Author Topic: Faster switching speed from mosfets in existing circuit/PCB  (Read 6752 times)

0 Members and 1 Guest are viewing this topic.

Offline blazini36Topic starter

  • Regular Contributor
  • *
  • Posts: 72
  • Country: us
Faster switching speed from mosfets in existing circuit/PCB
« on: December 13, 2018, 03:42:12 am »
This is a project that I've tried to get some help from a couple of electronics guys with that ultimately wound up with me wasting money on 2 bad board fabs. I decided to handle it myself and with a little help in the beginners forum was able to get a good circuit going. I designed some boards in Kicad and had them fabbed up and it's functionally working well. My only issue is that the switching speed is a little slow.

The whole purpose of the board is to hold 2 Meanwell DC LED drivers (LDH-45) and invert the logic to trigger them. The "PWM-DIM" pin of the driver acts as a remote off pin when it is sunk to ground. The inverted logic is necessary because the LED drivers are used for 2 machine vision cameras that output an "open drain" which in practice sinks to ground for strobe on-time. Without the inversion the drivers would be normally on and the cameras would effectively turn them off which is backwards of the intended function. I've only gotten this to work using an N-channel mosfet used as the schematic below. The fellas I had help me used BJTs and it just didn't work right.

I've incorporated a couple of switching regulators to handle some of the on-board components as well as some minor external power outputs. My main problem with this as it is is that it's slower than the original setup I had going but I'm not sure how much of it is attributed to the drivers themselves vs my "winging it" selection of component values. My original setup used 2 AC-DC Constant current LED drivers with  the cameras sinking the V- control side of a DC SSR to switch the LED driver's outputs. I wanted to tidy things up by using these DC-DC drivers and get rid of a bunch of AC driven power supplies hence this PCB. The AC Drivers/SSRs could switch fast enough that the exposure time of the camera's could be set at 100us. Using this design my exposure time has to be set to  800us which implies that there is a switching delay between the camera's output and the strobe turning on. I'd like to get this switching as fast as possible. I'm not sure my choice of mosfet and component values was really optimal, but it otherwise works well. Hopefully this can be improved just by swapping out components of the same footprint as I'd like to use my existing boards.



« Last Edit: December 13, 2018, 03:57:25 am by blazini36 »
 

Offline blazini36Topic starter

  • Regular Contributor
  • *
  • Posts: 72
  • Country: us
Re: Faster switching speed from mosfets in existing circuit/PCB
« Reply #1 on: December 17, 2018, 11:37:50 pm »
120 views, no suggestions?
 

Offline Circlotron

  • Super Contributor
  • ***
  • Posts: 3362
  • Country: au
Re: Faster switching speed from mosfets in existing circuit/PCB
« Reply #2 on: December 17, 2018, 11:48:07 pm »
My guess is that the LED drivers are some sort of switching regulator so they would have an inductor in series with the load. The larger this inductor the slower the turn-on speed of the load. There is also unknown circuitry between the control input and the load... What is the LED current? Might be more straightforward to just use a bigger mosfet + resistor to drive the LEDs directly. That may not be what you wanted to hear though.
 

Offline nick_d

  • Regular Contributor
  • *
  • Posts: 120
Re: Faster switching speed from mosfets in existing circuit/PCB
« Reply #3 on: December 18, 2018, 04:18:57 am »
Yes. I checked the datasheets etc but forgot to write the actual response. From memory, it gives a maximum PWM frequency in the datasheet for the LED driver, which sort of implies that the switching should be faster than that frequency (whose period is much less than 100us). But the LED is not necessarily switched synchronously with the PWM. They could simply be averaging your PWM and using the analog voltage to determine a voltage or current or duty cycle for the real LED drive. Since it is a module, you do not really know what they might be doing. The datasheet does not give a minimum turn-on time, suggesting that even the observed behaviour can't necessarily be relied upon.

It would be more sensible to have a suitable LED drive voltage available, as I think you did in the previous version of the project, and instantaneously switch this onto the LED when you get your signal. You can use something like an LM317 adjustable voltage regulator as a current regulator by adding a sense resistor, it doesn't have a shutdown pin though, so you'd need a transistor or a better regulator (or a dedicated LED driver without its own voltage generation).

Another option to consider is pulsed LED drive directly from a flyback inductor, which may well be what your module does, but if so you need to do it more quickly. It would be a cool circuit to build but possibly requires some research.

cheers, Nick
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 17426
  • Country: us
  • DavidH
Re: Faster switching speed from mosfets in existing circuit/PCB
« Reply #4 on: December 18, 2018, 08:20:10 pm »
The LDH-45 responds too slowly.

Since the LDH-45 is a constant current driver, I was going to suggest using a transistor to short the output however the datasheet has a warning not to do this.  What might work is switching a zener diode in parallel with the output to lower the output voltage lower than the LED forward voltage.

I would get rid of the LDH-45 and use something else more suitable instead of trying to get it to work in a way it was not intended.
 

Offline blazini36Topic starter

  • Regular Contributor
  • *
  • Posts: 72
  • Country: us
Re: Faster switching speed from mosfets in existing circuit/PCB
« Reply #5 on: December 19, 2018, 02:08:56 am »
My guess is that the LED drivers are some sort of switching regulator so they would have an inductor in series with the load. The larger this inductor the slower the turn-on speed of the load. There is also unknown circuitry between the control input and the load... What is the LED current? Might be more straightforward to just use a bigger mosfet + resistor to drive the LEDs directly. That may not be what you wanted to hear though.
I use the LDH-45B-1050 which is a 1050ma constant current driver. It maxes out at like 43V. The LEDs are heavily heatsunk so they can handle much more in small bursts. The AC-DC LED Drivers I used previously were 700ma but ran up to almost 100V. This was actually almost too much light because it had to be heavily diffused.

 

Offline blazini36Topic starter

  • Regular Contributor
  • *
  • Posts: 72
  • Country: us
Re: Faster switching speed from mosfets in existing circuit/PCB
« Reply #6 on: December 19, 2018, 02:59:16 am »
The LDH-45 responds too slowly.

Since the LDH-45 is a constant current driver, I was going to suggest using a transistor to short the output however the datasheet has a warning not to do this.  What might work is switching a zener diode in parallel with the output to lower the output voltage lower than the LED forward voltage.

I would get rid of the LDH-45 and use something else more suitable instead of trying to get it to work in a way it was not intended.

I'm not sure that I'm using the LDH-45 "in a way that is not intended". It is capable of modulating the output by means of PWM switching signal. As the datasheet says the PWM pin also functions as a "remote ON/OFF" both of which imply that it can handle a much faster switching signal than I am presenting it with. Though I'm not providing a constant frequency it can handle a much faster frequency than I'd ever throw at it, an average of 60hz is about as fast as it would ever see. There could easily be a delay in the driver but since the datasheet does not  indicate the delay ya gotta get one and play with it.  My first concern is that my circuit isn't optimal, but since no one has really mentioned seeing an issue in passive values I'll assume it's pretty good.

While shopping for that particular driver I figured I could kill 2 birds with one stone and get rid of the bulky AC LED driver and SSR and use the LDH to handle the switching. If the driver's switching is delayed there's really nothing to stop me from just leaving the LDH's PWM pin float so the output stays on and switching the output side of the driver with a bigger transistor or opto. Switching the output side of a AC CC driver worked fine. I guess they call optos that handle more than about 0.5A SSRs but I'm thinking I can just take one of these SSRs   https://www.digikey.com/product-detail/en/omron-electronics-inc-emc-div/G3VM-61GR2-TR05/Z5431TR-ND/5810877 and supply the Anode with V+ and use the camera's remote so switch the Cathode. I could switch VOUT+ from the LDH-45 with pins 3 and 4 of the SSR. This is technically pretty much the same thing I originally did using external components. I used 2 of these guys https://www.opto22.com/products/dc60s5 to switch the output side of 2 of these https://www.mouser.com/ProductDetail/MEAN-WELL/HLG-60H-C700A?qs=5gSubTaN56md2aeGwyPD7g%3D%3D. After proving that out I realize I don't need quite that much power, but I need to condense it all down to a single PCB.
« Last Edit: December 19, 2018, 03:13:24 am by blazini36 »
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 17426
  • Country: us
  • DavidH
Re: Faster switching speed from mosfets in existing circuit/PCB
« Reply #7 on: December 19, 2018, 03:28:16 am »
I'm not sure that I'm using the LDH-45 "in a way that is not intended". It is capable of modulating the output by means of PWM switching signal. As the datasheet says the PWM pin also functions as a "remote ON/OFF" both of which imply that it can handle a much faster switching signal than I am presenting it with. Though I'm not providing a constant frequency it can handle a much faster frequency than I'd ever throw at it, an average of 60hz is about as fast as it would ever see. There could easily be a delay in the driver but since the datasheet does not  indicate the delay ya gotta get one and play with it.  My first concern is that my circuit isn't optimal, but since no one has really mentioned seeing an issue in passive values I'll assume it's pretty good.

The datasheet does not say but I do not think the PWM input directly modulates the current output.  Instead, the LDH-45 produces an average output current based on the PWM input.  This might be as simple as a capacitor across the output or LC filter on the PWM input.
 

Offline blazini36Topic starter

  • Regular Contributor
  • *
  • Posts: 72
  • Country: us
Re: Faster switching speed from mosfets in existing circuit/PCB
« Reply #8 on: December 19, 2018, 05:37:06 am »
I'm not sure that I'm using the LDH-45 "in a way that is not intended". It is capable of modulating the output by means of PWM switching signal. As the datasheet says the PWM pin also functions as a "remote ON/OFF" both of which imply that it can handle a much faster switching signal than I am presenting it with. Though I'm not providing a constant frequency it can handle a much faster frequency than I'd ever throw at it, an average of 60hz is about as fast as it would ever see. There could easily be a delay in the driver but since the datasheet does not  indicate the delay ya gotta get one and play with it.  My first concern is that my circuit isn't optimal, but since no one has really mentioned seeing an issue in passive values I'll assume it's pretty good.

The datasheet does not say but I do not think the PWM input directly modulates the current output.  Instead, the LDH-45 produces an average output current based on the PWM input.  This might be as simple as a capacitor across the output or LC filter on the PWM input.
I don't think that's the case.  Watching the LEDs connected to the output, they obviously switch at the same speed as the signal on the PWM input. I typically use the cameras in triggered mode but setting them to "free running" operates them as a standard video camera. They will trigger the strobe at the same frequency or "framerate" that they are set to operate at. The frequency is changing but not the duty cycle (Though I can manually set that as well). The driver's react exacltly as the input. If they were modulating the output with PWM that implies that they are accepting a varying frequency or width and modulating only the width of the output. It's a $10 LED driver, so there's not a whole lot of thinking going on inside.

I believe the issue with the drivers is the charge time. It looks like the output takes a split second to rise and fall like it's due to internal capacitance. Now that I think of it, plugging the LED's into the PCB board with no VIN attached , The LEDs will light to waste off some stored power. That could be due to the caps on the VIN side of my circuit as well. After discussing it here it seems like I'd either need to do something to improve the rise and fall of the driver output (or input?), or just switch the output externally to the driver with the SSR I mentioned above.
 

Offline nick_d

  • Regular Contributor
  • *
  • Posts: 120
Re: Faster switching speed from mosfets in existing circuit/PCB
« Reply #9 on: December 19, 2018, 06:50:36 am »
Interesting observations. Observation beats theory :) And your reasoning makes a lot of sense. I agree that switching the other side of the LED may be an option. I suggest to breadboard it first and make sure all is OK. By disconnecting the LED you are asking the constant curreny source to provide an infinite voltage. It should go smoothly in and out of voltage limiting, but you do not know for sure until you try.

Another suggestion is to use a different but similar LED driver if you can find one with a shutdown pin as you require.

Keep us informed how it goes.

cheers, Nick
 

Offline fsr

  • Regular Contributor
  • *
  • Posts: 169
  • Country: ar
Re: Faster switching speed from mosfets in existing circuit/PCB
« Reply #10 on: December 19, 2018, 09:38:12 pm »
It seems like the 2n7002 are controlled by what seems to be a long wire. Could the slowness to turn be caused by something like wire inductance?
 

Offline blazini36Topic starter

  • Regular Contributor
  • *
  • Posts: 72
  • Country: us
Re: Faster switching speed from mosfets in existing circuit/PCB
« Reply #11 on: December 20, 2018, 12:40:11 am »
It seems like the 2n7002 are controlled by what seems to be a long wire. Could the slowness to turn be caused by something like wire inductance?
I don't think so. The cable that the remote wire is in is about 7 or 8 feet long, it's a 22 or 24awg 4 wire shielded cable. The cable that runs the VOUT to the strobe is just as long going back to the same place (2 wire 18AWG). The thing is that they are the exact same cables that ran the remote and VOUT with the AC-DC Driver and the External SSR. Unless the length of the wires is affecting the Mosfet differently than it did the SSR, I'd say it's probably not the cables.

I will say that I don't expect the exposure to be as fast as it originally was at 30-100us depending on the amount of light diffusion used. The power output of the DC-DC drivers is lower so the exposure time will be longer due to less available light. I would say I'd expect it to be ~400us max. ~900us implies a delay, and the exposure trigger is slightly out of sync with the strobes flash.

I had to cut this clip short but it illustrates the necessity for the quick strobe switching, This is on the original LED Driver setup.
https://www.dropbox.com/s/66sjfddnhj5u0mg/Video%20Sep%2007%2C%209%2014%2016%20AM_0.mov?dl=0
 

Offline blazini36Topic starter

  • Regular Contributor
  • *
  • Posts: 72
  • Country: us
Re: Faster switching speed from mosfets in existing circuit/PCB
« Reply #12 on: December 20, 2018, 12:48:08 am »
Interesting observations. Observation beats theory :) And your reasoning makes a lot of sense. I agree that switching the other side of the LED may be an option. I suggest to breadboard it first and make sure all is OK. By disconnecting the LED you are asking the constant curreny source to provide an infinite voltage. It should go smoothly in and out of voltage limiting, but you do not know for sure until you try.

Another suggestion is to use a different but similar LED driver if you can find one with a shutdown pin as you require.

Keep us informed how it goes.

cheers, Nick

RECOM makes some DC-DC drivers that I thought about trying but it seems they are only step down. I'm trying to consolidate everything to being run off a single 24v power supply so I need boost drivers. I did order some SMD SSRs, I will just assemble another one of the prototype PCBs and scratch out and jump some traces to make the SSR work on the output as a test before I route a new PCB for the SSRs. My soldering skills are much better than my breadboarding, tends to be a huge mess of jumpers lol
 

Offline fsr

  • Regular Contributor
  • *
  • Posts: 169
  • Country: ar
Re: Faster switching speed from mosfets in existing circuit/PCB
« Reply #13 on: December 20, 2018, 12:40:11 pm »
It seems like the 2n7002 are controlled by what seems to be a long wire. Could the slowness to turn be caused by something like wire inductance?
I don't think so. The cable that the remote wire is in is about 7 or 8 feet long, it's a 22 or 24awg 4 wire shielded cable. The cable that runs the VOUT to the strobe is just as long going back to the same place (2 wire 18AWG). The thing is that they are the exact same cables that ran the remote and VOUT with the AC-DC Driver and the External SSR. Unless the length of the wires is affecting the Mosfet differently than it did the SSR, I'd say it's probably not the cables.

I will say that I don't expect the exposure to be as fast as it originally was at 30-100us depending on the amount of light diffusion used. The power output of the DC-DC drivers is lower so the exposure time will be longer due to less available light. I would say I'd expect it to be ~400us max. ~900us implies a delay, and the exposure trigger is slightly out of sync with the strobes flash.

I had to cut this clip short but it illustrates the necessity for the quick strobe switching, This is on the original LED Driver setup.
https://www.dropbox.com/s/66sjfddnhj5u0mg/Video%20Sep%2007%2C%209%2014%2016%20AM_0.mov?dl=0
Well, the SSR could react differently, because they're optocoupled, so that any driving circuit is internal to the SSR. With a mosfet directly connected, you depend on the driving circuit to charge/discharge the gate fast enough.
 

Offline ealex

  • Frequent Contributor
  • **
  • Posts: 321
  • Country: ro
Re: Faster switching speed from mosfets in existing circuit/PCB
« Reply #14 on: December 20, 2018, 02:40:44 pm »
i don't think the LED drivers you are using are intended for what you want.
I can't find in their datasheet anything related to timing, etc.
the PWM dimming input can have a RC filter inside the module, and that adds delays, and then you have the control loop delays / internal soft-start / etc - you have a "black box" with an unspecified behavior.

if you only use the leds as a strobe light, why don't you have a constant voltage power supply, use some properly sized resistors for current limiting and a N-FET + proper driver for a low side switch ?
 

Offline blazini36Topic starter

  • Regular Contributor
  • *
  • Posts: 72
  • Country: us
Re: Faster switching speed from mosfets in existing circuit/PCB
« Reply #15 on: December 20, 2018, 11:31:28 pm »
It seems like the 2n7002 are controlled by what seems to be a long wire. Could the slowness to turn be caused by something like wire inductance?
I don't think so. The cable that the remote wire is in is about 7 or 8 feet long, it's a 22 or 24awg 4 wire shielded cable. The cable that runs the VOUT to the strobe is just as long going back to the same place (2 wire 18AWG). The thing is that they are the exact same cables that ran the remote and VOUT with the AC-DC Driver and the External SSR. Unless the length of the wires is affecting the Mosfet differently than it did the SSR, I'd say it's probably not the cables.

I will say that I don't expect the exposure to be as fast as it originally was at 30-100us depending on the amount of light diffusion used. The power output of the DC-DC drivers is lower so the exposure time will be longer due to less available light. I would say I'd expect it to be ~400us max. ~900us implies a delay, and the exposure trigger is slightly out of sync with the strobes flash.

I had to cut this clip short but it illustrates the necessity for the quick strobe switching, This is on the original LED Driver setup.
https://www.dropbox.com/s/66sjfddnhj5u0mg/Video%20Sep%2007%2C%209%2014%2016%20AM_0.mov?dl=0
Well, the SSR could react differently, because they're optocoupled, so that any driving circuit is internal to the SSR. With a mosfet directly connected, you depend on the driving circuit to charge/discharge the gate fast enough.
True, hence the reason for the post as I'm not sure my pullup values and such were optimal to operate the mosfet of that this mosfet was the best choice.  Though I tried to do my research my limited component level electronics knowledge makes me unsure of my component selection. I have many other considerations for the whole design of this thing so it's best I try to make the best of the PCB as relocating the drivers themselves to shorten the wiring is not as practical or easy as it seems.
 

Offline blazini36Topic starter

  • Regular Contributor
  • *
  • Posts: 72
  • Country: us
Re: Faster switching speed from mosfets in existing circuit/PCB
« Reply #16 on: December 21, 2018, 12:47:51 am »
i don't think the LED drivers you are using are intended for what you want.
I can't find in their datasheet anything related to timing, etc.
the PWM dimming input can have a RC filter inside the module, and that adds delays, and then you have the control loop delays / internal soft-start / etc - you have a "black box" with an unspecified behavior.

if you only use the leds as a strobe light, why don't you have a constant voltage power supply, use some properly sized resistors for current limiting and a N-FET + proper driver for a low side switch ?
Well they're not exactly intended for this purpose, there is nothing on the common market that is. A LED driver/power supply manufacturer (Autec) suggested the Constant Current route to operate the LEDs in "burst" mode. It has actually worked flawlessly in the previous implementation, so much so that it had to be bumped down a bit. The built in on/off-PWM thing was a feature I had to try for simplicity and I just wanted to make sure my circuit was giving it the proper chance. To be apples to apples I have to use a switching device on the output of the LDH, as you said it could have a soft start of sorts but that's not an issue if the output is switched instead.

A CV power supply would have to hit at least 48v to compare to the LDH driver. That would add another AC-DC supply to the mix unless I did a boost supply on the board from 24v but I don't know how that would react to 2 separate loads loads. Besides that I have no idea how to use a gate driver IC. It took me long enough to figure out how to use the Mosfet in the current design.The opto will be easy to implement if I can't get the exposure times down in the current design. If that doesn't work I'll probably make a post asking how to use a gate driver lol.
 

Offline fsr

  • Regular Contributor
  • *
  • Posts: 169
  • Country: ar
Re: Faster switching speed from mosfets in existing circuit/PCB
« Reply #17 on: December 21, 2018, 04:17:01 pm »
It seems like the 2n7002 are controlled by what seems to be a long wire. Could the slowness to turn be caused by something like wire inductance?
I don't think so. The cable that the remote wire is in is about 7 or 8 feet long, it's a 22 or 24awg 4 wire shielded cable. The cable that runs the VOUT to the strobe is just as long going back to the same place (2 wire 18AWG). The thing is that they are the exact same cables that ran the remote and VOUT with the AC-DC Driver and the External SSR. Unless the length of the wires is affecting the Mosfet differently than it did the SSR, I'd say it's probably not the cables.

I will say that I don't expect the exposure to be as fast as it originally was at 30-100us depending on the amount of light diffusion used. The power output of the DC-DC drivers is lower so the exposure time will be longer due to less available light. I would say I'd expect it to be ~400us max. ~900us implies a delay, and the exposure trigger is slightly out of sync with the strobes flash.

I had to cut this clip short but it illustrates the necessity for the quick strobe switching, This is on the original LED Driver setup.
https://www.dropbox.com/s/66sjfddnhj5u0mg/Video%20Sep%2007%2C%209%2014%2016%20AM_0.mov?dl=0
Well, the SSR could react differently, because they're optocoupled, so that any driving circuit is internal to the SSR. With a mosfet directly connected, you depend on the driving circuit to charge/discharge the gate fast enough.
True, hence the reason for the post as I'm not sure my pullup values and such were optimal to operate the mosfet of that this mosfet was the best choice.  Though I tried to do my research my limited component level electronics knowledge makes me unsure of my component selection. I have many other considerations for the whole design of this thing so it's best I try to make the best of the PCB as relocating the drivers themselves to shorten the wiring is not as practical or easy as it seems.
But in this case, whatever is driving the mosfets is over the remote1/2 connections. That part isn't on the schematic. I'm no expert by any means, but i think that you need some kind of buffer that can supply enough current to charge the gate capacitance as fast as you need. Well, or at least you need to test if you need it. A scope on the gate should reveal how much does it take to rise.
 

Offline blazini36Topic starter

  • Regular Contributor
  • *
  • Posts: 72
  • Country: us
Re: Faster switching speed from mosfets in existing circuit/PCB
« Reply #18 on: December 21, 2018, 04:36:13 pm »
It seems like the 2n7002 are controlled by what seems to be a long wire. Could the slowness to turn be caused by something like wire inductance?
I don't think so. The cable that the remote wire is in is about 7 or 8 feet long, it's a 22 or 24awg 4 wire shielded cable. The cable that runs the VOUT to the strobe is just as long going back to the same place (2 wire 18AWG). The thing is that they are the exact same cables that ran the remote and VOUT with the AC-DC Driver and the External SSR. Unless the length of the wires is affecting the Mosfet differently than it did the SSR, I'd say it's probably not the cables.

I will say that I don't expect the exposure to be as fast as it originally was at 30-100us depending on the amount of light diffusion used. The power output of the DC-DC drivers is lower so the exposure time will be longer due to less available light. I would say I'd expect it to be ~400us max. ~900us implies a delay, and the exposure trigger is slightly out of sync with the strobes flash.

I had to cut this clip short but it illustrates the necessity for the quick strobe switching, This is on the original LED Driver setup.
https://www.dropbox.com/s/66sjfddnhj5u0mg/Video%20Sep%2007%2C%209%2014%2016%20AM_0.mov?dl=0
Well, the SSR could react differently, because they're optocoupled, so that any driving circuit is internal to the SSR. With a mosfet directly connected, you depend on the driving circuit to charge/discharge the gate fast enough.
True, hence the reason for the post as I'm not sure my pullup values and such were optimal to operate the mosfet of that this mosfet was the best choice.  Though I tried to do my research my limited component level electronics knowledge makes me unsure of my component selection. I have many other considerations for the whole design of this thing so it's best I try to make the best of the PCB as relocating the drivers themselves to shorten the wiring is not as practical or easy as it seems.
But in this case, whatever is driving the mosfets is over the remote1/2 connections. That part isn't on the schematic. I'm no expert by any means, but i think that you need some kind of buffer that can supply enough current to charge the gate capacitance as fast as you need. Well, or at least you need to test if you need it. A scope on the gate should reveal how much does it take to rise.

The remote is an open drain output from the camera, which basically sinks to ground. The pull-up on the board sets the potential. Whether I’ve done that correctly or not I’m not 100% sure of. I have a scope but I’m no scope wizard. I take it this would be adjusted with the pull-up resistor on the gate
 

Offline fsr

  • Regular Contributor
  • *
  • Posts: 169
  • Country: ar
Re: Faster switching speed from mosfets in existing circuit/PCB
« Reply #19 on: December 21, 2018, 09:37:42 pm »
It seems like the 2n7002 are controlled by what seems to be a long wire. Could the slowness to turn be caused by something like wire inductance?
I don't think so. The cable that the remote wire is in is about 7 or 8 feet long, it's a 22 or 24awg 4 wire shielded cable. The cable that runs the VOUT to the strobe is just as long going back to the same place (2 wire 18AWG). The thing is that they are the exact same cables that ran the remote and VOUT with the AC-DC Driver and the External SSR. Unless the length of the wires is affecting the Mosfet differently than it did the SSR, I'd say it's probably not the cables.

I will say that I don't expect the exposure to be as fast as it originally was at 30-100us depending on the amount of light diffusion used. The power output of the DC-DC drivers is lower so the exposure time will be longer due to less available light. I would say I'd expect it to be ~400us max. ~900us implies a delay, and the exposure trigger is slightly out of sync with the strobes flash.

I had to cut this clip short but it illustrates the necessity for the quick strobe switching, This is on the original LED Driver setup.
https://www.dropbox.com/s/66sjfddnhj5u0mg/Video%20Sep%2007%2C%209%2014%2016%20AM_0.mov?dl=0
Well, the SSR could react differently, because they're optocoupled, so that any driving circuit is internal to the SSR. With a mosfet directly connected, you depend on the driving circuit to charge/discharge the gate fast enough.
True, hence the reason for the post as I'm not sure my pullup values and such were optimal to operate the mosfet of that this mosfet was the best choice.  Though I tried to do my research my limited component level electronics knowledge makes me unsure of my component selection. I have many other considerations for the whole design of this thing so it's best I try to make the best of the PCB as relocating the drivers themselves to shorten the wiring is not as practical or easy as it seems.
But in this case, whatever is driving the mosfets is over the remote1/2 connections. That part isn't on the schematic. I'm no expert by any means, but i think that you need some kind of buffer that can supply enough current to charge the gate capacitance as fast as you need. Well, or at least you need to test if you need it. A scope on the gate should reveal how much does it take to rise.

The remote is an open drain output from the camera, which basically sinks to ground. The pull-up on the board sets the potential. Whether I’ve done that correctly or not I’m not 100% sure of. I have a scope but I’m no scope wizard. I take it this would be adjusted with the pull-up resistor on the gate

According to this: https://www.infineon.com/dgdl/mosfet.pdf?fileId=5546d462533600a4015357444e913f4f , on the section called "GATE CHARGE", the way to calculate the turn on time, is based on the gate charge. Take notice of the following paragraph:

The advantage of using gate charge is that the designer can easily calculate the amount of current required from the drive circuit to switch the device on in a desired length of time because Q = CV and I = C dv/dt, the Q = Time x current.  For example, a device with a gate charge of
20nC can be turned on in 20 μsec if 1ma is supplied to the gate or it can turn on in 20nsec if the gate current is increased to 1A.  These simple calculations would not have been possible with input capacitance values.


In this datasheet, they specify 223 pC of "Total Gate Charge", so let's suppose that you want to switch this in 10 us, then current = Q/time = 223 pC / 10 us = 22.3 uA . Doesn't seems difficult to turn on at all.

A simulation seems to agree in that you should have short times to turn the state of the mosfet (see attachment). Of course,

The fact that the LDH-45 has analog and a PWM dimming could mean that the PWM input is just turned into a voltage internally, as others have already mentioned. Also take notice of the voltage limits on the PWM and Analog dim inputs, that seem to be exactly the same, so that further suggests that they're related. And they don't specify the time to turn on. Also, take notice that there is a graph that shows "output current vs PWM dim input", so, it isn't turning the leds on and off with the PWM duty cycle, but just modifying the output current according to it.

It seems quite possible that the LED drivers themselves are just slower to turn on. You can check that with an oscilloscope. Tie one channel to the PWM dim input, and the other to Vout. That way you can check how much time it takes the output to rise after you enable the driver. You should be able to do that with a digital scope with the trigger in manual mode, using the PWM dim input as trigger. (VOUT- is the same than GND? otherwise you could have trouble trying to measure the output, see Dave's video on "how to not blow your scope").

 

Offline blazini36Topic starter

  • Regular Contributor
  • *
  • Posts: 72
  • Country: us
Re: Faster switching speed from mosfets in existing circuit/PCB
« Reply #20 on: December 23, 2018, 04:37:44 pm »

[/quote]

The remote is an open drain output from the camera, which basically sinks to ground. The pull-up on the board sets the potential. Whether I’ve done that correctly or not I’m not 100% sure of. I have a scope but I’m no scope wizard. I take it this would be adjusted with the pull-up resistor on the gate
[/quote]

According to this: https://www.infineon.com/dgdl/mosfet.pdf?fileId=5546d462533600a4015357444e913f4f , on the section called "GATE CHARGE", the way to calculate the turn on time, is based on the gate charge. Take notice of the following paragraph:

The advantage of using gate charge is that the designer can easily calculate the amount of current required from the drive circuit to switch the device on in a desired length of time because Q = CV and I = C dv/dt, the Q = Time x current.  For example, a device with a gate charge of
20nC can be turned on in 20 μsec if 1ma is supplied to the gate or it can turn on in 20nsec if the gate current is increased to 1A.  These simple calculations would not have been possible with input capacitance values.


In this datasheet, they specify 223 pC of "Total Gate Charge", so let's suppose that you want to switch this in 10 us, then current = Q/time = 223 pC / 10 us = 22.3 uA . Doesn't seems difficult to turn on at all.

A simulation seems to agree in that you should have short times to turn the state of the mosfet (see attachment). Of course,

The fact that the LDH-45 has analog and a PWM dimming could mean that the PWM input is just turned into a voltage internally, as others have already mentioned. Also take notice of the voltage limits on the PWM and Analog dim inputs, that seem to be exactly the same, so that further suggests that they're related. And they don't specify the time to turn on. Also, take notice that there is a graph that shows "output current vs PWM dim input", so, it isn't turning the leds on and off with the PWM duty cycle, but just modifying the output current according to it.

It seems quite possible that the LED drivers themselves are just slower to turn on. You can check that with an oscilloscope. Tie one channel to the PWM dim input, and the other to Vout. That way you can check how much time it takes the output to rise after you enable the driver. You should be able to do that with a digital scope with the trigger in manual mode, using the PWM dim input as trigger. (VOUT- is the same than GND? otherwise you could have trouble trying to measure the output, see Dave's video on "how to not blow your scope").

[/quote]

That's very informative and thanks for verifying the validity of the passives, I didn't quite get a chance to wrap my head around Spice yet. I haven't had a chance to play with the scope yet, I really only use it for verifying the switching of proximity sensors on tooth wheels and encoders and checking for noise, but I will get into it as you suggested if I continue on this route. This board as is seems more feasible now that I've worked the lighting a little better to accommodate the lower voltage output of the LDH compared to the AC-DC driver. I'm down to about 460us which is probably acceptable, my only small gripe is that there is a very slight change in light output between camera frames. Again it's something that wasn't seen on the AC setup but I'm sure it has to do with slight inconsistencies in the LDH driver while using it in this way. Not quite a dealbreaker yet though.

I did breadboard an SMD SSR to test the LDH for switching the output and this looks like a pretty big fail, which is probably why the PWM, On/off function exists in the first place. Unilike the AC-DC drivers the LDH drivers do not recover the output well at all while being switched so that's not an option. I did consider ealex's comment a little more but it still seems it would be an issue having to pull 48v DC out of something.......Then it dawned on me that I've been a bit narrowminded again because of all the different turns I've had to take with this whole project. I use 12 LEDs wired in series on a PCB I had fabbed based on Testing with the previous methods and that works well. Not sure Why I didn't figure sooner that I can achieve the same thing running them in a series/parallel combination at 24v. I modified one of the LED PCBs for this. I still like trying to switch them with an Opto for isolation so I'm going to play around with that and see how that turns out.
 

Offline fsr

  • Regular Contributor
  • *
  • Posts: 169
  • Country: ar
Re: Faster switching speed from mosfets in existing circuit/PCB
« Reply #21 on: December 23, 2018, 10:23:26 pm »
You're welcome. I hope you find a way to make the circuit work as expected.
Of course, if you get the led board voltage under you PSU's voltage, you don't need any DC-DC converter at all.
If the voltage sags with the rapid change of the load, you can always add some capacitance close to the switching element to act as reservoir.
If you use two or more led strips in parallel, remember to use one resistor per string.
« Last Edit: December 23, 2018, 10:26:32 pm by fsr »
 

Offline blazini36Topic starter

  • Regular Contributor
  • *
  • Posts: 72
  • Country: us
Re: Faster switching speed from mosfets in existing circuit/PCB
« Reply #22 on: December 24, 2018, 05:00:51 am »
You're welcome. I hope you find a way to make the circuit work as expected.
Of course, if you get the led board voltage under you PSU's voltage, you don't need any DC-DC converter at all.
If the voltage sags with the rapid change of the load, you can always add some capacitance close to the switching element to act as reservoir.
If you use two or more led strips in parallel, remember to use one resistor per string.

I bread boarded 1 "channel" of the schematic with the opto,  and it seems to work well with the modified LED Board. Spent most of the day routing new boards, There's plenty of space on the PCBs for extra caps if necessary.

 

Offline spec

  • Frequent Contributor
  • **
  • Posts: 833
  • Country: england
  • MALE
Re: Faster switching speed from mosfets in existing circuit/PCB
« Reply #23 on: December 25, 2018, 10:38:31 am »
UPDATE #1 2018_12_25

Hi blazini36

There needs to be 56R gate stopper resistors connected directly to the gate terminals of any MOSFETs. This is to eliminate parasitic oscillations (especially with the long lead from the camera strobe output connected directly to the gates of the MOSFETs).

In general, you also need some decoupling using 100nF X7R dialectic ceramic capacitors (not surface mount, but the larger through hole type)

I take it that stobe 1 and strobe 2 inputs are from separate cameras and that a single camera strobe line is not feeding both NMOSFET gates?

Can you provide information on the camera strobe output: current sinking capability, resistance to 0V/max voltage when sinking current?

« Last Edit: December 25, 2018, 09:06:23 pm by spec »
 

Offline blazini36Topic starter

  • Regular Contributor
  • *
  • Posts: 72
  • Country: us
Re: Faster switching speed from mosfets in existing circuit/PCB
« Reply #24 on: December 25, 2018, 04:19:04 pm »
Hi blazini36

I haven't looked at your circuit in detail, but the first thing that stands out is that R2/R5 and R6/R7 have very high resistances (10k) and will slug the operation of the NMOSFET (2N7002), which has relatively high parasitic capacitances. A BJT, would be much better in this position.  The 2N7002 wide gate threshold range of 1V to 2V5 also complicates the issue. Depending on the current capability of the camera strobe switch, I would suggest lowering the values of R2/R5 to 1k or less if possible. Also, I would suggest removing the LED from across R2/R5 for the initial investigative stage at least.  And R6/R7 should be reduced to 1K. These modification should speed the switching on and off of the of the NMOSFET by 10 times.

There also needs to be 56R gate stopper resistors connected directly to the gate terminal of the NMOSFETs (between the junction of the camera strobe input and R2/R5 and the NMOSFET gates). This is to eliminate parasitic oscillations (especially with the long lead from the camera strobe output connected directly to the gates of the NMOSFETs) which may be another cause for effective delay in the NMOSFET switching.

In general you also need some decoupling using 100nF X7R dialectic ceramic capacitors (not surface mount, but the larger through hole type)

I take it that stobe 1 and strobe 2 inputs are from separate cameras and that a single camera strobe line is not feeding both NMOSFET gates?

I am not saying that speeding up the switching of the NMOSFETs will solve your problem but, if you are interested in pursuing this approach, please give details of the camera strobe switch, especially its current sinking capabilities and forward voltage/resistance.

https://www.diodes.com/assets/Datasheets/2N7002.pdf

Hi spec,

You're referring to the original schematic I posted right? The first iteration of this which I had someone help me with used an MMBT3904 rather than an N-fet. This only seemed to work with a diode and resistor between the remote and the base of the BJT. There were a bunch of other problems with that board so I just scrapped it and came up with the schematic I posted. The N-fet was pretty much because that's what I had laying around in a package I could bread board and it just worked.

Strobe 1 and 2 are completely separate for handling the 2 camera's individually. I'm attaching a page from the datasheet for the I/O on the camera's. Every one of these types of cameras I've seen use the same type of strobe output (what I call remote). Unfortunately they all seem to have lousy documentation on it but it states they can handle 200ma. Pin 10 is the output and through testing it seems that pin 12 is the reference.  I leave pin 12 connected to the common of the 24v power supply and my FPGA board sends 24v+ to pin 11 to trigger the camera. The camera internally handles its output on pin 10 as it can be adjusted through software. It's confusing (to me at least) the way it's listed but this is how I find it to work.

The adjustments you mentioned are pretty easy to implement, especially the 1K resistor as I can just swap that out on the current board. As I said a couple of posts ago it it seems I've been a bit narrow sighted using the Meanwell drivers rather than just rearranging the LEDs from 12 in series to 6x6 series parallel and just feeding them with 24v and sticking some high current resistors on the LED board. That approach just lets me use a high current switching device (an opto-mosfet in this case) and makes the boards much smaller. I've got the opto boards out for fab to test. If you wanna take a look at the last schematic I posted maybe you'll have some suggestions. Hopefully they'll only be component values as I'd have to fab new boards (again lol). The plus side to ditching the Meanwell drivers is that I can keep the boards under 100mm x 100mm and they're super cheap to get made.
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf