Caught out by the Coss - Effect of the output capacitance of a mosfet TX in PI.

[Ferric Toes]

I have been trying alternative mosfets as some newer devices appear to have a better spec than the IRF740 which I have used for many years. The latest device tried is SiHA22N60E which at first glance appeared to be very suitable. i.e. Vds 600V, Rds 0.23 ohms and Coss (output capacitance) of 90pF. By comparison the IRF740 has Vds of 400V , Rds of 0.55 ohms, and Coss of 330pF at 25V Vds and Vgs 0V.

Another transistor that works well is the P5NB40 which also has a Vds of 400V and a much lower Coss of 72pF at a Vds of 25V. The Rds is a bit higher at 1.5 ohms though. This device is good for short sample delay times as the lower capacitance in parallel with the coil/cable results in a higher ringing frequency and a higher value resistor for damping. However, for both devices the Coss is voltage dependant and rises for lower values of Vds. In the case of the IRF740 the figure rises to around 600pF and the P5NB40 to 110pF at 10V Vds. 10 to 15V Vds is that which is likely to be encountered on a PI metal detector.

In the case of the SiHA22N60E the value of 90pF is for Vds of 100V and which reduces to 73pF at 480V. Below 100V the Coss rises rapidly such that at 10V it is about 3000pF: an unacceptable figure for a PI transmitter. Ringing, with a 300uh coil and 1.5m of RG58 coax (100pF/m), is severe due to this high value of added capacitance. Damping of 150 ohms is required rather than the 1K for the P5N40. Substituting a P5N40 mosfet gives a smooth recovery as seen at a two stage x1000 gain preamp o/p.

P5NB40 with 1K damping. IRF740 with 1K damping. 22N60E with 1K damping. 22N60E with 150 ohms damping.

Looking at the Coss capacitance graphs, it does appear that running the mosfet at a higher Vds can be beneficial in minimising this unwanted capacitance. For the IRF740, 20V appears to be the value below which the Coss starts to rise fast. For the P5NB40, 15V would be a good voltage as the Coss is still below 100pF, but rises fast below 10V. These voltages may give rise to excessive current in low resistance coils unless low repetition rates are used. The next task is to raise the coil resistance and/or insert series resistance such that the mosfet can run at the higher voltage but maintain a reasonable pulse current and repetition rate.

Eric.

[KingJL]

Eric,
Try using an ultrafast diode in series with the MOSFET (be sure to NOT drain off the charge that accumulates at the junction of the MOSFET drain and the diode cathode). It raises the effective Vds felt during MOSFET OFF period. It does decrease your max coil current for a given coil/voltage, though that can be compensated for by raising voltage. I typically use the RF101L4S and typically limit coil current to 1A. If you use higher currents you could use higher current rated ultrafast diodes of the RF series. If you try to probe the Vds during the off period, be careful you do not drain off the charge. I typically see Vds voltages during the switched OFF times build to around 1/2 of the flyback voltage using the diode.
Just a thought, J. L. King

[green]

Eric,
Try using an ultrafast diode in series with the MOSFET (be sure to NOT drain off the charge that accumulates at the junction of the MOSFET drain and the diode cathode). It raises the effective Vds felt during MOSFET OFF period. It does decrease your max coil current for a given coil/voltage, though that can be compensated for by raising voltage. I typically use the RF101L4S and typically limit coil current to 1A. If you use higher currents you could use higher current rated ultrafast diodes of the RF series. If you try to probe the Vds during the off period, be careful you do not drain off the charge. I typically see Vds voltages during the switched OFF times build to around 1/2 of the flyback voltage using the diode.
Just a thought, J. L. King

Same, MUR460 diode works for me. Will almost eliminate capacitance effect for IRF740.

[Ferric Toes]

Thank you KingJL and green,
The only diodes I have that I thought might be suitable is BYV96E. This is an avalanche fast recovery rectifier diode with 1000V reverse voltage and an If(av) of 1.5A. On the mosfets that ring it initially seems to work, but then the ringing slowly appears and I end up with a situation that is no better than without the diode. My pulse current is 2A but the average is 0.4A. I have some MUR460 that will arrive today so perhaps these will work. I presume I can measure the voltage on the junction with the diode cathode and mosfet drain with a x10 probe? At the moment I see nothing except the back emf spike without any stored voltage. The two diodes that I have check out good on a test meter.

Eric.

[green]

Thank you KingJL and green,
The only diodes I have that I thought might be suitable is BYV96E. This is an avalanche fast recovery rectifier diode with 1000V reverse voltage and an If(av) of 1.5A. On the mosfets that ring it initially seems to work, but then the ringing slowly appears and I end up with a situation that is no better than without the diode. My pulse current is 2A but the average is 0.4A. I have some MUR460 that will arrive today so perhaps these will work. I presume I can measure the voltage on the junction with the diode cathode and mosfet drain with a x10 probe? At the moment I see nothing except the back emf spike without any stored voltage. The two diodes that I have check out good on a test meter.

Eric.

MPP with a MUR460 diode steps tp 300V and then decays to 100V thru the X10 probe.

[KingJL]

... The only diodes I have that I thought might be suitable is BYV96E. This is an avalanche fast recovery rectifier diode with 1000V reverse voltage and an If(av) of 1.5A. On the mosfets that ring it initially seems to work, but then the ringing slowly appears and I end up with a situation that is no better than without the diode...

This method is highly sensitive to the leakage currents of both the MOSFET and the diode as it is relying on small capacitive charges (mainly Coss) to create an effective Vds during switch off. I would have thought the BYV96E should work for the diode as I have used BYV28-200 Sinterglass diodes (reverse current 1-100uA depending on operating temperature) for my through holes. The main contributor is the MOSFET. I stopped using the IRF740 12 years ago mainly because I did not care for the Coss characteristics. I have since been using lower Coss MOSFETs (currently using the STD5NM60 (for SMD designs) in conjunction with the RF1014LS). The IRF740 has a reverse current of 25-250uA depending on operating temperature. I think that is why you see it initially work, but as the part warms up, the leakage current quickly negates/drains the capacitance storage effect. The STD5NM60 reverse current ranges from 1-10uA depending on temperature.

... I have some MUR460 that will arrive today so perhaps these will work...

I have not used the MUR460, but it's reverse current can range between 10 and 250 uA depending on operating temperature (may be questionable, especially at operating temperature).

I presume I can measure the voltage on the junction with the diode cathode and mosfet drain with a x10 probe?

The only time I successfully observed the drain/diode junction was with a 100x probe and an antique Hewlit Packard Oscilloscope with the PI TX at an operating PRT of 1000 usec, PW of 160 usec, and peak current 2 A. The voltage at the junction after flyback was about 190-200V and decreased/drained to about 75V at the end of the prt. I believe that the MOSFET was an FPQF7N60. Since my antique HP bit the dust and I am relegated to the newfangled solid state low impedance scopes, I trust in observation of the preamp output as I change the damp R to get a feel for what is going on.

Kind regards, J. L. King

[Ferric Toes]

The MUR460's arrived and I tried one out this evening; first with the 22N60B which has the highest capacitance at low voltages. Fantastic! it worked, and remained working. I tried a considerable assortment of mosfets of different voltages and Rds-on resistance. All worked fine with the diode, IRF740 too. A FQA24N60C seems a good candidate and another one tried was 17N80. I have made a jig that I can just plug in mosfets without turning off the power supply each time. Got quite a shock when holding the tab of the 17N80 and accidentally touching a ground point.

I've never been convinced that a series diode has much benefit, but I certainly can see it now.

Eric.

[KingJL]

The MUR460's arrived and I tried one out this evening; first with the 22N60B which has the highest capacitance at low voltages. Fantastic! it worked, and remained working. I tried a considerable assortment of mosfets of different voltages and Rds-on resistance. All worked fine with the diode, IRF740 too. A FQA24N60C seems a good candidate and another one tried was 17N80. I have made a jig that I can just plug in mosfets without turning off the power supply each time. Got quite a shock when holding the tab of the 17N80 and accidentally touching a ground point.

I've never been convinced that a series diode has much benefit, but I certainly can see it now.

Eric.

Depends on the diode... I've seen people try 1004's...of course to no avail. Glad it worked. Nice touch with the jig!

[Carl-NC]

Got quite a shock when holding the tab of the 17N80 and accidentally touching a ground point.

Haha I did the same thing today.

[Ferric Toes]

This is a new investigation for me, and fascinating it is. I managed to get some sensible pictures of what is going on at the junction of the mosfet drain and the series diode. At least they make sense of the voltage that is stored in the output capacitance of the mosfet. The top trace is the voltage, where one large division = 100V. (100M resistor to a x1 probe) It rises to about 450V and then drops to about 380V at the start of the next pulse. The transistor used is a 24N60C which has 650Vds and a Coss of 100pF for Vds above 100V. Rds is typically 0.14 ohms. The coil current is shown in the lower waveform and measured across a 0.1 ohm resistor in the coil return. peak I is 2.4A. Without the diode this transistor would not be useable due to the high Coss at low voltages. The supply here is 10V where Coss is about 3000pF, which you certainly don't want across the coil. I am surprised that there appear to be no high current, low Rds, mosfets on the market that have a built in series diode. The High Coss must be a disadvantage in other applications? I am wondering if the slow rise to maximum Coss voltage is due to capacitances in the probe circuit. 100M and 1pF gives a TC of 100uS. I have a high voltage x100 probe coming next week which should give a more accurate picture.



Eric.

[KingJL]

... I am surprised that there appear to be no high current, low Rds, mosfets on the market that have a built in series diode. The High Coss must be a disadvantage in other applications? ...

I am too! When I had first affirmed that the diode worked to reduce the effects of Coss, I started searching the internet for "diode in series with mosfet drain" and found very little information. The only results that even came close were discussing the slow recovery of the MOSFET body diodes. To address this they would put a diode in series with the MOSFET to negate the body diode(s) and then use a fast recovery diode from the top of the series diode to the MOSFET source to effect a fast recovery MOSFET. One such discussion is here. I had at one time found a short paper on this, but have since lost it. But nowhere did I find a discussion of increasing effective Vds, thereby reducing Coss.

Another factor at play here could be the fact that the reverse voltage seen by the diode pn junction also decreases the diode capacitance. Could these series capacitances further decrease the effective capacitance of the total circuit?

But like you, I feel the high Coss must be a disadvantage to other applications.

[dbanner]

I noticed that the power supply boards of these modern flat screen tv's are always with such diodes. Ultra fast switching rectifiers and no surprise, power mosfets as their companions in the circuit. Probably driving the backlight HV converters on the older designs.

[KingJL]

Eric, I first stumbled upon the "diode in series with MOSFET drain" back in 2008 when I discovered the SD2000 schematics traced by Cortana Zed which were posted on Geotech. I did not fully comprehend the SD2000 arrangement until much later. But in late 2008 - early 2009, I acquired some BYV28-200's (the diode used in the SD2000) and started experimenting with my Hammerhead boards. I noticed that I was able to use significantly higher values of damp resistance, resulting in the ability to sample earlier. At the time, I attributed this to the diode capacitance (when it is reverse biased by the stored charge at the junction of the MOSFET drain and the diode cathode) being in series with the output capacitance.

Since 2009, I have periodically revisited the SD2000 schematic to further understand this phenomenon. I found the SD2000 use of the diode to be more sophisticated than was apparent at first glance. The SD2000 not only captured the charge from the flyback, but regulated the value to provide a constant high effective Vds during MOSFET off period. The regulator maintained the voltage at the junction of the MOSFET drain and the diode cathode to a maximum of about 185V during the entire off period (coupled with the source being at -12V, that resulted in an effective Vds maximum of ~195V). Just using the simple diode in series without regulation results in the stored Vds falling off during the off period due to the stored charge bleeding off. I believe the regulation circuit of the SD2000 also protects the MOSFET from the effects excessive gate voltage Vdg and Vgs buildup created by the Cdg - Cgs capacitive divider network within the MOSFET. But that is an assumption.

After realizing this complexity of the SD2000 MOSFET drain diode implementation, I started diving deeper into the MOSFET Coss and it's makeup and realized that the majority of the diode advantage was the maximizing of a given MOSFET's Coss characteristics. I still feel that the diode capacitance in series with the Coss is a factor in the overall improvement.

Since I first started experimenting with the "diode in series with MOSFET drain" in 2009, I and others have commented on the resulting Rd improvements that can be realized. For the most part our observations were discounted as imaginary gains or artifacts of other phenomenon.

Eric, I applaud you for carrying out these detailed experiments and posting the results. You are the first person with metal detector design gravitas to acknowledge, explore, comment, and give credence on this phenomenon. Minelab understood it in 1995!

[Ferric Toes]

Eric, I first stumbled upon the "diode in series with MOSFET drain" back in 2008 when I discovered the SD2000 schematics traced by Cortana Zed which were posted on Geotech. I did not fully comprehend the SD2000 arrangement until much later. But in late 2008 - early 2009, I acquired some BYV28-200's (the diode used in the SD2000) and started experimenting with my Hammerhead boards. I noticed that I was able to use significantly higher values of damp resistance, resulting in the ability to sample earlier. At the time, I attributed this to the diode capacitance (when it is reverse biased by the stored charge at the junction of the MOSFET drain and the diode cathode) being in series with the output capacitance.

Since 2009, I have periodically revisited the SD2000 schematic to further understand this phenomenon. I found the SD2000 use of the diode to be more sophisticated than was apparent at first glance. The SD2000 not only captured the charge from the flyback, but regulated the value to provide a constant high effective Vds during MOSFET off period. The regulator maintained the voltage at the junction of the MOSFET drain and the diode cathode to a maximum of about 185V during the entire off period (coupled with the source being at -12V, that resulted in an effective Vds maximum of ~195V). Just using the simple diode in series without regulation results in the stored Vds falling off during the off period due to the stored charge bleeding off. I believe the regulation circuit of the SD2000 also protects the MOSFET from the effects excessive gate voltage Vdg and Vgs buildup created by the Cdg - Cgs capacitive divider network within the MOSFET. But that is an assumption.

After realizing this complexity of the SD2000 MOSFET drain diode implementation, I started diving deeper into the MOSFET Coss and it's makeup and realized that the majority of the diode advantage was the maximizing of a given MOSFET's Coss characteristics. I still feel that the diode capacitance in series with the Coss is a factor in the overall improvement.

Since I first started experimenting with the "diode in series with MOSFET drain" in 2009, I and others have commented on the resulting Rd improvements that can be realized. For the most part our observations were discounted as imaginary gains or artifacts of other phenomenon.

Eric, I applaud you for carrying out these detailed experiments and posting the results. You are the first person with metal detector design gravitas to acknowledge, explore, comment, and give credence on this phenomenon. Minelab understood it in 1995!

Hi James, Some years ago I did try a series diode when working on an industrial detector to find broken off needle tips in fabric. The mosfet I was using then was a IRF110 at low pulse currents (0.1A) but very high rep rates.. The IRF110 has a low Coss anyway and a series IN4148 did not make a noticeable difference. The limitations were tackled elsewhere i.e. low capacitance coil design, virtually no cable between electronics and coil, and a much faster front end amplifier. I managed to get down to 1uS delay but can't remember whether this was with, or without, the diode. Since then I have not bothered, until now.

The reverse voltage capacitance of the MUR460 is about 5pF at 100V and I suspect it gets less for higher voltages, the same as Coss does. As the example above, I expect my limitations are now with other aspects of the circuit such as coil, cable and preamp. I am still using a 5534 but two stages with 20x followed by 50x. I found some guitar coax cable that has 60pF/m capacitance and I currently have 1.5m of that with a 300uH coil wound with 1/0.16 PTFE insulated wire. I have to measure the resonant frequency of this arrangement too, which is relevant.

Eric.

[Ferric Toes]

Being in an experimentive mood, I thought I would see what effect adding more capacitance to the Coss would have. I added 1000pF in the form of a 350V polystyrene capacitor between source and drain of a 24N60 mosfet, and then a humble IRF740. The result was surprising. Below are scope shots for the IRF740. First, without the 1000pF and the second, with. With the capacitor there is no measurable drop between pulses of the 400V, and we are getting the full 400V instead of 275V as in the first shot. The 24N60 had similar results but a slightly lower voltage. The receiver waveform showed no change in either situation. Counter-intuitive but it appears to work.



Eric.