robrob 6/17/2017 12:24 AM
Help Me Understand the LM386 Chip Amp
I have a How the Ruby Amp Works webpage and I would like to go into some detail of how the LM386 chip functions on the guitar audio signal. As many of you know I'm a tube guy and I'm pretty weak on solid state theory.

The annotated schematic below is my first attempt to describe what each transistor in the LM386 is doing to the audio signal. Please look it over and let me know what I have wrong.

Keep in mind in the Ruby Amp pin 3 (+ input) is grounded and only input pin 2 is used.

The red line is the guitar signal path through the chip.

Is it correct to call all 10 of the transistors in the LM386 op-amps?

[IMG]https://robrobinette.com/images/Guitar/Ruby/LM386_Annotated_Schematic.jpg[/IMG]
 
Enzo 6/17/2017 7:08 AM
No, they are transistors. The whole thing is more or less an op map though.

Step back from the page a couple feet, so as not to over-analyze it. This is a simple power amp circuit, miniaturized into the IC package. What you called a power amp is what we would call a voltage amp stage or VAS. The two output transistors connected to pins 4 and 6 are like any other SS amp output, and they are current amplifiers, they take the voltage signal from the previous stage and allow a greater amount of current to flow into a load for the same voltage.

There is no grid to leak.

The fact this is tiny transistors inside the IC doesn;t change that it is a basic solid state amp. So I recommend you read Teemu K's book he has graciously posted for your download here:

https://www.ssguitar.com/index.php?topic=711.0
 
Pedro Vecino 6/17/2017 8:02 AM
I found this analysis, in Spanish. Very interesting.
ElectroSmash - Análisis del Amplificador de Audio LM386.
 
robrob 6/17/2017 8:02 AM
Thanks Enzo for the help. I revised the schematic and renamed some components. A browser "hard refresh" may be needed to see the updated schematic in the first post. Do you agree with the signal path as drawn?
 
robrob 6/17/2017 8:31 AM
Thanks Pedro, I did find and read the English version of that webpage before I posted here. There's not a whole lot out there on the LM386's internal function.
 
J M Fahey 6/17/2017 9:26 AM
I HATE having to write this, doubly so after your gorgeous drawings and visible efforts to do things right, but for the first time I see with my own eyes how weird ideas get injected into mainstream Internet.

The page is chock full of iffy, "weeellllll .... maaaayyyybee" or plain wrong stuff but since itīs gorgeously made and has a search engine friendly name, will be read, referred to and recopied all over the Net.

With pain in my heart:

1) Zobel network "high frequency cut"? No.

2) "The J201 JFET (Junction gate Field Effect Transistor) is an input buffer and keeps the guitar circuit from affecting the LM386 amplifier's input." : No.

3) "The LM386 is a chip amp that performs the functions of the preamp, power amp and output transformer." : weeellllll .... maaaayyyybee.
Perhaps a valid explanation in 1946 when nobody had ever heard of transistors, let alone chipamps, and you wanted to "explain the unknown based on something known".

4) nobody will build it on eyelet or turret because of size discrepancy, it simply canīt be done, whatīs proper for an experimenter is standard perfboard with .1" hole spacing, impossible with either turrets or eyelets.

Mind you, my first Op Amp based preamps were built on eyelet boards, but I used round 8 pin 741 amps with long legs
[IMG]http://i.ebayimg.com/images/g/6BkAAOSwuMFUkcx0/s-l300.jpg[/IMG]
which could be separated enough; forget doing that with DIP8 plastic packages.

Use an 8 pin DIP socket to mount the LM386 chip to the circuit board.
Easier said than done.

5) a buffer to keep the guitar circuit from loading down the signal entering the LM386 amp. No.

6) Think of the Gain Pot as a master volume No, itīs a gain control, very different thing.

7) The signal is then filtered by the 10 ohm resistor & 47nF cap to reduce very high frequencies (ice pick and beyond). Oh My God.

8) Again "Zobel as Hi Cut".

9) The Input Resistor sets the amplifier's input impedance and gives the AC guitar signal a path back to the guitar pickup. Not necessary.
No need for current to flow, applying pure voltage to Fet input is enough.
That resistor is there to give Fet gate a DC ground reference if no guitar is connected.

10) The J201 transistor keeps the guitar circuit from affecting the LM386 amplifier. No.

11) The Source Resistor attached to the transistor source leg functions like a cathode resistor by controlling the flow of electrons from ground. Not exactly to no.

12) You can bump this capacitor up to .1uF (micro Farad) for a Bassman Mod
"Bassman Mod" ??????? Grandiose name for a slight reduction in Bass cutting (which by the way is needed to prevent farting).

13) The Volume Pot functions as a variable voltage divider and controls volume by sinking excess guitar signal to ground. No.
You are focusing your explanation on current exclusively. Some things can be better explained by that, others by voltage.

14) The cascaded transistor amplifiers inside the LM386 are why the overdrive tone of this tiny chip amp sounds so good. Good or bad, cascading gain stages is the only way to build an amplifier.

15) The first op-amp You mean the first transistor?

16) The Voltage Amplifier functions as the "power amp" of the circuit. No. You have a voltage amplifier first, and a current amplifier/buffer second.

17) In this sense the Current Amplifiers do the job of a tube amplifier's output transformer.

No: transformers do NOT offer power gain; output transistors do.

18) Hi Cut 47nF Cap. Very high frequencies (ice pick and above) are sunk to ground. No.

19) The inherent resistance of the LM386 amp + 47nF cap form an RC (resistance-capacitance) low pass filter.
The inherent resistance of an LM386 is practically *zero*.

20) The size of the cap sets the filter's cutoff frequency and the 10 ohm Resistor reduces the effectiveness of the filter so some frequencies above the cutoff are passed on to the speaker. No.

21) The result is a deeper, meatier, more 5F6A Bassman-like tone. If you say so.
 
robrob 6/17/2017 10:40 AM
Juan,

Thank you for taking the time for the excellent critique of the webpage. This page went up today and is a work in progress so no one is reading it but you guys. I came here for input like yours to keep the info as accurate as possible. The webpage's purpose is to help tube amp guys understand the simple but good sounding little Ruby Amp.

1) I made the Zobel network correction.

2) Then what is the function of the J201 if it isn't an input buffer?

4) The eyelet layout helps tube guys compare the Ruby circuit layout to our standard eyelet boards. Yes there are better ways to build a Ruby but this would work.

9) I removed the "return path back to the guitar pickup".

12) The "Bassman Mod" is a standard Ruby Amp mod that's been around forever. I just try to explain what the mod does.

13) Removed the "sinks the signal to ground".

14) You're right, that sentence added nothing so I deleted it.

15) Yes, I meant transistor.

16) I revised the "preamp" and "power amp" labels to voltage amplifiers and current amplifiers.

Again, thanks for taking the time to give me the very valuable feedback.
 
Dave H 6/17/2017 12:29 PM
Quote Originally Posted by robrob View Post
2) Then what is the function of the J201 if it isn't an input buffer?
Quote Originally Posted by robrob View Post
The J201 JFET (Junction gate Field Effect Transistor) is an input buffer and keeps the guitar circuit from affecting the LM386 amplifier's input.
You have it back to front. The FET buffer is there to keep the LM386 circuit (actually the 10k Vol pot) from affecting the guitar circuit.

Isn't 220p too small for a bright cap on a 10k pot?
 
MarkusBass 6/17/2017 1:20 PM
Rob,

I'm with Juan - the text on this web page is extremely incorrect. I think that Juan discovered about a half (or even less) of the problems. Just a quick look at texts like "Grid leak", or "Signal path shown in red and negative feedback loop shown in blue. Each of the 10 transistors inside the chip is the equivalent of a tube triode." show that this is kind of a "vintage" explanation of how op amp works. This would be good in 1955 but not in 2017. Why do you think that the signal flows only through the upper output transistor and not the lower? Sorry for harsh words but on this forum most of people know how transistors and op amps work. And when they see such misleading information they just say that it's incorrect. You asked for opinions and you got it.

Mark
 
robrob 6/17/2017 1:33 PM
Thanks Dave, I'll make the edit.

The 220pF bright cap is what's used in the Ruby Amp standard "Bassman Mod." It seems to work for people.
 
Jazz P Bass 6/17/2017 2:45 PM
I agree with Enzo, the output stage signal flow is wrong.

The datasheet specifically states that the output voltage is centered at 1/2 of the supply.

http://www.ti.com/lit/ds/symlink/lm386.pdf

Therefore you will need a capacitor output to block the Vdc.

Follow the schematic with that in mind.
The upper half of the signal sine wave will turn on the upper NPN transistor harder.
The lower half of the sine wave will turn on the lower NPN transistor harder.
(the trick here is the PNP transistor)
Result: Power gain without phase reversal of the output.

Here is a nice read: Build a Great Sounding Audio Amplifier (with Bass Boost) from the LM386
 
robrob 6/17/2017 4:07 PM
the output stage signal flow is wrong.
Thanks Jazz, I revised the LM386 signal flow. How about now? You may have to do a browser hard refresh (shift key + refresh symbol) to see the updated schematic.
[IMG]https://robrobinette.com/images/Guitar/Ruby/LM386_Annotated_Schematic.jpg[/IMG]
 
Enzo 6/17/2017 5:14 PM
Um... I am computer illiterate, what is the refresh symbol? I am still seeing the original image.

It is a push pull, so both sides of the output are signal path.
 
Jazz P Bass 6/17/2017 5:28 PM
Aah. Refresh key: I hit F5 and it did indeed 'refresh it'.

I would get rid off the signal path on the PNP Emitter leg.

Also, I do not see the PNP transistor as a 'current amp'.
More like an inverting driver for the lower NPN.
 
robrob 6/17/2017 6:14 PM
Thanks Jazz, I made your suggested changes to the schematic and the supporting verbiage.
 
tedmich 6/17/2017 6:14 PM
https://en.wikipedia.org/wiki/Not_even_wrong
 
MarkusBass 6/17/2017 10:33 PM
Exactly. I see that you removed such terms like " "Grid leak" and "Current balance" (which were completely wrong). But there would be nothing wrong if you put there "current mirror" which is correct description of the op amp stage. You can learn about it here: https://en.wikipedia.org/wiki/Current_mirror

Mark
 
Dave H 6/18/2017 2:42 AM
Quote Originally Posted by tedmich View Post
It looks like "The Wrath of God" descended on Rob in this thread
 
robrob 6/18/2017 6:42 AM
Markus, I thought the current mirror acted to balance the two halves of the differential mirror, that's why I had "current balance" on the schematic. Does the current mirror function when only one input is in use like in the Ruby?
 
MarkusBass 6/18/2017 7:55 AM
Quote Originally Posted by robrob View Post
I thought the current mirror acted to balance the two halves of the differential mirror, that's why I had "current balance" on the schematic.
This is one of the "not even wrong" statements. There are no two halves and there is not differential mirror here. You shouldn't develop new terms to describe what is widely known. You could write "current equaliser", or "current double terminator" but why would you do that? You do this all the time: you call transistors - "tubes". Yes, somehow a transistor can be compared to a tube but no one says that a transistor is a tube (except you). There must be a reason for this
 
R.G. 6/18/2017 10:45 AM
@ robrob:

You're suffering from overthinking it. The LM386 is just one modestly specialized version of opamp, and ought to be looked at in that way by everyone who's not actually trying to dig into the transistor-ness of how the insides work.

The specializations are worth understanding.

- the LM386 opamp has been set up so that it can provide lots of output current, suitable for driving a speaker directly
- it has had biasing and gain-changing feedback resistors included inside so that making a few-components speaker amplifier is easy
- it has had its input circuitry tinkered with so that input signals are pulled to ground, and the inputs biased at ground by the included 50K input resistors
- pins were brought out to the outside to provide for attachment of large capacitors where needed, and for external gain modification

Other than that, it's an opamp. The input impedance is low, about the value of the 50K input resistors, so it needs an input buffer to avoid cutting treble from a guitar signal by loading the inductive guitar signal. That's what the input JFET is for. The output resistor/capacitor network doesn't shape frequency response in any audio fashion, it's there to maintain stable operation in the fact of possibly capacitive loads - another consequence if its being an opamp.

I highly recommend you re-cast your explanation in terms of a customized opamp being manipulated by other circuit elements. I can dredge you through the blow-by-blow of what the transistors do, but that's not going to help your readers any.
 
Dave H 6/19/2017 3:46 AM
The way the output is biased is interesting. The input transistor bases are grounded and their emitters are 2 x Vbe above ground which means there's almost the full supply voltage across the left side's 30k (2 x 15k) emitter resistor. The current mirror forces the right side emitter current to be equal the left side so the right side's single 15k resistor drops half the supply voltage and it's connected to the output so the output must be at half the supply voltage.
 
robrob 6/19/2017 4:20 AM
Markus, I meant to say "differential amplifier", not "differential mirror".

On the webpage I said each of the 10 transistors in the LM386 are the equivalent of a triode tube. Could we not replicate the LM386 circuit using triode tubes and an output transformer?
 
robrob 6/19/2017 4:22 AM
Thanks R.G., I'll try to incorporate that into the article.
 
MarkusBass 6/19/2017 4:36 AM
Quote Originally Posted by robrob View Post
Could we not replicate the LM386 circuit using triode tubes and an output transformer?
So you don't know it but yet you put this information on your web site. I doubt whether it's possible. Please try to do it and show us results.
I think that the information provided by R.G. is what people are looking for. And no one is interested in the information you are trying to provide. This is something that op amp designers could look at but since the information is highly incorrect, I doubt whether they would like to do so. And they already know all the technical details of this design.
 
robrob 6/19/2017 4:56 AM
That was a rhetorical question.
 
robrob 6/19/2017 4:58 AM
I want to thank (most of ) you guys for providing meaningful guidance. I took it to heart.

I took a public flogging but I learned a lot in the process.
 
robrob 6/19/2017 5:13 AM
In Chrome it's the circular arrow in the upper left corner of the browser.
 
Justin Thomas 6/19/2017 5:25 AM
Well, you <COULD> probably make a basic op-amp with triodes, but that's how we ended up with a pocket calculator the size of a skyscraper. And the Greenies would go apeshit if you tried.

Justin
 
Enzo 6/19/2017 7:40 AM
Here:
The Philbrick Archive

The famous Teledyne-Philbrick op-amp modules of the 1950s.


Could you do it in triodes? Well, there are only a few parts, draw it up as you think it would work.
 
R.G. 6/19/2017 8:22 AM
One could replicate the function (i.e. opamp) in tubes; Enzo beat me to the Philbrick implementation.

What you can't replicate as a tube analog part-for-part are the PNP transistors, as there are no "P-channel" triodes; these are just barely possible, if you constructed the tube insides out of anti-matter so the bits running from cathode to plate are positrons, but we have no significant amounts of antimatter to make this with, and I don't think we have yet made antimatter atoms more complex than anti-hydrogen. So there would be some additional work required to make the parts.

And maybe some special licensing to run the equipment to make the materials.
 
Mike Sulzer 6/19/2017 10:43 AM
My experience is that the sharp sounding distortion that you get with a huge amount of global feedback is not what you want for a guitar amp.

Quote Originally Posted by Enzo View Post
Here:
The Philbrick Archive

The famous Teledyne-Philbrick op-amp modules of the 1950s.


Could you do it in triodes? Well, there are only a few parts, draw it up as you think it would work.
 
SoulFetish 6/20/2017 2:02 AM
Rob, you may not have all of the facts, but you've got one hell of a chin!
You've taken a few licks in some threads from time to time, and you seem to keep a good attitude. Remember, circuit operation is a matter of science and isn't personal. Everyone here, at some point, didn't know their ass from their elbow.
Having said that, if you wouldn't mind me giving you a humble suggestion. in addition to vetting the information for accuracy, it may be prudent to have some patience with a circuit or topic and develop a solid understanding of it's function before posting an article to your website. I see your website come up often in search results, and you have some decent stuff there.