The smallest audio amplifier for everyday listening (SMD, discrete, fully symmetrical, low power)

For a single frequency analysis (for power dissipation, etc), make the R+jX (or R-jX) with a resistor and inductor (or capacitor). Set the R and L values to get the phase angle you want.

With a 3.07 ohms resistor + 1.02 mH inductor I get a 4 ohms load at 400 Hz, with 40 degree phase angle.
Now, if I put this at the amplifier output, the maximum V*I product is at 15V*2A. This is exactly the SOA of D44/45H. So it is too risky, I think ?
But the MJE15030/31's have SOA of 4A at 15V ! Much better.

Can I assume that this 4 ohms / 40* phase is the worst scenario for the output power transistors ?

I have a pair of speakers made by me long time ago, with PA-800 and PCT-300. They have LR4 pasive filters with 2.9KHz crossover. I tried to simulate the impedance and phase that my amplifier will see. With the TS parameters from their datasheets I calculate an equivalent model, using the instructions from this page. And the impedance and phase plot is like this... As you can see the minimum impedance is about 4 ohms at 400Hz. But the speakers ar rated as 8 ohms ! I don't know if it's an error in mai calculations or that is a normal behavior... And then, those 12 ohms at 2KHz, are strage too... The phase is 60* at 100Hz/8R.

Edit: Hmm... I simulated without the filters and now the lowest impedance is the bass speaker DC resistance, 6.5 ohms. So it means that the filters lower it to 4 ohms...

1748779294559.png
 
Last edited:
No, I don’t think the D44 is too risky, with it exactly at the S/B limited SOA. You will get a little lift even with a 100 ms pulse, and will be enough to cover temperature rise. But this experiment is EXACTLY why you don’t raise the voltage over +/-18 with those parts. The supply will likely dip under load anyway. At +/-25V you are squarely in the red, and that is risky.
 
  • Like
Reactions: Sigismund
Ok, now that we're done with transistors, let's move on to something else: the Zobel network and the output coil. The idea is to miniaturize these as well, using SMD components.

For Zobel I thought of using two 47nF capacitors in parallel, in 1206 package, C0G type, 50V.
And a 1206, 0.5W resistor. A full power sine wave of 20 KHz will dissipate 90mW into this resistor. So I selected a 500mW package to stay cold...

The damping resistor in parallel with the coil is also 0.5W 1206 package. Its dissipation is 30mW...
So far, so good... But now comes the hardest part: The Coil !
I know that here is always recommended to use a simple air wire-wound coil. But this is to big for what I want here...
Is there any chance that I can use a SMD coil made on ferite core ? Like this one: SM7850-1R0MT ?
It saturates at 7A and the maximum current throug it is 3A in my amplifier. It also has high Self Resonating Frequency of 145 MHz (I understand that this is good). But I don't know if there are other aspects that matter here, that such a coil doesn't meet...

1748855483414.png
 
Lets calculate around the coil. Assume we have 20% distortion in the inductive part. At 10kHz the reactance will be about 6m ohm.
6m ohm/8 ohm is about 1/1000. 20%/1000 is 0,02%. Hardly hearable but could be near for some people.
At 1 kHz it is 10 times less = not hearable. Hard to make a choice. We do want distortion a safety margin under the hearable don´t we?

I bought 10 Muratas from LCSC. They will last for a while. Cheap so i use COG. But they are large: 1206
5+ $0.1042
50+ $0.0865
150+ $0.0775
500+ $0.07
 
  • Like
Reactions: jan.didden
I have used ferrite cores in loadspeaker filters. They demand a lot of work in calculation and much larger ferrit cores than anyone would expect.
The choice of ferrite material is very important too. I didnt really gain in weight compared to an air core but the resistance was lower.
 
Linearity of the inductor matters more when the cutoff frequenc(ies) are within the audio band. As you go lower in frequency here the reactance is negligible. If it saturates and drops even more, so what? Yes, it will add some distortion. You just have to ask yourself if you can live with it, given the size reduction. How close to saturation are you really going to operate, on a 10 watt amp typically backed off 10 to 20 dB anyway? It’s not a 400 watt per channel PA amplifier.

The other way to make an inductor without growing the size much is run a couple of turns around the perimeter of the PCB. Then you have to live with the coupling (to everything, including enclosed ferrous metal) that results.
 
I have another idea. What if I make an air coil, but not in one layer (as I did here (1uH coil)), but on 3-4 layers to shorten its length ? The size will be acceptable...

I was thinking the amount of inductance you could get in an SMD footprint size would only be in the nanohenries. Going multilayer (more than two) wouldn’t help much without blind/buried vias. A bunch of through holes in the center to make connections doesn’t help with the size. Got a 2.5D EM simulator handy? It would be interesting to see what you can come up with. Takes longer to set these problems up than it does to run them these days.
 
would only be in the nanohenries.

I'am not sure I understand, are you talking about a coil made from PCB traces or someting ?

My current output coil that I used in my previous amplifiers, has a PCB footprint of 10 x 17 mm (one layer).
I used this calculator and I made a 1.6 uH coil with 3 layers that take only 8mm circle of PCB space, with about 8mm height also (I put it verticaly).
I consider this a fair compromise between the size and the quality... From 10x17mm to 8mm round, is someting....
DC resistance will be around 20m ohhms (with 0.8mm diameter cooper wire (for a max peak of 3A).

screenshot.21.png
 
Yes, I was talking about making it out of PCB traces. That’s how you make inductors out of nothing at all. Unfortunately, inductance can be pretty limited, and so can current carrying ability. If you are making it a centimeter or two on a side, it looks like you can get something workable. If you have that much space you might as a well take what you normally wind around a two watt resistor and stand it up like a Tesla coil. Add a little epoxy and its not going anywhere. But in an 0805 or 0603, you’re OS-OOL (without ferrites). I’m used to customers b*tching about having to put something like that outside the package, and at the same time raising a stink about the RF PA having spurs.
 
Be aware not to wind coils on low ohm metal film resistors. They can affect the inductance in two ways. The metal will function as a shortcircuit turn inside the coil and maybe function as a magnetic core.
Wirewound resistors dont make a shortcurcuit but maybe can function as a magnetic core.

I wind my coils on an 8mm drill foot end. First one layer then take it out and fix it with glue. After the glue has hardened the second layer + glue. Wait and then the third layer. Close to 13 mm outer diameter with my 0,75 mm cupper thread.
 
  • Like
Reactions: Sigismund
The old school way to do it was to wind a single layer of turns around a 10 ohm 2 watt carbon comp. Takes a 5 watt MOX to get that form factor these days. A wirewound resistor will still couple to the inductor because it’s a coil too. Precision isn’t required, and one can always take it up or down on the turn count and/or trim R to get it to measure the way you want it to.

If you want a pure inductance, or as close to one as you can get, you need to pull out whatever you used for a former. Then you are just going to put a resistor in parallel with it, so one wonders what the point was.