Share: What is the use of the 100 Ω resistor before the MOSFET gate?

qwqwqw2088

Share: What is the use of the 100 Ω resistor before the MOSFET gate? [Copy link]

The story begins when
young application engineer Neubean wants to prove through an experiment whether it is really necessary to put a 100 Ω resistor in front of the MOSFET gate in order to obtain stability. Gureux, an application engineer with 30 years of experience, supervises his experiment and provides expert guidance throughout the process.

Introduction to High-Side Current Sensing

The circuit in Figure 1 shows a typical example of high-side current sensing.

Figure 1 High-side current detection

Negative feedback attempts to force a voltage, VSENSE, across the gain resistor, RGAIN. The current through RGAIN flows through a P-channel MOSFET (PMOS) into resistor ROUT, which forms an output voltage referenced to ground. The overall gain is:

The optional capacitor COUT across resistor ROUT serves to filter the output voltage. Even if the drain current of the PMOS quickly follows the sensed current, the output voltage will exhibit a single-pole exponential trajectory.

The resistor RGATE in the schematic separates the amplifier from the PMOS gate. What is its value? An experienced Gureux might say, "100 Ω, of course!"

Try different resistance values

We found our friend Neubean, also a student of Gureux, thinking hard about this gate resistor. Neubean was thinking that if there was enough capacitance between the gate and source, or if the gate resistor was large enough, it should cause stability problems. Once it was determined that RGATE and CGATE would adversely affect each other, it could reveal why 100 Ω or any gate resistor value was a reasonable answer.

Figure 2 shows an example of an LTspice simulation used to highlight the circuit’s behavior. Neubean performed simulations to demonstrate stability issues, which he believed would occur as RGATE increased. After all, the poles from RGATE and CGATE should eat into the phase margin associated with the open loop. However, to Neubean’s surprise, all RGATE values exhibited no issues in the time domain response.

Figure 2 High-side current detection simulation

It turns out that the circuit is not simple

While studying the frequency response, Neubean realized that he needed to clarify what the open-loop response was. If combined with unity negative feedback, the forward path of the loop starts at the difference and ends at the resulting negative input. Neubean then simulated VS/(VP – VS) or VS/VE and plotted the results. Figure 3 shows a frequency domain plot of this open-loop response. In the Bode plot of Figure 3, the DC gain is small and no phase margin issues are found at crossover. In fact, the plot as a whole looks very strange because the crossover frequency is less than 0.001 Hz.

Figure 3 Frequency response from error voltage to source voltage

The result of breaking down the circuit into a control system is shown in Figure 4. Like almost all voltage feedback op amps, the LTC2063 has high DC gain and a single-pole response. The op amp amplifies the error signal and drives the PMOS gate, passing the signal through the RGATE–CGATE filter. CGATE and the PMOS source are connected together to the –IN input of the op amp. RGAIN is connected from this node to a low impedance source. Even in Figure 4, it might appear that the RGATE–CGATE filter should cause stability issues, especially if RGATE is much larger than RGATE. After all, the CGATE voltage, which directly affects the system RGATE current, lags behind changes in the op amp output.

Figure 4 Functional block diagram of high-end detection circuit

Neubean offers an explanation for why RGATE and CGATE do not cause instability: “The gate-source is a fixed voltage, so the RGATE – CGATE circuit is irrelevant here. You just adjust the gate and source as follows. It’s a source follower.”

Gureux, his more experienced colleague, said, “Actually, that’s not the case. This is only the case when the PMOS operates normally as a gain block in the circuit.”

Inspired by this, Neubean thought about the math - what if we could directly simulate the response of the PMOS source to the PMOS gate? In other words, what is V(VS)/V(VG)? Neubean quickly ran to the whiteboard and wrote down the following equation:

in,

The op amp gain is A and the op amp pole is ωA.

Neubean immediately spotted the important term gm. What is gm? For a MOSFET, looking at the circuit in Figure 1, Neubean's mind suddenly lit up. When the current through RSENSE is zero, the current through the PMOS should be zero. When the current is zero, gm is zero because the PMOS is effectively off, unused, unbiased, and has no gain. When gm = 0, VS/VE is 0, the frequency is 0 Hz, and VS/VG is 0, the frequency is 0 Hz, so there is no gain at all, and the graph in Figure 3 is probably valid.

qwqwqw2088

Trying to find instability issues with the LTC2063

Armed with this insight, Neubean quickly tried some simulations with non-zero ISENSE.

Figure 5 is a gain/phase plot of the response from VE to VS, which spans above 0dB to below 0dB and looks much more normal. Figure 5 should show a lot of PM at about 2kHz with 100Ω, less PM at 100kΩ, and even less at 1MΩ, but not unstable.

Figure 5 Frequency response from error voltage to source voltage under non-zero sense current conditions

Neubean went to the lab and used the LTC2063 high-side sense circuit to get a sense current. He plugged in a high RGATE value, first 100 kΩ and then 1 MΩ, hoping to see unstable behavior, or at least some kind of ringing. Unfortunately, he saw none. He tried increasing the drain current in the MOSFET, first increasing ISENSE and then using a smaller RGAIN resistor value. Still no results to make the circuit unstable.

He went back to simulation and tried to measure the phase margin with non-zero ISENSE. Even in simulation, it was difficult or impossible to find instability or low phase margin issues.

Neubean went to Gureux and asked him why he hadn’t been able to make the circuit unstable. Gureux suggested that he look at the numbers. Neubean was used to Gureux’s inscrutable comments, so he looked at the actual pole formed by RGATE and the total gate capacitance. With 100 Ω and 250 pF, the pole was 6.4 MHz; with 100 kΩ, the pole was 6.4 kHz; and with 1 MΩ, the pole was 640 Hz. The LTC2063 gain-bandwidth product (GBP) is 20 kHz. When the LTC2063 has gain, the closed-loop crossover frequency can easily slide below any effect of the RGATE– CGATE pole.

Yes, there may be instability issues

Realizing that the op amp dynamic range needs to extend beyond the RGATE – CGATE pole, Neubean selects an op amp with a higher gain-bandwidth product. The LTC6255 5 V op amp can be added directly to the circuit and has a relatively high gain-bandwidth product of 6.5 MHz.

Neubean eagerly ran a simulation using the current, LTC6255, 100 kΩ gate resistor, and 300 mA sense current.

Neubean then added RGATE to the simulation. When RGATE is large enough, an extra pole can make the circuit unstable.

Figure 6 and Figure 7 show the simulation results for high RGATE values. When the sense current is kept constant at 300 mA, the simulation becomes unstable.

Figure 6: Time domain diagram with ringing

Figure 7. Normal Bode plot after increasing current (VE to VS), poor phase margin performance. Experimental results

Experimental Results

To understand if the current would behave strangely when sensing non-zero current, Neubean tested the LTC6255 with varying load currents and three different RGATE values. ISENSE transitioned from a base of 60 mA to a higher value of 220 mA as the momentary switch cut in more parallel load resistance. There was no zero ISENSE measurement because we have already shown that the MOSFET gain is too low in that case.

In fact, Figure 8 ultimately shows that stability does suffer when using 100 kΩ and 1 MΩ resistors. Since the output voltage is heavily filtered, the gate voltage becomes a ringing detector. Ringing indicates poor or negative phase margin, and the ringing frequency shows the crossover frequency.

Figure 8 RGATE = 100 Ω, current transient from low to high

Figure 9 RGATE = 100 Ω, current transient from high to low

Figure 10 RGATE = 100 kΩ, current transient from low to high

Figure 11 RGATE = 100 kΩ, current transient from high to low

Figure 12 RGATE = 1 MΩ, current transient from low to high

Figure 13 RGATE = 1 MΩ, current transient from high to low

Brainstorming Time

Neubean realized that although he had seen many high-side integrated current sensing circuits, unfortunately, engineers had no power to determine the gate resistor because it was integrated into the device. Specific examples are the AD8212, LTC6101, LTC6102, and LTC6104 high voltage, high-side current sensing devices. In fact, the AD8212 uses a PNP transistor instead of a PMOS FET. He told Gureux, "It really doesn't matter because modern devices have solved this problem."

As if he had been waiting for this moment, the professor almost interrupted Neubean and said, “Let’s assume that you want to combine very low supply current with zero-drift input offset, such as a battery-powered instrument installed in a remote location. You might use the LTC2063 or LTC2066 as the main amplifier. Or you want to measure low-level currents through a 470 Ω shunt resistor as accurately as possible and with as little noise as possible; in that case, you might need to use the ADA4528, which supports rail-to-rail inputs. In these cases, you need to deal with MOSFET drive circuits.”

so……

Apparently, it is possible to make the high-side current sensing circuit unstable as long as the gate resistance is too large. Neubean talked about his findings to the helpful teacher Gureux. Gureux said that in fact, RGATE does have the potential to make the circuit unstable, but the reason why this behavior was not discovered at first was because the question was not formulated correctly. There needs to be gain, and in the current circuit, the signal being measured needs to be non-zero.

Gureux responded, “Sure, ringing occurs when the pole erodes the phase margin at the crossover. But your addition of a 1-MΩ gate resistor is ridiculous, and even 100 kΩ is crazy. Remember, it is good practice to limit the output current of the op amp to prevent it from diverting gate capacitance from one supply rail to the other.”

Neubean agreed, “So, what resistor value do I need?” Gureux confidently replied, “100 Ω.”