TPA3251D2EVM: TPA3251D2 design review

Hi Anders,

Thats a great design concept.  a tri-amped speaker with one IC!  

regarding the crossover side of things, I'll direct you to a TI design that was done for an analog active crossover a few years back.  Its a two way design, but much of the design concepts should be similar.   What is the end application?   If its more of a reference monitor sort of thing you may need to consider some additional filters to even out the frequency response.  that of course depends on the speakers you end up choosing.

As for the amplifier.  You haven't chosen values for the inductors in the LC filter,  so its hard to comment on that.  what I will say is that we typically choose inductors in the 6.8-10uH range. Make sure they have good linearity as that will play a role in the performance of the amplifier. check out our application report on LC-filter design for more info there.  Also check out our LC filter designer tool. The tool is an excel spreadsheet that will help you choose the right value inductor and capacitor depending on your desired cutoff frequency, speaker impedance, etc..   Its really helpful. because you've chosen the .68uF caps,  you could use a similar value to the EVM for the hf,  but you could pull the cutoff frequency lower  for the mid and low frequencies.

One thing I noticed was your use of C66 and C67 on the BTL outputs.  These are unnecessary.  The DC blocking caps are required for single-ended outputs only.  with BTL there is an identical DC offset on each side of the BTL amplifier,  so there is no DC current through the speaker.  also,  the TPA3251 has an intelligent ramp up circuit that prevents popping on power-up and shutdown.  so no worries there. 

for the other caps on the outputs like C70 and C74. make sure those are ceramic capacitors,  or if you really want to use metalized film,  just make sure they are higher voltage 160-250V.  The lower voltage film caps have a higher ESR and the ripple voltage can actually cause them to blow up. 

you don't have values for C1 and C2.   make sure that you make them 1000uf minimum and place them near the PVDD pins on the Device when you do the layout.  use the EVM layout as a guide there. 

For your low frequency inputs.  You need to drive that section differentially.  check out the EVM schematic to see how we convert from single ended to differential. using two opamp stages.   alternately you could use a nicer opamp like the OPA1632 which has a differential output.  Thats a really nice IC.

Lets stop there for now.  take a look at those application reports and tools I mentioned and resubmit your schematic with the changes,  or feel free to ask questions too.  

best regards, 

-Steve 

Hello Anders,

please feel free to post your updated schematic after you make the adjustments. I'm happy to take another look at it.

best regards,
-Steve

Hi Steve.

Thank you very much for your valuable feedback. I agree that the concept is great ;)
I have added values to the components to reflect the BOM that will be used for assembly. They are mainly taken from the EVM BOM to be aligned with TI recommendations. The LC filter inductors are 7uH, and the capacitance on the PVDD supply is 3300uF. And I confirm that the capacitors used in the output are ceramic.

I have added the OPA1632 to the design make the LOW signal input differential. Thank you for this one.

TPA3251D2_AND_FILTER_V2.sch.pdf

Anders,

I think one of the strengths of TI's High power audio amplifiers is that they can do a tri-amped speaker with one device. We have a lot of customers doing biamped too of course.

The OPA1632 is a great choice for the differential output opamp. We will actually be using that device on an upcoming EVM. It really is a great opamp. The BTL channel inputs should be AC coupled though, you had this prior to adding the opa1632, just make sure you add them back in. I recommend using 10uf capacitors to prevent any low frequency rolloff. The 1uf is fine for the higher frequency content

The value of your cstart cap should be higher. The EVM uses .47uf when using 2SE+1BTL mode. This will prevent and audible pop in the SE channels when you bring the device out of reset.

Also you still have C66 and C67 in there, is there a reason? you can remove these. in the EVM there is a Shunt that bypasses those capacitors by default, if a user wants to use SE mode, they need to pull the shunt. but for BTL those caps are uncessary, and will only add distortion.

you may have taken C1 & C2 a little too high, there is a lot of inrush current with such large caps, and your power supply will need to be able to handle that, if you design for it, no problem. But if you use 2,200uf that is plenty, 1000uf is what we consider to be acceptable.

regarding your inductor choice, the MA5173-AE is a fine inductor, but for really high performance, consider the Wurth 7443630700. This inductor has the highest performance of any we have tested so far. you can see the performance plots in the LC-Filter application note I linked to in my first post.

feel free to make use of the Clip/fault circuits from the EVM too. It can be nice to have that kind of feedback.

thats all the feedback i have for now.

What is the impedance of your speakers?

best regards,
-Steve Wilson

Hi Steve,

Thank you for your valuable feedback. It is all things, that will help to improve the design.
For C66 and C67 they are included in the design, to keep the opportunity available to use the amplifier as 4x single ended. When soldering the design, I will short the capacitor poles.

The design have now been produced, but with not a lot of success. The filters are perfect, and cut exactly where they were calculated to do it..
The amplifier though was not good. There are absolut nothing in the design that seems wrong, but there are no sound, and it seems totally dead. The 3v3 was supplied via an external supply (of course with a common ground connection), and on the first power up the power consumption were1A.. Way above expected, as the it is only used for config Mode doing power up.

Can you tell me if there is a need to have 10K pull up resistors between the mode pins? In the current design these pins are directly coupled to 3v3. When studying the design, are there anything that comes to your mind that could be an issue?

Thank you again for your inputs.
   - Anders

Anders,

I send you a connection request, lets take a look at the layout and see if anything pops out.

best regards,
-Steve

Anders,
Just checking back in with you. I believe that if you send me the layout it may shed some light on the issue. unless you've already solved it.

just let me know.
best regards,
-steve

Though I agree with the replies in that the concept is great, I wonder whether it will work well.

In particular, I wonder about the mode selection (pins M1 and M2).  I'm not sure I understand what the reason can be to "tell the IC" that it is working on single-ended vs. differential mode;  but there is a configuration setting for each, so there may be a reason, and then it may not work well when mixing.

You configure it as single-ended (four independent channels), and one can certainly combine two independent amplifiers into a bridged configuration (which is what you're doing).  Again, this makes me wonder:  why is there a configuration setting for single-ended (M1 = M2 = VDD) and one setting for BTL (M1 = M2 = 0) ??   Does it affect the DC protection feature?  The way the internal circuits handle the feedback?

I'm interested to hear any comments on this.

Thanks,
Carlos
--

Hi Carlos,

You are correct that academically speaking, that the function of channels A and B in SE mode wouldn't be any different from BTL mode given identical inputs. This is strictly looking at them as amplifiers however, inside the device there is a little more going on, and the differences are subtle but important.

1. Protection circuitry. SE mode has DC speaker protection and Pin to pin short circuit (PPSC) protection disabled. PBTL has DC speaker protection disabled.
The behavior of the protection circuitry is also different for PBTL in that if there is a local error detected in channel A, it will shut down all 4 channels, where as in SE or BTL this would just shut down channels A and B.

2. There is a phase offset in the PWM of the channels in different modes. In PBTL all 4 PWMs are locked on the same cycle, in BTL A&B are locked on to each other while C and D are shifted, and in SE all the outputs are shifted out of phase with one another, this helps with EMI.

Having the outputs locked in PBTL allows the device to be used in Pre-filter PBTL mode. I.E. the outputs are connected in parallel before the LC filter. This could not be done in BTL mode because of the phase shift. you essentially get shoot through and it will destroy the device. ( Ask me how i know ).
This is because the OC requires a few cycles, and since the PPSC is only for related output pins, (A&B or C&D) the error is not detected.

3. Finally, in PBTL the inputs of A and B are routed to the C and D amplifiers. Thus you only need 1 differential input. the inputs to Channels C and D should be grounded.

So there are good reasons to have the Mode pins, but you are correct that the different amplifier configurations do not inherently necessitate different modes... there are other considerations however, that make the different modes necessary.

That may have been a long winded explanation but I hope it answers your question.

best regards,

-Steve Wilson

Steve. I have been following this thread with some interest as I have completed our new design for an amplifier using much of the EVM layout guidlines along with the SE to DIFF opp amp circuit. Our design is BTL. The point I am interested in is the observation you made regarding the C start cap value. The EVM schematic shows 100nf for BTL mode but for us we are getting the amplifier fire up and play music for a few seconds then shut down. The current draw is no more than 120ma during the working period dropping to 5ma when shut down kicks in.

Power supply is 20v from a benchtop supply at present but on the PCB we have incorporated the made lner power supply to provide the PVVD at 26V with large 6800uF 80V caps across the rails.

We have left the reset pin unconnected but other than that we have prety much duplicated the EVM schematic and PCB layout. Its quite a dissappointment at present as we are not sure how to resolve the issue. would appreciate some assistance as we are waiting on the design to finalize a product for prduction this summer using a PCB full of TI products.

hope you can help. If I have posted this question and broken any rules I apologise as this is only the 2nd time I've posted.

I have schematic and layout gerbers if you would like to see them.

cheers

John

Hi John, 

I'm sorry to hear you're having some problems.  I am certain that we can find the problem and get your project back on track.  I've sent you a connection request, so you can send the schematic/layout in a private message.   I will say that The C-start cap shouldn't cause the amplifier to shutdown,  its controls the DC ramp on the output.  because there is a DC offset on the outputs of each amplifier section, there is a controlled ramp up,  which eliminates the pop at startup.  

For the sake of keeping the threads separate,  Just copy and paste your last post into a new thread, and I will respond there. 

best regards, 

-Steve Wilson

Hi Steve.
Long time no see.. It have been busy months, and then this project was put on hold.
The first version of the design did not work at all, and we have modified the design according all your recommendations. There is still the capacitors on the bridged output, but they will not be mounted. They are good to keep here as their size can be tricky in case the design become 4x SE in a later revision.
As the first revision did not work, I hope you can spend some time on the layout as well as the design. The active audio filters worked perfect, but we had absolute no sound from the amplifier.
The change we made is:
- Removing the 3v3 LDO and use the onchip version (DVDD)
- Added the differential op amp.
- Added access to fault and clip pins on TP.

Anders, 

I'm glad to hear from you.  I was wondering how this project was going.  I'm sorry to hear that you are still running into problems.  But I hope we can get them solved for your quickly. 

You mentioned that you have added access to the Fault and Clip pins. when you turn the amp on,  what do those pins measure?  Do you get PWM output at all?  Ie. is the amp idling? 

One other thing, you may consider a different solution for the reset pin. When you turn the amp on and off, you can get some rather large pops unless you turn the device on and off while it is in reset.   you can take a look at our EVMs  there is a power supervisor IC that we use for that function, and it works quite nicely. 

-Steve

Hi Steve.
Rev02 has not yet been realized, so I do not know the values of the Fault and Clip pins. When I measured them on the first version, they were active which of course is a problem.
I had noise on the output, and nothing that really made sense in respect to the inputs.. I will consider an alternative solution for the rest, as this annoying speaker pop is worth to avoid.
Did you receive my layout and design files on a private message?

- Anders