EEVblog Electronics Community Forum
Electronics => Projects, Designs, and Technical Stuff => Topic started by: ricko_uk on November 21, 2020, 07:07:34 pm
-
Hi,
I need to make a high-power current driver using a high power mosfet in the classic configuration (see attached picture "OpAmp Current Driver.jpg"):
The power mosfet I am planning to use is this: https://www.littelfuse.com/~/media/electronics/datasheets/discrete_mosfets/littelfuse_discrete_mosfets_n-channel_linear_ixt_40n50_datasheet.pdf.pdf (https://www.littelfuse.com/~/media/electronics/datasheets/discrete_mosfets/littelfuse_discrete_mosfets_n-channel_linear_ixt_40n50_datasheet.pdf.pdf)
Because the mosfet is a large one and has large gate capacitance, I am thinking of using the LM7321 which according to the datasheet has "unlimited capacitance drive": https://www.ti.com/lit/ds/symlink/lm7322.pdf?ts=1605984341900&ref_url=https%253A%252F%252Fwww.ti.com%252Fproduct%252FLM7322 (https://www.ti.com/lit/ds/symlink/lm7322.pdf?ts=1605984341900&ref_url=https%253A%252F%252Fwww.ti.com%252Fproduct%252FLM7322)
I have tow separate designs:
a) one working up to 45KHz
b) and another up to 250KHz
Three questions:
1) are there any linear mosfet drivers for large mosfets perhaps designed for this type of application? I could not find any. Maybe I have been searching for the wrong words
2) would it generally work using the op-amp I mentioned above and that mosfet or (because the MOSFET is a large one) do I need to add a discrete BJT voltage follower as shown in the screenshot attached (Driver 2.png)?
3) Specifically to the two versions, would it work for both i.e. up to 250KHz? If not would it work at least for the lower frequency one at 45KHz?
Many thanks :)
-
I am thinking of using the LM7321 which according to the datasheet has "unlimited capacitance drive"
Unlimited capacitive drive only means that the operational amplifier is stable driving capacitive loads but this is achieved at the cost of lower bandwidth limiting performance.
1) are there any linear mosfet drivers for large mosfets perhaps designed for this type of application? I could not find any. Maybe I have been searching for the wrong words
There are operational amplifiers which have increased output current capacity. Multiple operational amplifiers could be used in parallel. The LT1010 150 milliamp power buffer can be used with an existing operational amplifier.
2) would it generally work using the op-amp I mentioned above and that mosfet or (because the MOSFET is a large one) do I need to add a discrete BJT voltage follower as shown in the screenshot attached (Driver 2.png)?
I would likely include a class-ab bipolar buffer anyway like the diamond buffer in your example.
3) Specifically to the two versions, would it work for both i.e. up to 250KHz? If not would it work at least for the lower frequency one at 45KHz?
Well, let's see. You do not give any requirements except frequency so:
Optimistic load capacitance is about 1200 picofarads. The high input capacitance is divided by the transconductance. But note that this assumes that the load voltage does not change increasing the effective reverse transfer capacitance which is why cascode transistors are sometimes added to current sources.
Assuming 50 milliamp drive capability from the LM7321, maximum slew rate into 1200 picofarads is 40 volts/microsecond which is about double the slew rate of the LM7321 so buffering is not absolutely required. LM7321 bandwidth falls below 3 MHz with such a heavy capacitive load so bandwidth could be considerably improved with a high current buffer.
So you may get away with using the LM7321 to directly drive the IXTH40N50L2, but I would plan on using a bipolar diamond buffer anyway, and with 600 milliamp 2N4401/2N4403 transistors or better instead of 200 milliamp 2N3904/2N3906 transistors. The higher current is not needed but the higher power dissipation and current gain are a good idea.
Power dissipation in the LM7321 does not appear to be a problem either.
The dual LM7322 could be used to drive the heavy load in parallel.
-
Thank you David, as usual, for the detailed explanation. Much appreciated! :)
-
Optimistic load capacitance is about 1200 picofarads.
I think that you forget a 0.
The input capacitance on IXTH40N50L2 is more like 12000 pF.
-
Thank you David and Sorin,
I assume you both refer to Figure 11 (graph attached)? If so then it does look like 12nF instead of 1.2nF. Does that change things then or can I still use the op-amp directly? Perhaps with a small resistor to limit the slew rate?
Or would the resistor affect the behaviour at 250KHz?
Thank you
-
If so then it does look like 12nF instead of 1.2nF. Does that change things then or can I still use the op-amp directly?
Yes of course. You will have 10 times smaller bandwidth 50mA / 12nF = 4.17 V/µS
For 250kHZ @ 7Vp-p (around 10A on IXTH40N50L2) you need 11 V/µS
so you can not drive this MOSFET directlly, if you do so you will have a bandwidth of around 100kHZ.
-
Why is IXTQ40N50L2 Ciss so high?
I am comparing it with IGBT IXGH32N60C TO247 used here.
IXGH32N60C Turn Off SOA corner @ 125 C, 5V/ns : 64 Amp 480 V ; Ciss = 2700 pF
I drive them 2V ( off) to 16 V (on) from the SG 3525A into 27 Ohm gate resistors,
I_gate_pk approx 500 mA
These run up to 80 kHz, not tried yet at 100 kHz.
Not an equivalent rating on Data sheet, but
IXTQ40N50L2 Single 25 us pulse SOA corners @ 75 C :80 Amp 220 V down to 32 A 500 V ; Ciss = 10500 pF
-
expect UHF parasitic oscillations in the FET, you need a 100 ohm gate series R, placed as close as possible to the gate lead.
Jon
-
Have you considered using a smaller transistor, probably more of them in parallel? FQA9N90C for instance, 1/4 the capacitance and only 1/2 the power rating.
Do you actually expect to get the full power rating, anyway? You'll need a liquid cooled copper heat spreader to even approach the datasheet value. Very powerful transistors tend to be rather uneconomical.
At high frequencies, Cdg, and its nonlinearity, dominates the control loop response. Minimizing it is critical. You may want to consider LDMOS (RF) transistors, or even SiC (minding that SiC have very small dies, as they're optimized for switching performance; that they handle what power they do, is impressive).
I don't have an intuitive feel for how to compensate this under various load conditions, and what the maximum bandwidth, given that capacitance range, and subsequent compensation, is. I have compensated these circuits before, only on an iterative basis and not with optimization towards ultimate bandwidth.
I believe Win Hill (of Art of Electronics fame) is, or has, worked on a circuit with similar motivation; I forget, it may be in the
X Chapters.
Tim
-
Optimistic load capacitance is about 1200 picofarads.
I think that you forget a 0.
The input capacitance on IXTH40N50L2 is more like 12000 pF.
As I perhaps too briefly explained in my post, the input capacitance is bootstrapped by the transconductance in the common drain configuration, reducing it by roughly 10 times in this case. This is also why bandwidth of a common drain amplifier is inversely proportional to input capacitance but also proportional to transconductance.
Switching applications usually use the common source configuration so have to deal with the full input capacitance.
The reverse transfer capacitance remains and is multiplied by the change in voltage across the load but that was not specified and often in a current source application, the load has a low impedance minimizing this. A cascode can be used to reduce this effect.
-
Hmm, is any load impedance really "low" at 250kHz and, presumably, 10s of amperes?
I don't think application has been mentioned yet, and this may be a good point to ask oneself a nice, deep "why?".
Tim
-
Thank you all :)
The application is for general experiments of various ideas related to ignition, ESD discharge etc. Just building various general bits of equipment that allow me to drive, test and measure various solutions.
That is why often the specs are lose, being at the early stages I don't yet have much reference points.
Thank you again for all input! :)
-
You might find some inspiration in this application note from Linear Technology / Jim Williams:
https://cb.wunderkis.de/wk-pub/lt-app/AN133%20-%20A%20Closed-Loop,%20Wideband,%20100A%20Active%20Load.pdf
-
Interesting application note. Thank you
-
I had forgotten about Jim William's implementation of this idea. I have some comments.
I do not think A3 is strictly needed if careful layout is used to make a single point ground at the current shunt. It does not add to precision but it does break the potential ground loop between the high current and control circuitry.
I did not discuss it in detail but Williams' mentions "Q1’s inherent nonlinear gain characteristic". Q1's transconductance is lower at lower drain currents and this transconductance is what divides Q1's large input capacitance, which means the effective input capacitance varies annoyingly with output current. Since the frequency compensation has to account for the worst case conditions at low drain currents when effective input capacitance is high, performance is necessarily compromised at higher drain currents. This is why I recommend using the low output impedance buffer to drive the effective input capacitance whether it is strictly needed or not; the lower output impedance reduces the effect of the effective input capacitance changing with transconductance which changes with drain current.
Log amplifiers have a similar problem with frequency response needing to be suitable for worst case conditions compromising performance. High performance solutions might use frequency compensation which varies with output level (via varactor diodes or transconductance amplifiers) but adding the buffer is simpler in this case.