"Small Audio Amplifier" from Ludens.cl

[Kipper4]

I Was there. If anyone else here was, perhaps you can tell me if we had a good night.
Cheers beer.

[tubegeek]

I'll tell you what *I* remember from that night:

[tubegeek]

I ended up building a quick and dirty dual power supply after all: I found I had everything I needed including a small dual secondary transformer (nominal 12V+12V, delivers about 15V+15V under this light load) as well as a full wave bridge that I was never going to use for anything else (I usually use 4 fast/soft discrete diodes when I build a FWB rectifier.) I also came up with a nice little heatsink from my junk vintage parts collection and mounted both of the output transistors to it so they would stay the same temperature as each other.

Schematic (updated thanks to highwater's keen eyes):


Partway through the build: power supply & output transistors:


It's working well, just some hum that'll probably disappear once it's in a box. Sounds pretty OK with a nasty little speaker from a junked salvaged Danelectro mini combo. I'll replace the previous amp I built in the 8" combo cab once I get it boxed up.

I reduced the 100R B-E resistors down to 47R because the NE5532 dual opamp draws 2x the current of the original NE5534 and I figured I ought to keep the voltage drop about the same as in the original design, it measures roughly 360 mV.

The odd value of the feedback cap came from a bit of tinkering with the gain. I estimated that I wanted about a maximum gain of 80 and that ended up being too much for my bench amp needs. Originally I had had a 10K pot in series with the 4K7 feedback resistor on the first opamp for gain adjustment. I found that I didn't want even that much gain with the 10K pot turned down to 0, so I took it out. I had had a 27K feedback resistor against 4K7 in the second opamp originally and I knocked that down by putting 68K in parallel with the 27K. I'll probably just put a 1M pot at the input and leave the gain fixed the way it's drawn in the schematic.

With the feedback at its original values, I was getting more distortion than I wanted and I thought the power supply caps might be too small. I don't seem to have any 1000uF/35V caps on hand, just the 330uF/35V. With the increased feedback maybe the smaller caps will be OK, I'll decide once I get a better speaker connected. It's hard to tell what's the speaker farting out and what's amp distortion at this point.

Anyway this is certainly close and it will be a useful item.

[highwater]

I can't help but think that you'd be better-off with a 5534 by itself for the output stage... the transistor drive* is based on how much current the opamp is drawing, and the input stage contributes to that. When the other half was a rail-splitter, that was almost definitely a problem. Now, they're both moving the same signal around, in the same direction... but it's beyond my pay-grade whether that makes it completely OK, or just close-enough.

Also, you have the lower 47Ω resistor in the wrong place on the schematic... OTOH, if you're anything like me, that only increases the odds that it's in the right place on the real thing.


*:  See Fig 10.15 in this Analog Devices app-note. Pretty-much the same idea, but with a fancier current-mirror than you and Jonny are using.

[tubegeek]

Now, they're both moving the same signal around, in the same direction... but it's beyond my pay-grade whether that makes it completely OK, or just close-enough.

Good question. Let's see whether anybody smarter than me replies (present company excluded.)

Looking at this setup as a voltage-to-current converter (as it is described in that interesting app note!) brings up another question that I had.

Should I have the ubiquitous .1 ceramic bypass cap between pins 4 and 8 in this design? I've never been sure and I keep forgetting to ask the assembled experts. I do have it and I *have* had it all along. So I just updated the schematic to make that explicit.

Also, you have the lower 47Ω resistor in the wrong place on the schematic... OTOH, if you're anything like me, that only increases the odds that it's in the right place on the real thing.

Thanks - oops. Only in the drawing though, I fixed it so it is correct now. I think that happened when I did some photoshop cleanup of my notebook drawing. One of the things I like about this circuit is how symmetrical it is - it definitely makes it harder to mess up the build.

*:  See Fig 10.15 in this Analog Devices app-note. Pretty-much the same idea, but with a fancier current-mirror than you and Jonny are using.

Thanks for the link, I'll take a look.

[jonny.reckless]

It's looking good now!

I'd put as large smoothing caps on the power supply as you can reasonably fit. I'd be tempted to go as high as 2200uF per rail to keep the ripple as small as possible, which will help transient current delivery into the load and also reduce the 120Hz hum a little bit. It will also mean the 120Hz "riding on the wave" you get on the load under clipping will be less noticeable.

Also keep the wire from the transformer to the smoothing caps separated from the wire from the smoothing caps to the amp ground, with no "shared copper". The input half of the circuit has half wave rectified current pulses flowing it in which will appear as rather harsh sounding hum if allowed to get into the feedback circuit. Traditionally you want to "star" ground a design like this. You should be able to get this totally silent with a bit of effort in the grounding and wiring layout.
This approach is pretty simple and is usually what I use:

Here's an example of star grounding for an audio amp:

[jonny.reckless]

I can't help but think that you'd be better-off with a 5534 by itself for the output stage.

Seconded. You're using the power supply pins of the op-amp as the "output" of a phase splitter with current outputs. It might work with a dual op-amp since both halves share the same signal in this instance, but I wouldn't recommend it.

Also, the 100nF "decoupling" capacitor you have from pin 8 to 4 is performing a subtly different function here. Since the cap will shunt high frequency current into the op-amp, it is acting as a form of loop compensation capacitor. I usually don't fit the 100nF "decoupling" cap but instead use 100pF from base to collector of each output transistor if necessary. Either approach is OK to tame HF oscillations, as it acts as a dominant pole on the whole NFB loop formed by the op amp, its power supply pins, and the output transistors.

[tubegeek]

OK, OK. Next time I order parts I'll get a few 5534's and I'll try another one of these - I don't think I'll actually take this one apart though. I think I'll just knock down the gain of this one even further. It actually does NOT sound better connected to the "better" speaker - I think that's the power supply caps showing their true colors though. I must have something suitable that I can fit, I'll keep digging through my Big Box O'Electros until something turns up. There's got to be at least a pair of 1000/35V salvaged from something in there.

Re: ground scheme: power cord ground connects to the secondary center tap, and the diodes and filter caps are grounded right at that connection. A ground bus made from a strip of vero board bolstered by some bus wire takes off from there and everything audio is grounded outside the current pulse loop, onto the vero. The vero's a combination of a low inductance planar ground bus with some extra added copper further reducing resistance. The input cable's shield connection is at the most distant end of the bus, at the edge of the perfboard. The chassis grounds ONLY at the input jack - not a UL-Listed approach, but it's working fine and I was careful about shorts etc. It's a pretty good layout, I think. The whole signal path is probably less than 3 inches of wire and passives all added together, and it's certainly quiet enough that I can hear the noise difference between single coils and humbuckers quite clearly.

I can try the alternate bypass cap scheme pretty easily, I think - the dual base-collector caps will probably end up on the back of the perf though.

Unplugged this amp hums pretty loudly. It's half-boxed-up in a chunk of "C"-channel with an open top still. With the guitar plugged in it's quite acceptable.

Thanks again for all the advice/help/experience. Interesting design with the "voltage to current converter" aspect of it, just gradually getting my head around it.

[tubegeek]

I made the following changes: dropped the first stage gain down to unity, and replaced the filter caps with 2200uF. At only 25V rating they may not last forever, but forever is quite a while anyway. It's what I had in my bin that I could find easily. These changes are for the better, I think. I'm wondering whether I have the perf-surgery skills to change the layout to parallel both op-amp halves together in order to fake a single op-amp for the sake of the current-conversion feature: this gain level suits me OK as of now. There are two other possibilities I am considering:

a) swap the gains between stages so the input amp provides the ~4x gain and running the 2nd (output) op-amp at unity gain to increase the feedback on that stage.

This would only require another 2 resistor swap operations: put the 1K feedback resistor back up to 4K7 and remove the parallel 27K||68K pair and replace it with another 4K7. I'm pretty sure I can do both of those without ruining the perfboard.

b) up the gain of the first opamp to something like 10x, and then replace the link between pins 3 & 7 with a volume pot. Convert the 2nd op-amp to unity gain as above.

Still to try: remove the 100nF from pin4 to 8 and provide 2 100pF caps directly on the transistors Base-to-Collector.

[tubegeek]

Still to try: remove the 100nF from pin4 to 8 and provide 2 100pF caps directly on the transistors Base-to-Collector.

Did that. Maybe sounds better, I dunno. Looking at the scope, the waveform is clean as a whistle with the output as large as I could possibly stand it. Going to leave it alone for a bit, I think.

[Rob Strand]

Did that. Maybe sounds better, I dunno. Looking at the scope, the waveform is clean as a whistle with the output as large as I could possibly stand it. Going to leave it alone for a bit, I think.

For devices with outputs going to the real world I usually try to break it by putting different loads on the output.    Low Z, high Z, parallel caps like 10n, 100n caps (maybe in between values if I smell a problem) across the output and at the end of a cable, series inductors, different cables, speaking load emulator.   I also check clipping the amp with the different loads.    Sometimes you will see oscillations, overshoots and weird behaviours.   When you try different compensation schemes some will be tougher than others.   In some cases you have to watch out as one compensation scheme might be tougher against different loads but the amp bandwidth is lower so you aren't comparing apples to apples.

DC power supplies can suffer the same instabilities.  I know  some HP or Tektronix power supplies used to oscillate with certain loads - not good!

[jonny.reckless]

You'd need a Zobel network to make (pretty much) any feedback amp unconditionally stable into a capacitive load. I never bother with them with home builds, since I've never seen a speaker that has an impedance curve like that (it's pretty much always inductive at the top end, even with a passive crossover), although I always used a Zobel when I was designing power amps commercially.

You can tell a lot about the dynamic behavior of an amp by looking at how it handles square waves, especially into a reactive load like a speaker. Ironically amps with some ringing and overshoot often sound more "exciting" for music instrument (not HiFi) applications.

[tubegeek]

A very minimal update: I replaced my bench speaker with a lovely alnico 12" I had in my basement, mounted in an open back box that's just a little bigger than the the basket.
This amp sounds even better through the new speaker. (So does my Supro 1616T clone mini-head.) Very pleased to have built it. I'm considering making a 2 channel version to run the audio for my TV through - if I can figure out how to place a couple of "real" speakers nearby without uglifying my living room too much.

Haven't been on this site much lately - been building guitars (from kits) and tweaking/modding them.

LOTS of time on my hands lately... stay safe everyone.

[Mark Hammer]

Where do you get a Grohl-model kit?

[tubegeek]

Where do you get a Grohl-model kit?

Well spotted! That's a Trini Lopez Standard kit, that guy Grohl's a Johnny-come-lately. But I like his taste in guitars!

Got it from thefretwire.com.

Not so sure I'd completely recommend this kit. You'd need to go into it with your eyes open.

1) The headstock of the kit neck is all wrong. I did some serious carpentry to enlarge it and reshape it to make it much more accurate.

2) The neck is not a mortise & tenon joint: it's shaped more the way a bolt-on is shaped, the mortise is the full width and depth of the neck. The problem is that it hasn't got a proper heel after all is said and done. I had to trim off the body binding in the area where the heel cap was supposed to be and then make and shape a smooth rounded heel cap to match a real 335/Trini shape.

3) Haven't wired it up yet but the humbuckers are most likely ceramic and will most likely be replaced. Playing it acoustically, well, now that I did the bridge swap (see #6) it really sounds VERY promising. To be fair, thefretwire does offer pickup upgrades for a bit of extra dough.

4) Sourced my own vintage-style 6-on-a-plate tuners, it came with some lousy Fender-style individual tuners. The headstock was ready for a Strat shape, but that really wouldn't acommodate the Firebird-style shape.

5) Thefretwire.com did kindly throw in a stop tailpiece, studs and inserts so I could make it a stop-tail. Had to drill for that part myself, but they were agreeable to my request.

6) But the "ABR-1"-style bridge from the kit was really crappy - I upgraded the bridge & stop bar & mounting hardware to a Gotoh replacement set.

7) neck still feels like it's going to need some more work - possibly just a level & crown but I'm not sure. Not something I'm experienced with so I'm nervous to take it on.

So, it was a hell of a lot of work to get it to this point. Plan is to finish it in a "non-standard" color, British Racing Green.

Somehow I've found myself with a lot of time on my hands in recent weeks, so this project has made more sense than it otherwise might. I've wanted a custom-colored stop-tail Trini ever since I saw John Hiatt playing a black one in the '80's. I was too cheap to buy one back then, before people discovered that they were not only exactly the same guitar underneath as an ES-335, but they were ten times as cool-looking too. So this is the best I'm going to do, but once I realized I could, well, I had to. I have maybe $250-275 into it now which I can certainly justify.

If you want more details, look here, I've put a full saga and some in-progress pictures up:

http://www.kitguitarsforum.com/board/viewtopic.php?f=53&t=9156

To get slightly back on topic, the 12" alnico is SWEET with this crazy little amp. Been enjoying giving the combination a workout and I even made the effort to mount the amp board neatly into its hunk of C-channel.

I may not be getting out much, but I'm having some fun. Hope all is well with you and yours, Mark!