How does your PVCC and AVCC looks like? Can you see you switching frequency on those rails? Measure at caps and also on the IC pins.
Thanks for the reply--sorry I missed the earlier post on this. Also thanks for the recommendation on cap replacement. I've been looking at various possibilities for those and found a lot of them are just too big in diameter to fit. I'm hoping better caps might help with the hiss.
I received three of the dual chip PBTL boards from china, hoping I may improve them enough to tri-amp a set of 3-way, 2R speakers. They sound better than they should (initial test at 12v) considering the poor layout and bad choice of caps throughout. And yeah, all three of them have gains set at 36db, with both chips misconfigured master. Not a deal killer so far since I expected all that... until I began to look the PCB over.
The OUTPR and OUTNR (and OUTPL & OUTNL) pin pairs are tied with very narrow traces leaving the chip pins in a Y trace, and then joining into a single-width trace (same trace width as the input signal traces and others) for several millimeters, before entering onto the wider traces containing the LC output filters.
There lies the deal killer because the outputs are choked into those narrow traces and so is limited to 25-30 watts or so, at any PVCC and speaker load. The PCB is not useable.
So now I'm stuck with crash coursing Kicad. First impressions of that clunky mess are less than desirable as well... but I'll get there.
Thanks to all who contribute here!
The OUTPR and OUTNR (and OUTPL & OUTNL) pin pairs are tied with very narrow traces leaving the chip pins in a Y trace, and then joining into a single-width trace (same trace width as the input signal traces and others) for several millimeters, before entering onto the wider traces containing the LC output filters.
There lies the deal killer because the outputs are choked into those narrow traces and so is limited to 25-30 watts or so, at any PVCC and speaker load. The PCB is not useable.
So now I'm stuck with crash coursing Kicad. First impressions of that clunky mess are less than desirable as well... but I'll get there.
Thanks to all who contribute here!
Yeah... thats it. I actually thought about that just for a minute! 😱
But the board would still be useless if I stopped patching it up there, as there are many other stupid decisions that found their way onto the PCB. After looking it over carefully my decision is to salvage the two chips on each board and toss the rest of it out. Assuming the TI chips are genuine...
I am learning a lot from you folks here. The humor is great too! Thanks.
Can you showed us which dual chip board that you have purchased please. Thanks!
Just curious, do you have measurements to support this assertion?
With all the shortcomings that you pointed out and yet, the sound of these boards are still "acceptable", these boards are really "amazing", aren't they?
The three dual PBTL boards I received are identical to those in this post, #8437: http://www.diyaudio.com/forums/class-d/237086-tpa3116d2-amp-844.html#post4532722
The silkscreen label on the boards say: XH-M190 V3
No point in measuring anything because there is insufficient trace widths there, as I explained. Advertised output is not possible unless/until the traces leaving the output pins are jumpered with added wire. Scratching the mask off and adding solder to the tops of the traces won't help enough either in my opinion... still not physically possible to flow enough current to get anywhere near the chip's rated (100w @ 4R, 125w @ 2R) power shown on the charts in the DS. My previous estimate of 25-30 watts max is very fair.
They sound decent with PVCC=12volts, but there is no point in trying to push them any higher until the PCB is swapped out. Improvements that I had planned to make to improve the sound will be a waste of my time and money.
And yeah.... amazing... can't argue with that.
EDIT: I just measured the trace width leaving the OUTPR and OUTNR pins (Left outputs are identical width) where they both converge at the "Y" trace using my digital calipers. That single trace on the PCB must pass the total current exiting the chip, onto the LC output filter loop (where there are fatter trace), then to the loudspeaker load. I measured that single trace bottleneck to be 0.35mm wide. Now imagine whether those PCBs are etched 1 oz copper
... or more likely less...?
The silkscreen label on the boards say: XH-M190 V3
No point in measuring anything because there is insufficient trace widths there, as I explained. Advertised output is not possible unless/until the traces leaving the output pins are jumpered with added wire. Scratching the mask off and adding solder to the tops of the traces won't help enough either in my opinion... still not physically possible to flow enough current to get anywhere near the chip's rated (100w @ 4R, 125w @ 2R) power shown on the charts in the DS. My previous estimate of 25-30 watts max is very fair.
They sound decent with PVCC=12volts, but there is no point in trying to push them any higher until the PCB is swapped out. Improvements that I had planned to make to improve the sound will be a waste of my time and money.
And yeah.... amazing... can't argue with that.
EDIT: I just measured the trace width leaving the OUTPR and OUTNR pins (Left outputs are identical width) where they both converge at the "Y" trace using my digital calipers. That single trace on the PCB must pass the total current exiting the chip, onto the LC output filter loop (where there are fatter trace), then to the loudspeaker load. I measured that single trace bottleneck to be 0.35mm wide. Now imagine whether those PCBs are etched 1 oz copper
... or more likely less...?
Last edited:
Hmmm, that might explain partially why my Breeze with a 16V SMPS seems meatier than the dual chip with a 24V Meanwell. I thought it was all in my head!
I'm rather new to this so sorry if this is a stupid question. I'm looking at purchasing;
TPA3116 2 1CH Class D 2 50W 100W AMP CSR4 0 Bluetooth Module Amplifier | eBay
As I understand, the LR channels provide the theoretical maximum 50w at 4 ohms.
Is the subwoofer channel providing the 100W at 2 ohms?
TPA3116 2 1CH Class D 2 50W 100W AMP CSR4 0 Bluetooth Module Amplifier | eBay
As I understand, the LR channels provide the theoretical maximum 50w at 4 ohms.
Is the subwoofer channel providing the 100W at 2 ohms?
The tpa3116's when "watercooled" probably can, for the YJ amp to come close, apart from cooling not being close to sufficient, it needs BT signal amplified enough for tpa3116 to reach max output. In another thread someone has a YJ tas5630 BT amp, that is a 2x250watt 4 ohm amplifier in theory, maximum measured output is 5 watt into 4 ohm, (he uses a 48v PSU). So any YJ amp might be way way way way way way off.
Only RC mod on analog pin in will help noticably with hiss. Changing the potmeter to a lower value one will also help noticeably. All these mods are mentioned in detail earlier 🙂
Best very real improvement mod IMHO is removing the RC split on any ampboard that comes with it standard.
High frequency ripple reduction on AVCC pin is the goal I believe, TI more recently does advise the split without the R part, so AVCC/PVCC joined further from pins, decoupling caps in between chip and point where they join, bigger decoupling cap in AVCC leg.
The RC split seems to bring increased levels of DC on output, at least that is what I found, comparing several ampboards with the two that have the RCsplit standard and also comparing DC levels between those two RCsplit boards with same boards where I removed the RCsplit.
The RC split seems to bring increased levels of DC on output, at least that is what I found, comparing several ampboards with the two that have the RCsplit standard and also comparing DC levels between those two RCsplit boards with same boards where I removed the RCsplit.
Interesting! How much dc on outputs are we talking about? I've not noticed any negative sideeffects with the RC mod personally. Its one of the best mods for nearfield use in my mind.
Outputlevel matters (inputlevel too), at listening levels highest I saw was 708mV DC when at same time board with removed RCsplit reached 36mV.
What values does the R+C had?
I actually don't see a relationship between RC-filter and DC-offset due to imbalance.
Where did they, do you have a link for this?
I actually don't see a relationship between RC-filter and DC-offset due to imbalance.