PWM noise generator design advice

[reyntjensm]

Hello everyone,

I already tried to solve my LTspice problems within another thread here on the forum but i guess my design is bad. So for this reason i decided to start a new thread with some more information about my project. I hope somebody can give me some feedback on how to design this thing.

I'm working for a research group and they asked me to manufacture a "Noise generator". It will be used by a university research group. At the university, they have a giant setup with 20 electrical cabinets, each cabinets must represent an industrial building. All the cabinets are connected with a low voltage distribution net. The idea is to generate noise within one cabinet and check how the noise will behave within the distribution net and what it will do to the other users.

So they asked me to make a switching device for a high power resistor, the idea is to switch a 20amp load at 230vac within a frequency range from 2-500kHz. So i thought to use a PicoScope to generate the PWM signal, with this signal i want to switch 2 mosfets so i can control the load and generate some noise. I would like to have some galvanic isolation between the PicoScope and the noise generator. My first idea was to use a simple optocoupler but they are all to slow within that frequency range. I tried different and more complex devices like a HCPL3140 and a 6N137. I first need to simulate the design within LTspice before i can start to design the PCB. And this is where things get difficult, i'm struggling to get the LTspice simulation running.

Also the design should be as simple as possible since we want to manually calculate al the voltages and currents so we can verify the LTspice simulation and make a good scientific report on this project. So for this reason i can't get too complex with the galvanic isolation since this would be too hard to calculate( like the 6N137 which has internal logic gates etc).

So for this reason i think an optocoupler or something similar is not the way to go, i hope somebody can give me some tips so i can get this thing to work with the least amount of components.

Our supervisor suggested to work with an optical fiber connection to send the PWM signal from the PicoScope to the noise generator, but i think this will be too complex and i have no idea how to start with this.
Attached you can find my current LTspice simulation which doesn't work and always results with the same error....

Thank you all!!

[Phoenix]

So they asked me to make a switching device for a high power resistor, the idea is to switch a 20amp load at 230vac within a frequency range from 2-500kHz.

This is not a trivial design and if successful will generate an incredible amount of "noise" on the line without any additional filtering! What kind of duty cycle is required for the frequency?

You need to consider:
A fast enough transistor. Probably SiC 650V or GaN 600V and 30A+ rating.
Gate drive power supply to suit the transistor.
A gate drive IC fast enough to be effective at 500kHz (2us period), with enough current sourcing/sinking.
Getting the heat into a big heatsink and fan.

And probably the hardest part - a method to validate your setup is behaving as required in practice.

I would suggest not trying to simulate this in LTSpice; it's not the right program. In fact this will be very hard to simulate in any usable fashion as layout will have huge effects.

[Kleinstein]

The 6N137 opto-couplers have a relatively high coupling capacitance. In addition to the chip itself the 2 sides are still relatively close to each other. To be sure the measured signal is not coming from the control path, I would also suggest some fiber optics. There are relatively common fiber optic parts for use with audio equipment. They should not be so much more complicated to use. With the near perfect isolation one can leave the fiber optic part out from the simulation.  So the fiber optics would even simplify the simulation.

[Phoenix]

The 6N137 opto-couplers have a relatively high coupling capacitance.

The HCPL-3140 in his LTSpice sim is listed as 60pF input-output capacitance - similar to a 6N137. I like the modern RF coupled gate drivers like the UCC23513 with 0.5pF and 150kV/us.

[bill_c]

Very few industrial loads are resistive, fewer still are switched with MOSFETs. If you want industrial noise, use a heavy duty contactor and switch something inductive like a motor or transformer. Also get a 3phase variable frequency drive (cheaper usually makes more noise) and ignore all the special shielding and special grounding instructions (but do still ground everything). Use EPVC conduit and plastic enclosures so that noise can go everywhere.  Add a few single phase capacitor start motors, maybe a180V DC motor with an SCR drive.  But that still isn't going to simulate the repair guy with a welder using the metal building structure and/or the conduit/ground wire to the machine as a return for the welder output at the other end of the shop.

Example:
https://www.automationdirect.com/adc/shopping/catalog/motor_controls/iec_contactors_-z-_starters_-z-_overloads/fuji_contactors_-z-_overloads/32_to_50_amp/sc-e2g-24vdc
You will need to go a few sizes larger than rated since you will be switching often, these IEC ratings are based on a limited number of on/off cycles.
https://www.ebay.com/itm/7-5KW-10HP-220V-Variable-Frequency-Drive-Inverter-CNC-VFD-VSD-Single-To-3-Phase/353142083678?

[reyntjensm]

So they asked me to make a switching device for a high power resistor, the idea is to switch a 20amp load at 230vac within a frequency range from 2-500kHz.

This is not a trivial design and if successful will generate an incredible amount of "noise" on the line without any additional filtering! What kind of duty cycle is required for the frequency?

You need to consider:
A fast enough transistor. Probably SiC 650V or GaN 600V and 30A+ rating.
Gate drive power supply to suit the transistor.
A gate drive IC fast enough to be effective at 500kHz (2us period), with enough current sourcing/sinking.
Getting the heat into a big heatsink and fan.

And probably the hardest part - a method to validate your setup is behaving as required in practice.

I would suggest not trying to simulate this in LTSpice; it's not the right program. In fact this will be very hard to simulate in any usable fashion as layout will have huge effects.

The idea is to program all different type of PWM frequencies and duty cycles to see the impact on the distribution net.
With an LCR meter we measured all the cable connections within the grid.
This is also the reason why we need to simulate and calculate the influence of the noise generator on the grid.
The idea was to make the noise generator as simple as possible so we can calculate the resistance/capacitance and inductance and add this to the properties of the distribution net. Measure it again an see if the agrees with our calculations( i'm not the mathematically guy, i only need to make the PCB;) ).
When i have a working circuit, i can do the layout and re-simulate with the effect of my layout included? Since i only need to go up to 500kHz and the amount of components should be as less as possible, i think the layout doesn't matter that much? Could it be possible to measure the noise generator PCB with a LCR meter? I don't think so since it includes active components like the mosfet's etc?

The 6N137 opto-couplers have a relatively high coupling capacitance. In addition to the chip itself the 2 sides are still relatively close to each other. To be sure the measured signal is not coming from the control path, I would also suggest some fiber optics. There are relatively common fiber optic parts for use with audio equipment. They should not be so much more complicated to use. With the near perfect isolation one can leave the fiber optic part out from the simulation.  So the fiber optics would even simplify the simulation.

This is indeed a good point. Do you have a good starting point/components i can start googling with?

Very few industrial loads are resistive, fewer still are switched with MOSFETs. If you want industrial noise, use a heavy duty contactor and switch something inductive like a motor or transformer. Also get a 3phase variable frequency drive (cheaper usually makes more noise) and ignore all the special shielding and special grounding instructions (but do still ground everything). Use EPVC conduit and plastic enclosures so that noise can go everywhere.  Add a few single phase capacitor start motors, maybe a180V DC motor with an SCR drive.  But that still isn't going to simulate the repair guy with a welder using the metal building structure and/or the conduit/ground wire to the machine as a return for the welder output at the other end of the shop.

Example:
https://www.automationdirect.com/adc/shopping/catalog/motor_controls/iec_contactors_-z-_starters_-z-_overloads/fuji_contactors_-z-_overloads/32_to_50_amp/sc-e2g-24vdc
You will need to go a few sizes larger than rated since you will be switching often, these IEC ratings are based on a limited number of on/off cycles.
https://www.ebay.com/itm/7-5KW-10HP-220V-Variable-Frequency-Drive-Inverter-CNC-VFD-VSD-Single-To-3-Phase/353142083678?

We are going to use a big 5kW breaking resistor ( 20 ohm, we measured it with an LCR meter and it's almost perfect resistor).
I know equipment like frequency drives/PV inverters and CFL bulbs generate a good amount of noise. But this is not what we are looking for. We want to have good control of the noise we generate and try every different configuration possible. With a frequency drive or whatsoever we can't control the noise emitted by the equipment, but this is exactly what we want to do so it's not a good option. This is also the reason why we use a resistive load and not a capacitive or inductive load...

[reyntjensm]

Somebody suggested me another circuit with a pulse transformer. Does anyone have something to say about this schematic?

[KK6IL]

Digital Isolators have much greater bandwidth than optocouplers. I've used the Si8660 family of parts with success.
<https://www.silabs.com/documents/public/data-sheets/Si866x.pdf>
Google "Digital Isolator" for other products.

[jonpaul]

Bonjour, What frequency bands are you concerned with? Which region, EU, USA, China? What standard of compliance" Pulsed transients or RF noise?

I Strongly suggest to consult standard texts on noise and EMI, and compliance engineering

 
EMI and  noise generators are commercially available.

You need LISN and noise injection isolation networks, not cheap.

Commercial labs have 100s thousands or millions EU of equipment to make these tests.

Standard line EMI filters have both CM and DM noise surpression at any desired attenuation.

Finally there is a safety issue in your proposed experiments!

Danger of shock and death.

Do you have an experienced electronic engineer/technician used to working on 240V circuits?

Just the ramblings of an old retired EE

BON CHANCE!

Jon

[T3sl4co1l]

Can you ask a higher-up to come here and reply to questions?  The entire methodology is problematic, and as you say:

i only need to make the PCB;)

It seems there are challenges here, that you can't hope to meet.  Either they expect a mediocre result, or they're setting you up to fail.  Neither option looks great.


I am curious about this:

We are going to use a big 5kW breaking resistor ( 20 ohm, we measured it with an LCR meter and it's almost perfect resistor).

In what sense is it "perfect"?  At what frequencies was it measured?

Tim

[reyntjensm]

Bonjour, What frequency bands are you concerned with? Which region, EU, USA, China? What standard of compliance" Pulsed transients or RF noise?

I Strongly suggest to consult standard texts on noise and EMI, and compliance engineering

 
EMI and  noise generators are commercially available.

You need LISN and noise injection isolation networks, not cheap.

Commercial labs have 100s thousands or millions EU of equipment to make these tests.

Standard line EMI filters have both CM and DM noise surpression at any desired attenuation.

Finally there is a safety issue in your proposed experiments!

Danger of shock and death.

Do you have an experienced electronic engineer/technician used to working on 240V circuits?

Just the ramblings of an old retired EE

BON CHANCE!

Jon

They want to see how the noise emitted from a high power load will contribute trough the network. So only injecting noise into the network is not good enough. It has to be a fast switching load, in this case. A braking resistor. We have a LISN here but it's not so easy to model this within the network.

Can you ask a higher-up to come here and reply to questions?  The entire methodology is problematic, and as you say:

i only need to make the PCB;)

It seems there are challenges here, that you can't hope to meet.  Either they expect a mediocre result, or they're setting you up to fail.  Neither option looks great.


I am curious about this:

We are going to use a big 5kW breaking resistor ( 20 ohm, we measured it with an LCR meter and it's almost perfect resistor).

In what sense is it "perfect"?  At what frequencies was it measured?

Tim

Thank you for your input. We measured this between 4hz and 8Mhz and it was the best big resistor we have available.

If somebody is still interested, i attached the current schematic.

[jonpaul]

Schematic: Add HF bypass cap across 30V supply, 10 uF will not work for MHz transients or parasitics!

Schematic:

Add series damping resistors 10-51 Ohm in series with every FET

Use Resistor, Globar, Carborundum  for load  20 Ohm,  search ebay.

Layout bridge and DC PSU for minimum inductance.


Good luck,


Jon

[Wolfram]

Add the real impedance of your mains feed to your simulation, and see what happens to the current through the resistor and the voltage across the transistors when switching at 500 kHz.

The noise current you inject will be highly dependent on the actual impedance of the grid and anything else connected to it , so your results will depend on a lot of uncontrolled variables. I fully agree with Jonpaul, EMI generation and propagation on the grid is an established and well-researched field, if you want to do things radically differently then this needs to be justified technically.

[jonpaul]

Wolfram, many thanks!

To originator, reyntjensm:

Please note Mains impedance is a function of distance to distribution transformer, wire gauge and length, other connected loads, etc.

Most interférence  is in form of impulses (distribution switching, lightning or breaking inductive load) or wideband EMI created by SMPS, and electronic lighting ballasts.

The allowable EMI spectrum and testing transients are defined by the regulatory and standards industries eg CISPR and IEC 61000-6-4: Emission standard for industrial environments

IEEE defines very well the LISNs and transient waveshape and frequency of occurence.

The design of EMI filters for compliance and transient protection is very well studied and documents for many decades.

Notice the the EMI spectrum starts at 150 kHz or 450 kHz and extends to beyond VHF frequencies.

Testing below 500 kHz is not very effective.

Suggest that reyntjensm first research the many standards and texts on EMI and power transients before expending money  time and energy on "reinventing the wheel"

With Kind Regards,

Jon

MSEE, (power electronics, 1968, 1971)

[reyntjensm]

Wolfram, many thanks!

To originator, reyntjensm:

Please note Mains impedance is a function of distance to distribution transformer, wire gauge and length, other connected loads, etc.

Most interférence  is in form of impulses (distribution switching, lightning or breaking inductive load) or wideband EMI created by SMPS, and electronic lighting ballasts.

The allowable EMI spectrum and testing transients are defined by the regulatory and standards industries eg CISPR and IEC 61000-6-4: Emission standard for industrial environments

IEEE defines very well the LISNs and transient waveshape and frequency of occurence.

The design of EMI filters for compliance and transient protection is very well studied and documents for many decades.

Notice the the EMI spectrum starts at 150 kHz or 450 kHz and extends to beyond VHF frequencies.

Testing below 500 kHz is not very effective.

Suggest that reyntjensm first research the many standards and texts on EMI and power transients before expending money  time and energy on "reinventing the wheel"

With Kind Regards,

Jon

MSEE, (power electronics, 1968, 1971)

Thank you for your feedback. We have an EMC lab and test chamber within the university, so i think the guys are knowing what they are doing. The reason why they want to measure between 2-500kHz is because there hasn't been a lot a research between those frequencies on a full scale grid.
I've added the source impedance's and it does effect my rise times but not in a huge way.
Why would i need damping resistors for each fet? This will just make the rise times longer?

[reyntjensm]

For anybody who would still be interested in this topic. This is the circuit i made with good results.