TAS5611A with dynamic speaker load

Hello,

The Q of the Class-D output filter is determined by the impedance of the load. If the load's dynamic impedance drops the Q of the filter can rise and cause significant ringing and overshoot. This overshoot can damage the output of the amplifier by exceeding the maximum ratings. Schottky didoes are recommended to clamp this overshoot.

The output filter can be deigned to be critically damped with the lowest load impedance so that overshoot is minimized. However, this may sacrifice bandwidth at higher load impedances as the Q will decrease.

See this app not on Class-D LC filter design: www.ti.com/.../sloa119b.pdf

Best Regards,
Matt

Thanks for quick answer! What current rating of Schottky-diodes will be prefered? Are 3-Ampere SMD types ok?

3-Ampere should be a good starting point especially if the filter Q is well controlled.

While using the TAS5611A with filter damping RC (zobel) and schottky diodes direct to the OUT_A,B,C,D and stable constant 30V DC supply voltage the output shuts down for about 50ms while playing drum pulses in speaker load of about 3.4Ohms Rdc (PBTL, 22uH & 1uF LC filter). The ready signal goes low and the output tends to zero for these event. An overload case is not present because the /SD and /OTW flags stay high.

My feeling is the dynamic load introduces local thermal heating the TAS5611 is not capable of spreading to the heatsink...

Would it be helpful (without changing circuit parameters) to use TAS5613A or TAS5630B to get rid of this effect? Remeber the output MOSFETs may be the same with Rdson 60mOhm (typ.).

Kind regards 

If you are experiencing a thermal issue, the TAS5613A or TAS5630B shouldn’t offer too much of an improvement.

 

Do you have a schematic you can share?

Do you have impedance plots of the load?

Have you tried to measure if schottky diodes are in fact conducting?

 

You may want to remove the schotty diodes as a control and see if you are experience the same issue. It may also be beneficial to check the temperatures with a thermocouple. Does the heatsink/amplifier get hot before you experience the shutdown issue?

 

Measuring the peak current and voltage that triggers the shutdown event may also help pinpoint the issue.

 

Best Regards,

Matt

We are quite sure it is a thermal effect because it is independent of schottky diodes present at the outputs or a RC network for LC-filter dampening. Removing these things doesn't change a bit while it makes difference how long the amp is on an heating itself. When TAS5611 is cold, the error won't happen while after a minute of full power output the density of drop outs increase. We also changed the thermal interface material and the pressure and found a drop dependence.

(What makes me wonder the ready flag is changing for the moment the drop out appears while the SD and OTW signals are stable...? It cannot be overcurrent with Roc=22k and +30VDC supply)

I know about the tolerances you propose for the TAS series amps. I think about changing the TAS5611APHD to TAS5613A-DKD with bigger package and better heat slug, because it is easier to bring up to the heatsink. Hope this will give better results...

BTW: the amp is working in PBTL mode. Do you know if there are issues in contrast to normal BTL?

Kind regards

Have you tried measuring the current at the output? If you use a current probe around the wires going to one side of the load you can use a scope to view the peak current during the event. It would be good to rule this out before evaluating thermals.

Also if you have a burst generator, you can see how much headroom you actually have before OC trip. The burst will keep the temps low and allow accurate OC testing without the effects of temperature.

My concern was that the shottky diodes were clamping due to a ringing issue at a specific frequency and causing large currant the ground tripping the OC. However, it sounds like they are not this issue since there is no change without them.

Yes you are correct the thermal performance of the TAS5613A-DKD will be improved over the PHD package but we should rule out a current trip first.

Best Regards
Matt

The current when the drop out begins to occur is about 6 Ampere peak at fequencies of about 30 to 50Hz (Remember typical T.I.-PBTL app. with 4 pcs MSS1038-22uH-Coilcraft & 2 pcs 1uF). Typical for base drum like 'Hell freze over; Hotel California (Eagles)' .
There are other things should be taken into account. My feeling it is a mixture of more than one issue: I have read about the behavoir of TAS5613A in another thread where the chip was not able to reach more than 55W to 2Ohm (Roc=22k). Reason for limitation was low Isat of output coils! When using other coils with much higher Isat, the error disappeared and power was floating...
Reason could be decrease of inductance at high peak current and therefore loss of inductance (TI suggests min 7uH). So ripple current will increase then. This is the next point I will try out - new coils are ordered.
I have build up an amplifier with TAS5613A some years ago for full band application where I put in 30mm toroid core inductors with air gap and low THD. Stereo BTL application and never heard of problems...
Another point to be checked is the switching frequency adjust. TI suggests 10k at pin FREQ_ADJ to GND. I have used 15k to come down to 320kHz for inductor core loss and switching power reduction (subwoofer application). This also increases ripple current. In conjunction with inductance drop, this could be tha main cause. Self brewed... ;-)
I will check this next and come back with results.
Thanks and regards.

Hi, assembled back to suggested values Lout=10uH, bigger coils and fsw=400kHz, but no success.

Then I have made some measurements on TAS5611A when driven into clipping with speaker loaded (woofer Rcd=3.2Ohm 6inch vented cabin). This can be seen on the first picture. Interesting to see when TAS5611A is in clipping, the switching frequency drops down from 400kHz to 130kHz.

The second picture shows the output voltage at the speaker after the amp was overdriven (some millisecond ago...). At the end of the diagram the dark blue curve shows a spike in the curve at zero crossing. Right there the dark green curve shows the BST voltage (bootstrap) where you see the output voltage and the gate drive same time. Shortly after that spike the amp stops switching (READY pin goes low for about 50ms). It restarts then, but a drop out can easily be heard.

The lower part of the diagram shows a zoom of the BST voltage (time slice when spike on output voltage accurs) where an abnormal switching behavoir takes place: the low phase of the PWM is disturbed by small pulses! And 300us later the drop out occurs.

Interesting to notice, when the max overdrive level is reduced (smaller amp input voltage) or the supply voltage is increased by 3V (more headroom) the drop outs will no longer happen!

This leads to stringend input level limitation, doesn't it? Interested in your opinion...

Switching inductors was a good idea. The MSS1038-22uH-Coilcraft inductors have a low I-sat rating and can cause lots of issues if saturated. We have been using these with very good results: www.coilcraft.com/ma5172.cfm

As for the glitch issues:
From your waveforms it appears that there is shoot-through from the high to low side FET. The spike at the zero crossing of the blue output signal and the bootstrap voltage spike hint at this. The bootstrap signal should be stable and should not contain spikes. Any shoot through would immediately trip the OC trip point and this current would not be visible on the output.
To help diagnose:

1. Confirm that channels are paralleled after the inductors. You hinted that you were using 4 inductors so I assume this is the case, but I want to confirm.

2. Please capture PWM and bootstrap waveforms for all outputs A B C and D. Please display PWM out and bootstrap for channels A and C together with the same time scale. Likewise, capture B and D together. Preferably just a couple of cycles in the view for good detail.

3. You can also try to measure current directly at the output pins of the amplifier for A-C and B-D. You may be able to capture a current spike.

4. If possible please share a schematic. I will send you a PM so you can do so privately.

5. Make sure your heatsink is grounded. This will help reduce noise although this may not be a noise related issue.

Regards,
Matt

Hi Matt, I have already used bigger inductors of the ETQ series wit 10uH and Isat=8A, but the drop remains. So I decided to look for other issues...

For your information there is a limiter on the driver circuit limiting the output voltage of TAS56 to 16.5Vrms. And I had measured the peak current when the drop occured and it was about 6A. So saturation was not my first thought.

And there is also the other effect to be measured: start of double pulses in PWM signal and short after this shut down (diagram I have sent) and this happens when the overdriving pulse is long gone...

But anyway I will try out much bigger inductors and also try sharp limiting the peak input voltage corrsponding to about 27V peak output (5.5V headroom).

Kind regards, Bernd

Hi again, I would expect an overcurrent error from the TAS56 even in case of coil saturation and not something else...

But thanks anyway.

Hi Matthew, let me summarise the circuit: TAS5611A with 33VDC in PBTL mode, 10uH output inductors and Schottky diodes direct to the OUT_x pin, gate drive buffered locally 3R3&100nF and power inputs decoupled with four 2uF MLCCs (distance 1.5mm). Increasing the Isat of the output inductors didn't give any cure. A 3.2 Ohm speaker chassis is the target load and connected in a vented cabinet for all my measurements. I promise to you it will be interesting...

Here you see the output voltage of one summing point (speaker plus) in blue. The green curve shows the output current. You can see voltage and current are not in phase when the drop occurs:

The next picture shows the output voltage of out H-bridge output and the current throught one inductor in the moment the drop takes place (bottom=zoom):

You can see the output ripple current of 2A(pp) and the means value tends to zero...

For the moment the drop comes in here is the OUT_A and OUT_C and it's zoom and further zoom (bottom). Failing pulses takes place:

Then a little bit more to the point of interest:

In the zoom you can see the outputs, connected via two inductors to one capacitor, begin to switch against each other in a more dramatic way, phase shifted. 

But now here is something new! I got a mains fail detection circuit (diode clamp, RC filtered) connected to the secondary side of the switchmode power supply. In the case of mucis pulses in conjunction with real speaker load and heavy voice coil excursion the speaker is feeding back it's energy to the power supply (known as bus pumpimg). The TAS5611A like all H-bridges can't avoid this. The designer himself has to take care of this effect. So the power supply receives current from the speaker back and because it is a system with a feedback, the supply skips PWM pulses (a lot) to avoid overvoltage at the power voltage node. Here you can see the output voltage of the mains-fail circuit (green curve) one millisecond before the drop rushes in:

That voltage goes near zero for about 2ms and so the system does what it should in case of mains-fail: The /RESET of TAS5611A is asserted and the TAS goes to reset state while signal is still present. This is the drop out and the long search of something really stupid simple signal.

The reset causes the TAS56 to finally stop for a while and the overvoltage of the supply is no longer present so the mains-fail signal rises again and the TAS-Reset is set to inactive (high). Music will be played on...

There is one thing wondering me why the TAS56xx does not stop switching at the outputs OUT_A,B,C,D and is drifting into chaotic behavior when /RESET is asserted?

After re-dimensioning the clamp circuit the amplifier now works perfect and I would like to thank you for your help and I am really sorry for wasting your time. But maybe it was interesting to you, too... ;-)