What is the purpose of the series diode?

[Qiaozhi]

Very educational, thanks!

The (theoretical) conclusion is that the diode would be useful/noticeable when the value of Coss is near the coil's capacitance or larger, which is not the case for quick and dirty home-made coils above 300uH with capacitance in the order of 200-300pF.

That seems to be the conclusion.
I haven't tested this yet, but I suspect there's two possible scenarios:

1. A fast coil is being used, and the MOSFET's Coss is comparable to the coil's parasitic capacitance. In this case a series diode can be used to block the effect of Coss, and hence achieve a faster coil current decay.
2. The coil's parasitic capacitance is significantly higher than Coss, and including the series diode causes a longer decay time. The reason here is that there's a higher flyback voltage without the diode, and this naturally produces a shorter decay time. So the effect of the higher flyback voltage dominates over that produced by blocking Coss.

Of course, there's little point playing around with the diode anyway, if your MOSFET is going into avalanche mode. Then it's more important to either replace it with a higher Vds MOSFET, or reduce the TX-on time to prevent it avalanching.

Regarding third party libraries, this is the one I used is mcemos.lib (attached). Copy the .lib file to the /sub directory and the .asy file to the /sym directory. You'll get a new symbol called "MCEMOS", usit by changin the name to the mosfet of interest in the library.

Installed and working. Thanks!

[Teleno]

Installed and working. Thanks!

If you look at the source, I have replaced all the lines:

.SUBCKT IRFXXXXX 10 20 40 40

with

.SUBCKT IRFXXXXX 10 20 40

so that a three-terminal MOSFET symbol can be used.

I haven't corrected the floating LS error that shows up all over the library yet.

[Qiaozhi]

If you look at the source, I have replaced all the lines:

.SUBCKT IRFXXXXX 10 20 40 40

with

.SUBCKT IRFXXXXX 10 20 40

so that a three-terminal MOSFET symbol can be used.

You must have done that with a script.

I haven't corrected the floating LS error that shows up all over the library yet.

There should actually be 3 small inductors in the model, one for each of the bond wires. The effect of these inductors will only be significant at high frequencies, so they probably meant to take them out, but neglected to remove LS.

[green]

That seems to be the conclusion.
I haven't tested this yet, but I suspect there's two possible scenarios:

1. A fast coil is being used, and the MOSFET's Coss is comparable to the coil's parasitic capacitance. In this case a series diode can be used to block the effect of Coss, and hence achieve a faster coil current decay.
2. The coil's parasitic capacitance is significantly higher than Coss, and including the series diode causes a longer decay time. The reason here is that there's a higher flyback voltage without the diode, and this naturally produces a shorter decay time. So the effect of the higher flyback voltage dominates over that produced by blocking Coss.

Of course, there's little point playing around with the diode anyway, if your MOSFET is going into avalanche mode. Then it's more important to either replace it with a higher Vds MOSFET, or reduce the TX-on time to prevent it avalanching.



Installed and working. Thanks!

Of course, there's little point playing around with the diode anyway, if your MOSFET is going into avalanche mode. Then it's more important to either replace it with a higher Vds MOSFET, or reduce the TX-on time to prevent it avalanching.


Spice simulation. Both 1 amp peak current. 1MHz coil. Coil volts has to decay to 10mv before amplifier can start to come out of saturation. 2.5usec with diode, 6.5usec no diode.

[Teleno]

Of course, there's little point playing around with the diode anyway, if your MOSFET is going into avalanche mode. Then it's more important to either replace it with a higher Vds MOSFET, or reduce the TX-on time to prevent it avalanching.

At lower Vds the Coss is larger and the effect of the diode more noticeable.

Example with avalanching MTP30N06VL (60V mosfet), decay is over 2us faster with diode.

[homefire]

So what are the Delay Gains with the Prescribed Q1 IRF740 ???

[Teleno]

So what are the Delay Gains with the Prescribed Q1 IRF740 ???

If simulations are right you'd gain 3-4us.

[Qiaozhi]

As the original question was about the diode in the MPP, I have simulated that part up to the preamp output. See attached simulation file.

As you can see, there is a definite improvement in the ability to early sample when the diode is present.
For this simulation you need the following parameters:
MOSFET = IRF740
Coss effective = 211pF
With diode: Rd = 1200, Rs = 3R3, preamp comes out of saturation at 6.7us
Without diode: Rd = 480, Rs = 4R15, preamp comes out of saturation at 8.8us

Then, replace the MOSFET for an IRFIBF30G:
Coss effective = 103pF
With diode: Rd = 1200, Rs = 3R3, preamp comes out of saturation at 6.5us
Without diode: Rd = 1200, Rs = 4R15, preamp comes out of saturation at 6.5us

In this second example, there is no difference with or without the diode.
Note that you need the MOSFET libraries posted by Teleno in an earlier thread.

In the next few days I will test example 1 using a real MPP board.

[bbsailor]

That seems to be the conclusion.
I haven't tested this yet, but I suspect there's two possible scenarios:

1. A fast coil is being used, and the MOSFET's Coss is comparable to the coil's parasitic capacitance. In this case a series diode can be used to block the effect of Coss, and hence achieve a faster coil current decay.
2. The coil's parasitic capacitance is significantly higher than Coss, and including the series diode causes a longer decay time. The reason here is that there's a higher flyback voltage without the diode, and this naturally produces a shorter decay time. So the effect of the higher flyback voltage dominates over that produced by blocking Coss.

Of course, there's little point playing around with the diode anyway, if your MOSFET is going into avalanche mode. Then it's more important to either replace it with a higher Vds MOSFET, or reduce the TX-on time to prevent it avalanching.



Installed and working. Thanks!

Qiaozh,

A fast coil has a characteristic that many people overlook. The coil discharge TC (calculated by the coil inductance divided by the Rd value) should be 5 times shorter than the target TC to fully stimulate it. That means that higher damping resistor values cause a faster pulse discharge TC and stimulate some smaller targets better as some of the data in this thread shows.

My own rule of thumb is that for each 100 pf in the coil circuit capacitance I can reduce, I can sample about 1 uS faster. This includes all the things that cause capacitance including the MOSFET COSS that can be easily reduced by using a series diode but only as long as the other sources of capacitance are optimized (reduced) as much as possible. Seeing the difference in the adjusted Rd measured value with and without the diode should tell you if the series diode has an effect, as with the diode it should be about 200 ohms higher assuming that other higher capacitance values are not swamping this improvement.

With a mono coil, the opamp input resistor and the clamping diodes appear to be in parallel with the Rd value until the damped flyback voltage falls below about 0.6V. Then, below this 0.6V point it is the Rd value alone that governs the coil discharge TC.

One of the largest sources of capacitance is the length of the coax cable between the coil and PI machine. Some forum members have found creative ways to minimize this.

The biggest difference between diode or no diode is when all the other sources of capacitance have been minimized. It is reasonable for a 200pf MOSFET COSS elimination to gain 2uS in earlier sampling.

Also, look at how well you are stimulation a particular Time Constant target. Larger and thicker targets need longer TX pulses to fully stimulate/saturate them and are not changed much by Rd values. Smaller targets of around 2.5 uS are right on the edge of detection at 10 uS but at 6uS can be easily detected.

A 2.5 uS target needs a .5 uS coil discharge TC to fully stimulate it. For a 300 uH coil that is 600 ohms. But 600 ohms in parallel with 1000 ohm input resistor down to 0.6V is only 375 ohms. If you can get the Rd value to be 1200 ohms than you have an effective 545 ohms better able to stimulate those targets right on the edge of detection. On some of Eric Fosters PI designs, he uses a 2.2K ohms input resistor just for this reason; to better stimulate smaller targets. Just making the input resistor 1.2K ohms will put the effective Rd value right at 600 ohms.

I hope this puts things into a new perspective.

Joseph J. Rogowski

[Qiaozhi]

Hi Joe,

Actually I'm already aware of the points you raised, but I expect others may not be so clear on the subject.

The discussions so far in this thread have not been concerned with target response, so much as trying to understand why the diode appears to improve the ability to early sample. In my last simulation, there was a noticeable improvement for the IRF740 with the diode installed. But zero difference when the MOSFET is swapped for an IRFIBF30G. The IRF740 has an effective Coss of 211pf, whereas the IRFIBF30G has only 103pF. The results seem to agree with the theory that Coss needs to be in the same ballpark as the other parasitic capacitances in order for the diode to have any effect. Also, with the IRFIBF30G MOSFET, it was possible to sample very slightly faster (even without the diode) than the IRF740 with the diode.

As far as I can see, this implies that the diode only eliminates about 50% of the MOSFET capacitance.

I plan to test this on an MPP board when I get some time. My previous experiment (with a different circuit) showed that inserting the diode actually produced a slower response, which was contrary to what was expected. I need to revisit that experiment to understand what's going on.

[Teleno]

My previous experiment (with a different circuit) showed that inserting the diode actually produced a slower response, which was contrary to what was expected. I need to revisit that experiment to understand what's going on.

My explanation.

The damping factor of a parallel RLC circuit is:

; (higher damping factor = slower decay)

If R remains the same, lowering C increases the damping factor (= slower decay)

If you only add the series diode the response deteriorates, you must also increase the damping resistor R.

Simulation adding the diode and keeping R the same:




Simulation adding the diode and increasing R:

[Qiaozhi]

My explanation.

The damping factor of a parallel RLC circuit is:

; (higher damping factor = slower decay)

If R remains the same, lowering C increases the damping factor (= slower decay)

If you only add the series diode the response deteriorates, you must also increase the damping resistor R.

I agree. That's why I also increased the damping resistor in my simulation. For each change, I set Rd for critical damping.

Another thing you can try, is to replace the MOSFET with an ideal switch, and put a 211pF capacitor across it. With the diode, the preamp comes out of saturation just slightly under 6.5us, whereas the IRF740 was measured as 6.7us. Interestingly, lowering the parallel capacitance to 103pF (to match the IRFIBF30G) also results in 6.5us.

I think the final conclusion (at least from a SPICE simulation) is that the series diode does improve the ability to early sample. How much you gain depends on a number of factors.
What I don't understand at the moment is why my practical experiment produced results that were opposite to what was expected. I need to revisit that soon. Perhaps I hadn't optimised the damping resistor properly, or something else is going on.

[Qiaozhi]

More food for thought ...

Here's some tests and measurements made on a real MPP board, and comparisons made with simulation:

I used a convenient mono coil that was available with an inductance of 432uH and a dc resistance of 2R6. The damping resistor tool was used in conjunction with an oscilloscope connected to the output of preamp #1 to find the value at critical damping. This was found to be 540R. The diode was shorted out. However, the actual point where critical damping is achieved is somewhat subjective. So I do not consider this method to be very accurate where its purpose is to determine the total capacitance in the TX circuit.

For reference, the total capacitance:
where the total capacitance represents the coil, cable, MOSFET and other parasitic capacitances.

A better method is to remove the damping resistor altogether, and measure the resonant frequency of the resultant ringing. In this case:




From the above, it is clear that the diode provides a reduction in total capacitance. Also, the reduction can easily be determined as 143pF.
Since Coss effective of the MOSFET (IRF740) is 211pF, this indicates that the diode effectively removes 67.8% of this capacitance from the TX circuit.

With no diode, the total capacitance is 305pF. Therefore, the capacitance associated with the coil, cable and other parasitics, is 305pF - 211pF = 94pF.

Inserting these values into the simulation, shows a very small improvement in early sampling of approximately 200ns with the diode.
This measurement is confirmed with the real MPP circuit, where it is extremely difficult to see any improvement.

My conclusion is that inserting the diode doesn't make things worse, but (depending on many factors) there could quite likely be minimal improvement. Practical measurements do however show that the diode "blocks" the MOSFET capacitance, resulting in a higher resonant frequency, and the requirement to increase the value of the damping resistor. In cases where you're trying to sample as early as possible, and "every little helps"; then a series diode can only improve things.

It appears, in my initial experiment which prompted this thread, that I must have incorrectly adjusted the damping resistor when the diode was in circuit. The result was that the sample delay needed to be increased, leading me to question the logic of fitting the diode in the first place.

Finally, I did not adjust the value of the series resistor between the two scenarios during these tests. But doing so would have reduced the improvement even further.

[green]

More food for thought ...

Here's some tests and measurements made on a real MPP board, and comparisons made with simulation:

I used a convenient mono coil that was available with an inductance of 432uH and a dc resistance of 2R6. The damping resistor tool was used in conjunction with an oscilloscope connected to the output of preamp #1 to find the value at critical damping. This was found to be 540R. The diode was shorted out. However, the actual point where critical damping is achieved is somewhat subjective. So I do not consider this method to be very accurate where its purpose is to determine the total capacitance in the TX circuit.

For reference, the total capacitance:
where the total capacitance represents the coil, cable, MOSFET and other parasitic capacitances.

A better method is to remove the damping resistor altogether, and measure the resonant frequency of the resultant ringing. In this case:




From the above, it is clear that the diode provides a reduction in total capacitance. Also, the reduction can easily be determined as 143pF.
Since Coss effective of the MOSFET (IRF740) is 211pF, this indicates that the diode effectively removes 67.8% of this capacitance from the TX circuit.

With no diode, the total capacitance is 305pF. Therefore, the capacitance associated with the coil, cable and other parasitics, is 305pF - 211pF = 94pF.

Inserting these values into the simulation, shows a very small improvement in early sampling of approximately 200ns with the diode.
This measurement is confirmed with the real MPP circuit, where it is extremely difficult to see any improvement.

My conclusion is that inserting the diode doesn't make things worse, but (depending on many factors) there could quite likely be minimal improvement. Practical measurements do however show that the diode "blocks" the MOSFET capacitance, resulting in a higher resonant frequency, and the requirement to increase the value of the damping resistor. In cases where you're trying to sample as early as possible, and "every little helps"; then a series diode can only improve things.

It appears, in my initial experiment which prompted this thread, that I must have incorrectly adjusted the damping resistor when the diode was in circuit. The result was that the sample delay needed to be increased, leading me to question the logic of fitting the diode in the first place.

Finally, I did not adjust the value of the series resistor between the two scenarios during these tests. But doing so would have reduced the improvement even further.

Interesting test. The coil with cable resonance(94pf, 432uH) should be 789kHz. I've been testing resonance exciting the test coil and cable with an operating PI. Wondering if you could remove the 432uH coil and cable from your test circuit. Connect a different coil and cable to the test circuit. While monitoring amplifier out for resonance bring the 432uH coil and cable near the operating coil to see how close to 789kHz the resonance is using this method.

[Qiaozhi]

Interesting test. The coil with cable resonance(94pf, 432uH) should be 789kHz. I've been testing resonance exciting the test coil and cable with an operating PI. Wondering if you could remove the 432uH coil and cable from your test circuit. Connect a different coil and cable to the test circuit. While monitoring amplifier out for resonance bring the 432uH coil and cable near the operating coil to see how close to 789kHz the resonance is using this method.

I do not have another coil with the correct plug available, except one that has the damping resistor built into the coil shell. So I disconnected the 432uH coil and connected it to the scope, and provided external excitation to the coil with a different PI. The resonant frequency was 588.2kHz, which equates to a capacitance of 170pF.

If the total capacitance is 305pF, coil + cable is 170pF (from measurement above), Coss effective is 211pF; then the parasitics would be -76pF. As you can see, there is a discrepancy in the values obtained, as you cannot have negative parasitic capacitance.

At the moment I believe the total capacitance value of 305pF is correct, so the error is most likely in the evaluation of Coss, which varies with applied Vds. The calculation of Coss effective is basically a fudge to determine a fixed capacitance that would give the same charging time as the output capacitance of a MOSFET while Vds is rising from zero to 80% Vds with Vgs = 0V. Perhaps this value is a bit iffy in these circumstances, and needs to be a dynamic value, rather than static.