Mosfet Selection / Damping Resistor

[porkluvr]

... a huge problem arises because it is next to impossible to find tin plated, stranded, Teflon insulated wire. At the temperatures necessary to apply and cure Teflon insulation, individually tinned strands will fuse together. If you want stranded tin plated wire, you had better look for some other insulation besides Teflon.

Halo has informed me that Surplus Sales of Nebraska lists some Teflon insulated, tin plated, stranded wire on their website (in both 18 and 26AWG). (And, with 600V worth of insulation thickness.) I had read (and regurgitated) that that combination of characteristics was nearly impossible to find.

SSN is one of my favorite places to window shop and I had looked very hard at their offerings before I made post #13... so those offerings are either fairly new (or newly edited), or else maybe my reading comprehension has ... gone south.

So anyway that was slightly off topic. Maybe nobody will notice... or care... and maybe SSN will still have some left for me by the time I get my order in. $$

[AdrianM]

Thanks for the link mikebg - it also had some really useful stuff about putting equations up on these forums using LaTex! I did mention getting a 'ball-park' figure for the damping resistor which is useful when planning construction. The formula you supplied is of course a rearrangement of the one I used, when (the DC resistance of the coil) is input as zero.

when

A typical coil that I've made has 1 Ohm, 620uH and 200pf which means I get a calculated damping resistor value 880.59 compared to 880.34 if I neglect the coil resistance. Oh well

[mikebg]

Regarding the value of the resistor for critical damping it can be calculated in advance as R=sqrt(L/C)/2

where:
L=inductance of coil measured out of circuit with LCR meter
C=intrinsic capacitance of coil

Adrian,
C is not only the intrinsic coil capacitance. It is an equivalent capacitance of network connected across an ideal (capacitanceless) TX coil having parameters r and L only. Note that the equivalent capacitance is not connected directly across coil inductance L. The coil resistance “r” appears in series as damping element.
Here is the equivalent circuit and correct formula for analysis of critical damping resistor R:
http://www.geotech1.com/forums/showpost.php?p=125342&postcount=11
I use term “analysis” instead “calculation”, because no need to waste time calculating critical resistance. You simply shoud adjust damping resistor in an operating metal detector.

[AdrianM]

mikebg, as per the posts you refer to, the capacitance CDS for both these devices diminishes to negligable amounts so long as VDS remains high (hence the benefit of using an additional series diode such as D2 in the schematic below). The drain remains near the flyback peak voltage until the next switch cycle - so long as no bleed resistor is added! Gate capacitance is the most challenging aspect of this device, but there are some very capable drivers available for peanuts these days.

On my test harness I swap in and out mosfets and most of the other parts shown in the schematic to make A-B comparisons and select parts accordingly. The flyback transient says it all.



I'm not sure I understand you point about the formula's. Having made a coil, its parameters are required in order to calculate a ball-park figure for the damping resistor. To get the most out of a real front-end the value needs tweaking for sure, but with the coil in-circuit, quite an accurate measurement of its intrinsic capacitance for the (purposes of damping) can be made by resonating it with an external pulse while the diode (D2 above) is reverse biased or removed from circuit. The inductance of the coil and connecting lead doesn't change appreciably either in or out of circuit. From my experience using this technique the required critical damping resistor can be calculated pretty accurately, but I could never resist a tweak while looking at the scope!

[mikebg]

Coil capacitance can't be measured directly because the coil inductance will throw a meter. Instead it can be inferred by resonating the coil with a variable frequency generator (or pulse from another nearby working PI metal detector) then plugging the resonant frequency into the following equation:

C=1/(4 x pi^2 x F^2 x L)

I've been meaning to write all this down somewhere I won't lose it - here seems to be a good a place as any, and if I've got it wrong I'm sure someone'll be along soon to correct it

Adrian,
If you want to be a successful ham designer you must try to find and correct alone your own mistakes. One designer has more to think, to compare and to use known formulas mostly for an analysis instead to calculate useless values of components, which in practice will be others. The designer should not waste time in testing and measuring unnecessary parameters.
Such is the case with the damping resistor. We need formula for analysis of damping resistance, not for coarse calculation and measurement of coil capacitance. The coil capacitance is only a part from equivalent capacitance of TX network. But let we start with your favorite MOFETransistor. Compare its capacitances with capacitances given in posting # 21:

[AdrianM]

My current favourite is the STW20NK50Z SuperMESH power MOSFET

■ Extremely high dv/dt capability
■ 100% avalanche tested
■ Gate charge minimized
■ Very low intrinsic capacitances
■ Very good manufacturing repeatability
Application
■ Switching applications
Description
The SuperMESH™ series is obtained through an
extreme optimization of ST’s well established
strip-based PowerMESH™ layout. In addition to
pushing on-resistance significantly down, special
care is taken to ensure a very good dv/dt
capability for the most demanding applications.

Most of the parameters seem pretty useful for PI use...

Equivalent output capacitance VDS =0, VDS = 0 to 640 V - 187 pF
RDSon 0.23 typical
As always though, it needs a good gate driver to get the most out of it.

Regarding the value of the resistor for critical damping it can be calculated in advance as R=sqrt(L/C)/2

where:
L=inductance of coil measured out of circuit with LCR meter
C=intrinsic capacitance of coil

Coil capacitance can't be measured directly because the coil inductance will throw a meter. Instead it can be inferred by resonating the coil with a variable frequency generator (or pulse from another nearby working PI metal detector) then plugging the resonant frequency into the following equation:

C=1/(4 x pi^2 x F^2 x L)

I've been meaning to write all this down somewhere I won't lose it - here seems to be a good a place as any, and if I've got it wrong I'm sure someone'll be along soon to correct it

[bklein]

Wow, so I wasn't very original with the concept!
Oh, BTW, the TI Tina I had was old. TI recently updated their version to Version 9 from Version 7 so it probably works ok - too late (1:30am) now for me to give it a try.

Barry

[mikebg]

Bklein,
Something like your idea has been done in the late 19th century. I even published an ancient circuit diagram of the time. However its experimentation several years ago using components from old phones was irrelevant.

http://www.geotech1.com/forums/showpost.php?p=101743&postcount=23

Perhaps this principle should be experimented with modern components. Here is the principle:

http://www.geotech1.com/forums/showpost.php?p=101590&postcount=17

Here is an engraving published in 1917, which seems to indicate the realization of this idea: metal detecting by two operators. However many contemporary designers think the rotogravure is wrong because is more important the first operator to carry headphones.

http://www.geotech1.com/cgi-bin/pages/common/index.pl?page=metdet&file=history/pm17/index.dat

In the gravure only the second oprator wears receivers (headphones).

[bklein]

Interesting that I was thinking something similar.
For one thing, I was starting to play with the miscel.exe program and look at inductor discharge with parallel resistor. Making the resistor smaller lengthens the time to zero voltage (across it) after the tx pulse. Other things come into play but this surprised me.
.

I'm retracting this part. It does cross zero volts sooner but rings as you said. I finally got the Tina simulator to work using the circuit in the SurfPI schematic thread and could play with circuit values and see what's going on. The original Tina simulator I got from TI was too downlevel and the scope would not work for me. I went to www.tina.com and got their 30 day demo and it works great. As suggested there you can play with all the component values realtime and see the outcome.

[bklein]

Bklein, the correct question for damping process in time domain is:
TO RING OR NOT TO RING?
I think the damping resistor should be designed using frequency domain, because for an amateur radio designer, the term PI means wide band metal detector.
http://www.geotech1.com/forums/showp...5&postcount=24

Interesting that I was thinking something similar.
For one thing, I was starting to play with the miscel.exe program and look at inductor discharge with parallel resistor. Making the resistor smaller lengthens the time to zero voltage (across it) after the tx pulse. Other things come into play but this surprised me.
On the frequency domain thing, send a noise burst and get the response. Convert the response to audio with appropriate filters for our brains to better judge differences in frequency content enough to tell what material the target is... Maybe this could be as "simple" as stretching the return signal response out to an audio period and repeating the waveform.
I asked previously somewhere I think whether anyone had taken a spectrum analyzer and compared the various target return profiles with it. For those that can still hear fairly well this may work. For those of us that went to loud rock concerts or flew airplanes for hours on end without headphones maybe not.

[mikebg]

This may be well accepted fact, but I don't see how it matters. Maybe you guys could explain it a bit more. If I short the coil after TX I would think the dissapation is at highest rate. If the shorting is done right at the coil you could isolate the feedline and its capacitance. You could be left with ringing afterwards that a traditional damping resistor would quell, but I wonder if at that point you could utilize a variable resistance/gain circuit or other means to normalize it. (?)

Bklein, the correct question for damping process in time domain is:
TO RING OR NOT TO RING?
I think the damping resistor should be designed using frequency domain, because for an amateur radio designer, the term PI means wide band metal detector.

http://www.geotech1.com/forums/showpost.php?p=102465&postcount=24

[bklein]

At exponential interval of damping process, we can not increase the power of dissipation because the resistance for critical damping is independent of flyback voltage. That's why breakdown damping is very important for fast TX network.

This may be well accepted fact, but I don't see how it matters. Maybe you guys could explain it a bit more. If I short the coil after TX I would think the dissapation is at highest rate. If the shorting is done right at the coil you could isolate the feedline and its capacitance. You could be left with ringing afterwards that a traditional damping resistor would quell, but I wonder if at that point you could utilize a variable resistance/gain circuit or other means to normalize it. (?)

[Tepco]

This is not actually correct.In order to achieve 4A current in typical 300uH coil, you will need about 120uS pulse without much ohmic losses etc.(IRF740 will reach breakdown voltage whit 20-30uS pulse under similar conditions.) At this point energy stored in coil will be about 2.4mJ, (or 2.4 milliwatts per second) coressponding to 1.2W power at 500pps. And when it is released in typical 1uS pulse, peak power will be (ideally) 2.4KW. Dumping and breakdown rating of IRF will only influence that time. Actually, device whit higher breakdown voltage (whitout you can reach that limit) and lower capacitance will reduce this time, and even increase "peak power" ever more. Point is, you cannot just multiply breakdown voltage and coil current as posted here. (what you will get whit IXYS 4KV rated device then ,16KW or what?). But this "peak power" is just a math, and not so important compared to other issues (energy stored in coil and speed of release of is).

German designers did use all sorts of things in those detectors, including 4Ohm RdsON mosfets, way wrong dumping resistor values etc, only neon sign to dump the circuit is missing... I built one, another member of this forum built another, and, actually, when all boobietraps from schematics are eradicated, these are quite good, useful and nice built detectors.

At exponential part of ringdown process, (it is not actually exponential, this is RLC not LR condition and R is usually quite nonlinear) you can actually increase peak power avoiding breakdown and any nonlinearity related to it, dumping resistance is dependent on flyback voltage when breakdown is reached. But, peak power aside, optimal circuit will be one that can get rid of coil energy as soon as possible, and start sampling, what we trying to measure here are millivolts, after hundreds, or even thousands volts of flyback "bang". As far as i realized, name of the game here is speed, not peak power, coil voltage or current, pulse width or frequency,these just have to be optimized in right manner for maximum speed under given (coil) conditions.

Someone correct me please if i misfired something here.

And, i have to make another post related to those pro&con series diodes. I didn't wanted to raise that issue (but i did) and finally i had few hours free to make some measurements again, under different (breakdown or nonbreakdown conditions,diferrent coils etc) and i believe results will be interesting to anyone, and how i get managed to be half -right and half-wrong on it. That is for now, got to work...

(I believe most typical problem for most of us on this forum is typical lack of time to devote to this hobby)

[mikebg]

Yes, but those inner dissipation (including breakdown) is, as is, mean uncontrolled.

What we need is controlled breakdown. To reach this MOSFET inner dissipation is of small use.

So we need outer damping resistor: to get control over breakdovn, to shape breakdown curve in such way that we get useful signal prominent as much as possible.

Do we need damping resistor? Yes, we need it: for better results (sensitivity).

WM6, to realize fast TX network we need fast dissipation of energy stored in reactive elements L and C. That means we should achieve high momental power of dissipation
p(t)=u(t).i(t)
at damping process.
At breakdown interval with RF740 we have constant u(t)=Vbr=400V. If we have coil current 4A at moment t=0 (when switch turns off), this gives momental power of dissipation p(0)=400x4=1600W. When breakdown ceases, too litle energy remains in tank circuit for exponential dissipation (with damping resistor). Instead damping with MOSFET breakdown, German designers use VDResistor or HV Zener diode to realize damping with breakdown process:

http://www.pulsdetektor.de/

At exponential interval of damping process, we can not increase the power of dissipation because the resistance for critical damping is independent of flyback voltage. That's why breakdown damping is very important for fast TX network.

[Tepco]

!!!!!!!!!!!!IMPORTANT!!!!!!!!!!!!!!!!
First of all, i have to apologize to this forum about my previous writings, concerning series diode-MOSFET performance, and part, or all of my explanation here may be wrong. I did measured this dozen times, but in setup and under conditions when this trick cannot make any improvement (due to other tricks already used). As you may guess from my previous posts all my PI' s are based on cascode connected switch whit healthy 2KV flyback. But i failed to measure this under ordinary conditions (Surf-like detector etc). So it is not correct that under some conditions some improvement cannot be achieved doing this. I will not become advocate for this solution, changes are small, this will not make revolution in PI, but i was wrong claiming this will always be slower. In my current setup (not MD, trying to measure something else) i have too much noise and layout issues, and i don't have DSO at hand, but at first opportunity i will post waveforms under different conditions, coils etc to make comparison, and to see what can be gained (or loosed) doing this.
Until then, don't listen to me (we are not saints really). And, BTW, forums like this are ideal place to make mistake, get criticized and corrected, this is why i'm here.

Another part is, from previous post, using another MOSFET to short the coil. Not going to work, you need to get rid of all energy stored in the coil, shorting it this will not happen. Shorting the coil in part of TX pulse (burst) is the part of multipulse technology, see relevant Minelab patents etc. Sorry, at least i should be right on this one. Appologise accepted-YES-NO CANCEL?