Post Go back to editing

CM and DM separator

Good morning,

I am trying to simulate on ltspice a circuit to separate common mode and differential EMI noise. The circuit is an active one that uses Analog op amps.

We have and input buffer (A1 − A3, Analog Devices AD8000), a difference amplifier (A4 − A6, Analog Devices AD8099),  and the buffer A7 (Analog Devices ADA4817). The CM and DM are modeled by an equivalent three phase source that comprise both CM and DM noise. 

The circuit works fine on ltspice if I use ideal op amps. But when I use the spice models of the op amps that I downloaded from Analog's website I have a problem with the AD8099. Because it has a high oscillation on the output if I connect the recommended compensation network on the datasheet , I think the problem is in the compensation network. I dont have the oscillation if i use higher value of capacitances connected  the compensation pin.
I attach the ltspice model.

 CM_DM_separator_thirdpartymodel.asc

Parents
  • Hi,

    I think the problem is probably the finite output impedance of the amplifiers A1-A3. If you look at Figure 30 of the AD8000’s datasheet (for gain=2, which seems to match your case), you’ll see that the output impedance can be quite high at high frequencies. This means that at high frequencies the inverting input of the next amplifier, the AD8099, sees more than 499 Ohms. Accordingly, the non-inverting gain (the noise gain) can be lower than the specified minimum of +2 (the datasheet says ‘External compensation allows gains from +2 to +10 with minimal trade-off in bandwidth’). Therefore, the amplifier can be unstable. You can try to add another impedance between the inverting input of the AD8099 and ground. If it is a capacitor-resistor series combination, then you can choose the equivalent impedance to be high at the signal frequency not to affect the gain, but lower at higher frequencies to get gain higher than 2. Alternatively, even a single resistor can also be effective: this will increase the non-inverting gain (you can lower the amplitude at the non-inverting input accordingly by reducing R1/3), but will leave the inverting gain unchanged. Keep in mind that the outputs of the A1-A3 amplifiers are also loaded by the 100 Ohm and 33 Ohm resistors.

    According to the simulation you use rather low frequency signals, so I just wonder if you really need such very fast amplifiers, they are very sensitive to stability margins.

Reply
  • Hi,

    I think the problem is probably the finite output impedance of the amplifiers A1-A3. If you look at Figure 30 of the AD8000’s datasheet (for gain=2, which seems to match your case), you’ll see that the output impedance can be quite high at high frequencies. This means that at high frequencies the inverting input of the next amplifier, the AD8099, sees more than 499 Ohms. Accordingly, the non-inverting gain (the noise gain) can be lower than the specified minimum of +2 (the datasheet says ‘External compensation allows gains from +2 to +10 with minimal trade-off in bandwidth’). Therefore, the amplifier can be unstable. You can try to add another impedance between the inverting input of the AD8099 and ground. If it is a capacitor-resistor series combination, then you can choose the equivalent impedance to be high at the signal frequency not to affect the gain, but lower at higher frequencies to get gain higher than 2. Alternatively, even a single resistor can also be effective: this will increase the non-inverting gain (you can lower the amplitude at the non-inverting input accordingly by reducing R1/3), but will leave the inverting gain unchanged. Keep in mind that the outputs of the A1-A3 amplifiers are also loaded by the 100 Ohm and 33 Ohm resistors.

    According to the simulation you use rather low frequency signals, so I just wonder if you really need such very fast amplifiers, they are very sensitive to stability margins.

Children
No Data