Signal Tracer - Requesting input on requirements

[sam_sam_sam]

Re: Signal Tracer - Requesting input on requirements

I have one comment about the capacitor across the other pads ( in the second picture) is you might want to put heat shrink tubing on the component leads so nothing bad happens ( just a thought )

Nice work

[redwire]

Re: Signal Tracer - Requesting input on requirements

It shouldn't get really quiet like that. Unless the speaker GND is shorting out the input signal or something?
I couldn't make out all the wiring, this pic might help. The hookup wire is a bit thick and awkward, I take wire out of old junk stuff like inkjet printers are reuse it. In case this starts to drive you nuts.

The PCB has the speaker ground (JST) trace go around and over to the input jack, and the IC GND is at the the other end of the trace. Sigh. It's not the greatest.

I would the leave the 2.2uF cap where it is, and move the DC power GND up to where the DC jack GND is or C6(-).
Then the signal ground is from the pot GND that goes to pin 4. The input jack GND is far away so I would prefer to not use it. If you must use those input jack pads, I would cut the trace to it and the LED/or to it and at the corner, and run a jumper from the pot GND at pin 4 to the input jack pad. Getting rid of the long loop this board has there, look at the path from input jack GND to the IC lol.

Ideally you want to have the four major grounds: DC power, loudspeaker, IC, signal input - all connect to one point or the star. It's not usually a big deal but some LM386's I built up get finnicky and I have to redo things. The IC has no markings on top?
R2 4k7 is only in if really strong signals coming in, I would leave it out.

It's normal for this to pickup noise and hum when nothing is connected to the input. Putting your hand near it will bring AC hum. It should always be quiet if the pot is turned down though.

[clearchris]

Re: Signal Tracer - Requesting input on requirements

It shouldn't get really quiet like that. Unless the speaker GND is shorting out the input signal or something?

Thank you!!! I have been trying to trace out what has been going on, and I'd unplug the speaker from the board so I could flip the board and take a look, turns out the speaker plug was reversed. Woops.

I couldn't make out all the wiring, this pic might help. The hookup wire is a bit thick and awkward, I take wire out of old junk stuff like inkjet printers are reuse it. In case this starts to drive you nuts.

The PCB has the speaker ground (JST) trace go around and over to the input jack, and the IC GND is at the the other end of the trace. Sigh. It's not the greatest.

I would the leave the 2.2uF cap where it is, and move the DC power GND up to where the DC jack GND is or C6(-).
Then the signal ground is from the pot GND that goes to pin 4. The input jack GND is far away so I would prefer to not use it. If you must use those input jack pads, I would cut the trace to it and the LED/or to it and at the corner, and run a jumper from the pot GND at pin 4 to the input jack pad. Getting rid of the long loop this board has there, look at the path from input jack GND to the IC lol.

Ideally you want to have the four major grounds: DC power, loudspeaker, IC, signal input - all connect to one point or the star. It's not usually a big deal but some LM386's I built up get finnicky and I have to redo things. The IC has no markings on top?
R2 4k7 is only in if really strong signals coming in, I would leave it out.

It's normal for this to pickup noise and hum when nothing is connected to the input. Putting your hand near it will bring AC hum. It should always be quiet if the pot is turned down though.

R2 4k7 was replaced with one of the ceramic caps shown in your schematic.
I cut the ground trace in the top left and top right corners. I ran a jumper wire straight to the cap from pin 4, and ran another jumper wire from pin 4 for the led and where I grounded the two diodes.
Then I soldered all the grounds to a large blob attached to pin 4. Pic attached. Not proud of this.
The chip is marked "16CXY LM386 M-82". It doesn't show unless you get the camera angle just right.

I don't know if this information is useful, but I dragged out my computer speakers and a probe, and traced the signal on this new one. Signal was clean to the pot, but at the pot wiper terminal, I couldn't hear anything, possibly because the pot reduced the signal. No noise at pin 3 on the lm386 either. At pin 5, plenty of noise, even with the volume turned way down. From my readings, this is likely still a grounding problem.

So my options seem to be:
1. Shielded cable for the pot (do I ground the shielding?)
2. Shorten the pot cables to the point I'll have to completely redesign the case
3. Take the 10uf cap out to reduce the gain to 20x instead of 200x.

Also, I noticed that all the other LM386 noise filters on the output have the cap first, then the resistor connected to ground, yours is the only one I have seen that has the resistor (R3 10R) first and the cap (47uf) attached to ground. Does the order matter here?

[clearchris]

Re: Signal Tracer - Requesting input on requirements

I have one comment about the capacitor across the other pads ( in the second picture) is you might want to put heat shrink tubing on the component leads so nothing bad happens ( just a thought )

Nice work

If you liked that, you should see the new pic! If I could get this working, I'll probably coat it with some liquid tape. Either way, it will be 4 point mounted in a protective plastic box with 10mm of space underneath so it should be safe once installed...

[clearchris]

Re: Signal Tracer - Requesting input on requirements

So my options seem to be:
1. Shielded cable for the pot (do I ground the shielding?)
2. Shorten the pot cables to the point I'll have to completely redesign the case
3. Take the 10uf cap out to reduce the gain to 20x instead of 200x.

I shortened the cables on the pot to around 1 inch, no change in noise. Then I took out the 10uf cap to change gain to 20x instead of 200x, replaced the 1M pot with a 10k pot and bypassed the 4R7 resistor. Still noise.

Also, I noticed that all the other LM386 noise filters on the output have the cap first, then the resistor connected to ground, yours is the only one I have seen that has the resistor (R3 10R) first and the cap (47uf) attached to ground. Does the order matter here?

Unless this makes a difference, I'm going to have to rebuild the circuit on perfboard or get a different amp. I think I'll get both the perfboard and a new amp on order and see which I can get working faster.

Does anyone have a suggestion for a more user friendly style amp module? Even if I don't use it myself, when I release the design, I'd like to have a more friendly option available, I'm not sure I would recommend the LM386 to anyone else unless they really wanted the challenge to level up their skills.

[redwire]

Re: Signal Tracer - Requesting input on requirements

I'm not sure what the noise is, can you describe it a bit more.
The grounding is as good as it can get, so it's something else.

Because you are powering this from a wall-wart, the SMPS might be making a lot of common-mode noise/interference that the LM386 is picking up. Try run it off a battery 9V or something.

The LM386 output 10R/47nF the part order does not matter, it's a series circuit.

[eccerr0r]

Re: Signal Tracer - Requesting input on requirements

are we talking about white noise or picking up ambient/PSU noise?

IIRC the LM386's noise floor is fairly high, but it's been a while since I had a LM386 on its own...

[clearchris]

Re: Signal Tracer - Requesting input on requirements

It's white noise. Same level of white noise if the volume is all the way down or turned all the way up (audio input hooked up, but no sound playing). My compuer speakers have white noise if I put my head next to them with the volume all the way down, becomes slightly louder if I turn them up. Can't hear it from 1ft away. The LM386 has constant white noise that I can hear, with the speaker facing away, from at least 5ft away. The noise is loud enough I don't think I could use it to chase down noise in a system I would be troubleshooting.

I peeled the components off the board and put them in my breadboard (pic attached) for troubleshooting. I swapped out the caps that came with the kit for nichicon / kemet and I do think the sound is a touch better, but the noise didn't change. I didn't add in the protection diodes, the led or the resistor from pot to pin 3. When I add the 10uf cap between pins 1 and 8 (200x gain), the noise is overwhelming, so my breadboard layout is probably worse than the PCB, but results without the cap (20x gain) are similar.

A minor miracle has occurred, the electronics shop has some JRC LM386D in stock. From what I have read, that's pretty much the gold standard (if not fake), I'll go pick one up and see how it sounds...

[eccerr0r]

Re: Signal Tracer - Requesting input on requirements

Problem with the chip itself not your design, at least if you make sure you're not introducing more noise for it to amplify...

[clearchris]

Re: Signal Tracer - Requesting input on requirements

I went and picked up the new chip, it wasn't JRC, but it was a natsemi lm386-4 (the 1w version). I put it in, total silence at 20x gain and zero volume. At 200x gain I start to get some AC hum from nearby lights, etc. I did have some chirping, but I think it was either from my PS or from my desk lamp, further testing is required. No chirping in the dark using a battery. Candle light signal tracing anyone?

I can work with this!

So where do I go from here?

1. Dig up some shielded cables.
2. Test longer shielded cables for the volume pot
3. Reroute some wiring in the box or redesign the box
4. Research if it's feasible to add a potentiometer for adjusting the gain.

On point 4, from the datasheet, a 1.35k resistor sets the gain at the lowest setting of 20. There aren't any 1.35k pots available and the nearest ones don't accept a knob. So 2k is the closest available. Using an audio taper 2k pot could yield the most usable range.

I'm probably going to end up reprinting the box, but we'll see.

Thoughts anyone?

BTW, thanks again for everyone's help!

[redwire]

Re: Signal Tracer - Requesting input on requirements

Glad you found it, crypto guys need pure white noise for random number generation, that bad IC could be worth a fortune lol.

The 1.35k resistor is built into the IC (across pins 1-8) to set its gain at 20.
A lower impedance between pins 1-8 increases the gain. Max gain 200 is with a 10uF cap and no resistor there. Gain of about 52 (34dB) is with a 10uF and 680R for example.

You could have a high/low gain switch that just puts in the 10uF+resistor or not.
The input pot lowers gain, the amplifier should be quiet with the pot set low and no long input cables.

[eccerr0r]

Re: Signal Tracer - Requesting input on requirements

I have a 2sc458 that I coupled with a LM741 specifically to make white noise
Just need a power source and hook up an amplifier to it...and it's pure hiss...

[clearchris]

Re: Signal Tracer - Requesting input on requirements

I have a 2sc458 that I coupled with a LM741 specifically to make white noise
Just need a power source and hook up an amplifier to it...and it's pure hiss...

Yeah, this chip would have worked great, except then it picks up AC hum too, probably adding too much non-randomness to it. Last I heard of this (and it has been a long time) people were cracking open smoke detectors, taking out the radioactive bit and attaching it to a sensor...

Glad you found it, crypto guys need pure white noise for random number generation, that bad IC could be worth a fortune lol.

The 1.35k resistor is built into the IC (across pins 1-8) to set its gain at 20.
A lower impedance between pins 1-8 increases the gain. Max gain 200 is with a 10uF cap and no resistor there. Gain of about 52 (34dB) is with a 10uF and 680R for example.

You could have a high/low gain switch that just puts in the 10uF+resistor or not.
The input pot lowers gain, the amplifier should be quiet with the pot set low and no long input cables.

Yeah, no kidding. Super loud, only reason I didn't suspect it sooner is that everyone online seems to complain about hiss and noise with this chip and people respond to ground the chip better. But yeah, after attaching everything to the back, and it not making the slightest bit of difference...

A switch isn't a bad idea, I might even have something suitable on hand instead of trying to get a 2k reverse audio taper pot. Initial testing shows that having longer leads on the 1 and 8 pins doesn't affect the sound much if at all.

I have been playing some songs through this setup so I can compare when I add the longer shielded leads, and I'm now pretty impressed with this setup, it sounds far better than my logitech computer speakers. I did buy a supposedly decent speaker, IIRC, I spent under $15 for a pair, but you never know what you are going to get when you import electronics...

BTW, does anyone have any opinions on which side (or both) of the shield from the shielded wires should be grounded?

[clearchris]

Re: Signal Tracer - Requesting input on requirements

So a theoretical question. Why would you, by default, want to set your gain to a static value and weaken the input signal instead? Seems to me that you would want to leave your good input signal, and adjust the gain to the minimum level required to get the sound output you want, since the amplification by the chip is what introduces noise/distortion. What am I missing here?

[eccerr0r]

Re: Signal Tracer - Requesting input on requirements

mainly for an adjustable gain circuit, as gain changes, distortion changes?
Better off attenuating input and have a fixed gain - with no distortion.

[sam_sam_sam]

Re: Signal Tracer - Requesting input on requirements

I have learned a lot with this post keep up the good work

[clearchris]

Re: Signal Tracer - Requesting input on requirements

mainly for an adjustable gain circuit, as gain changes, distortion changes?
Better off attenuating input and have a fixed gain - with no distortion.

So the usual is to set your gain to a set amount with no distortion for a certain type of input. That makes sense. In this case we could be dealing with different types of input though and specifically wanting to chase down subtle noises or just general troubleshooting at full levels...

I have learned a lot with this post keep up the good work

Yeah, me too. I was really expecting this to be one of my simpler projects, and it's turning out to be a lot more complicated.

At least I found some perf board. Does anyone have a good schematic/layout tool they use? I haven't looked into it yet, but I think I'll use something like that to get the arrangement to how I like it first...

[eccerr0r]

Re: Signal Tracer - Requesting input on requirements

Err... no, you design your circuit for the max gain you need and fix the circuit so that it does not distort at that gain. Then you can assume for any input it will not distort - unless you're overdriving it but this is not a problem because you accounted for it already - else it wouldn't be the nondistorting max gain you need...

Anything can be made arbitrarily difficult...

[clearchris]

Re: Signal Tracer - Requesting input on requirements

Anything can be made arbitrarily difficult...

This may be my new motto.

Well, I decided against using my crusty 20 year old perfboard, and the new perfboard arrived 2 days ago. It's going together pretty well. Unfortunately, my garage has to be over 100F, so I'm doing it in shifts. I should have something soon.

If I'm really lucky, I might get it into the same general form as the old board with an adapter and cutting the end of the board. We'll see...

[clearchris]

Re: Signal Tracer - Requesting input on requirements

I had to spend some time out of town, I'm back on the project. It's all together on the perfboard and I used shielded cable for the connections to the pot. I'm not picking up any noise from the flourescent lights any more.

However, I'm also not getting much amplification either. While connected to the 3' speaker If I turn up the volume past a quiet speaking voice volume, the audio starts dropping out at parts. I can turn it up with headphones connected I can raise the volume much further, to where the headphones start distorting.

I suspect something was damaged when I was breadboarding. I'm going to take the caps off tomorrow and test/replace, anyone have any other possible suspects?