Hi,
I've been trawling posts and forums to debug some noise, squeals, and oscillations on some LM386 circuits I built.
I've synthesised all of this knowledge and calculated a load of component values. I wanted to run this past some people more knowledgable than myself and see if I'm approaching things correctly. I've not breadboarded this yet as I don't have all the components and was looking for some feedback before I order.
Also, please note that I am a complete amateur here. I’m more than capable of plugging numbers into calculations but I don’t fully understand all of the concepts behind these. I’d never heard of Xc for example.
Here's the datasheet for anyone who needs it - https://www.ti.com/lit/ds/symlink/lm386.pdf
Here's a schematic:
So, first of all, C1, the input DC blocking capacitor. The 10K volume pot in parallel with the IC's 50K input impedance results in 8.3K input resistance.
I'm aiming for Xc lower than 8.3K/10 at 100hz (this is low enough for my application, 20hz seems excessive for me). A 2.2µF cap gives Xc of 0.72KΩ at 100Hz so this seems good to me. This is higher than any other LM386 amplifier design I’ve seen online though (ranging from 47nF to 1µF)?
Next, C2. I went by the datasheet which shows 50µF as the best frequency response (Fig 2. Power Supply Rejection vs Frequency). The graph is for 6V supply though and I’m using 9V so I’m unsure what impact that might have. Again, 50µF is higher than the e.g. 0.1µF I’ve seen elsewhere.
For C3, I’m connecting to a miniature 8Ω speaker (well actually vibration speaker element so a speaker coil driver without the cone). The chip’s output resistance is 0.1Ω so I have a total output resistance of 8.1Ω. To work out the capacitor I’m aiming for Xc less than 8.1Ω/10 at 100Hz. A 2,200µF cap gives me 0.72Ω at 100Hz. This is much higher than used in many other circuit. Size and cost seems a more likely factor behind this and I have seen 2,200µF recommended elsewhere.
C4 and R1 form part of the Zobel network. Rather than take the standard values seen in many circuits (generally 10R and 47nF) I calculated these. My speaker has a voce coil/DC resistance of 7.4Ω. recommended resistor value is this times 1.25 and so 9.25Ω. 10Ω resistor is fine. For the cap the calculation is Le (voice coil inductance) / R (9.25Ω above) * R (well R squared but I can’t do superscript 2). Le for my speaker driver is 0.06mH. I dived this by 1000 to get Henries (and multiplied the result of the calculation by 1000000 to get µF) resulting in 0.61µF. I understand that going higher should be fine so a 1µF cap is an easy value.
C5 and C6 are power supply filtering/decoupling. I couldn’t find a source to calculate this so went with a range of advice and settled on 100µF and 0.1µF.
I’m leaving pins 1 and 8 open as 20 gain should be enough for me.
I’m using the inverting input (pin 2) as it’s suggested that it sounds better.
As my calculations result in some higher value capacitors than I’ve generally seen I’ve plumped for some polarised elctrolytics where I don’t usually see them (C1 and C2) so I’m unsure about polarity here.
Phew, so that’s the schematic. I’ve also laid out a board design based upon all of this and taking into account best practice I’ve seen. There’s especially an issue with keeping in, out, and power grounding separated and so I’ve done my best here. I’m using non-standard symbols (heavily drawn from Peter Blasser’s paper circuits) but the one’s which should be hardest to figure out are the circles with black triangles next to them. These are for the input and output jack grounds.
I appreciate that there’ll be a range of factors such as space and cost and so I’m just aiming for something more ‘ideal’ than practical at this stage.
So, anyone who actually understands all of this properly able to comment?
Thanks for your time.
I've been trawling posts and forums to debug some noise, squeals, and oscillations on some LM386 circuits I built.
I've synthesised all of this knowledge and calculated a load of component values. I wanted to run this past some people more knowledgable than myself and see if I'm approaching things correctly. I've not breadboarded this yet as I don't have all the components and was looking for some feedback before I order.
Also, please note that I am a complete amateur here. I’m more than capable of plugging numbers into calculations but I don’t fully understand all of the concepts behind these. I’d never heard of Xc for example.
Here's the datasheet for anyone who needs it - https://www.ti.com/lit/ds/symlink/lm386.pdf
Here's a schematic:
So, first of all, C1, the input DC blocking capacitor. The 10K volume pot in parallel with the IC's 50K input impedance results in 8.3K input resistance.
I'm aiming for Xc lower than 8.3K/10 at 100hz (this is low enough for my application, 20hz seems excessive for me). A 2.2µF cap gives Xc of 0.72KΩ at 100Hz so this seems good to me. This is higher than any other LM386 amplifier design I’ve seen online though (ranging from 47nF to 1µF)?
Next, C2. I went by the datasheet which shows 50µF as the best frequency response (Fig 2. Power Supply Rejection vs Frequency). The graph is for 6V supply though and I’m using 9V so I’m unsure what impact that might have. Again, 50µF is higher than the e.g. 0.1µF I’ve seen elsewhere.
For C3, I’m connecting to a miniature 8Ω speaker (well actually vibration speaker element so a speaker coil driver without the cone). The chip’s output resistance is 0.1Ω so I have a total output resistance of 8.1Ω. To work out the capacitor I’m aiming for Xc less than 8.1Ω/10 at 100Hz. A 2,200µF cap gives me 0.72Ω at 100Hz. This is much higher than used in many other circuit. Size and cost seems a more likely factor behind this and I have seen 2,200µF recommended elsewhere.
C4 and R1 form part of the Zobel network. Rather than take the standard values seen in many circuits (generally 10R and 47nF) I calculated these. My speaker has a voce coil/DC resistance of 7.4Ω. recommended resistor value is this times 1.25 and so 9.25Ω. 10Ω resistor is fine. For the cap the calculation is Le (voice coil inductance) / R (9.25Ω above) * R (well R squared but I can’t do superscript 2). Le for my speaker driver is 0.06mH. I dived this by 1000 to get Henries (and multiplied the result of the calculation by 1000000 to get µF) resulting in 0.61µF. I understand that going higher should be fine so a 1µF cap is an easy value.
C5 and C6 are power supply filtering/decoupling. I couldn’t find a source to calculate this so went with a range of advice and settled on 100µF and 0.1µF.
I’m leaving pins 1 and 8 open as 20 gain should be enough for me.
I’m using the inverting input (pin 2) as it’s suggested that it sounds better.
As my calculations result in some higher value capacitors than I’ve generally seen I’ve plumped for some polarised elctrolytics where I don’t usually see them (C1 and C2) so I’m unsure about polarity here.
Phew, so that’s the schematic. I’ve also laid out a board design based upon all of this and taking into account best practice I’ve seen. There’s especially an issue with keeping in, out, and power grounding separated and so I’ve done my best here. I’m using non-standard symbols (heavily drawn from Peter Blasser’s paper circuits) but the one’s which should be hardest to figure out are the circles with black triangles next to them. These are for the input and output jack grounds.
I appreciate that there’ll be a range of factors such as space and cost and so I’m just aiming for something more ‘ideal’ than practical at this stage.
So, anyone who actually understands all of this properly able to comment?
Thanks for your time.
Last edited:
I don't know that I fit your specification, but as a hobbyist, I've used, and still use, the LM38X chips a lot. My experience is that:
First, the data sheet circuits and values work pretty well, I'm not sure you are going to improve on them much, or even at all.
The minimum component count circuit probably will be noisy; it pretty much always needs a bypass capacitor on pin 7. I'm not sure what the data sheet suggests - I use 10uF. Try to get the bypass cap as close as you can to pin 7. In the picture, you can see I've used a non polarized ceramic cap.
I put a cap between the wiper and the input I'm using, typically 10 uF, and ground the other input, and call that part done. In the picture, you can see I've used a non polarized ceramic cap.
It should not oscillate if the input and output are kept separate, and you should not need a lot of heroic decoupling. I've attached a picture of a completely stable high gain variant, and you can see it only uses a 10nF disc ceramic cap at the B+.
A non polarized ceramic cap can be used in place of an electrolytic between pins 1 and 8, it just was not convenient to do that here, after the thing was built ...
I use between 100 - 470 uF on the output. It will look funky on a scope without the output network, but plenty of consumer stuff gets along just fine without it.
I've attached the pages from the 1980 National Semiconductor Audio Radio Handbook as they pertain to the LM386, as there is some information there, that is not in the data sheet.
This pic shows the data sheet circuit for the LM386 as used in a broadcast receiver.
Hope this is useful. The LM386 probably does not get much love on a hi fi forum, but I think it is a terrific little chip.
Attachments
The factory advice for the LM386 is "CHEEP!"
First: this is for a Pocket Transistor Radio (remember those?) with 2-inch speaker. Anything under 300Hz will just be fuzz. The competition used those sub-dollar transformers.
An input cap is not always needed. If it is, it can sit between pot and chip. Now it feeds 50k. 0.2u would pass to 20Hz. 0.1u is quite reasonable and inexpensive.
The supply rejection just gets better, 50u is not "optimum". But how dirty is your supply? Battery only needs a few uFd here (to set audio gain against 10k resistor). AC operation needs filtering for the main B+, and if that is commercially reasonable (<5%) then 50uFd is enough added filtering for the input stage.
Why would this cap vary much with supply voltage?
At the output: 10X minimum size is over-generous. That's real money!! Yes a livingroom amplifier might use 2000uFd but never a penny-pinching pocket radio. Couple-hundred uFd will usually do. The only time I'd think more is headphone amp, low-Z, and then only if you find GOOD LM386. (Many have excess crossover, being tuned for minimum power drain in dial-up modems.)
First: this is for a Pocket Transistor Radio (remember those?) with 2-inch speaker. Anything under 300Hz will just be fuzz. The competition used those sub-dollar transformers.
An input cap is not always needed. If it is, it can sit between pot and chip. Now it feeds 50k. 0.2u would pass to 20Hz. 0.1u is quite reasonable and inexpensive.
The supply rejection just gets better, 50u is not "optimum". But how dirty is your supply? Battery only needs a few uFd here (to set audio gain against 10k resistor). AC operation needs filtering for the main B+, and if that is commercially reasonable (<5%) then 50uFd is enough added filtering for the input stage.
Why would this cap vary much with supply voltage?
At the output: 10X minimum size is over-generous. That's real money!! Yes a livingroom amplifier might use 2000uFd but never a penny-pinching pocket radio. Couple-hundred uFd will usually do. The only time I'd think more is headphone amp, low-Z, and then only if you find GOOD LM386. (Many have excess crossover, being tuned for minimum power drain in dial-up modems.)
Last edited:
Thanks for taking the time and fair comments so far.
I think I misunderstood the graph and pin 7 power supply rejection. So... the frequency refers to the frequency of the noise you are trying to deal with? I initially took it with regards to audio frequency response of the chip. My bad.
I get that lower specs work 'fine' for many applications. As I tried to get across above, though probably failed, I'm coming at this from a rather theoretical perspective. Is this how you 'should' calculate for input resistance at a set frequency etc. Does the attempt at star grounding on the PCB look about right? Have I calculated the zobel network correctly? (I'm curious as to why the spec sheet value for the cap is quite different?).
@PRR you mention 0.1u between pot and chip. Are there disadvantages to this? Why isn't that common? Maybe it is but not on the schematics I've seen.
I don't know if 2,200u is that excessive on the output cap. I've seen it suggested in plenty of other places and it matches up with my understanding of the calculations. Here's a post on this very forum in which one of the respondents suggest it (LM386 amp Capacitor suggestions). Yes there's cost impact for mass production stuff but for DIY etc it's a small increase.
Finally, I think the biggest difference here from the pocket radio example is I'm using it with miniature vibration speakers. yes, they've got significant physical limitations but the way they interact with the surface they're in contact with 'may' be give better bass response than a similar speaker with a cone.
Anyways, thanks again, I'm probably just going to have to build and see if it was worth the extra effort.
I think I misunderstood the graph and pin 7 power supply rejection. So... the frequency refers to the frequency of the noise you are trying to deal with? I initially took it with regards to audio frequency response of the chip. My bad.
I get that lower specs work 'fine' for many applications. As I tried to get across above, though probably failed, I'm coming at this from a rather theoretical perspective. Is this how you 'should' calculate for input resistance at a set frequency etc. Does the attempt at star grounding on the PCB look about right? Have I calculated the zobel network correctly? (I'm curious as to why the spec sheet value for the cap is quite different?).
@PRR you mention 0.1u between pot and chip. Are there disadvantages to this? Why isn't that common? Maybe it is but not on the schematics I've seen.
I don't know if 2,200u is that excessive on the output cap. I've seen it suggested in plenty of other places and it matches up with my understanding of the calculations. Here's a post on this very forum in which one of the respondents suggest it (LM386 amp Capacitor suggestions). Yes there's cost impact for mass production stuff but for DIY etc it's a small increase.
Finally, I think the biggest difference here from the pocket radio example is I'm using it with miniature vibration speakers. yes, they've got significant physical limitations but the way they interact with the surface they're in contact with 'may' be give better bass response than a similar speaker with a cone.
Anyways, thanks again, I'm probably just going to have to build and see if it was worth the extra effort.
A lot of my work uses contact microphones and so the stuff on ceramic/crystal cartridges is great. Thanks for this.
I just did a design of lm386 audio amplifier circuit and get a problem, Design steps: First, think about the chip based on its function, thus designing the circuit Second, to understand the nature of the chip, characteristics. Third, use DXP software to draw the circuit schematic. I could not use DXP software, how to fix?
I adhere to the idea of add a cap between pot wiper and ic input. Base current flowing in the pot sooner or later become it scratchy.
For a century, schematics were done in pencil/pen on paper.
https://www.instructables.com/Preparing-a-Schematic/ Step 1: Draw a Schematic
How to draw an electronic schematic video
https://imgs.xkcd.com/comics/circuit_diagram.png cartoon
TMI: https://industrial-electronics.com/drafting-for-electronics-11.html
Old guys' yammer: https://electronics.stackexchange.com/questions/308878/how-were-schematics-drawn-before-cad
Last edited:
Since the LM386 topic has become active again, attached is a link to a recent, very thorough, article on use of the -4 variant of the chip, complete with measurements:
http://qrp-popcorn.blogspot.com/2022/11/another-look-at-lm386.html?m=1
Lots of other interesting stuff if you're interested in ham radio.
Win W5JAG
http://qrp-popcorn.blogspot.com/2022/11/another-look-at-lm386.html?m=1
Lots of other interesting stuff if you're interested in ham radio.
Win W5JAG
Hi All - since this thread is active and comprehensive, maybe someone could help me understand exactly what the value of the capacitor at pin 7 affects. I'm really green at this sort of thing. Anyway, I'm making little guitar amplifiers using these (with 9v batteries) and one example schematic uses a 47uf cap, while another uses a .01uf cap. My understanding is that this pin is to filter out noise from the power supply side, and would essentially act as a low-pass filter. Is that correct?
I'm not good enough at electrical engineering to figure out what the cutoff frequency would be for these different values, or if it even matters that much (e.g., if it's just trying to dump high-frequency noise on the supply voltage side, then maybe the value isn't that important?).
Thanks in advance.
I'm not good enough at electrical engineering to figure out what the cutoff frequency would be for these different values, or if it even matters that much (e.g., if it's just trying to dump high-frequency noise on the supply voltage side, then maybe the value isn't that important?).
Thanks in advance.
It is in the datasheet!!
Ah.... sorry.... get an OLD datasheet. The new kids at TI have totally botched the LM386 sheet. Here's a Jan 2000 copy:
https://www.futurlec.com/Datasheet/Linear/LM386N-1.pdf
Top of page 4.
Since you usually have 100/120Hz junk on the power rail (if wall-powered), you want more than 10uFd. 22u and 47u are often the same price.
Also this is bypass for the gain-setting network. The gain is simple when bypassed but less-simple if you omit the cap.
Thanks for that info! I've looked at the data sheet a bunch but never realized what PSSR was or that this graph related to the pin 7 cap.
Since I'm here, I'll keep asking questions until I overstay my welcome...
What does it mean to "bypass the gain setting network"? Does it mean that it's a way for power supply noise to "bypass" the gain network?
I adhere to the idea of add a cap between pot wiper and ic input. Base current flowing in the pot sooner or later become it scratchy.