Fasten seat belts. TDA8932 pessimistic review.

My 2pcs of 4x18650 holders just came. Would it hurt the 35V capacitors a lot if I power TDA8932 from 2x4 18650 packs in series giving out 33.6V fully charged and 28V nearly depleted? I believe the answer is definite yes, but it's more logical to use full bays than using 6 18650's and leaving two empty bays.
Thanks. Photo just for reference of what I'll use:
32191124 1793088654062982 203911642621149184 n — imgbb.com
 
My 2pcs of 4x18650 holders just came. Would it hurt the 35V capacitors a lot if I power TDA8932 from 2x4 18650 packs in series giving out 33.6V fully charged and 28V nearly depleted? I believe the answer is definite yes, but it's more logical to use full bays than using 6 18650's and leaving two empty bays.
Thanks. Photo just for reference of what I'll use:
32191124 1793088654062982 203911642621149184 n — imgbb.com

I have used my blue TDA8932 boards with five 100uF/35V SMD capacitors at 32V for hours. In my case none of the capacitors failed.
True, the closer to the voltage rating the shorter life.
 
Member
Joined 2008
Paid Member
The Pessimist v1.0.1

Continuing from post #572 above, I received the boards and the parts. I built up two, both worked on the first try! All in, took about 2.5 hours to build these up: roughly an hour for SMD work, 30 minutes for bare-minimum through hole components, 30 minutes for sanity check, and another 30 minutes for the rest of the TH components. Then I did another final test, and they work just great. (At least for short 5ish minute tests!)

I thought I had the right screw terminals on hand, but I don't. So I'm unable to test both at the same time without doing more work than I want to. I ordered the needed screw terminals. I also had to order a new inductor, you can see in the picture the housing of one is cracked. It basically crumbled in my hand when I took it out of the bag.

I inadvertently got a red and black theme going on! (Great Iron Maiden song BTW.)

I recall back when I played with the cheap boards a while ago, the TDA8932 chips got pretty hot to the touch quite quickly. IIRC, I was feeding them 24v. Today I was feeding my boards 13.8v, with a 4ohm speaker, and the chip wasn't even warm. Granted I didn't run it very long. But you can see in the "family photo" below, my boards are considerably bigger than the cheapies. In theory, that should allow for better thermals, given a much larger ground plane to act as the heatsink.
 

Attachments

  • family_shot.jpg
    family_shot.jpg
    230.3 KB · Views: 511
  • single_board_closeup.jpg
    single_board_closeup.jpg
    149.1 KB · Views: 496
Member
Joined 2008
Paid Member
For anyone who's interested in "The Pessimist" PCB (show above), here are the various support files:
  • BOM - see attached zip, which contains the BOM in CSV and ODS formats. You can also view the BOM in Google docs. I also have a Mouser project for this, PM me if you want me to send you the link.
  • KiCad files - this was created with the nightly builds (v5 "beta"), and uses the new footprint libraries (currently have to install these manually from GitHub).
  • Gerber files - this is the exact zip file I submitted to Elecrow.
  • Schematic in PDF format.

*** Note: I designed this with 4R speakers and a 12v or 13.8v power supply in mind. I see no reason why it wouldn't work with 8R speakers and a higher voltage power supply, but adjustments to some component values will be necessary.

Lastly, I have more boards than I will ever use. I'm happy to give these away for free, as long as you cover shipping. PM me if interested!
 

Attachments

  • tda8932_btl_ds_ref_v1.0.1_kicad.zip
    92.3 KB · Views: 87
  • tda8932_btl_ds_ref_v1.0.1_BOM.zip
    13.5 KB · Views: 89
  • tda8932_btl_ds_ref_v1.0.1_Schematic.pdf
    63.5 KB · Views: 128
  • tda8932_btl_ds_ref_v1.0.1_Gerber.zip
    72.6 KB · Views: 71
Thanks for doing this Matt - quite a bit of work and it looks like a very good job so it's much appreciated. I'm looking forward to getting some boards from you whenever is convenient - no hurry :) Thanks again!

BTW, I can't see how to connect pins 31 from multiple amps together to sync the clocks?
 
Last edited:
I have used my blue TDA8932 boards with five 100uF/35V SMD capacitors at 32V for hours. In my case none of the capacitors failed.
True, the closer to the voltage rating the shorter life.

I'd say 32V was the highest sane voltage for a 35V cap, however, 33.6V seems needlessly high when Xappys could just add a couple of 8 amp diodes in series to drop it to 32.2V. Plus he's running the amp from a battery supply with no regulator so there might be increased ripple. And then he might want to charge the batteries in their cases so he'd need (current controlled) ~34V to do that. And then we don't know if the batteries have protection circuits of any kind - I bought 18650 with built in protection so they are a little longer than standard but can be charged in their case.

I expect it'll work for a while but I've had China caps fail and seen videos of li-ion fires so at the least I'd add those diodes.
 
Member
Joined 2008
Paid Member
Thanks for doing this Matt - quite a bit of work and it looks like a very good job so it's much appreciated. I'm looking forward to getting some boards from you whenever is convenient - no hurry :) Thanks again!

Thank you! I'll keep this thread updated of testing status. So far only had time to do a quick sanity check that each works individually. Hoping to do a longer stereo test this weekend.

BTW, I can't see how to connect pins 31 from multiple amps together to sync the clocks?

D'oh, you caught an error! :eek: Total brain fart on that one. I remember thinking to myself, looks like pin 31 can stay unconnected. Now I look at the datasheet and it's blatantly obvious that's incorrect! So for this board revision, if you want to sync two boards, you'll have to solder directly to pin 31. This will be fixed in a subsequent revision.

Having said that - I'm curious to see if I notice an issue with un-synced boards. I've been using the TI TPA31xx boards in PBTL mode (i.e. two chips, one each for left and right channels) for years and never noticed any issues. Regardless, for completeness, the next board revision will have pin31 exposed in a more easy-to-use way.
 
as yet unnamed stereo BTL TDA8932 amp board

I at last got around to routing this and its gone off for manufacture.

Its a very conservative 15+15W stereo amp using both input transformers (on-board) and output trafos (toroids off-board). Power rail is 35V. The analog signal rail (pin8) is regulated with a discrete reg and LC filtered.

Any Kicad users know how to get the TDA8932 to show up in the 3D rendering? At present I'm calling up the SOICW-28 pin 3D model but nothing is showing.
 

Attachments

  • TDA8932.jpg
    TDA8932.jpg
    67.5 KB · Views: 404
Member
Joined 2008
Paid Member
Looks good!

Any Kicad users know how to get the TDA8932 to show up in the 3D rendering? At present I'm calling up the SOICW-28 pin 3D model but nothing is showing.

If you look at the 3d images of my boards (above), you'll see mine are also missing the 3d model for the tda8932 IC. I haven't tried messing with it, though, I just chalked it up to a beta issue that will be worked out (I'm using the nightly builds). For me, the 3d models also aren't working (or are missing) for the screw terminals.

But anyway - I recommend doing a raytrace render for screenshot purposes. That's what gives that "photo realistic" look. It is time-consuming (a couple minutes or so on my fairly modern PC), but worth it IMO for pretty screenshots.

Edit: what regulator are you using for pin8? Is that a tl431 I see?
 
Last edited:
If you look at the 3d images of my boards (above), you'll see mine are also missing the 3d model for the tda8932 IC.

Do you mean the pics in post #572? Perhaps my eyes are playing tricks on me but I'm definitely seeing an IC in those. Seeing you'd got one was what made me think there's a solution to this somewhere, but perhaps its because you have a more up to date version of Kicad than me?

But anyway - I recommend doing a raytrace render for screenshot purposes. That's what gives that "photo realistic" look. It is time-consuming (a couple minutes or so on my fairly modern PC), but worth it IMO for pretty screenshots.

Thanks for the tip, I shall look into that.

Edit: what regulator are you using for pin8? Is that a tl431 I see?

Yes, using a TL431 to generate a stable reference voltage (about 14V) then (based on my experience with lingDAC's PSU) filtering that output rather heavily (two series RCs) to remove noise before sending the filtered output into a CFP as pass transistor.
 
Member
Joined 2008
Paid Member
Do you mean the pics in post #572? Perhaps my eyes are playing tricks on me but I'm definitely seeing an IC in those. Seeing you'd got one was what made me think there's a solution to this somewhere, but perhaps its because you have a more up to date version of Kicad than me?

D'oh, I guess I did have a 3d image. :eek: I'm already working on a revision, and have since lost the 3d model.

Which version of KiCad are you using?

I'm using v5 beta (i.e. nightly builds). I'm also using the new libraries and footprints (pulled from GitHub).

The old v4 had an SOJ-32 footprint in the SMD_Packages library. This was an exact match for the TDA8932 footprint. Looks like that footprint has not been ported to the new footprints/libraries yet, see this post on the KiCad Forums.

I attached my little library of a few random things, but it also includes the SOJ-32 footprint from the old v4 libs.

I found a 3d model in the new libs that is a close match: Package_SO.3dshapes/TSOP-II-32_21.0x10.2mm_P1.27mm.wrl. You have to rotate it an scale it a bit (see screenshot), but it's close enough IMO.
 

Attachments

  • lib-workpc.zip
    9.2 KB · Views: 52
  • SOJ-32_3d_model.jpg
    SOJ-32_3d_model.jpg
    91.9 KB · Views: 727
Member
Joined 2008
Paid Member
Stereo test is working great! I ran them for over an hour on the basement speakers, no issues. That gave me the confidence to try them on my office speakers. They sound pretty good to me - good enough that I think they deserve an actual case.

The only caveat so far is a bit of hiss with my ear close to the tweeter. It wasn't noticeable at all on the low-sensitivity basement speakers, but my office speakers are more sensitive (87dB). It's not noticeable at normal listening distances. I don't think I've had any amp that didn't exhibit this behavior to some degree, though. But there is something satisfying about a completely silent background (if you have to ask, "is it turned on?" I consider that a good thing!)

I'm definitely glad I implemented the gain reduction. In my office (10x11) with the 87dB speakers, they have plenty of volume with more headroom than I'll ever need. And that's with only 13.8v supply.

With regards to heat, so far there isn't any. I know this chip got hot on the tiny cheap boards with 24v (or more) supply. But with 4R speakers and 13.8v supply, the chips and inductors are tepid - that is, just a bit above room temp, but warm would be too strong of a word. (Temps measured with the always-accurate, infinitely-precise touch-my-finger-to-it technique. ;))

I'm using balanced inputs. But the amps play with the cold signal shorted to ground (with an expected reduction in volume). So these should work with single-ended inputs.

As mentioned above, pin31/OSCIO is not connected to anything. I'm running both amps as master, no sync between them. I can't hear any obvious aberrations, which is consistent with my experience with TI TPA31xx amps.

I did realize the silkscreen on the input side is mislabeled. It reads SLAVE|GND|SLP|GND|MUT|GND|POS|GND|NEG, but it should be: SLP|GND|MUT|GND|SLAVE|GND|POS|GND|NEG. (In other words, the signal inputs are correct, just the slave, mute, and sleep pins are messed up.)

Now that I know the design is working, I'm working on a revision that makes room for the "deliciously overkill" VER2923 inductors. I also expose pin31/OSCIO on the input terminal block, and make the slave setting a jumper. I used 1206-sized components where tight spacing wasn't necessary, just to make assembly a bit easier. And the terminals for DC power and speaker outputs got a bump in size (from 3.5mm pitch to 5.08mm pitch). A rendered image is attached.

So anyway - I have several spare boards, the offer stands - if anybody wants some, PM me, they are free if you pay shipping (or stop by if you're in the Chicago area!).
 

Attachments

  • stereo_test_20180519.jpg
    stereo_test_20180519.jpg
    178.8 KB · Views: 729
  • tda8932_btl_ds_ref_raytrace-20180518_edit.jpg
    tda8932_btl_ds_ref_raytrace-20180518_edit.jpg
    304.1 KB · Views: 713
Which version of KiCad are you using?

Still on the original one I downloaded, 4.0.7.

The old v4 had an SOJ-32 footprint in the SMD_Packages library. This was an exact match for the TDA8932 footprint.

Thanks, I found that and it worked very nicely with 0.5X scale on Y and 0.6X on X, together with a 90o Z rotation. Then a 0.05" offset to get the correct position.

I played with different rendering options, couldn't see a ray tracing one so is that also a feature that's not present in 4.0.7 or have I just missed it?
 
Member
Joined 2008
Paid Member
Still on the original one I downloaded, 4.0.7.
...
I played with different rendering options, couldn't see a ray tracing one so is that also a feature that's not present in 4.0.7 or have I just missed it?

Ah, I did a quick web search, it appears the raytrace feature was added to the nightly builds (i.e. v5 beta). I've been using the nightly builds for so long now it's hard to remember the differences between v5 and v4.


The only caveat so far is a bit of hiss with my ear close to the tweeter. It wasn't noticeable at all on the low-sensitivity basement speakers, but my office speakers are more sensitive (87dB). It's not noticeable at normal listening distances.

So, after spending more time with this, the hiss is a bit louder than I'd like. It's generally not noticeable. But on quiet passages of some recordings (particularly older classical works), it becomes one of those things that once heard can't be "un-heard". But on more modern/"cleaner" recordings, it's really only noticeable on quiet passages with my ear a few inches from the tweeter. Thus, observation #1: it's at least somewhat recording-dependent.

Another thing I noticed is that it's dead-quiet when no music is playing. To me that says the noise is not from the amp itself, but from the source component. So observation #2: it's likely coming from the upstream device.

As I wrote that, I had a "duh" moment: why not try with a different DAC? Previously I had been using my tda1387 balanced out RPI HAT DAC. I swapped out that DAC with my Allo Boss (another RPI HAT DAC, though single-ended). No hiss!

It even passed the "Bolero" test. Ravel's Bolero is a great song to do testing on, since it starts out so quiet, and very slowly builds in volume (it's basically an extended quiet passage, conveniently right at the start of the track). It allows me to crank the volume to neighbors-call-the-cops levels, but without without actually souring relations with my neighbor.

In short: good news for this tda8932 design, bad news for my tda1387 DAC.

Even despite the hiss, I was wanting to say that this amp really seemed to have a "polished glass"-like detail to it. Now that the hiss is gone, I definitely feel that's the case. To my mediocre ears, it manages to be detailed but smooth at the same time. (Although, I'm missing the "organic character" of the tda1387 DAC, as the Allo Boss feels a bit analytical in comparison.)
 
Member
Joined 2008
Paid Member
The Pessimist v1.1.0

OK, after a day of work (pretty much eight hours of labor), I have my revision 1.1.0 boards built, put into a case, and playing sweet, sweet music!

Several people have emailed me about getting boards. I was holding off until this revision was validated. I will respond to you in the next day or two. If anyone else is interested in some "naked" PCBs, please let me know. Boards are free, you pay only shipping. I have both v1.0 (see previous posts) and v1.1 boards. There are some minor bugs with the v1.0 boards (listed above), but they are much cheaper to build. I've thus far only tested that the v1.1 boards play music, so there could be undiscovered bugs. But they are much more expensive to build, mostly due to the big Coilcraft VER2923 inductors.

The power supply is an Antek AS-1215 transformer (15V/100VA) paired with a Sigma11 PSU set to 14V output. The Sigma11 was recycled from a previous project. Case is from the diyAudio store, the Galaxy 2U, 230mm x 280mm.

Here's some "amp porn" for those of you who are into it (I know I am).
 

Attachments

  • board_detail_birdseye.jpg
    board_detail_birdseye.jpg
    573.9 KB · Views: 269
  • pair_boards_overhead_front.jpg
    pair_boards_overhead_front.jpg
    542 KB · Views: 607
  • pair_boards_overhead.jpg
    pair_boards_overhead.jpg
    564.8 KB · Views: 623
  • in_case_all_birdseye.jpg
    in_case_all_birdseye.jpg
    595.1 KB · Views: 326
  • in_case_all_angle_view_1.jpg
    in_case_all_angle_view_1.jpg
    610.2 KB · Views: 306
  • in_case_all_angle_view_2.jpg
    in_case_all_angle_view_2.jpg
    668.2 KB · Views: 283
  • case_rear_panel.jpg
    case_rear_panel.jpg
    525.8 KB · Views: 251
  • case_closed_3-4_view.jpg
    case_closed_3-4_view.jpg
    392 KB · Views: 226
Member
Joined 2008
Paid Member
v1.1.0 Board Info

See post #585 above for v1.0.1 files (schematic, gerbers, KiCad, BOM). A quick recap of known bugs/issues with this version:
  • pin31/OSCIO is not connected to anything. If you want to sync two amps, you'll have to manually solder directly to the chip pin. (Note: I'm running both amps as master, no sync between them. I can't hear any obvious aberrations, which is consistent with my experience with TI TPA31xx amps. YMMV.)
  • The silkscreen on the input side is mislabeled. It reads SLAVE|GND|SLP|GND|MUT|GND|POS|GND|NEG, but it should be: SLP|GND|MUT|GND|SLAVE|GND|POS|GND|NEG. (In other words, the signal inputs are correct, just the slave, mute, and sleep pins are messed up.)
Also, I originally suggested Coilcraft RFS1113 inductors, but I actually changed to RFS1412: they fit the footprint perfectly (13mm diameter, 7mm lead spacing), have a bit better specs, and aren't much more expensive.

---------------

Now I'll switch gears and talk about the newer revision, v1.1.0. This is a minor update to the previous version. See post #598 above for pictures of my completed amp build using this board. Changes from previous version:
  • Use of Coilcraft VER2923 inductors. These are big and expensive, but should alleviate worries of inductor-limited performance.
  • The previous version used strictly 805-sized SMD components; this revision uses a mix of 805 and 1206. I tried to use 1206 when it wasn't necessary to get super close to the tda8932 chip.
  • Fixes input label silkscreen.
  • Slave setting is now a jumper.
  • Pin31/OSCIO is now exposed on the input terminals.
  • Use bigger (5mm pitch) terminals for speaker output and DC power in.
  • Slightly bigger board, bigger top-layer ground pour (should help with thermals).
  • Could be new, yet-undiscovered bugs. ;)

Note1: the attached BOM file is a zip archive that contains two CSV files (one for each PCB revision), and a single ODS file that has both BOMs on separate sheets. Make sure you're looking at the right sheet! I also have Mouser projects saved for these BOMs, PM me and I can send you the link.

Note2: I'm using the KiCad nightly builds (aka v5.0 beta). You'll need a similarly new version and the latest (non-legacy) libraries and footprints to use these files.

---------------

Lastly, a word on the output filter. Class-D amplifiers (such as the tda8932, TI tpa311x), as switching amplifiers, generate noise on the output signal. This noise is well above the audio band, so in theory your speakers should filter the noise. However, having this high-frequency noise on your speaker wires can cause EMI to nearby devices. So most designs call for an RLC filter to remove this noise. The "R" is your speaker; the LC are typically on the board itself.

I recommend reading this primer on class-D output filters: EE Times: Understanding output filters for Class-D amplifiers, John Widder and Yun Tao Zhao, STMicroelectronics, Jan 9, 2008. This topic has been discussed many times here on diyAudio. Here is one short but sweet thread: Some rambling calculation on TPA3116/8 filters. Lastly, this website has useful online RLC calculators.

My design is based on the datasheet reference schematic, Fig 39, page 40. The output filter components are "Llc" and "Clc" in that schematic. On my boards (both revisions), these components are L1, L2, C8, C12. Also in the datasheet, Table 14 on page 24 gives L and C values for common speaker impedances. I included this table on the PCB silkscreen.
 

Attachments

  • tda8932_btl_ds_ref_v1.1.0_bare_board_top.jpg
    tda8932_btl_ds_ref_v1.1.0_bare_board_top.jpg
    551.3 KB · Views: 231
  • tda8932_btl_ds_ref_v1.1.0_bare_board_bottom.jpg
    tda8932_btl_ds_ref_v1.1.0_bare_board_bottom.jpg
    481.9 KB · Views: 222
  • tda8932_btl_ds_ref_v1.1.0_gerber.zip
    85.5 KB · Views: 86
  • tda8932_btl_ds_ref_v1.1.0_kicad.zip
    126 KB · Views: 86
  • tda8932_btl_ds_ref_v1.x_BOM.zip
    19.4 KB · Views: 108
  • tda8932_btl_ds_ref_v1.1.0_schem.pdf
    67.5 KB · Views: 148