Hello to all,
I built a prototype of the amplifier with circuit TAS5630.
I used a total of 3 circuits TAS5630 for 5.1 channel sound. Differential outputs are interference-free, beautifully clean sound.
But, by SE Outputs is heard a audible noise (around 1kHz). Wenn I touch the heatsink, the audible noise is lower, but always heard.
I´m using one heatsink for all 3 ICs, but separate with the thermal insulating foil.
Questions:
- Must be Heatsink connected with GND? (I have connected GND with PE over USB)
- Are the cooling areas from TAS5630 connected with GND or similar?
- Any other idea, why is the audible noiese presented?
best regards, Radim.
Hi, Radim,
We recommend grounding the heat sink. The thermal pad on the top of the TAS5630 is loosely connected to ground inside the package. By solidly grounding it externally, you block any noise coming in.
-d2
-----
Don Dapkus
Audio Applications Engineering Manager
Dallas, TX USA
Radim, Don is correct, the heatsink must be grounded solidly. But I noticed you said "I´m using one heatsink for all 3 ICs, but separate with the thermal insulating foil."
This is a problem, because the substrate of TAS5630 is connected to the Power Pad on the top of the device. The Power Pad must be electrically connected to the heatsink so it can shunt any currents that enter the substrate to ground. Without this connection any currents that enter the substrate can flow randomly and cause problems like noise.
We recommend using a high-conductivity thermal compound instead of a thermal pad and especially instead of anything that insulates. We prefer a compound called Ceramique from a company called Arctic Silver, because it has the correct properties. I have inserted a data sheet for this compound.
Can you try this and let me know if it improves the noise?
Best regards,
Steve.
Hi Steve,
thank you for your reply.
Problem is, that I can´t use your Céramique Product or any other products without adhesive sides. I´m using Thermally conductive foil both sides adhesive from Fischer Electronic. Look at Datasheet. and through this foil I glue the heatsink to three TAS5630.
What I can do is, that I can connect the Heatsink through wire to Ground. It is better then nothing?
I have problems with noise only by unbalanced outputs, so I did another PCB, where I use only balanced outputs and it works without any noise.
Radim, it is very risky to operate TAS5630 without a path to ground from the Power Pad, whether this is causing the noise or not. It can lead to damage to the device.
I am also now concerned because it sounds like the 3 TAS5630 in your system are supporting the heatsink mechanically, and this also is very risky. Torque from motion of the heatsink can damage the device mechanically.
We recommend only mounting the heatsink with a good mechanical connection to the PCB as well as a good electrical connection. Without this the ICs can be damaged electrically or mechanically.
Did you attach a copy of the data sheet you described? I did not see it, and I would prefer to receive it from you so I can be certain that I have the correct data sheet.
You can try grounding the heatsink with a wire, but please make that a very strong connection, and let me know what you find. However, without the data sheet I can not know if the thermal foil is conductive between the heatsink and the Power Pad. If it is not, grounding the heatsink is not likely to help.
I attach the datasheet conductive foil. This foil is only a thermal conductive with electrical insulation of 5kV ...Question. May I connect through one heatsink all three thermalpads, when heatsink is grounded? or do I have to use a separate heatsink for each chip ...
(When omitted mechanical hazard.)...
Radim, there are 2 types of adhesive pads in the data sheet. The one that uses Kapton "isolates" (5kV/mm, but complete thickness of the pad is only 0.127mm). The other that uses aluminum foil does not "isolate", but it does not provide a figure for electrical conductivity. I am not sure what you are using, but you mentioned the aluminum foil earlier.
If that is what you are using, there may be enough conductivity and capacitance through the aluminum foil pad to connect TAS5630 adequately to the heatsink electrically.
You should not need separate heatsinks for the different ICs. It should be possible to attach a short heavy wire near the center of the heatsink and connect it to ground very close. If in fact the aluminum foil pads connect TAS5630 adequately to the heatsink electrically, this will be good.
- Please understand this: it is very risky to operate TAS5630 without this sort of ground connection from the Power Pad through the heatsink to ground. You must provide such a connection, and it must be conductive, not isolating or insulating. Otherwise stray switching currents into the substrate of the device may interfere with its operation, and this can cause effects including erratic operation and maybe noise, and even failure.
I do not understand the phrase you wrote earlier, "(When omitted mechanical hazard.)...". But I must emphasize that it is risky to rely on fastening between the heatsink and the TAS5630 to support the heatsink. This is not an application mode that TI can support. There is too much risk of mechanical damage to the device.
sorry for my english... German would be better for me, but ...
i am using the isolated version of foil with 5kV isolation, because under heatsink are other parts, such resistors, capacitors etc... But you write, that it is wrong.
So I can use isolated foil under the entire heatsink (for isolation of other components) and in places where are TAS5630`s i make a cut. With this will be the TAS5630`s and their Thermalpads connected with heatsink ... The whole heatsink I screw with two screws to the PCB and at the same time I connect its to the ground. Right?
I will write you, when the changes will be done.
thanks a lot. Radim.
Radim, when you cut away the thermal pad between the heatsink and the 3 TAS5630, please use something like Ceramique between the heatsink and the IC Power Pads. This will complete the connection thermally and electrically.
I did not realize you had included screws to mount the heatsink. I would like to advise 2 things about the screws.
- If the screw heads contact ground on the PCB with a strong connection to ground at the TAS5630, you should not need a wire. The screws will make a good enough ground. (This is what we do in our EVMs.)
- It is necessary to have a screw at each end of each TAS5630 (for your system, 4 screws total). Without those fasteners the PCB will bend and will not be parallel to the IC Power Pads, so the connection between them will not be complete - there will be gaps and they will defeat the connections. So please use a screw at each end of each TAS5630. Maybe this presentation can make this clear.
7536.TAS561-3x-Hsink-connect-mount-120404.pdf
Radim, have you had any success in providing grounding from the PowerPADs through the heatsink? I hope the presentation I sent helped.
I would like to go back to something you said earlier, that you heard a "noise around 1kHz" at your SE outputs.
- What are the operating conditions when you hear this "noise"? Are you driving other outputs in your system, and do you have loads on those outputs?
- If you are driving other outputs, how does the frequency match the frequency you are using to drive those outputs?
I wonder if this is not simply crosstalk within the IC that you are using for SE outputs. Can you provide a scope graph of the different waveforms in your system? It will be important to identify the vertical scaling. Also, it will probably be necessary to filter the 400kHz ripple in the SE output waveforms to make them clear.
I have isolated all other components from heatsink and all three TAS5630 are yet connected together with Heatsink with GND. It seems to be OK, but one channel of 6 seems to be demaged. Sometimes I hear from this Channel noise, and sometimes is OK with great sound. It makes with grounded Input too. So i suppoused, that Output parts are not good.
Now I do not have much time. (unfortunately).
to SE Outputs:
I left the design with SE Outputs. Now i am using only balanced outputs and they are good. The noise by SE Outputs was worst, wenn all three IC were enabled. So I think, that problem is due wrong Layout. The noise was the same with and without load and with/without input signal.
is there any evalboard, where SE Outputs are using? (with Layout).
what kind of signal exactly do you want to see? Output´signal on load?
I attached my Schematic. Probably you find my other mistakes :)
with Regards, Radim.
Radim, I do not see a problem in your schematic. I am sorry, but we do not have an SE layout. The best references for PCB layout are the existing EVM, TAS5630PHD2EVM, and notes in the data sheet about PCB layout. I think these will be essentially the same for BTL and SE outputs.
I would like to see the signals that you describe as "noise around 1kHz" and other outputs. I need to know what is operating when the noise occurs, all channels. It will probably be necessary to filter the "noise" signals strongly to eliminate components like switching ripple that are outside the audio band. This may be possible with RC filters but it may also take LC filters. I have made combined RC+LC filters, with the following design. These worked well.
4666.Class-D-meast-RC+LC-filter-120411.pdf
I want to understand if what you are hearing is broadband audio noise or crosstalk between channels. At 1kHz with PHD package like you are using and a good PCB layout, crosstalk to SE channels is around -60dB from other outputs, depending on load and frequency. It is a little higher at 100Hz and a little lower at 10kHz. These levels are low but still audible if they are within the passband of the loudspeaker used for the listening.
Radim, I thought more about what you call noise and about the fact that the noise got worse as you enabled more TAS5630. I have a few more suggestions.
- If there is high ripple on PVDD caused by audio and switching currents into loads, this can increase the crosstalk. Are you feeding inputs to all 3 TAS5630 at the same time, or are some of them quiescent? If the "noise" gets worse only when you drive more of the TAS5630, PVDD ripple could be the cause. If PVDD to the SE channels is drawn from a PCB bus and this is a cause of the problem, it might be possible to reduce it by taking the PVDD supply to PVDD_C and PVDD_D directly from the power supply where it is regulated better.
- I noted in your schematic that your 3 TAS5630 are not sync'd to the same switching frequency, so they switch at slightly different frequencies. It is possible that beats between the different switching frequencies are the cause of what you hear. To sync the TAS5630, pick one as the Master. There is no change to wiring for that device. For each of the other 2 TAS5630, connect the FREQ_ADJ pin to VREG. Then drive the OSC+ and OSC- of those devices from OSC+ and OSC- of the Master device, with resistors like 100 ohms in series with each connection.
- It is possible that what you hear is caused by ground interference in the lines from your SE source opamps to TAS5630 SE inputs . You can try disconnecting the SE inputs before the 10uF input caps C133 and C136 and grounding the disconnected side of each cap to AGND at the device. If ground interference is a cause of the problem this will reduce or eliminate it. (Maybe you have already tried this.)
I am looking forward to seeing waveforms that could help answer these questions.
Radim, I would like to check status of this case. Have you been able to resolve the question with SE outputs, or determine if it is made worse by PVDD ripple or other effect? It is possible it is normal TAS5630B crosstalk.
i´m so sorry about this case... I had to stop this project and start with another. I will try to measure the signals, but in couple of next week i have to pay attention to another project...
but, how i wrote already. I left the design with SE Outputs --> i will always use only differencial outputs.
i write you, when i make a measurement.
Radim, thank you for your reply. I expect that what you heard was normal crosstalk in the IC, so it cannot be resolved easily.
If you do not intend to use the SE configuration I will not ask you to do anything more on this case. If you still decide to check this in future, I will be very happy to hear your results.