Totem-pole gate driver VS Active pull down VS Fast Fet turn off

[Mr.Jaick]

why can't we lower the Pulse width to avoid avalanche instead of higher resistance or lower drive voltage?

[Ferric Toes]

Having gone through a considerable number of different Mosfets, I am now revising which ones are good for a rapid switch off and those that are not. This is a result of looking at plots of Tx current waveform, high voltage flyback, and gate drive. One thing it is best to avoid is avalanching as this generates a lot of uncontrolled ringing noise. There are many variables which influence the result and the coil is obviously one of them. My coil is 10inches diameter, 330uH and 5.4ohms, with 1.2m of 75ohm Vandamme coax giving a total of 5.8ohms. In series, there is the 0.1ohm current sensing resistor, plus a series 4.7ohm metal oxide resistor to improve Tau and reduce coil current to 0.9A. This was necessary to bring the high voltage pulse below avalanche using a 12V supply to the Tx. Resistive damping to the coil is quite critical and was adjusted to give the fastest response without coil ringing.

As to the Mosfet, I found that those with a Vds of >500V were necessary. A commonly used one is the IRF740, but this avalanched and rang in my test circuit. Types with very low On resistances did not work well, but the STP9NK50 with a typical 0.75ohms Rds(on) worked fine with very low ringing and 500V Vdss. The ringing we are talking about and measured on many Mosfets had a period of 80nS which equates to 12MHz, so it's not coil ringing. Another candidate is the STP10NK60. All of these Mosfets are in the same ballpark as regards voltage and current ratings.

One puzzling feature on the blue trace for coil current, is the positive going spike which crosses the zero line to indicate a reverse current and then returning to -0.8A before decaying more slowly to zero which is the damped coil response. Some Mosfets have a much bigger excursion into the reverse current which results in higher amplitudes of the 12MHz ring. Green commented on this a while back and I thought is was an artifact of the back emf pulse. However the two do not coincide in time, except that the voltage starts to rise rapidly as the coil current approaches zero. Maybe this blip in the coil current is caused by one or more parasitic capacitances in the Mosfet. The true coil current decay is the curve that ends between 1.5uS and 2.0uS and ideally continue smoothly to join the 0.9A level.

The top green trace shows the back emf peaking just below 500V and then decaying to around 350V, then with a slow decay towards 200V when it is shorted to ground by the turning on of the next pulse. this voltage is stored on the Coss of the Mosfet and blocked by the series diode, which is now the MUR460 rather than the HER208 used in previous plots. Changing to the MUR460 seemed to make a small difference, but will look into that later.

The two best Mosfets are P9NK50 and P10NK60. The latter having more margin before avalanche sets in. Both Mosfets have the blip in the current trace ending close to zero.

The P9NK50 then had the supply voltage increased to 13.5V with the result that the flyback exceeded the avalanche voltage. Now the ringing appears. . Using the 600V device you can increase the voltage to 15V before ringing starts.

A device I recently advocated, 17N80, because of its very low Rds(on) and Vds of 800V didn't come out so well and there was a longer delay time before switching started. Notice the much larger amplitude of the spike in the blue trace and that we are still at least 300V away from avalanche.

Lastly, how does the good ol' IRF740 fare? Avalanche at 450V and ringing with a big negative spike.

Remember that all these plots are with extra resistance in the coil circuit and only 0.9A maximum coil current. Lower coil resistance or higher supply voltage will likely cause avalanche to occur, sometimes to the point that the Mosfet will be turned on a second time for a brief period.

Where we have delays of >10uS these effects will be long gone and may not matter, but the original question was for fast switching a device and these plots show show what is happening at early times. One can only proceed further when we have a preamp attached that will operate in its linear region much quicker than we have at the moment.

Eric.

[dbanner]

I apologise, I shall retreat.

[dbanner]

Remember, we are essentially flashing a target. The feedback of which we try to make sense of.
Notwithstanding
The prf, after a certain frequency doesn't improve the response.
That's when I realized......

[dbanner]

Walltr is absolutely correct.
I've observed, which doesn't come without much effort, a strange behavior. Now I need more analysis power, other wise I cannot make heads or tails.

[green]

Played with real circuit adding 8ohms series resistance. Signal less but not as much as I was thinking. Noise level doesn't seem to change(avalanche, snubbed or adding resistance to limit peak current). Maybe because I have separate Tx and Rx coils.

[waltr]

Been thinking detection distance should decrease. Tried adding resistance in spice, signal lower but looks like can sample sooner. Need to try with real circuit.

Yes, the sooner sample will improve the SNR since the signal is larger.
One could also then have a longer sampling time which will also improve SNR. This is very helpful for higher conductive targets but not so much for small lower conductive targets.
MY HH PI pulse rate is 1600Hz with smallish caps in the integrator (100nF). The standard PI integrator design, R's & C's verse sample rate, is another none linear function. Best way to analyze is a spreadsheet doing regressive calculations per time step.

As I said, when I built the HH it was not great. Adding an R in series with the coil improved sensitivity. Did additional experiments with different values of series R to increase coil current and distance did not improve significantly. I documented this in my HH2 build thread.
My HH works quit well and have been surprised at how it detects very small nails and lead.

[green]

This has been some good reading.

So keeping MOSFET out of avalanche is a good thing.
Another way to do this is to reduce the peak coil current with a series resistor. This also decreases the coil Tau and allows longer TX on pulse (good for high conductive targets).

I ended up adding 10 Ohms in series with the coil on the Hammer-head I built. Without the resistor the detection was not quite stable.
Later did distance tests and found that detection distance didn't decrease with lower coil current. Possible due to MOSFET going into avalanche and first sample being cleaner.

Been thinking detection distance should decrease. Tried adding resistance in spice, signal lower but looks like can sample sooner. Need to try with real circuit.

[Monolith]

Now if I blast you with a series of equations you will accuse me of cutting and pasting. Well I can assure you, I'm not a cut and paste kind of guy.

Well, here is a cut and paste. I cut out this small part, because there are many pages of it in the public domain. It is from the legendary genius who designed more commercially successful detectors than anyone in the world. Dave Johnson.

SPECTRAL CHARACTERISTICS OF THE TRANSMITTER
If the transmitter voltage were a square wave, say for instance like the Fisher CZ's, the voltage of the harmonics would be inversely proportional to the harmonic number. This puts about 90% of the voltage energy at the fundamental. The current waveform is a triangle wave, the integral of the the voltage waveform.
In a CCPI system, the current waveform is approximately a square wave-- that is, the first derivative of a triangle wave. Over a broad range, the voltage of the harmonics is proportional to their frequency. This relationship no longer holds at frequencies whose period is less than about twice that of the transmit pulse duration (Phases I and III). There is a null in the spectrum for harmonics which have a period about that of the transmit pulse duration, and the higher frequencies roll off at a 1/f rate.
The square wave (as used by the Fisher CZ) is good for finding objects within a somewhat restricted size (i.e. equivalent conductivity) range, for instance US coinage and aluminum pulltabs. The transmitted spectrum of a CCPI machine is particularly well suited to finding objects of unknown size-- for instance gold prospecting, which requires high sensitivity to tiny nuggets while retaining good sensitivity to larger nuggets as well.

[dbanner]

If we ramp the current, notwithstanding the limitations of the devices we use, then the target ought to have been fully energized, to respond, in a manner which is percievable to detection, hence a metal detector. We know this to be rudimentary.however....

[dbanner]

I challenge anyone to plot I vs V with such a complex intra dependent waveform. The closest I've seen is Mr. Green's analysis. Yet he fail to realize the obvious.

[dbanner]

Now if I blast you with a series of equations you will accuse me of cutting and pasting. Well I can assure you, I'm not a cut and paste kind of guy.

[dbanner]

Well, maybe not in the milky way, but my point being that one should not try to over analyze what is staring at you straight in the face.
No amount of code can get us out of this Quagmire.
Now we enter the realm of quantum physics.

[dbanner]

An ideal TX would be the one that senses the target. That being said, I don't mean to oversimplify.
Drake's equation states.....
Yes waltr, but it isn't that obvious to those whom art not skilled in the art.
What may seem intuitive to some may fly straight over one's head.

I offer a solution.

An exponenential curve, lest we forget.
GB is merely an obstacle, to surpass.....

[Monolith]

Tests I've done, signal strength proportional to peak current. If current is reduced to half, how much increase in pulse rate should be needed to get S/N back to where it was before current was decreased.

Waltr and dbanner give a good answers.

Maybe, the way to look at this is to make a simplified version, like an ideal version. Ideal TX, ideal target, ideal integrator etc. Then the ideal answer is simply: half the TX current = double the repetition rate.
Since an ideal world does not exist, we might simplify a very complex answer by choosing just the most important factors of the many, many that are involved.
Should we start with: what would be an ideal TX?