Pulse Width Modulator

Hi Sabrina,

This is a pretty open-ended question, and this may not be the correct forum - this is primarily for supporting Circuits from the Lab and other released reference designs. 

But let's see if we can figure out where it ought to go.

• Is the coil represent by V5?
• And how long are the wires, and is the coil grounded or floating? The differential inputs to the AD8421 must have a low-impedance path to the circuit ground somewhere - right now the only path for the amplifier's bias current is the two 10M resistors. If the wire is long, it will pick up lots of common mode noise, which the AD8421 may or may not be able to handle.
• You have a very high gain, and your pulse amplitude is 5V in the simulation. You will saturate the amplifier and it may take an indeterminate amount of time to recover.
• What is the point of the PWM circuit? Does it turn the output of the amplifier into a PWM representation of the pulse? A PWM circuit would typically have an output frequency that is much higher than the highest frequency present in the input signal - is this the case?
• Is your objective to simply detect the presence of a pulse, not measure its amplitude or width (since you're shaping the pulse with an LTC6993?)

If that's the case - and you're truly looking to detect a pulse that can be as small as 1mV or as large as 30V, you may be better off with a medium gain amplifier that saturates gracefully - (like, put back-to-back diodes across the feedback resistor), then use a comparator to actually detect the pulse and trigger the LTC6993

With that, I'm going to move this to the op-amp and comparator forum, they can answer much more authoritatively.

-Mark

Hi Mark,

Thank you for replying to my question. Sorry I'm late to reply because I just recovered from illness.

1. Is the coil represent by V5?

I sent an exciting pulse with a period of 1 us, a voltage of 30 VDC and a current of 3 A to the wire.

Then the exciting pulse will flow into the wire until it hits the float magnet, and the exciting pulse will be reflected back to coil sensing.

I use a coil sensing to detect the return signal. I don't know what the voltage and current of return signal, I assume 1 until 5 VAC.

2. And how long are the wires, and is the coil grounded or floating?

The length of the wire is about 3 meters, and the wire is connected to ground at bottom side.

Coil is floating connected to input of AD8421.

3. You have a very high gain, and your pulse amplitude is 5V in the simulation. You will saturate the amplifier and it may take an indeterminate amount of time to recover?

I would reduce the gain by changing the value of the RG resistor on the AD8421 to avoid saturation in real hardware. Thanks for the input.

4. What is the point of the PWM circuit?

I saw the LTC6993 datasheet, the trigger example is a modulated carrier. That's why I use PWM because it triggers LTC6993-4.

Can the LTC6993-4 be triggered with a 1us square pulse (not in the form of a modulated carrier)?

5. Is your objective to simply detect the presence of a pulse, not measure its amplitude or width (since you're shaping the pulse with an LTC6993?)

I want to detect the presence of the return signal with high precision, which will later trigger the microcontroller.

The return signal will of course have noise and I have to be able to distinguish which is the noise and which is the return signal. How to do it?

I added the ltspice files. It's works for simulation, but I don't know if it works in real hardware too.

Thanks,

Sabrina

V(vin_coil) - blue

V(amplitudo_amplifier) - green

V(period_stretch) - red

V(carrier) - tosca green

V(pwm) - pink

V(one_shoot) - grey

V(one_shoot_inverse) - dark green 

LTspice XVIIl: LevelSensor.zip

1. Is the coil represent by V5?

I sent an exciting pulse with a period of 1 us, a voltage of 30 VDC and a current of 3 A to the wire.

Then the exciting pulse will flow into the wire until it hits the float magnet, and the exciting pulse will be reflected back to coil sensing.

I use a coil sensing to detect the return signal. I don't know what the voltage and current of return signal, I assume 1 until 5 VAC.

• Comment: do you have any actual measurements from a real circuit? This would be extremely helpful.

2. And how long are the wires, and is the coil grounded or floating?

The length of the wire is about 3 meters, and the wire is connected to ground at bottom side.

Coil is floating connected to input of AD8421.

• This is contradictory - How is the coil both floating and grounded? You should show the ground connection in the LTspice simulation - as it is right now, it's still floating. D1 and D2 look like they're intended to prevent the common mode from exceeding +/-33V, but the AD84212 is powered from +/-15V. Also while this may provide some protection from damage given the abs max ratings of the inputs, it looks like the CMRR of the AD8421 is spec'd as +/-10V when powered from +/-15V. If the coil "floats" outside of this range, behavior may not be specified. (A question for the apps engineer for this part.)

3. You have a very high gain, and your pulse amplitude is 5V in the simulation. You will saturate the amplifier and it may take an indeterminate amount of time to recover?

I would reduce the gain by changing the value of the RG resistor on the AD8421 to avoid saturation in real hardware. Thanks for the input.

4. What is the point of the PWM circuit?

I saw the LTC6993 datasheet, the trigger example is a modulated carrier. That's why I use PWM because it triggers LTC6993-4.

Can the LTC6993-4 be triggered with a 1us square pulse (not in the form of a modulated carrier)?

• Refer to the datasheet. The minimum recognized TRIG pulse width is 5ns. As drawn now, the PWM circuit doesn't really do anything aside from generate a square wave when the output is enabled. Also you've got a carrier period of 100ns, but the LT1011's response time is as high as 250ns. You may not get any response at all.
• What datasheet figure are you referring to?

5. Is your objective to simply detect the presence of a pulse, not measure its amplitude or width (since you're shaping the pulse with an LTC6993?)

I want to detect the presence of the return signal with high precision, which will later trigger the microcontroller.

The return signal will of course have noise and I have to be able to distinguish which is the noise and which is the return signal. How to do it?

• Does the circuit need to reject an initial large signal, then distinguish a much smaller signal at a later time? This is not a trivial task - you may need to "blank" the amplifier while the excitation pulse is present.

Designing your complete circuit may be out of scope for this forum - we'll certainly try to help as much as possible, but it would be good to have some high-level requirements, backed up with at least basic data from an actual circuit - what does the coil's output waveform look like, both during the initial excitation pulse, and the desired response pulse? Amplitude, rise time, fall time, etc.? What is the desired output waveform of the circuit in response to this pulse?

Another point worth considering - how many of these do you need to build? If the answer is one, then can you simply use benchtop test equipment? Also is this a completely new idea, or are there existing circuits that you can refer to?

One application that came to mind, if in fact you do have the problem of rejecting a large excitation pulse, is a proton precession magnetometer. A large pulse is used to excite a mass of material, and a much smaller waveform is measured some time later. Do a web search on this topic and you may find some ideas to leverage.

-Mark

1. Is the coil represent by V5

Do you have any actual measurements from a real circuit? This would be extremely helpful.

I read an article about that sensor, the return voltage is around 1-5 vac depending on the amount of electric current sent to the wire, it will affect the return signal read by the sensor coil.

2. And how long are the wires, and is the coil grounded or floating

This is contradictory - How is the coil both floating and grounded? 

Look at the image of the sensor above, there is a difference between wire guide (blue)  and coil sensing (red).

The wire guide is the wire which the exciting-pulse is inserted. After the exciting-pulse is inserted, the exciting-pulse will flow through a 3 meter long wire. When the exciting-pulse hits the float-magnet, the exciting-pulse will be reflected back. 

The reflection back from the exciting-pulse is read by the sensing coil. There is an op-amp symbol there. For the sensing coil both ends are connected to the input of the AD8421 op-amp.

You should show the ground connection in the LTspice simulation - as it is right now, it's still floating. D1 and D2 look like they're intended to prevent the common mode from exceeding +/-33V, but the AD84212 is powered from +/-15V. Also while this may provide some protection from damage given the abs max ratings of the inputs, it looks like the CMRR of the AD8421 is spec'd as +/-10V when powered from +/-15V. If the coil "floats" outside of this range, behavior may not be specified. (A question for the apps engineer for this part.)

When I use ground in the coil, it means I have to change the AD8421 to single supply. Uses only one input and ground the other inputs. Is that true?

4. What is the point of the PWM circuit

Refer to the datasheet. The minimum recognized TRIG pulse width is 5ns.

I saw the envelope detector circuit in the LTC6993 datasheet, they use a trigger in the form of a modulated carrier. From that, I concluded that the LTC6993 requires a trigger in the form of a modulated carrier.

You are right. I just found out that the LTC6993 can also work with a 1 pulse trigger with minimum 5ns period. 

As drawn now, the PWM circuit doesn't really do anything aside from generate a square wave when the output is enabled. Also you've got a carrier period of 100ns, but the LT1011's response time is as high as 250ns. You may not get any response at all.

You are right. I think PWM is not needed here. I just need a circuit that can read the return pulse and ignore the noise pulse.

5. Is your objective to simply detect the presence of a pulse, not measure its amplitude or width (since you're shaping the pulse with an LTC6993?)

Does the circuit need to reject an initial large signal, then distinguish a much smaller signal at a later time? This is not a trivial task - you may need to "blank" the amplifier while the excitation pulse is present.

I think the pulse feedback that is read in the sensing coil will not be the same as the pulse sent to the wire guide, there is noise there. But, I will try it first by making a pulse sender circuit with a mosfet. After that I will tell you what the output looks like, the voltage, the electric current, and the noise making it easier to read the original signal from feedback pulse.

Thanks for the explanation and guidance.

Best Regards,

Sabrina

Hi Mark,

I get reference from http://ojs.bbwpublisher.com/index.php/JERA/article/view/164/159 

Best Regards,

Sabrina

Great! So someone has built a working circuit you can start from. What I would encourage you to do is carefully identify where this circuit is deficient, given your requirements. (More precision? Shorter reflection times? Etc.) Then work through these limitations one by one. Simulation can get you part of the way there, but there is no substitute for bench-testing of the actual circuit.

I'll leave you with a couple of observations:

• The AD620 is in a gain of nearly 500, which will result in a BW of approximately 240kHz. Is this fast enough? Also - IF the amplifier ever saturates - then "all bets are off" on how quickly it will recover, ready to cleanly measure the next pulse (or reflection pulse. Also - simulation will NOT accurately predict this recovery time.
• While the front-end is an AD620 instrumentation amplifier, the coil1 input is single-ended - like I was asking above, the coil is clearly grounded right at the signal chain input. So why not use a conventional op-amp in a non-inverting configuration?

Another reference that might be useful for you - that illustrates the difficulty of measuring tiny signal artifacts, while rejecting large voltages, is Application Note 120:

https://www.analog.com/media/en/technical-documentation/application-notes/an120f.pdf

Figure 13 is a very high-gain, "soft saturating" amplifier in which no amplifier ever saturates, and is always "ready" to enter the active region. You could borrow sections of this circuit, and follow the output with a comparator to detect your pulse.

But this is far down the design process - again, a reasonable approach would be to start with that circuit you referenced, given that it is known to work.

-Mark