New PI technology proposals.

[Carl-NC]

Segmented PI Coils

I am breaking this out separately, please see the post in the Coils section.

- Carl

[Carl-NC]

Andy,

An inductor with half the turns has 1/4 the inductance. Therefore your individual coils should be 75uH. Then, to account for their interaction, you need to make them mutually coupled. Aziz has his schematic set up correctly.

- Carl

Are you guys including the inevitable capacitance when modeling this? The exercise is rather futile if you aren't.

A capacitance free coil, if it actually existed, wouldn't need a damping resistor. The energy would simply dissipate through it's own resistance.

[Andy_F]

I am sorry. Correctly to model so (inductance in the "improved" variant should be in 4 times less, than in classical). But also so anything good is not present.

[Andy_F]

To ease the imagination of the idea, I will put the schematics of this idea. Then you can make your own thoughts.
...
The following picture shows the MONO coil configuration.

The offered variant, IMHO, the senseless.

On the scheme two variants - classical and "enhanced". The green line corresponds to a classical variant, the red line corresponds to "enhanced". With other parameters being equal "enhanced" variant works obviously worse.

[Aziz]

If we cannot achieve the good balance of the CT coil, here is an another proposal:

We have two coils:
- Transmitter coil: (less windings, less inductivity, less capacitance). This can also be a center-tapped coil but driven as MONO coil and optional damped on each half.

- Center-tapped receive coil: (more windings, more inductivity).
This will be used in a differential mode. The center-tap is connected to ground or left unconnected (differential line). Each half will be critical damped. The receive coil can have a much higher flyback voltage as it doesn't have any impact to the MOSFET's avalanche breakdown voltage (galvanic decoupled). Transmitter and receiver coil will share the same geometrics and acts as a MONO coil (same coupling k=1). If the receive coil has some minimum distance to the transmitter coil, there will be less capacitance from the transmitter to receiver coil. The center-tapped receive coil can have a twisted pair coil cabling to omit coax cable capacitance. The transmitter coil cable should anyway not be shielded. Maybe, the damping of the transmitter coil could be underdamped to force the receiver coil to damp more.

Aziz

[Aziz]

Hi Carl,

Aziz,

I did more careful measurements today. The mono coil is settling to 100mV @ 6.0us. The CT coil is settling to 100mV @ 4.8us. So that's a 1.2us improvement. In both cases, those times are measured from the 555 pulse that drives the MOSFET. There is ~0.5us delay from that edge to the actual current edge, so subtract 0.5us from both numbers to get actual settling.

Sims suggest more improvement. I think any sims involving coils will be severely limited by the coil model. I've dealt with this in designing RF chips, and a distributed model is necessary to get good results. When mutual inductance is involved (as it is with the CT coil) the complexity is just about beyond feasibility. The lumped model didn't even come close to predicting the severe oscillations I got with the original (differentially damped) CT configuration.

Sims also suggest that splitting the coil into multiply damped segments will improve settling. But this is also with a lumped model, and I suspect that a distributed model would show the limitations of this technique. I'm in the process of making a 4-segment coil to test, and will post those results as well.

- Carl

thanks for your effort. It's looking very interesting. I think, the ringing of the CT coil could be lowered, when the lower side of the half coil is not reaching the avalanche breakdown voltage of the MOSFET and diode. So the current dissipation on the coil halves could be the same (balanced dissipation).

We also need not much amplification of the front-end. This will increase the S/N (noise amplification gets less relevant). The effective target signal quantity is higher, the earlier we sample.

The picture below visualizes this. The surface area of the net target signal gives a higher potential for the enhanced coil (MONO coil).

Aziz

[Aziz]

Hello Stefan,

Aziz,

Just a quick note regarding the high flyback voltage you develop using the center tapped coil arrangement. As I see it, the driven 75uH coil is clamped by the avalanche effect of the MOSFET to about 400 volts. The other 75 uH coil as it is connected in series and bifilar has an effective inductance of 225 uH ( total both coils of 300 uH ) is unclamped other than the loading resistor so this part of the coil will develop higher voltages. This also means that the decay curve will be a lot less as the majority of the coil is unclamped. You can prove or disprove this on the standard mono driven circuit by using a MOSFET which has an avalanche of 800 volts or more.

So your circuit is showing that it is the unclamped coil that is giving the gains in reduction of delay.

Stefan

if we solve the ringing effect of the center-tapped coil by a careful coil and front-end design, we will have:

1. more effective higher flyback voltage (twice more), which can exceed the MOSFET's avalanche breakdown voltage. Only the half coil (half voltage) will be seen by the MOSFET and diode.
If we avoid the breakdown voltage, than the balance of the half coils should be much better. If we have a breakdown on the MOSFET, the balance will be disturbed due to different coil currents! It may cause the ringing effects observed from Carl!
The higher the flyback voltage, the better will be the target response. It may interesting to use higher voltage MOSFET's and diodes (800V/1000V).

2. a higher target response (full coil will induce the target signals). We have anyway a higher target response by damping each halves of the coils individually. So the center-tapped coil is predetermined for cascaded critical coil damping.

3. differential signal line: no coax cabling is needed (lowering the total capacitive load)

We should look, whether we can improve the use of center-tapped coils. If we can achieve this, we will have more benefits.

Aziz

[Carl-NC]

...and a blocking diode between the MOSFET and the coil

As I mentioned in another thread, the diode has almost no effect.

I built the 4-segment coil today, will try to test it tomorrow.

- Carl

[Carl-NC]

How is applying Tx to 1/2 of the coil doing anything other than requiring more current during the transmit pulse? I understand the CT coil with the extra damping resistor would shorten the time for decay. But, I don't see how the Tx only on 1/2 of the CT coil makes a difference.

Using the half-coil during the ON period means that a higher current will develop. This would produce a higher B-field but, as I said before, is negated by the N/2 factor. So it's possible that the TX B-field is the same.

During the OFF decay, the additional half-coil is added, and each half-coil is individually damped. I suspect this is where the real speed-up occurs. I have more experiments in mind to verify.

- Carl

[bbsailor]

Aziz,

I did more careful measurements today. The mono coil is settling to 100mV @ 6.0us. The CT coil is settling to 100mV @ 4.8us. So that's a 1.2us improvement. In both cases, those times are measured from the 555 pulse that drives the MOSFET. There is ~0.5us delay from that edge to the actual current edge, so subtract 0.5us from both numbers to get actual settling.

Sims suggest more improvement. I think any sims involving coils will be severely limited by the coil model. I've dealt with this in designing RF chips, and a distributed model is necessary to get good results. When mutual inductance is involved (as it is with the CT coil) the complexity is just about beyond feasibility. The lumped model didn't even come close to predicting the severe oscillations I got with the original (differentially damped) CT configuration.

Sims also suggest that splitting the coil into multiply damped segments will improve settling. But this is also with a lumped model, and I suspect that a distributed model would show the limitations of this technique. I'm in the process of making a 4-segment coil to test, and will post those results as well.

- Carl

Carl,

Here are some observations about your coil Rd segmenting experiment.

The value of the damping resistor (Rd) is based on three main capacitance adding sections of the PI design and one phantom parallel resistance in parallel with Rd being the opamp input resistor (Rin).

1. The coil with its coil shield has a range of capacitance of from 75 to 100 pf.

2. The coax coil cable can range from 45 pf for a short cable to about 200 pf for a 7.5 ft cable.

3. The MOSFET COSS that the TX circuit sees typically in the range of from about 50 pf to about 1000 pf.

If 2 and 3 above have higher capacitance then segmenting the damping resistor will have a minimal impact, as any benefit is being over shadowed by the higher capacitance in these areas. However if you have a short, low capacitance coax cable of about 30" with 45 pf and a low COSS MOSFET and a blocking diode between the MOSFET and the coil, then you can probably shave off the equivalent of half of the capacitance damping in 1 above for a saving of about the equivalent to 50 pf.

In a mono coil, with a 1K Rin value, the discharge of Rd if effectivly in parallel with Rin for almost 5 TCs of discharge until about 0.7V plus the time until the opamp comes out of saturation. There is a tradeoff between raising the value of Rin to raise the effective parallel value of Rd and Rin at the expense of the higher value of Rin having more noise.

While segmenting the damping resistor into multiple segments can speed up the coil damping, it does so at the possibility to have a higher effective value of Rd only related to 1 above while 2 and 3 still needs to be damped.

I believe you will find when you actually try a segmented Rd in a coil and trim the segmented Rd values for optimum damping that you will find the effective value of Rd to be about 1.3 times higher than the single Rd producing a slightly higher peak flyback voltage. I suspect that you may be able to sample about 1.5 uS faster and easily get under 10uS. Even though Rd is higher it is still in parallel with Rin, so its increased effect is reduced somewhat but still worth optimizing along with optimizing everything else.

Going to the trouble of segmenting the Rd value is most useful when you combine it with a short, low capacitance coax, low COSS MOSFET with a high enough voltage to not clamp the peak flyback, and using a diode between the MOSFET and coil to attempt to minimize the MOSFET capacitance.

This Rd segmenting would have the most potential speed-up effect when used in a DD coil along with the other things I mentioned above. I hope your experiment bears out my observation above.

bbsailor

[UWLocator]

I have to be missing something here.

How is applying Tx to 1/2 of the coil doing anything other than requiring more current during the transmit pulse? I understand the CT coil with the extra damping resistor would shorten the time for decay. But, I don't see how the Tx only on 1/2 of the CT coil makes a difference.

Rip

[Carl-NC]

TX on the half-coil.

[UWLocator]

Aziz,

I did more careful measurements today. The mono coil is settling to 100mV @ 6.0us. The CT coil is settling to 100mV @ 4.8us. So that's a 1.2us improvement. In both cases, those times are measured from the 555 pulse that drives the MOSFET. There is ~0.5us delay from that edge to the actual current edge, so subtract 0.5us from both numbers to get actual settling.

- Carl

Carl, were your results with the CT coil using the Tx only on 1/2 of the coil and Rx on the full coil? Or, were both across the full coil?

Rip

[Carl-NC]

Aziz,

I did more careful measurements today. The mono coil is settling to 100mV @ 6.0us. The CT coil is settling to 100mV @ 4.8us. So that's a 1.2us improvement. In both cases, those times are measured from the 555 pulse that drives the MOSFET. There is ~0.5us delay from that edge to the actual current edge, so subtract 0.5us from both numbers to get actual settling.

Sims suggest more improvement. I think any sims involving coils will be severely limited by the coil model. I've dealt with this in designing RF chips, and a distributed model is necessary to get good results. When mutual inductance is involved (as it is with the CT coil) the complexity is just about beyond feasibility. The lumped model didn't even come close to predicting the severe oscillations I got with the original (differentially damped) CT configuration.

Sims also suggest that splitting the coil into multiply damped segments will improve settling. But this is also with a lumped model, and I suspect that a distributed model would show the limitations of this technique. I'm in the process of making a 4-segment coil to test, and will post those results as well.

- Carl