing C7 (Cc in datasheet) and series R to reduce potential quarsi-saturation oscillation (470p + 330R) but I have to consider value of C3 (Cf?) as well?
3. Nixie62 suggests change input RC filter to more proper response at low frequency.
4. Jack pointed out that phase shift issue with current C3 value. Suggests lower value...
Not sure where to start. I'm trying to learn more about TINA-TI to analyze stability
I avoid coupling capacitors wherever possible. I have a DC speaker protection circuit. As long as the output offset is less than 150mV, I don't see any problem.
It's not at all clear that common-mode distortion is dominant enough that removing it would improve anything, given the low signal levels.