The complementary IRF240/9240's are fairly common output-stage devices in a number of Nelson Pass designs (Aleph's, FirstWatt, BurningAmp, etc.). If you don't mind borrowing output stage topology, DiyAudio offers PC boards sets for a number of Nelson Pass designs:
Ready-to-stuff PC boards greatly simplify implementation; just provide the power-supplies, heatsinks and chassis...
Ready-to-stuff PC boards greatly simplify implementation; just provide the power-supplies, heatsinks and chassis...
Mosfets do not require gate current. They only require charge in order to control the internal capacitances. This greatly depends on speed and frequency of conductance changes and changes in internal capacitances, which depends on operating voltages. That being said, the driver stage must be able to supply this 'charge on demand' to the gates of the mosfets when needed. I would say 10mA per pair is the minimum I would go. I typically run ~15mA/pair. 250mA per pair is way overkill. What is the voltage rails and what type of output stage configuration are you running? This makes a difference in determining the most suitable driver transistor. For example, for a source follower Cgs is bootstrapped so the effectiveness of Cgs to the circuit is about 1/10th, roughly. You would only be driving the change in Vgs across Cgs and this depends greatly on mosfet Gm.
check this link
Using HEXFETs in High Fidelity Audio
from the website>>
We will do calculations using the gate charge method, which IRF recommends (AN-944). Looking at the data-sheets, we find the IRFP240 has a total gate charge (Qg) of 70nC and the IRFP9240 has a Qg of 44nC. Don't add these yet! We will find each device's needs individually. The general formula to determine gate current is ...
I = 100Qgf where I is current needed, Qg is the total gate charge in Coulombs, and f is frequency of operation.
The multiplication factor of 100 gives the headroom needed for accurately reproducing a square wave (or high frequency sinewave), since the gate driver needs a lot of current to quickly switch the MOSFET from OFF to ON. Although the requirement for this is minimal (the CD format is incapable of anything even approaching a square wave above a couple of kHz), it has become an expectation that power amps should be able to reproduce perfect square waves at 10kHz as a minimum.
When we plug our figures in for the IRFP240, we get I = 100 * (70E-9 * 50,000) = 350mA per device
For the IRFP9240 we get I = 100 * (40E-9 * 50,000) = 200mA per device
Multiplying each figure by five (because there are five devices of each polarity) gives us 1.75 amps for the upper driver and 1 amp for the lower driver. So a Class-A driver would need bias set to 3.5 amps to get the job done with a reasonable safety margin.
The value for R7 will depend on the linearity of the driver transistors. I had to guess and check with my ammeter to get a good value. This can range anywhere from 100 Ohms up to perhaps 5k. Make sure you check the idle current before calling the design done! These drivers (Q7 and Q8) may need a heatsink. Also note the capacitor in parallel with R7. This should be of a high value (i.e. 100uF or more), and 470uF works fine for my 10 MOSFET stage shown here. It helps with discharging the MOSFET gates by providing a path for the gate current.
These current figures seem quite high, but keep in mind this current will only last a very short time compared to the signal, and virtually no current is needed to keep the devices either in the OFF or ON state. The current to reproduce a sinewave will be a bit lower, since it is a smooth curve, but this much headroom will drastically lower distortion. This is why we cannot practically use a class A driver, unless, of course we use one pair of output devices.
For some comparison, below is a HEXFET setup driven by a class A driver at 13mA bias:
Figure 2 - Spectrum of HEXFET with 13mA Class-A driver
The large notch is at the second harmonic, and the small bump to the right is the fourth harmonic. Barely any third harmonic is seen. This shows 0.25% distortion at the second harmonic at ¾ power, and 250mA amp bias. Not very good for a true hi fi, unless we are making a valve-like amplifier. Even this will not show the same effects as a true valve amp - the nature of the distortion components will almost certainly be different.
Adjusting the bias to 1 amp removes nearly all distortion, yet now we are approaching a heater... I mean class A.
After fixing the problem by adding a class AB driver, distortion was greatly decreased ...
Figure 3 - Spectrum of HEXFET with Class-AB driver
As the picture shows, the second harmonic was reduced considerably, while the fourth harmonic is below the noise floor. This shows 0.04% distortion solely on the second harmonic at ¾ power and still with 250mA bias. This greatly improved the amplifier. At one watt, the distortion is not measurable at all, unlike with the class A driver. Reducing the gate resistors to 4.7 Ohms to get more current through does nothing noticeable, so the use of 10 Ohm resistors is fine. There is no evidence of “notchâ€� distortion or any other nasty odd harmonic, only a “niceâ€� second harmonic added in. Also note that this amp was built on a breadboard. A compact and nicely wired PCB should decrease distortion even further.
Using HEXFETs in High Fidelity Audio
from the website>>
We will do calculations using the gate charge method, which IRF recommends (AN-944). Looking at the data-sheets, we find the IRFP240 has a total gate charge (Qg) of 70nC and the IRFP9240 has a Qg of 44nC. Don't add these yet! We will find each device's needs individually. The general formula to determine gate current is ...
I = 100Qgf where I is current needed, Qg is the total gate charge in Coulombs, and f is frequency of operation.
The multiplication factor of 100 gives the headroom needed for accurately reproducing a square wave (or high frequency sinewave), since the gate driver needs a lot of current to quickly switch the MOSFET from OFF to ON. Although the requirement for this is minimal (the CD format is incapable of anything even approaching a square wave above a couple of kHz), it has become an expectation that power amps should be able to reproduce perfect square waves at 10kHz as a minimum.
When we plug our figures in for the IRFP240, we get I = 100 * (70E-9 * 50,000) = 350mA per device
For the IRFP9240 we get I = 100 * (40E-9 * 50,000) = 200mA per device
Multiplying each figure by five (because there are five devices of each polarity) gives us 1.75 amps for the upper driver and 1 amp for the lower driver. So a Class-A driver would need bias set to 3.5 amps to get the job done with a reasonable safety margin.
The value for R7 will depend on the linearity of the driver transistors. I had to guess and check with my ammeter to get a good value. This can range anywhere from 100 Ohms up to perhaps 5k. Make sure you check the idle current before calling the design done! These drivers (Q7 and Q8) may need a heatsink. Also note the capacitor in parallel with R7. This should be of a high value (i.e. 100uF or more), and 470uF works fine for my 10 MOSFET stage shown here. It helps with discharging the MOSFET gates by providing a path for the gate current.
These current figures seem quite high, but keep in mind this current will only last a very short time compared to the signal, and virtually no current is needed to keep the devices either in the OFF or ON state. The current to reproduce a sinewave will be a bit lower, since it is a smooth curve, but this much headroom will drastically lower distortion. This is why we cannot practically use a class A driver, unless, of course we use one pair of output devices.
For some comparison, below is a HEXFET setup driven by a class A driver at 13mA bias:
Figure 2 - Spectrum of HEXFET with 13mA Class-A driver
The large notch is at the second harmonic, and the small bump to the right is the fourth harmonic. Barely any third harmonic is seen. This shows 0.25% distortion at the second harmonic at ¾ power, and 250mA amp bias. Not very good for a true hi fi, unless we are making a valve-like amplifier. Even this will not show the same effects as a true valve amp - the nature of the distortion components will almost certainly be different.
Adjusting the bias to 1 amp removes nearly all distortion, yet now we are approaching a heater... I mean class A.
After fixing the problem by adding a class AB driver, distortion was greatly decreased ...
Figure 3 - Spectrum of HEXFET with Class-AB driver
As the picture shows, the second harmonic was reduced considerably, while the fourth harmonic is below the noise floor. This shows 0.04% distortion solely on the second harmonic at ¾ power and still with 250mA bias. This greatly improved the amplifier. At one watt, the distortion is not measurable at all, unlike with the class A driver. Reducing the gate resistors to 4.7 Ohms to get more current through does nothing noticeable, so the use of 10 Ohm resistors is fine. There is no evidence of “notchâ€� distortion or any other nasty odd harmonic, only a “niceâ€� second harmonic added in. Also note that this amp was built on a breadboard. A compact and nicely wired PCB should decrease distortion even further.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.