AMC3330-Q1: High voltage divider circuit does not work

Hi Tang,

I'm a little confused on your description of the waveforms.  To me it looks like there is a ~650kHz noise signal on the A-phase line (Green waveform) and that correctly gets divided down through your voltage divider circuit and shows up across the INP/INN pins.  I also see a pretty higher amplitude noise signal at a ~150kHz across your INP/INN pins.  I've marked your scope capture to show what I'm saying.

Do you know where the ~650kHz and ~150kHz noise signals are coming from?

Would you be able to provide scope captures of the device output as well as the high side supply signals (HLDO_IN, HLDO_OUT)?

Hi Ahmad,

Saleh Ahmad said:

there is a ~650kHz noise signal on the A-phase line

The high voltage (green waveform) is generated by a three-phase two-level DC-AC converter, and the switching frequency is 160kHz. Since there are 6 IGBTs, the frequency of ripple may look like 6*160kHz=960kHz.

Saleh Ahmad said:

correctly gets divided down through your voltage divider circuit

The ratio of divider is 200:1. For the fundamental component (50Hz), the divider works properly.

However, it does not work when dealing with 160kHz component. As can be seen, the peak-to-peak amplitude Vpp of ripple voltage across UP and UN is about 2V. And the Vpp of INP-INN is about 1.5V. The scale is not right.

Saleh Ahmad said:

I also see a pretty higher amplitude noise signal at a ~150kHz across your INP/INN pins

Yes, this is the switching-frequency component, which is not scaled down by the voltage divider. This confuses me, and so i ask this question.

Saleh Ahmad said:

Do you know where the ~650kHz and ~150kHz noise signals are coming from?

As I mentioned above, the ripple voltages at 160kHz and multiples of 160kHz are generated by a power electronics converter.

Saleh Ahmad said:

Would you be able to provide scope captures of the device output as well as the high side supply signals (HLDO_IN, HLDO_OUT)?

Actually, I remove the AMC3330 chip, and the voltage waveforms of UP-UN and INP-INN remains unchanged. Do you know the reason? 

is this a common mode component? Is it necessary to connect HGND to PE through 1MΩ//1nF? 

Hi Tang,

tang wy said:

However, it does not work when dealing with 160kHz component. As can be seen, the peak-to-peak amplitude Vpp of ripple voltage across UP and UN is about 2V. And the Vpp of INP-INN is about 1.5V. The scale is not right.

I don't think the 160kHz component on the INP/INN pins is coming through the voltage divider.  

From the scope capture it looks like the green waveform only has the higher frequency noise on it (~650kHz) and not the 160kHz signal.  The blue waveform looks like it has both the 160kHz and ~650kHz signals.  

tang wy said:

Actually, I remove the AMC3330 chip, and the voltage waveforms of UP-UN and INP-INN remains unchanged. Do you know the reason? 

Just to verify,  both waveforms look the same without the AMC3330 device on the board?

tang wy said:

Is it necessary to connect HGND to PE through 1MΩ//1nF?

No,  the HGND only needs to be connected to the INN pin and DCDC_HGND pin.

How are you measuring the waveforms captured above?  What kind of probes are you using?  Also,  have you checked the output of the AMC3330?

Thanks.

Hi Ahmad,

Saleh Ahmad said:

Just to verify,  both waveforms look the same without the AMC3330 device on the board?

Yes, the waveforms of INP-INN remain the same no matter the AMC3330 is on the PCB or removed. As given below, the left figure shows the waveforms across INP-INN without AMC3330, while the right figure shows the waveforms with AMC3330. They look exactly the same.

It should be noted that i have changed the switching frequency from 160kHz to 16kHz. Thus the ripple frequency is 16kHz in these figures.

Saleh Ahmad said:

How are you measuring the waveforms captured above?  What kind of probes are you using?

I use a Tektronix MSO58 oscilloscope and a high voltage differential probe THDP0200 to measure the waveforms captured above. The experiment setup is shown below.

Saleh Ahmad said:

have you checked the output of the AMC3330?

Yes, I have checked the output of AMC3330. The voltage across OUTP-OUTN is twice the input voltage INP-INN, which means the AMC3330 functions well, the problem lies in the input of AMC3330.

Hi Tang,

Do you see this noise show up on all your voltage divider measurement points?

Can you remove the 0Ohms on the site without the AMC3330 and see if the noise still shows up across the INP/INN pins?

Thanks.

Hi Ahmad,

Saleh Ahmad said:

Do you see this noise show up on all your voltage divider measurement points?

I have measured the voltage across different points, as shown below, named as U1, U2, U3, U4. U1 represents the high voltage, while U2/U3/U4 look almost the same. The 50Hz component is scaled down, and the high-frequency component remains almost the same.

Saleh Ahmad said:

Can you remove the 0Ohms on the site without the AMC3330 and see if the noise still shows up across the INP/INN pins?

I have removed the 0Ohms and measured the voltage across INP/INN pins, and the high-frequency components are very low.

Can you tell me the value/material of the resistors that TI uses to divide the HV for AMC3330 during test? Do you recommend any specific resistors? I think this problem may be caused by the nonideal characteristics of resistors.

Actually, I try to use 4*1MOhm + 20KOhm to divide the HV（theoratical scale = 200:1, the same as before）, and the actual high-frequency scale is 8:1. Although the scale is still not correct, it is better than  the actual high-requency scale of 1:1 when I use 2*100K + 1K to divide the HV.

Hi Tang,

tang wy said:

Can you tell me the value/material of the resistors that TI uses to divide the HV for AMC3330 during test? Do you recommend any specific resistors? I think this problem may be caused by the nonideal characteristics of resistors.

Actually, I try to use 4*1MOhm + 20KOhm to divide the HV（theoratical scale = 200:1, the same as before）, and the actual high-frequency scale is 8:1. Although the scale is still not correct, it is better than  the actual high-requency scale of 1:1 when I use 2*100K + 1K to divide the HV.

I'm not sure if we have specific recommendations for the resistors to use for the divider, but I'll check and get back to you on this. Also, sorry for not realizing this earlier, but typically you do want the higher value resistors for the voltage divider to at least minimize the power losses.  Could you also try seeing the effect of removing the 1nF cap with the 4MOhm+20kOhm divider?

Thanks.

Hi Ahmad,

Saleh Ahmad said:

I'm not sure if we have specific recommendations for the resistors to use for the divider, but I'll check and get back to you on this.

I think this information would be useful. looking forward to your reply.

Saleh Ahmad said:

Could you also try seeing the effect of removing the 1nF cap with the 4MOhm+20kOhm divider?

When 2*100K + 1K divider is used, the paralleled capacitor is 1nF. When 4MOhm + 20kOhm divider is used, the paralled capacitor is 47nF, so that the RC time constant remains the same (1K*1nF ≈ 20K*47pF).

Therefore, the reduction of high-frequency component is caused by divider, instead of a lower bandwidth of RC filter (which remains the same). 

According to your suggestion, I have compared the waveforms of INP-INN with/without Capacitor, as given below. With capacitor removed, some noises at higher frequency can be observed, which is not important.