A Simple TPA3250 BTL Board

Member
Joined 2008
Paid Member
I recently assembled a 3e-audio tpa3251 amplifier. I was really happy with the result, and thought it would be nice to have a similar amp board built around the tpa3250. I don't need a lot of power (20-30 watts is more than enough), and the tpa3250's power pad down design allows the PCB to be used as a heatsink, allowing for a smaller board. 3e-audio appears to have made a tpa3250 board at one point, but it is no longer available.

I haven't made time for diyAudio projects in the last year or so, and combined with the desire to have a tpa3250 board, I thought I'd take a stab at designing my own. The first PCB I ever designed was for the tpa3118 chip; I called it the Hot Doug, as an homage to DUG's tpa3116/8 board. (I built an amp around DUG's boards over five years ago, and it's been in continuous operation in our living room ever since.)

Also noteworthy is the Gmarsh-designed Wiener Pro prototype based on the tpa3250. He doesn't appear to be active on diyAudio any more, unfortunately.

So here I present the "Hotter Doug" tpa3250 board. This is by far the most complex board I've designed; in the interest of trying to minimize problems, it is essentially a slightly simplified version of the TI TPA3250D2EVM (TI's own evaluation module/reference design). Design notes:
  • Fairly compact: 92x73 (mm).
  • Component references exactly same as TI EVM.
  • I originally wanted to use the "deliciously overkill" VER2923 inductors, but decided instead to use the Coilcraft ua801x series for cost and space savings.
  • Uses only 805 or bigger SMD components. This is my personal threshold for what is comfortable to solder by hand. (Also allows for more caps to be C0G.)
  • Strictly BTL (stereo), does away with all EVM jumpers except SE/DIFF input.
  • Omits the EVM's supervisor IC (for managing the RESET pin) in favor of continuous 3.3v on that pin (copied the idea from the tpa3220evm-micro).
  • Nearly continuous bottom layer groundplane, stitched to top layer groundplane, to make whole PCB act as a heatsink for the tpa chip.

My main goal with this version is just to have something that works. Unless I made a copy error, the schematic should be valid, since it's a direct clone of the TI EVM. Component density is high (for me anyway), so my biggest concern is soldering errors. I plan to use my frying pan/skillet method. Assuming everything works, some ideas for future revisions:
  • Take measurements, investigate post-filter feedback.
  • See if there's a simpler/smaller/cheaper way to derive the 12v and 3.3v power rails. The EVM uses a lot of board space to do this (LM5010 SMPS for PVCC to 15v, LM2940 LDO for 15v to 12v, TLV1117 LDO for 12v to 3.3v).
  • Look for a simpler/smaller/cheaper way to do the input buffer (that allows single-ended or differential input, and supplies the necessary differential input to the TPA chip).

But that's getting ahead of myself! At this point, I think I'm good on the layout, other than minor tweaks---unless somebody points out potential issues that I'm not aware of! I'm slowly working on a BOM and Mouser project. After getting some feedback, and finishing the BOM, I'll post the Kicad files as well as the bill-of-materials.

Thanks for looking and any feedback!
 

Attachments

  • tpa3250_btl_amp_v1.pdf
    111.1 KB · Views: 169
  • pcb_bottom_3d_raytrace.jpg
    pcb_bottom_3d_raytrace.jpg
    162.6 KB · Views: 124
  • pcb_bottom_3d_render.jpg
    pcb_bottom_3d_render.jpg
    134 KB · Views: 540
  • pcb_top_3d_raytrace.jpg
    pcb_top_3d_raytrace.jpg
    356 KB · Views: 496
  • pcb_top_3d_render.jpg
    pcb_top_3d_render.jpg
    370.3 KB · Views: 513
  • pcb_bottom_solid.jpg
    pcb_bottom_solid.jpg
    405.8 KB · Views: 500
  • pcb_top_solid.jpg
    pcb_top_solid.jpg
    276 KB · Views: 508
Member
Joined 2008
Paid Member
Good job!
I did it differently, separate powerline - GVDD, AVDD, and opamps.
I do not see PFFB parts on PCB.

Thanks! Did you post your implementation?

Having the other power rails proved somewhere else (i.e., off-board) was something I thought about. But I wanted a simple "plug and play" board, where I just have to supply the main power, and the board will take care of the rest. Perhaps a simple future enhancement is to have terminals for user-provided power rails, and a jumper or something to disable the on-board regulators.

There are no PFFB parts - that's a "maybe" feature for a future revision. I want to keep it as simple as possible this time around.
 
Member
Joined 2008
Paid Member
It's taken me a few months, but I finally have a board up and running!

Before sending the board off to be fabbed, I changed my mind and broke it up into three separate boards: (1) actual tpa3250 amp board, (2) power supply, and (3) op-amp front-end.

My first attempt at the power supply board was a failure, because I used the wrong footprint for the LM5010. I fixed it, opting for the crazy-small WSON-10 package. When I saw how small that chip was, with pins that are barely visible to the naked eye, I thought, "no way I'm going to be able to solder this!" My way of doing surface mount soldering is to use a frying pan on my cooking range --- clearly not a precision operation!

But to my pleasant surprise, it actually worked!

That gave me the confidence to populate the actual amp board. That was fairly straightforward. Though, again I was concerned about solder quality of the tpa3250 chip; it has lots of pins that are tiny and very close together. After I soldered that board, I inspected for solder bridges on the amp IC. I did in fact see one or two, and used my regular soldering pen with some solder wick braid to remove the bridges. I did a final check using my phone, taking zoomed-in pictures of the chip pins, making sure the solder job looked reasonable.

After that, I used my DMM in resistance mode to spot check the board for obvious problems, e.g. shorts to ground.

I didn't find any obvious problems, so I went ahead and did an actual test... success! At this point, I haven't built the op-amp front end, and used a DAC with differential output as a direct source for the amp. I didn't hear any crackling or obvious distortion; I also used a Kill-a-Watt mains power meter to get an idea of the total power draw. It showed about five watts at idle and very low listening levels. So overall, I think it's mostly correct. However, a couple observations might indicate that something isn't quite right:
  • I didn't measure the actual temperature of the tpa3250 chip, but put my finger on it. Even at idle/low playback volume, it was surprisingly hot. I could only comfortably keep my finger on it for a few seconds. I seem to recall the older tpa311x chips staying pretty cool to the touch in similar circumstances. When I cranked the music up a bit, the chip was too hot to touch for more than a second. And even in this "cranked" state, the AC power meter only showed about 13 watts being drawn, so clearly no where near this chip's rated output power. (Maybe I doomed myself by naming the board "Hotter Doug" --- certainly lives up to the name!)
  • Also when the playback was cranked, I did see the clipping indicator LED light up. Interestingly, it lit up in sync with the music's rhythm!

I was using a pair of 8-ohm Overnight Sensations speakers for testing. These should be a pretty easy load for any amplifier.

I can post Gerbers, BOM, KiCad files if there's any interest. Though I intend to make some revisions. At a minimum, I want to get away from screw terminals and use something like JST connectors. I also want to build up the op-amp front-end. After further testing, I'll likely discover other issues that need to be resolved.
 

Attachments

  • tpa3250_amp_board_20201025.jpg
    tpa3250_amp_board_20201025.jpg
    425 KB · Views: 173
  • tpa3250_psu_20201025.jpg
    tpa3250_psu_20201025.jpg
    293 KB · Views: 130
  • hotter_doug_live_test_20201025.jpg
    hotter_doug_live_test_20201025.jpg
    677.9 KB · Views: 131
  • tpa3250_amp_no_front_end_v1_render.jpg
    tpa3250_amp_no_front_end_v1_render.jpg
    529.2 KB · Views: 201
  • tpa3250_psu_v1_render.jpg
    tpa3250_psu_v1_render.jpg
    241.5 KB · Views: 129
  • tpa3250_amp_no_front_end_v1_schematic.pdf
    77.8 KB · Views: 128
  • tpa3250_psu_v1_schematic.pdf
    36.4 KB · Views: 72
  • tpa3250_ic_pin_solder_detail_1_20201025.jpg
    tpa3250_ic_pin_solder_detail_1_20201025.jpg
    530.9 KB · Views: 151
  • tpa3250_ic_pin_solder_detail_2_20201025.jpg
    tpa3250_ic_pin_solder_detail_2_20201025.jpg
    372 KB · Views: 145
Good job on tackling the layout yourself.
I rather bought an amp in a lazy moment.
But maybe I'll take your kicad files and go from there.
I'd love to have a compact 4-channel amp.

As it seems you could easily use a big thermal pad and mount the bottom side to a solid Aluminium plate or similar.
That works quite well when the surface is that big and the pad is not too thick. Just do not tighten the screws too much!
 
Member
Joined 2008
Paid Member
This is Overtemperature Warning

OK, well that seems consistent with how hot it felt to the touch.

Do you have any thoughts on what is likely causing it to get so hot? I feel like an implementation error is unlikely, since it actually works.

My suspicion is that the power pad on the bottom of the IC didn't get a good solder to the PCB. And/or the ground pins don't have a good solder connection to the PCB.

Another thought is that I damaged the chip during the "frying" procedure.

But if the power pad and/or ground pins are the problem, then joensd's suggestion (a pad to thermally mate the PCB to a metal chassis), probably won't help, right? Since, if this reasoning is correct, then the chip isn't correctly transferring heat to the PCB.

I wasn't sure how much solder paste to put on the exposed ground of the PCB (that is intended to mate to the amp IC). I was afraid of using too much and having it run out and create solder bridges on the actual pins. Now I'm wondering if I used too little?

Dunno, just thinking out loud. Open to suggestions!

Thanks!
 
Do you have any thoughts on what is likely causing it to get so hot? I feel like an implementation error is unlikely, since it actually works.

My suspicion is that the power pad on the bottom of the IC didn't get a good solder to the PCB. And/or the ground pins don't have a good solder connection to the PCB.
You guessed it, one of the two!

Layout/implementation/GND routing can still cause the chip to oscillate and generate excessive heat (although it works).
Chances are also that the pad is "cold" soldered to the PCB and there is a huge thermal resistance in that spot. Did you use solder paste?

Can you visually make out if the connection of the thermal pad is good or not?
How hot does the chip get when it idles? (no input signal)
Does it work properly when you play music with little volume?

In the forelast picture on the top right I see a big blob of solder, right?
The cap doesn´t look shorted but watch out for solder pellets or similar and inspect the whole board.
 
Last edited:
Member
Joined 2008
Paid Member
@joensd - thank you for the assistance! The KiCad files are attached to this message. Let me know if you want the BOM or any other supporting materials.


Layout/implementation/GND routing can still cause the chip to oscillate and generate excessive heat (although it works).

Agreed, though I think I'm going to focus on solder issues first...


Chances are also that the pad is "cold" soldered to the PCB and there is a huge thermal resistance in that spot. Did you use solder paste?

Can you visually make out if the connection of the thermal pad is good or not?

I did use solder paste, Mechanic XG-Z40 Sn63/Pb37.

I can't tell by looking if the thermal pad connection is good or not, at least not with my naked eye. I think we have a USB microscope somewhere, I think that would be helpful here!


How hot does the chip get when it idles? (no input signal)
Does it work properly when you play music with little volume?

I've thus far only done the one test. But based on that: I didn't really check the temperature when idle. Most of the testing was done at fairly low volume, and it sounded OK. During this relatively quiet playback, the IC temperature was such that I could touch it, but only for maybe five seconds before it got uncomfortable.

When I turned the music up a bit, enough to trigger the clip/over temp indicator, the chip was too hot to touch even for a second.

I do have an IR thermometer, wishing now I'd have used that, rather than relying on touch!


In the forelast picture on the top right I see a big blob of solder, right?
The cap doesn´t look shorted but watch out for solder pellets or similar and inspect the whole board.

I don't think those caps are shorted. The angle of the picture makes it look like there's a big blob of solder between them. But there's actually a fairly comfortable gap between them. The one cap was soldered by hand, since it did move too much during "frying", and initially was touching the other one. FYI, those are the 33nF bootstrap caps.


is there a short between the IC pins?
in a more closer look I also see lots of solder pellets under the IC
which is somewaht expected when using the "pan-fry-method" ;-)

Those close-up pics of the chip pins were taken using my phone with 4x zoom - those solder pellets are not visible to my naked eye!

But I agree, with the zoomed-in pics, the chip pin soldering does look a little sketchy.

I'm thinking I can check those chip pins for shorts. If there's an obvious short I'll try to touch it up by hand. If I don't see an obvious problem, I guess I'll try to remove the tpa3250 and "re-fry" it.

Thanks again!
 

Attachments

  • tpa3250_kicad_20201027.zip
    1 MB · Views: 62
@joensd - thank you for the assistance! The KiCad files are attached to this message. Let me know if you want the BOM or any other supporting materials.
You´re welcome. Thanks for the files!
I did use solder paste
That´s good because usually when the amount is right the soldering works out quite well; even with a "hot plate".
I think I'm going to focus on solder issues first...
Yes, that´s what I would do. Maybe you find that microscope.
Otherwise look for solder pellets or excessive solder that goes beyond its pad.
From the side you should be able to judge if the IC sits flat on the pad.
Considering how much solder you applied you should also know if it has a good connection.

Take your time. "Refrying" the chip can cause more problems than it solves...
All of the solder and parts will swim; vias can tear; the chip can take damage...


If it´s really the OTW the chip should "measure" about 125° which is substantial.
I´d expect maybe 1W idle dissipation with very low volume and 24V-PSU but that causes not even a rise of ~10K. (Junction-to-board thermal resistance ~ 6.5K/W). You should measure the idle dissipation and note voltage&current.

Did you check all the supply-voltages? What´s your main supply-voltage?
An oscilloscope would be great of course to see if there´s oscillation.
 
Member
Joined 2008
Paid Member
I just set it up again, this time with better testing tools.

When I "cranked" the music, enough to get the CLIP led to trigger, I measured the temperature of the top of the tpa3250 IC with my IR thermometer: about 32.3 degrees C (90 degrees F). The IR thermometer said about 38 degrees C (100.4 F) on the PCB right next to the tpa3250 IC.

It's a cheap/old IR thermometer, so I doubt it's accuracy is all that great. And I'm assuming the actual circuitry of the chip is probably warmer than the plastic package. And FWIW, to the touch, the IC feels hotter than the PCB.

I put my DMM in current measuring mode, and put the leads in series with the +24V wire that goes to the amp board (PVDD). The current danced around quite a bit with music playing. I watched it for a while with the volume up (enough to trigger the CLIP light), and the max I saw was about 750 mA. Typically though it stayed under 300 mA.

At idle, I measure about 40 mA on +24V line.

At idle, I also measured the 12V and 3.3V rails (between my power supply PCB and the amp board). On the 3.3V rail, at idle, I read about 11 mA. I'm pretty sure this is just used as a reference voltage for the RESET pin, and also the power-on LED.

Now, when I measured the 12V rail at idle, I'm actually seeing about 101 mA. This seems too high to me. Note: that's with my DMM in DC Amps measuring mode; if I switch to DC milliamps mode, it reads 0.90 mA, which is way too low.

To be clear, all the above measurements were done with the ammeter in series between the different PSU rails (of my PSU PCB) and the amp board.

Lastly, I measured the current between my multi-rail PSU PCB and the main AC-DC power supply (a Meanwell EPP-200-24 SMPS). At idle, I measured 140mA.

In terms of physical inspection: I did find that USB microscope, but it's a cheap one, and I didn't have much luck with it. My wife's phone's camera actually does 10x zoom, and I found that to be usable. I spent some time with rubbing alcohol, cleaning up all the "blobs" left over from the soldering process. It certainly looks nice and clean. All the connections look reasonable to me. The TPA chip itself does seem to have a tiny gap between it and the PCB - not enough to let any light in though, maybe half the thickness of a hair. It's just barely not completely flush to the PCB.

So I'm not sure. My thoughts are that the 12V current draw looks suspicious; and I still haven't ruled out a poor solder connection between the IC and PCB.
 
Member
Joined 2005
Paid Member
For inspection, you can get some pretty cheap, USB optional microscopes now that have a reasonably sized LCD screen, but mostly I use a head magnifier with multiple magnifications and a dual adjustable LED light, they're very cheap and commonly used. If you're willing to spend a little extra, you can get surgical/dental loupes with LED illumination, you can get them with magnification of 2.5x to 3.5x or higher. Head worn gear is great because it gives you more room to work on the board. Unrelated tip: if you can't read a number on an IC, sometimes the inks are U.V. reactive. Also, some inks are organic and a tiny drop of Luminol will clearly show you the number under U.V.

Hot air stations and tips are pretty cheap now and are very easy to use. Make sure there are plenty of vias beneath the IC and remember to use solder paste! Clean up any rosin/resin, if any, when done. I use rosin on both the IC and the board to ensure a good solder joint. There are also some very good conductive and non-conductive thermal epoxy's around, but it can be tricky to use enough but not too much. Sorry, I'm not in the shop to give you the brands/numbers, but I found them without looking too hard.

A more extreme solution to pad down IC's is to fab the IC footprint with a cutout so the IC fits in to it, and mount the IC upside down, then machine the bottom of a heatsink to contact the pad only and machine it to stand off the board enough to clear components (there's always the heat pipe option too). Tape isn't very effective to mount the heatsink, the epoxy solution works well,or you can spring mount the heatsink to the board with screws (don't over tighten!) and use thermal paste. The advantage to spring mounted heat sinks is that if you accidentally bump the heat sink, there's less chance of damage to the IC/connections/PCB traces.

Just some tips and ideas. For personal use I tend to overkill for thermal stability because Class-D type amps constantly change in temperature and it affects the sound quality, distortion, and lifespan (thermal fatigue).
 
Member
Joined 2008
Paid Member
Thank you for the tips and ideas! I haven't revisited this project in a while. I have spare boards and parts, I think I'll eventually just build up another, as my guess is that my problem has to do with a poor solder connection between the IC's thermal pad and the PCB.

For inspection, you can get some pretty cheap, USB optional microscopes now that have a reasonably sized LCD screen, but mostly I use a head magnifier with multiple magnifications and a dual adjustable LED light, they're very cheap and commonly used. If you're willing to spend a little extra, you can get surgical/dental loupes with LED illumination, you can get them with magnification of 2.5x to 3.5x or higher. Head worn gear is great because it gives you more room to work on the board. Unrelated tip: if you can't read a number on an IC, sometimes the inks are U.V. reactive. Also, some inks are organic and a tiny drop of Luminol will clearly show you the number under U.V.

We have one of those USB microscopes. The main problem is finding decent software for it. It doesn't have it's own custom software, the manual just recommends a few free apps; I could only get one of the suggested ones to work, and even then, saying it "works" is being a little generous.

But my wife's phone's camera actually has 10x zoom, and I found that works surprisingly well.


Hot air stations and tips are pretty cheap now and are very easy to use. Make sure there are plenty of vias beneath the IC and remember to use solder paste! Clean up any rosin/resin, if any, when done. I use rosin on both the IC and the board to ensure a good solder joint. There are also some very good conductive and non-conductive thermal epoxy's around, but it can be tricky to use enough but not too much. Sorry, I'm not in the shop to give you the brands/numbers, but I found them without looking too hard.

Indeed, there are lots of vias below the thermal pad. I might have to get a hot air station.


A more extreme solution to pad down IC's is to fab the IC footprint with a cutout so the IC fits in to it, and mount the IC upside down, then machine the bottom of a heatsink to contact the pad only and machine it to stand off the board enough to clear components (there's always the heat pipe option too). Tape isn't very effective to mount the heatsink, the epoxy solution works well,or you can spring mount the heatsink to the board with screws (don't over tighten!) and use thermal paste. The advantage to spring mounted heat sinks is that if you accidentally bump the heat sink, there's less chance of damage to the IC/connections/PCB traces.

In this particular case, I deliberately chose the pad-down version of the chip so that I could avoid worry about a heatsink. There is another chip in the TI TPA line (3251 I think) that is the exact same circuit, but has the thermal pad facing up, for use with an external heatsink. But I used the 3250 specifically for the thermal pad down, so I can use the PCB as a heatsink.

Thanks again for the tips and ideas!
 
Founder of XSA-Labs
Joined 2012
Paid Member
Hot plate from bottom and hot air from top makes superior solder joint bond. You should use all paste on passives as well and reflow at same time. Apply lower temp no lead paste to allow lower temp hot plate not to damage parts. Plate heat at 125C and hot air at 350C apply local heat to spot melt only where air is pointed. Joints will be clean and smooth fillets. I mentioned this on your other thread. With good home technique your joints can look and perform better. Less question of what went wrong. Resist temptation to solder SMT parts with an iron. Only use iron for cleanup.

Here is how I solder 0603’s with “Henna” nozzle, and hot air:
XRKaudio SMT Method Demo - YouTube