Audio Grade Electrolytics and the Ever Elusive ESR...

[Bunkey]

Why do you think you want zero or near-zero source impedance??

You're speaking to somebody who just spent the best part of an hour deciding which M3 bolt would look best for attaching the heatsink to the TO-220 casing 

Edit: Better omit that before I call my own sanity into question.

I just obsess over every little detail but as long as I can keep the project moving forward it tends to be advantageous. Never truly satisfied if I feel I've left something on the table.
If it's any consolation I am a lot less inclined to pick out the perfect capacitor now than I was before. You're keeping me right.

you also need a local Vcc-to-ground bypass cap right at the chip; generally a 10uF or so and a 100nF will do it.

Whether a chip's datasheet says so or not, it's good practice to put a power supply bypass cap as close to the chip as you can get it.

I'd actually breadboarded the amp like this before I started to overthink everything so I'll keep that in place and add another to pin 7.
So the ESR of this bypass cap(s) is going to parallel that of the 220uF cap at the regulator side - and that's why you would suggest a small resistance in-line at the output?

PS. Rob, what are you using to set up and simulate those graphs? I'd quite like to play around with some of these values myself. Cheers.

[Bunkey]

Ahh I've managed to copy your setup on LTspice, Rob

Just not sure how to get it to graph multiple values like that.. I'll do some poking around.

[Rob Strand]

That's good too, but you also need a local Vcc-to-ground bypass cap right at the chip; generally a 10uF or so and a 100nF will do it.

Whether a chip's datasheet says so or not, it's good practice to put a power supply bypass cap as close to the chip as you can get it. Kemet had a great paper on capacitor basics, titled "capacitors_basics_applications.pdf" if you can find it.

There's actually a longish thread started by Paul Marossy a couple of years ago.   He had a design he had made in the past with LM386's.   He built it with "modern" LM386's and had no end of oscillations issues.   After trying many things out of "oscillation fix toolbox" and addressing some finer points he didn't have much luck.   After some perseverance, Paul found the only thing that worked was big caps *right across* the supply pins.

Ahh I've managed to copy your setup on LTspice, Rob

Just not sure how to get it to graph multiple values like that.. I'll do some poking around.

The ".step" function does that.   The details are,

- Set the part value using an expression.  In my example that's the cap ESR.  The expression is best written in braces {}.
  FYI, while I've shown a single variable inside the braces you can actually write whole expressions like {2*ESR+5}.
- The expression contains a variable, in this case ESR.   Spice uses the term 'parameter' for variables.
- The step function assigns different values to the ESR parameter on consecutive spice runs.
  .step param ESR list 10m  100m 100m
   "list" means the ESR values come from the list of values following.
    Spice lets you define other ways of stepping, like linear steps or log steps where you just put in the range of values.
- FYI, you can step more than one variable.   However, the output can be confusing.
- The line,
   .param ESR 300m
    does nothing when step is present.
    This lets me comment out the step function and the ESR value will get set to a known good value; otherwise ESR is undefined.
    Sometimes stepping creates too much clutter, so I turn it off.

Oh, you need to click the .OP icon to add a .step.

[PRR]

> "capacitors_basics_applications.pdf" if you can find it.

I found a 61-page "power point deck" (800kB PDF file) by that name and company on the Euro Space Agency sub-site:
https://escies.org/download/webDocumentFile?id=62173

Some amusing graphics:

[duck_arse]

You're speaking to somebody who just spent the best part of an hour deciding which M3 bolt would look best for attaching the heatsink to the TO-220 casing 
Cheers.

this has me interested. what did you choose, and why?

[R.G.]

> "capacitors_basics_applications.pdf" if you can find it.

I found a 61-page "power point deck" (800kB PDF file) by that name and company on the Euro Space Agency sub-site:
https://escies.org/download/webDocumentFile?id=62173

Yep, that's it. It gets across in simple language the ideas that:
> Capacitors can be modeled as ideal caps with small series inductors and inductors to a HIGH degree of accuracy
> ESR and ESL isolate the capacitance from their loads
> For effective decoupling, you have to put capacitors physically local to their load, perhaps in a string of local decouplers.

[Bunkey]

You're speaking to somebody who just spent the best part of an hour deciding which M3 bolt would look best for attaching the heatsink to the TO-220 casing 
Cheers.

this has me interested. what did you choose, and why?

Sometimes I say things which make sense or seem amusing to me at the time and then I log back in and read what I've written previously and I regret saying anything.
I probably look like a complete idiot to some people, myself especially. I seem to take myself far too seriously but there's not much I can do about that.

If it's a genuine question then I was just working out the total thickness of the assembly vs the various nut and bolt combinations available so I didn't have a lot of material sticking out the back of the sink and looking clunky. It's also close to the side of the enclosure. The complete regulator assembly needs to be squeezed into a specific area so piecing it together whilst considering things like ground routing, heat dissipation, physical component size and reducing the amount of wiring floating around the place which might cause hum issues is quite a task. I'd like its appearance to reflect the effort I've put into designing it too. I have a bolt supplier who does nicely formed A4 stainless bolts I use on some of the engines and things I build in my other life, so I've gone through their catalogue and picked out a cup head type with a hex socket that looks a bit better quality than a standard A2 phillips head bolt. The A4 alloy also has a nice blue hue to it as well as being more resistant to corrosion, maybe you know this already.



Welcome to my process.

Sketch at bottom - The hashed thing is the sink. The protoboard on the card is what I'm building it onto placed on a layout sketch of the amplifier; you can see the spacial requirements. The regulator assembly being breadboarded on the left is exactly what I don't want in terms of that bolt sticking out the back (uses a different sink too).

Also if anyone spots a problem in the schematic then please let me know. As it's a DC input from a wall wart there is no 'earth'. I figured having the switch connected right at the amp's power terminal with the central ground point for the amplifier between it and the additional 1800u reservoir/smoothing cap would help to mitigate noisy audio ground issues arising from not great switch contacts (it's a flip flop type out of a Les Paul).

I'll probably stick a fuse right at the +ve terminal too and maybe a reverse biased diode across the input for polarity protection?

[Bunkey]

Sorry I felt I was becoming the butt of a joke and got a bit upset. I don't understand half of what you guys are talking about.

Check this thing out though, freaking awesome with the new power supply:

Low gain and volume here. Breaks up into nice compressed light fuzz with the gain set a bit higher, really dynamic when playing and the guitar volume control cleans it all up.
The clipping is all pre-amp gain with seperate volume preceding the LM386 - that's set to the internal gain so there's clean headroom for the whole volume range. Could do with implementing the bridged pair for a bit of a lift at low gain settings though.

Still gotta tweak a few things in the tone shaping and mess with the cap values on the regulator.
Obviously it's quite cobbled together too but man it's working well at this stage.

I've got a 5mm shit ply speaker baffle in that cabinet atm (2x 4" speakers) and I think it'd benefit from a solid(er) 6mm birch ply baffle instead. The thin stock gives the cab more life but there's a strong resonance in the mids at higher volumes and reckon the baffle is most definitely the culprit. The eq is sweet at low volume so it'd make sense.
This thing goes loud even with just the one 386 in it! Quite surprising.

I'll maybe turn the gain up tomorrow and get a better clip.

Might copy this over to a dedicated build thread if there's interest - don't mean to talk at you guys, I just like to document my work.

Cheers and thanks for the help you've given me so far.

[Rob Strand]

Low gain and volume here. Breaks up into nice compressed light fuzz with the gain set a bit higher, really dynamic when playing and the guitar volume control cleans it all up.

The break-up sounds good.

In the end it was all worth while.

[duck_arse]

really, if we can't talk the relative merits of screw types in a forum like this, what hope is there for the wider world at large. and I assure I am not doing any butt jokes just at the moment.

two points to make - the black/blackish screw will dissipate heat more like a heatsink [than a silver screw], which is also black for thermal reasons, and, surface area - a longer screw will present a greater surface area to dissipate heat, and right at the source of the heat. but the thread would be to the fins-side of the heatsink, no?

an edit three point - thermal grease. are you using any between tab and heatsink? or a sillicone washer of some modern description?

[Bunkey]

really, if we can't talk the relative merits of screw types in a forum like this, what hope is there for the wider world at large.

the black/blackish screw will dissipate heat more like a heatsink [than a silver screw]

an edit three point - thermal grease.

I should get you on board with this haha.

We employ the same theory in spec racing engines where you can't swap out the standard radiator for a more efficient unit but there's nothing in the regulations to stop you painting it black. I do wonder whether a layer of matte black enamel has more of a physical insulating property than it's worth in effective dissipation of radiation though. Can't see the same issue with an anodized heatsink.

Ended up getting a tube of the adhesive type so even if I can't settle on a mechanical fastening we should be ok. 30ml is a surprising amount and there are a lot of unsuspecting items going to get thermally bonded over the course of its shelf life I think.

[Bunkey]

A few interesting observations:

My breadboarded circuit made use of a Vishay 256 cap I had spare whilst I was sizing up the parts I need for the finished amplifier.
I really wanted to use the Panasonic FC caps that I'd spec'd for this supply, and designed the board around, but every time I put one in either the 220u C2 position or indeed the 1800u reservoir there's something stressed about the sound (a kind of upper-mid bloom) that's not particularly inspiring to play. I've gone back to a Vishay 256 1200u reservoir cap and a KZ on C2 - The amp's voicing is the same but it's much more relaxed and musical sounding.
I wondered if it was just an effect of increasing the capacitance - Generally speaking increasing the capacitance increases the definition and clarity of what's being played - but then auditioning a 3300u Vishay against the 1800u Pana has exactly the same outcome in how musical that sounded. In fact in terms of what sounds good I would say there is more of a difference between the Vishay caps and the Pana FC than there is between using a 1200u or a 3300u for the reservoir - Either Vishay sounded musical in it's own right here; the Pana was industrial by comparison.

This experiment has catagorically confirmed for me that caps aren't just caps (whether they're in the signal path or not) and the sound of a particular amplifier is more than just its circuit topology.

When you've got a really simple circuit like this it becomes obvious there's a considerable distinction between various different capacitors of a given value.


As for ESR to dampen the resonant peak of the supply, I've got a few 100mOhm and 1ohm wirewound resistors that I'm going the try in series with the low-esr KZ cap and in series with the output as RG suggested, to see if there's a tangible difference in this application.

[Bunkey]

...of course the one thing that is different between the Vishay and the Panasonic reservoir caps is the ESR - 150mOhm and 30mOhm respectively.

Having just tried the 100mOhm resistor in series with the low impedance (of unknown value) KZ output cap, I'm inclined to leave it there based on what I can hear - but that's about as much of a conclusion as I'm willing to draw for now.

Edit: Definitely got a bit more presence with the added series resistance on the regulator output cap. Neither is more musical than the other like was the case for the reservoir caps though; it just cleans up the highs a bit, as though the guitar isn't competing with the supply at those frequencies or something.

[Bunkey]

There are no definitive answers in this game it seems. One point I can make is that in practice a 22u Pana FC is working best for the Cadj bypass - better than a 10u listed on the datasheet or any values above 22u that I've tried.