Really really small TPA3132 amp

[IO390]

This is my first post here other than the introduction, so hi )

So earlier this year, a friend of mine had a cheap class D amp from ebay and it broke (turns out the battery wasn't charged...). He was pretty fed up with cheap ebay stuff so he asked me if I could build a better one.

A browse through TI and I settled on the TPA3132. This is basically the same as the TPA3116 (I think) but in a QFN package. The first set of boards were 1.5x2.5" in size and were just the datasheet schematic. I was completely lazy and just stuck everything in one ground plane and didn't pay much attention to the layout.

So we built a few of the boards and even with the sloppy layout they sounded MUCH better than the ebay ones, and were already much smaller. He had paid for all the parts so he took the boards and left me one to play with. I gradually cut the thing up and tried different components to improve it as one of my projects on the side.

Fast forwarding to now, I've completed the assembly of a new amp, which measures 1.5x1", by about 3/4" high when populated, uses the TPA3132 with ferrite bead output filters, has far more responsible use of ground planes among other improvements. I also shoehorned screw terminals and connectors onto it, so it can be used with no soldering. BTW the input is single ended, as the idea of this thing is just to be plugged into a phone or something.

By comparison to my composite LM3886 amp, the difference in sound quality is only just noticeable (to my ears). It doesn't affect any of my FM radios whatsoever, but obviously still radiates (depends mostly on speaker wire length). It also barely gets warm when running on max volume. Works great off a laptop power brick, or a battery if you want a portable set of speakers.

As for the specs, the gain is 32dB with an input impedance of 15K. The output power is 2x25W(8Ω) 2x50W(4Ω), and it runs on anything between 4.5V-22V.

I realise not everyone needs such a small amp, but I've built 10 so I'll keep one for me and sell 9, if anyone is interested. Would be perfect for a set of portable speakers. Note that they are hand assembled, so some of the boards have a bit of excess flux etc... on the non-reflowed components, but all work well. I've made up 500mm long 3.5mm jack to the 3-way female header sockets to mate with the amp. I couldn't fit a normal jack socket on the board as I ran out of room.

Anyway this thread isn't just trying to sell these things, any and all questions/criticisms are welcomed. But if you are interested in one, let me know.

(Reading it back it looks like I have something against items from ebay, what I mean is cheap crap from ebay )

Edit: All boards now gone. Also posted on EEVblog so they went very fast. Thanks for all the interest. I'm now working on a new version (building another 10 of them) with full LC output filters and some other minor changes.

[Saturnus]

What are the physical dimensions in real numbers (metric)? Inches typically swing between 24mm and 28mm.

[IO390]

I always thought an inch was 25.4mm. I design my boards in imperial as that's pretty much the standard for board design, although things are changing a bit now.

With the callipers, the dimensions (rounded to 1mm) are 26Wx38Lx18Hmm.

[IO390]

I ought to add that one of the biggest improvements in sound quality was separating the AVCC and PVCC supplies. On the old boards the analogue supply was tied right to the power supply pins. On the one I played with I cut the trace and put the analogue supply through a ferrite bead and added extra decoupling for it. So I incorporated it here.

Note that there is a footprint for a SOT-23-6 package on the board. This is because I had the bright idea of supplying the analogue through a separate 5V reg. I only tested this on an old board after I had sent off the gerbers for the new ones. The AVCC and PVCC inputs need to be within about 0.5V of each other, or the output develops serious noise.

[Saturnus]

That means it's almost exactly the same size as the old amp32 from 41hz (slightly larger height but only by 2mm). The important thing is the width. It cannot be more than 26mm, not even by 0.1mm, if people are to use them as replacements for the amp32. If that's the case I might know some people that'll take them of your hands pretty quick.

[IO390]

Thanks for the information. I've included some photos for you, those are nice calibrated callipers so that's spot on.

In hindsight I should have put those two electrolytics on the other side of the board, that would have shaved nearly 6mm off the height. But that's fairly easy and can be done in the next revision.

[IO390]

I have two claimed, that's all the ones I have built so far. I've just ordered amp chips to finish the rest, they will be here next week and I can finish the rest.

If there is demand, or if anyone thinks there might be, please let me know and I can get another batch built ASAP so no one has to wait.

[den_hr]

I'll most likely take two - PM sent...

[AudioLapDance]

Originally Posted by IO390

... TPA3132 with ferrite bead output filters...

Short speaker wires ...

Originally Posted by marce

...
There is NO such thing as a filterless class D even these low power devices rely on the speaker/headphone coil to act as the inductor, with a couple of caps before...

Originally Posted by marce

What I think some are missing here is the use of these devices, in small portable electronics, TVs, small PC monitors etc. where the device is pretty much next to the driver (this being the inductor in the filter) or within a certain length (as the not from the double linked TI stuff shows) to limit the antenna effect of the speaker wires and the bandwidth broadcast. Further these devices are designed to be used in the same housing as the driver, allowing further EMC measures to be taken such as shielding to minimise any EMC problems. Being in the same enclosure means that the "switching loops" are within the enclosure and can be shielded, not going out into the open.
Most headphone drivers in small portable devices use headphone amps similar to the one linked to below, again the long headphone wires can act as very good antennas and the length of the cable determines the transmitted bandwidth.
So you ain't going to see a filterless class D driving some long speaker cables to a remote speaker anytime soon, without....
a filter

see

http://www.diyaudio.com/forums/class...ml#post4108928
and beyond

esp

Switcher EMC design
http://www.eetimes.com/document.asp?doc_id=1272171

[IO390]

I'm using about 2M long speaker wires here, no problems at all.