Of course, the footprints "should" support any output filter values you want to use. I assumed 4R speakers because that's what I'll be using.
I'd not recommend this chip in BTL mode driving a 4ohm speaker. I've tried that and didn't get great results but then I was constrained to having a 20V supply. Perhaps its OK at lower voltages, I see the DS shows measurements at 12V supply with 4R load in BTL.
The larger the caps at the input the more propensity for picking up the output switching hash. I'd stick with X7Rs if I were you, but hey this is your design
Will v3.0 have input transformers?
No need to keep the 100nF when there's a much larger 'lytic there.
Yes, they become 'no fits' (NF).
What was(were) the non-great result(s)? Something specific?
At any rate, I'm planning on using a 12v supply. The main reason I renewed my interest in this chip is because I realized my power needs are in the single-digit wattage range. The DS suggests this chip will reliably do 15w into 4R load with 12v supply, so it seems to suit my needs perfectly.
Also, the DS gives me the impression that much of the current limitations are for thermal reasons (not surprising). It makes me wonder if the physically smaller chip (HTSSOP32) would be better, as it has a thermal pad, and can therefore use the PCB's groundplane as a heatsink. My thought is: more (bigger) heatsink means better heat dissipation, which maybe gives more power output headroom.
The larger the caps at the input the more propensity for picking up the output switching hash. I'd stick with X7Rs if I were you, but hey this is your design
For sure. I'm thinking I'll put SMD solder pads on there, so the user can choose.
Can the output switching hash be eliminated and/or reduced? Take a look at the TI tpa3118 datasheet, their reference implementation circuit has a EFI snubber circuit (page 26, figure 37). I wonder if something like that would be beneficial on the tda8932?
Also, tomchr's lm3886 done right has EMI filters on the signal inputs (basically an RC circuit). I wonder if something like that could be adopted for the tda8932 as well?
How much EMI suppression can I expect when running truly balanced inputs? I suppose it's a question of does the EMI affect the hot and cold signals exactly the same?
Will v3.0 have input transformers?
Unlikely. In my mind, given their cost and physical size, transformers are best suited for a separate board, an "analog front end" so to speak. Purely personal preference here.
I have some nice Jensen transformers and some cheaper Edcors, both unused. I actually tried to use the Edcors with the cheap tda8932 boards a while ago. But it seems the cheap Chinese tda8932 boards aren't wired for balanced input. That is another one of my motivating reasons for designing my own board: I can use balanced input (and play with those transformers again).
Does that pin serve a similar function as pin7 on the tda1387?
Also: for output inductors, I'm designing for the Coilcraft RFS1113. These are reasonably cheap, readily available (at least in the USA), not-terrible ESR, and have a flat inductance versus current curve up to the max of this chip's output power. I'm actually using a footprint that's slightly bigger than the RFS1113, so there's a little room for experimentation.
You can read this post where doctormord tested several inductors for the TI tpa3118, and found the Codaca CSCF2014-6R8M to be the best. He then went to to say those appear to be only available in large batch purchases, so the Wurth 7443630700 appears to be the same. Those Wurth's are pretty expensive (at least in the USA), so I'll save them for version 2.
Hi guys,
I'm finished with building my portable speaker which uses bass reflex alignment paired with Dayton DSA135 8ohm woofer and Dayton ND20FB 4ohm tweeter. Each driven by it's own TDA8932. The crossovers are 2nd order, passive, custom built, with zobel network included. Now comes the modding part.
• The speaker gets it's power from batteries (4x18650, so 16.8v on full charge) boosted by xl6009 boost converter.
• Then this xl6009 boost converter's power (boosted to something in between 20-26vdc) reaches the woofer's TDA8932.
• Later tweeter's TDA8932 gets voltage stepped-down from booster's voltage back to 11vdc by LM2596s buck converter (tweeter is 4 ohms and higher sensitivity, while woofer is 8 ohms, hence the need of step down).
• My questions are basically in the photos below: do I need to replace boost converter's (red board) output capacitors and step-down converter's (blue board) input capacitors, leaving boost converter's input caps and step-down converter's output caps untouched to get less ripple?
Photos of modules
Thank you. I'll post full build details once I'm finished modding.
I'm finished with building my portable speaker which uses bass reflex alignment paired with Dayton DSA135 8ohm woofer and Dayton ND20FB 4ohm tweeter. Each driven by it's own TDA8932. The crossovers are 2nd order, passive, custom built, with zobel network included. Now comes the modding part.
• The speaker gets it's power from batteries (4x18650, so 16.8v on full charge) boosted by xl6009 boost converter.
• Then this xl6009 boost converter's power (boosted to something in between 20-26vdc) reaches the woofer's TDA8932.
• Later tweeter's TDA8932 gets voltage stepped-down from booster's voltage back to 11vdc by LM2596s buck converter (tweeter is 4 ohms and higher sensitivity, while woofer is 8 ohms, hence the need of step down).
• My questions are basically in the photos below: do I need to replace boost converter's (red board) output capacitors and step-down converter's (blue board) input capacitors, leaving boost converter's input caps and step-down converter's output caps untouched to get less ripple?
Photos of modules
Thank you. I'll post full build details once I'm finished modding.
Last edited:
Yes, this was me replacing the chipamps (LM1875 clones) inside my 'Swan' 4inch active speaker. It made the sound too dynamic, like turning the contrast on a monitor right up to maximum. Hyper-real and rather fatiguing after a while.
That should be fine, interested to hear about how it sounds
This is indeed the same conclusion I came to - so long as the OCP (over-current protection) isn't triggered (4A) then the limits are thermal. You'll not be able to trigger OCP with a 4ohm speaker on a supply of 12V. The smaller chip will deliver more power provided a multilayer board is used (4 layers typically). On my design I eek more total power out at much higher supply voltage by running the analog circuitry (40mA draw from pin8) at a lower voltage than the output stage.
I help the output hash problem with using an output transformer. I can't quite see what TI's circuit is going to help with, perhaps reduces much higher harmonics of the switching freq? To get reduced switching hash, a higher-order output filter would do the trick.
I can only see what looks like a 'black box' there, no specific details. Certainly we can add extra caps at the inputs, after the 470R series Rs.
Its that and the CMRR of the input stage of the chip.
I thought about that but I have some 19mm width transformers which are very cheap so they're going on my PCB
Similar in that its a reference voltage. In TDA1387 the voltage reference is converted into a current reference, that's different.
I shall look at those, sometimes Coilcraft is a bit tricky to obtain over here so I wonder about others. I'm using on my board some 12mm TDKs which are very cheap (about 1RMB) but then I need a different value due to my use of OPTs.
Last edited:
If I were you I'd be running the buck converter direct from the 16.8V battery pack, not from the the boost. You'll get better efficiency that way and also a less noisy tweeter amp supply. Do you know what the minimum in-out voltage is for the LM2596?
You'll also need to attenuate the input to the tweeter amp to compensate for the higher efficiency.
I'd start out by leaving those modules as stock then see if you run into noise problems in practice. At which point, break out the low ESR caps and get soldering
.That's not so simple in practice
. While buck converter is in the internal circuitry, the boost converter, however, is connected to the 18650 battery holder and is external. That means I have very little options to connect the batteries straight to the buck converter unless the batteries would be internal, but I've already learnt from the past that it's better having li-ion batteries each as seperate unit, than having something such as 6s2p pack, which is a pain in the neck to find an appropriate charger for. As a plus, tweeter's buck converter gets it's power straight from laptop PSU (no need for booster) when I'm at home, so it's fine for the sound in the environment where it truly matters.I'd start out by leaving those modules as stock then see if you run into noise problems in practice. At which point, break out the low ESR caps and get soldering
I'm going to solder few caps for TDA8932s (pin8 and pin12) anyway, so it would be two works at once. I would be happy if you could answer which caps of booster and buck modules may I replace. Also, when I'm adding capacitor to pin12 where should the second leg go? If soldering a cap to pin8 should the second leg go specifically where power negative wire is connected or is there any place on smaller caacitors or resistors nearby the main chip? It's a blue Sanwu board:
Some circles on the picture or wordy description for places where the second legs of added pin8 and pin12 capacitors should be connected would be greatly appreciated. I'm not as experienced to experiment without knowing how each component affects the circuit. Thanks.
I found NXP's AN10436 - Application Notes for the TDA8932. Lots of useful info in there! Many of the topics in the dialog above are covered, for example: output RF filter, different chip package thermal considerations, inductor selection, PCB layout considerations, etc. Definitely worth a read for all TDA8932 enthusiasts.
I have tried aircoils and it works extremely well! The sound is very detailed and easy and transparant I bought some 0,1 mH air coiles for crossover filters and unwinded them till 22 uH. They keep cool (28 degrees Celcius). I will try them on my other amplifiers as well. I don't want chokes any more. I intend to use the wire that I unwinded to make aircoils myself using PVC pipe.
My only concern is the 'leakage field'. I am thinking of wrapping the aircoils in aluminium foil, would that help?
Last edited:
Do those go before or after the DC-blocking caps? Also, in BTL mode, where the inputs are wired in anti-parallel, is it necessary to put those resistors in front of all four pins, or is two sufficient?
Also, after giving this more thought, I decided the 36dB default gain (for balanced input) is way too high (for me). I'm copying the gain-reduced input circuit from AN10436, figure 25, page 27. (Similarly covered in the datasheet on page 25.) Basically, it's a voltage divider.
When using an arrangement like this, would you still recommend resistors in front of the input pins? Or will the gain-reduction voltage divider also help with RFI suppression as well?
See attached schematic - you can see those 470R resistors right before the chip pins on the right. Is that right? Or should they be omitted entirely, or moved elsewhere?
Thanks!
Attachments
If you're determined to replace caps then input caps can be replaced at will (provided the input cap isn't the output cap of another converter) but output cap changing may well affect the stability of a module. So do it with care, if you don't have a scope to check for oscillation then don't do it.
Pin 9.
Pin 9 again or very close by.
The schematic you've shown is the way I'd do it - each pin having its own R. We want the R to be as close to the pin as possible. Yes the external attenuator will definitely help with the RFI attenuation. I see you've added a shunt cap to further attenuate the HF noise.
The schematic you've shown is the way I'd do it - each pin having its own R. We want the R to be as close to the pin as possible. Yes the external attenuator will definitely help with the RFI attenuation. I see you've added a shunt cap to further attenuate the HF noise.
Thanks Richard!
Scope creep got the best of me. Changes from the previous post:
- I went ahead and added the TI TPA-style RF filter on the output (easy enough to omit or remove if desired).
- Added the gain-setting resistors. As drawn, this will set the gain around 26dB, instead of the default 36dB.
- Added 470R series resistors in front of input pins 2, 3, 14, 15 per Abraxalito's suggestion.
- Added a 470uF/6.3V electrolytic to pin12 (voltage ref).
- Added two 10mm electrolytic caps for pin8 (analog voltage supply).
- Reduced input DC-blocking cap footprint. Won't be able to fit polypropylene caps, but can still use smaller film. Also have footprints for 1210 SMD caps which can be used instead.
- Removed 16mm PSU decoupling caps in favor of two more 10mm caps. Easy enough to add off-board capacitance if this is deemed necessary.
- Inductor footprint assumed to be Coilcraft RFS1113. The footprint is actually slightly bigger, so there's a little room for experimentation or substitution.
Attachments
I found your older post saying the additional capacitor to pin12 should go in parallel with the old one and assuming pin8 addtional capacitor should go to pin9 I've made this quick sketch over the photo. Are the connections as pictured correct and rational and is the polarity of capacitors right? Yellow "things" are mimicking the added capacitors. If it's so, I'm gonna get soldering done tomorrow and write back with the impressions afterwards.
Pin Connections
Last edited:
1) You won't need the 4 electrolytics ( C1, C21-23) on the main supply. Suggest just one 2200uF and put the two smaller ones where it really matters, on pin8. Or just go with one 470uF/35V on pin8. I'm assuming this is one of your 80/20 designs and you don't want to go to the trouble of a separate regulator for the pin8 supply
(...)
TPA3118 has a similar structure in that there's an analog supply rail for the signal level circuits (pin17, AVCC) but there's zero additional decoupling for that pin on most boards. TI's application circuits don't suggest it needs anything extra. NXP on the other hand do add a 10R series resistor and a cap specifically for that pin (pin8).
I did mod my first TPA3116 with extra decoupling for pin17 and it gave a substantial improvement. Its a tricky mod to do though as you need to lift the pin from the PCB. Given the 10R resistor isolating pin8 from the main supply on the typical TDA8932 board, the addition of decoupling caps to that pin is much easier. They go from pin8 to GND, the value I'd suggest as a minimum is 470uF.
What is the purpose of that 10R in front of pin8/VDDA?
I assume when you add extra capacitance to VDDA, the 100nF to filter any HF noise that got picked up along the way.
But does the extra capacitance go "after" the 10R (i.e. in parallel with the 100nF)? Or "before" the 10R? In my schematic above, I put the extra VDDA caps "before" the 10R, and now that I think about it, that's probably wrong, as the result is simply more caps added to the main supply.
Looks like you found pin12 OK. The cap is physically much bigger than you need here (1000uF/35V vs 470uF/6.3V) but it should still work fine assuming it doesn't fall off under its own weight
Together with the 100nF on pin8 it forms a low-pass filter to keep the highest frequencies of supply line noise from reaching pin8.
I meant the extra caps to go 'after' the 10R, i.e. in parallel with the 100nF which is already there. This lowers the corner frequency of the RC filter, which is set way too high in the standard application schematic.
Looks funny even to me, I posted the picture and immediately noticed it by myselfLooks like you found pin12 OK. The cap is physically much bigger than you need here (1000uF/35V vs 470uF/6.3V) but it should still work fine assuming it doesn't fall off under its own weight
I already changed it with 2200uf 6.3V Nichicon. I also soldered Nichicon to Pin12 of tweeter amplifier and when my damn head recognized it's unneeded there I tried to unsolder it and peeled off the pcb trace. Luckily, after half an hour of soldering I fit a wire jumper between the damaged trace. Anyways, after that I soldered the 35V 1000uf straight from Pin8 to Pin9.I screwed back my backplate, turned it on - nothing, complete silence. I unscrewed it back, left it open - working as it should be. It comes out it was two backplate screws very precisely snapping the wires of both amplifiers, I guess I am one of those who could win a million in a lottery? Well, after that, happy it didn't go kaboom, I moved the amplifiers farther, and screwed it back together. The impressions?
• Bass has truly enrichened I gotta say! Even at low volumes (finished yesterday at 12'o clock ZZZ) it seems more powerful, as if somebody integrated DSP or EQ'd it
• The midrange and treble smoothened out and that's no braintrick - I checked it with "Stephen - Fly Down ft Mack (Live at Halfway House)". I know, I know, it's low quality youtube audio, but what I heard was a real change - the woman singing there has a really high pitched voice and before it was making me cover the ears, but now I'm listening happily one notch from maximum volume.
• One more change I noticed - it gets louder without woofer sounding like it's smacking the backplate. Before I could listen at the best two notches behind max, now I can confidently get one more notch or even maximum at some songs. I don't know how this is connected to the mods I've done, but I'm guessing it has to be additional Pin12 decoupling presenting more power to amplifier so it doesn't clip?
I know this has gotten long but I hope it may help somebody
Last edited:
Cool, thanks for the explanation. I corrected the schematic and updated the PCB layout accordingly. I went ahead and put in the order at the fab house! I also ordered a couple of the cheap blue Sanwu boards for comparison. The BOM cost for my PCB is over $50 (though i spec'ed expensive organic polymer caps). Still, even with regular electrolytics, my boards will be 10x the cost of the cheap boards from China. I doubt they'll sound 10x better, but knowing I'm running my own design has a value beyond dollars.
Now the waiting begins. Which is okay, because I just received your lingDAC, so I have something to play with in the meantime!