TAS5630 3ohms load issue

Hi, Jezequel,

This is a little odd. I wonder if your speaker impedance isn't really 3 ohms... Maybe at 100 Hz it's not really 3 ohms?

-d2

Hi Don,

The impedace I am using is a set of  100W and 50W wirewound resistors. I don't think they are inductive enough to produce instabilities.

Furthermore, I noticed that the output PWM follows the oscillations. Maybe the problem comes from the internal feedback loop?

The amplifier works fine with "normal" audio (music). The oscillations appear only with a constant low frequency wave (< 500Hz square or sine).

Is it possible to add some compensation to the feedback loop?

Thank you

I have just checked the other pins of the TAS5630.

Bootstrap supplies are OK (no variation of the voltage across C25, C27, C29 and C31).

PSU_REF is pretty noisy (maybe it comes from my probe).

C_STARTUP is constant (3.25V) at low audio volume, but the voltage decreases when the output current becomes too high (during the square wave slopes). When it goes below 2.66V, the oscillations appear. Is it a normal behaviour?

Changing OC_ADJ resistor value was not significant. Below 20k, the amplifier stays in shutdown mode. Above 30k, it runs OK until the volume becomes too high ; it then goes in shutdown and needs a power down to restart.

Touching OC_ADJ with a x10 oscilloscope probe shuts down the amplifier.

Increasing C11 (C_STARTUP) from 4.7nF to 72.7nF just smoothed the voltage shape, but had no effect on the behaviour.

Hello,

Any idea?

Pascal

Jezequel, I understand you are using one of our EVMs.  I believe you have made some modifications, including changing Rocp and adding 68nF at C-Start.

Have you made any other changes?

From the 65kHz frequency that you described, I believe what you call "oscillation" is actually LC filter ringing on the rising and falling edges of your square wave input.  I believe we need to be sure of the inductance of your load resistors, because there should not be significant ringing at that frequency with resistive loads.  Can you provide this, either with measurements or data sheets?

There are a few problems with the way you are wiring TAS5630.  A couple of these may be causing failures (IC shuts down but does not recover with power cycling).

- Rocp cannot be set below 24k because the overcurrent limit that results is too high for TAS5630 to protect itself.  (See data sheet SLES220B, page 11, ELECTRICAL CHARACTERISTICS, Ioc.  This is the minimum permissible value of Rocp.)

- In BTL mode TAS5630 cannot be operated safely into a 3ohm load, especially not a speaker load.  The minimum load impedance is 3.5ohms.  (See data sheet SLES220B, page 4, RECOMMENDED OPERATING CONDITIONS, RL(BTL).  This is the minimum permissible value of R.L in BTL mode.  Note that this is an absolute impedance, not a speaker impedance rating.)

- TAS5630 can never be operated with PVDD = 60V.  The maximum is 52.5V.  (See data sheet SLES220B, page 11, RECOMMENDED OPERATING CONDITIONS, PVDDx.  This is the maximum permissible value of PVDD.)

Best regards,

Steve.

Hello Steve,

All the changes I made on the EVM for test purposes have been removed. All the parts have their original value.

I measured 2.82µH (100 ohms at 5.64MHz) on the resistive loads, including four 1ohm wound resistors and 2 metre leads.

The 3 ohms load simulates the worst case speaker impedance. The nominal speaker impedance is 5.5 ohms.

However, our application must withstand this 3 ohms load.

60V is the maximum output voltage of my power supply. The amplifier is always powered with 48V or lower.

Here are screeshots of the measured signals:

The input signal is a 100Hz square wave, PVDD is 40V.

The blue signal is the PWM output (before LC filter), 10V/div

The pink signal is the LC filter output, 10V/div

The green signal is the power supply current, 5A/div

This first screenshot is when the amplifier is running normally.

This is what happens when I increase input amplitude, with 4 timebase settings:

Do you call this ringing?

Best regards

Pascal

Jezequel, I don't know what load you were using for the scope graphs or what the value of Rocp was, but the final graph tells us that TAS5630 is doing one of 2 things:

- pulse skipping near clipping; or

- cycle-by-cycle current limiting (CBC).

In either case TAS5630 reduces its switching frequency from the normal 400kHz to 1/3 of that, 133kHz.  In clipping this is done to extend the switching cycle to get closer to 100% duty cycle.  (100% duty cycle cannot be permitted because some off time is required to recharge high side output device bootstrap capacitors.)  In CBC switching pulses are truncated to hold output current at the limit.

It is not possible to say more without knowing Rocp and the loads.  Please tell me what those were in these measurements.

By the way, it appears that the ringing is LC filter ringing.  It has a period of about 15uS, so frequency of about 67kHz.  It must be excited by interruptions in output current when pulses are skipped.

Best regards,

Steve.

Hello Steve,

Rocp is 24k (R12 on the EVM)

Rload is 4 ohms

All the components values are those of the TAS5630DKD2EVM.

Is it possible to damp the LC ringings? I am expecting EMC issues.

Best regards

Pascal

Jezequel, I expect it is possible to damp the ringing, but I think we need to determine root cause before trying that.  I have a couple of questions.

- You are using an EVM and not your own circuit - is that correct?

- Can you check impedance of your 4ohm load over frequency from around 10kHz to 1MHz?  I think the answer may be there.

Best regards,

Steve.

Steve,

It's correct, I am using the EVM. I don't want to design a board while this problem is not solved.

Here is the 4ohms load impedance measurement (Excel sheet): 5808.Impedance measurement.xls

Impedance units are ohms.

Frequency units are hertz.

Voltage units are volts.

Best regards

Pascal

Hello,

Can you give me a solution?

Best regards

Pascal

Jezequel, your resistors seem OK.

Can you provide a scope photo of voltages at the 2 sides of a BTL output plus the current into one of the inductors, with the conditions you used before?  (PVDD 40V, load 4ohms.)  Please also include PVDD at the IC and your input voltage at the IC, and use the same time bases as before.  Maybe these will reveal what is going on.

Best regards,

Steve.

Hello Steve,

Here are the measurements you asked:

The blue and purple curves are the two BTL outputs (10V/div).

The orange curve is the current into the inductor of  "purple" channel (5A/div).

The green curve is the PVDD rail, AC-coupled, 0.5V/div. Most of the noise comes from the power supply.

First measurements without oscillations:

2nd measurements with oscillations:

The green signal is now INPUT_B:

The green signal is now INPUT_A:

Best regards

Pascal

Jezequel, thank you.  However, I was looking for the switching outputs, not the LC filter outputs, to compare the switching waveform in detail to the current generation.  This was not clear.  Can you generate those waveforms?  Please include PVDD but not the inputs, so it doesn't take so long - the inputs are fine.

The most important part of this is near the beginning of a square wave cycle, where the ringing begins.  I hope that reveals what is happening (which I think is TAS5630 response to a clipping condition, but we shall see.)

Best regards,

Steve.

Steve,

Here are the new measurements with the same setup as before.

Blue: PWM output

Purple: LC filter output

Orange: inductor current

Green: PVDD (AC coulped)

Rising edge of the 100Hz square wave:

Various screenshots of ringings:

Hope this will help.

Best regards

Pascal

Pascal, thank you for sending me the new graphs.  The graph with 5uS time base may tell us what is going on.  I cannot see both switching outputs of the BTL channel, and I think maybe the current measurement is out of phase with the switching voltage measurements, but still this looks like a clipping condition.  (I will guess that your PVDD is about 48V in these graphs.)  The clipping condition could be caused by setting the output target above what the IC can deliver, but I think it also could be caused momentarily by noise, including power supply noise.

TAS5630 high side bootstrap capacitors require some output off time every third switching cycle or so to remain charged fully enough to keep high side FETs fully on.  So when TAS5630 duty cycle approaches 100% the device starts skipping pulses and injecting a low pulse at 1/3 of the normal switching frequency.  Since output current (the same as inductor current) is only about 10A peak, it seems it is clipping rather than OCP (over current protection) that causes the pulse skipping.

Can you try an experiment?  Set PVDD = 40V and set input voltage to the level at which the ringing first starts.  Then increase PVDD to 42 or 44 V without changing the input level and see if the ringing stops.

Best regards,

Steve.

P.S.  When you send me more scope graphs, it is not necessary to include PVDD any longer.  Instead I would want to see the 2 switching outputs of a BTL channel, the inductor current and the LC filter output of one side of the channel.  Also, the most interesting points are those at which the current hits peaks.

Hello Steve,

I tried to increase PVDD but it has the inverse effect: increasing PVDD increases ringing.

Here are the graphs:

Blue: output A

Purple: Output B

Green: LC filter of output A

Orange: current into the inductor of output A

First measurement with PVDD=40V, just before ringing appears:

Same as above, with ringing:

Larger scale with PVDD=40V:

PVDD=41V, same input signal:

PVDD=42V:

What happens if I increase bootstrap capacitors values?

I'd also try another experiment: I will add a feedback from the amplifier outputs to the corresponding inverting inputs of the preamplifiers, then creating a closed loop. The datasheet says that the TAS5630 inverts the signal from input to output so I will connect output A to the positive input of the OPA1632 through a 120k resistor, and output B to its negative input. The loop gain will become 23dB instead of 30dB (actual open loop gain).

Pascal

Hi Steve

I'm evaluating the TAS5630 and read above in your post from 08-31-2011 1:32 PM that the maximal voltage must 52.5V not exceed. The datasheet writes an absolute maximum of 69V from PVDD to GND. Below is written under (2): These voltages represents the dc voltage + peak ac waveform measured at the terminal of the device in all conditions. What is ment with this?

Our powersupply has about 56V, seems to be a problem.

Thank you

Michael

Steve;

You forgot the minus sign. Lenz's Law: V = - L di/dt

Thank you, it seems that it's not easy to decrease the power supply... Do you see a chance with a voltage supressor, to limit the voltage? Maybe this has been already tested.

Best regards, Michael

Michael, a voltage suppressor on the output of a DC power supply would likely draw high current when activated.  I don't recommend this solution.  The answer is probably in power supply regulation.

I think you need to find a better regulated power supply.  If the DC value of the power supply output can go above 52.5V it will be very risky to use it with TAS5630.

Best regards,

Steve.

Steve, I tested it with a simple linear regulator circuit from 55V to 51V and it worked not so bad. Maybe i'll use a switcher in the future, if higher power rates are required.

Thank you for your inputs.

Best regards, Michael

Michael, the result with regulated supply is good, but please do not operate TAS5630 with PVDD > 52.5Vdc.  There is risk of damaging the device.

You should be able to regulate to 52.5Vdc or less with the regulator you have tried.  You may want to move to a switcher in future to minimize power consumption, if you are trying to meet governmental requirements.

Best regards,

Steve.

Completely contrary to Steve's theory that the problem you are seeing is due to pulse skipping of the PWM, I believe the problem is due to the undamped nature of the L/C filter on the output of the amplifier. The frequency of around 63kHz is telling.

You can "settle the bet" by putting your own R/C network directly on the output. For absolute perfect damping, the resistance should be around 4 ohms and capacitance should be (unfortunately) around 4 microfarrad, although resistors as high as 10 ohms and capacitors as small as 1 uF may also work. The unfortunate part is that the resistor will dissipate heat at high frequencies, but this is needed to add damping to the L/C filter.

Another way to test Steve's theory would be to select an alternative PWM switching frequency. These parts have special "AM Avoidance" frequencies which you can select. If Steve's theory about pulse skipping is correct, then this frequency will be reduced from 63kHz as you reduce the switching frequency of the PWM.

If my suggestion of using an R/C filter tames (or solves) the problem, then it will indicate that the impedance of your load (AT 63 KHZ!!) is high -- perhaps well over 10 ohms. This could be due to the inductance of your load, which of course includes cross-over networks.

It is certainly easy enough to add an R/C filter at the output and see what happens. 

Bill