DMOSFET for High Side drive

Go To Last Post
12 posts / 0 new
Author
Message
#1
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

I have an application that requires 8 channels of 1v to 12v at about 30 ma max.  This voltage is "settable" and varies slowly over time in response to process conditions.  This must be "high side" drive.

 

My attempt was using a  TLC59116 (Datasheet) to drive a TBD62783AFNG  8-channel DMOSFET gate array (Datasheet) but I discovered that the turn-off spec on the TBD62783AFNG was 2 usec and the (fixed) PWM frequency of the TLC59116 (97-kHz) meant that I had very little usable range of the PWM signal.

 

I selected the TLC59116 because the (ATMega1284P) application firmware already has support for this chip that is used elsewhere on the board.

 

I don't have any pins left over to 'roll my own' PWM.

 

I could switch out the TLC59116 for a PCA9685 (Datasheet) which has a PWM frequency that is settable from 24 Hz to 1526 Hz but I would like to avoid that due to requiring an additional support module in the firmware.

 

I looked for other gate arrays in the 'TBD' sequence but the others were either no faster or were for 'low side' drive applications which won't work here.

 

I would like to keep this to a two-chip solution due to board real-estate.

 

Does anyone know of a faster, high-side gate array I could use and retain the TLC59116?

 

It's ok if it's inverting or not because I can account for that in my PWM setting.

 

Thanks in advance for any insight you might offer ...

 

Chuck

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

requires 8 channels of 1v to 12v at about 30 ma max.

 

The first chip you mention is a sink driver (gnd side driven), so it doesn't even seem to fit your statement (it is not a high side driver).  Then you jump in to pwm??/

 

Why not simply use a small 8 output dac?  You can easily wire the outputs to emitter followers & drive hundreds of milliamps per channel (not very efficiently, since it is a linear drive).  But at 30ma, not goona be a lot of heat or wasted power per chan.

 

Of course if the output is moving slowly, you can easily set up a timer irq (say 25KHz) & use some counter variables  to toggle individual i/o pins as a multichannel pwm...then simply RC filter down to  maybe 10Hz   & buffer everything using only 2 quad opamps set for a gain of 3 (to get from 5V to 15v output range).

 

 

I don't have any pins left over to 'roll my own' PWM.

 

 

Not sure what you have free exactly....maybe just stick with trying a DAC. 

When in the dark remember-the future looks brighter than ever.   I look forward to being able to predict the future!

Last Edited: Wed. Aug 14, 2019 - 02:33 AM
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

ChuckH wrote:

I have an application that requires 8 channels of 1v to 12v at about 30 ma max.  This voltage is "settable" and varies slowly over time in response to process conditions.  This must be "high side" drive.

 

Does anyone know of a faster, high-side gate array I could use and retain the TLC59116?

 

That sort of Hi-Side driver, is not designed for speed.

What are you driving with this - does it need very fast PWM ?

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

avrcandies wrote:

The first chip you mention is a sink driver (gnd side driven), so it doesn't even seem to fit your statement (it is not a high side driver). 

 

Yes, it's used to drive the DMOSFET from the low side but the DMOSFET output is high side..

 

avrcandies wrote:

Why not simply use a small 8 output dac?  ....

 

Had not considered that - something like a DAC5578SPW has an I2C interface like the LED driver I'm using now so no more pins needed.  

 

One reason I was using the PWM was its efficiency.

 

I may see if I can find an I2C high-side PWM LED driver but I fear they are all low-side, but, then again, I can use the other low-side LED driver I mentioned that has the selectable PWM frequency so that I can live with the slower response of the DMOSFET.

 

Chuck

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Who-me wrote:

That sort of Hi-Side driver, is not designed for speed.

What are you driving with this - does it need very fast PWM ?

 

This is for setting the rail bias voltage for automatic train signals for ride-on railroads (My Web Site).

 

I need to vary the track bias voltage as environmental conditions change.  Wet track has more leakage lowering the track voltage and reducing the signal-to-noise ratio which can lead to "false occupancy" signals.  The track voltage is kept to a minimum for several reasons, among them: reduce rail and tie screw corrosion and keep power consumption down.

 

With my current signal controller this is done by manually selecting a resistor that works in the wettest conditions but this is tedious for the customer and causes excess power usage and corrosion.

 

The new version of my signal controller will do this automatically by monitoring the track voltages and increasing the bias voltage if it falls too low.  It will reduce the bias voltage as the track dries out and the (unoccupied) track voltage rises.

 

So the bias voltage I generate changes very slowly.

 

The PWM speed needs to just be fast enough that it can be integrated by reasonably small capacitors (RC filter) to keep 'ripple' from being detected as a track shunt (occupied rail).  The rail input goes through a conditioning network and then to an ADC input on the ATMega1284P which then uses an algorithm to eliminate the effect of wet/dry track, debris on the track (noise), etc.  There is more to it but that's the thumbnail version ...

 

At the moment it looks like my best move is to change to keep the DMOSFET and switch the FET driver to something like the PCA9685 with a slower PWM frequency but I'm still evaluating alternatives.

 

Chuck

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Yes, it's used to drive the DMOSFET from the low side but the DMOSFET output is high side.

Ok, now makes more sense.  

 

A dac with a buffer (see below) is your best bet...generate any voltage you want on the high side (sourcing).  If you find enough spare pins & slow voltage is ok, you can make multipwms & filter & buffer (basically save a dac chip).    

Assume roughly 12V worst case drop   12x0.03*8= 2.88 Watt...that's a lot of heat.  If Vcc=15V, even worse. This will require some power components at the final stage (if using linear output).  A switcher is more efficient, but 8 of them is "blah"in terms of parts & cost.

 

Could tie the dac into control (adj) pin of 8 small (or 4 dual) linear regulators..the dac then tweaks their output voltages. Let them take the heat & be dac controlled....they also are thermal/short protected.

 

 

 

When in the dark remember-the future looks brighter than ever.   I look forward to being able to predict the future!

Last Edited: Wed. Aug 14, 2019 - 05:44 AM
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

avrcandies wrote:

A dac with a buffer (see below) is your best bet...generate any voltage you want on the high side (sourcing).  If you find enough spare pins & slow voltage is ok, you can make multipwms & filter & buffer (basically save a dac chip).    

Assume roughly 12V worst case drop   12x0.03*8= 2.88 Watt...that's a lot of heat.  If Vcc=15V, even worse. This will require some power components at the final stage (if using linear output).  A switcher is more efficient, but 8 of them is "blah"in terms of parts & cost.

 

Could tie the dac into control (adj) pin of 8 small (or 4 dual) linear regulators..the dac then tweaks their output voltages. Let them take the heat & be dac controlled....they also are thermal/short protected.

 

I do not have any pins for generating my own PWM.  I use the I2C bus to drive the PWM driver.

 

I want to stay with PWM because power consumption is an issue.  One installation currently has power & data bus runs >2,000 feet, 28v @ 1.2A to power all 19 signal controllers, 60 LED signal heads, and 100 sampled tracks.

 

I may have found something that will allow me to stay with the PWM driver that is currently on the board.

 

BSS84 P-CHANNEL ENHANCEMENT MODE MOSFET  (datasheet)

 

in place of the 8-channel MOSFET I have now.

 

Anyone have other P-Channel FETs that have a voltage rating ~50v, a current capacity of about 20 ma and turn-on/turn-off times in the sub-us range?  (I'm new to FET selection ...)

 

The operating voltage is at 12v but I want the margin to (along with other surge suppression components) survive nearby lightning, etc.

 

Not worried about short-circuit protection because there is a 1k ballast resistor in the circuit.

 

Chuck

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

BSS84  ---I was looking to see how many milliohms this fet is rated at  (100 milliohms these days is rather high, unless rated for hundreds of volts)...but this fet is rated as 10000 milliohms.....hard to imagine one rated for more!! What a terrible part today, its the equiv of a 16 byte memory.  Spend 10 cents and get one creating less loss.  Of course, its limited drive might mean it won't deliver a punch under short circuits (instead the fet will get hot and act like a 10 ohm resistor).

 

Now if you get fets as part of a packaged chip functionality, those internal fets are often annoyingly high Rds.

 

here's some that might be good for your purpose:   Note that the Rds  will affect the PWM voltage accuracy (not sure how tight is needed)

https://www.renesas.com/us/en/www/doc/datasheet/el7457.pdf

http://ww1.microchip.com/downloads/en/DeviceDoc/MCP14A0303_4_5-Data-Sheet-20006046A.pdf

 

When in the dark remember-the future looks brighter than ever.   I look forward to being able to predict the future!

Last Edited: Mon. Aug 19, 2019 - 03:14 AM
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

avrcandies wrote:

here's some that might be good for your purpose:   Note that the Rds  will affect the PWM voltage accuracy (not sure how tight is needed)

https://www.renesas.com/us/en/www/doc/datasheet/el7457.pdf

http://ww1.microchip.com/downloads/en/DeviceDoc/MCP14A0303_4_5-Data-Sheet-20006046A.pdf

 

(Since I'm new at FET selection) I missed the 10 ohm on resistance.  But, since, in my case, there is a 1k current-limiting resistor between the 'driver' and the load is this really a concern?

 

Assuming 12v and a perfect (0-ohm) FET the max current is 12 ma.  With the 10-ohm FET we are talking max current of 12/1010 = 11.88 ma and 11.88 ma * 10 = 0.1188 volt drop across the FET.  Did I miss something?

 

The PWM voltage accuracy is not a concern in this application and I can adapt to a reasonable fluctuation/inaccuracy.

 

The parts you mention have an 18v rating which does not leave a lot of headroom when there may be shoot-through in the surge suppression circuit between the 1k R and the 'outside world'.  The 50v rating of the part I mentioned gives me more comfort.

 

Also, these parts appear to be pull-up/pull-down where I only want pull-up because, in my application, it basically comes down to the fact that I am measuring the load current/resistance by monitoring the voltage at the junction of the output side of the 1k resistor (also a small cap to Gnd, thus RC) - i.e.: the 'load' voltage.

 

Chuck

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

ChuckH wrote:
(Since I'm new at FET selection) I missed the 10 ohm on resistance.  But, since, in my case, there is a 1k current-limiting resistor between the 'driver' and the load is this really a concern?

 

Depends also on how fast you want to slow things down. 470ohm could work between the pin and the FET down to 22ohm..

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

18v rating which does not leave a lot of headroom when there may be shoot-through in the surge suppression circuit between

That's true, though the fet will act like a zener if overvoltaged & prob ok if the source of the overvoltage is low powerlimited.    There's plenty of fets that have low ohms, like 0.1 & high voltage...the your pwm accuracy is not mushed.

When in the dark remember-the future looks brighter than ever.   I look forward to being able to predict the future!

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Chuck, how many pins from the AtMega1284 can you use to control this 8 Channels High Side Inverted DAC from 1 to 12V at 30mA max?

 

The fact that you will use no more than 30mA doesn't mean you can use very small mosfet transistors, mostly if you are connecting this to the external environment world, like train rails.

Whenever you touch external world, mostly exposed long metals, your electronics could be fried very easy by any eletro discharge nearby, even a 500 yards lightning can damage your small transistors.

 

I would think to produce this 8 channels "power supply" much more sturdy and tough, able to provide 5A or more, even that your actual power supply that feeds the 8 drivers could not provide this current, but the transistors probably would need to deliver 10 times that current, say 300 to 500mA each, resistors to ground, with other protections like varistors for low voltage (20 to 30V) and such.  Also, I would NOT think to use MosFets on this application, they are too much fragile to discharges, you can rethink and use bipolars BJTs like TIP136 (PNP 8A 70V Darlington Beta 1000~15000, TO220).

 

You can choose any PNP from this list, considering 3A or higher

https://www.futurlec.com/TransPowerSpecTIP.shtml

 

Or the BD series, like the BD436 TO126 (I think it is plastic all over), 4A 32V Beta 40~140, from this table:

https://www.futurlec.com/TransPowerSpec.shtml

 

Both above runs up to 1MHz with no problem. 

 

I would not touch the external world without a 100mA fast fuse in series with the external wire, and two inverted diodes to ground (diode anode) and +VCC (diode catode), anything strange the fuse will open and isolate the wire.

 

At this point I would not worry about the internal resistance of the transistor, since your maximum current is very low (30mA), and even so, if you need, an output voltage feedback could read and regulate the voltage you need at ouput.

 

Also, using PWM is always a good idea to save power, and use smaller components, but, again, 30mA maximum output current is almost nothing to any power transistor to dissipate the extra voltage from the power supply.  Lets imagine your general power supply is 15V driving this 8 channel "power supplies", if driving 1.5V at 30mA (almost worst case), the linear regulator transistor would dissipate the extra 13.5V @ 30mA = 405mW... that is not that bad, a small heatsink will solve the problem (a single piece of aluminum bar can hold the 8 transistors at once), and you will not need any PWM, just need to produce the relative individual voltage to control the channels.

 

There are many ways to generate those 8x relative analog voltage from your AVR, and this is why I asked you how many pins from the AtMega1284 you have available for that.

One easy solution is using an extra Atmega328, communicating with your AtMega1284, receiving a string of 8 bytes, each byte from 0~255 meaning a voltage from 1.25~12V at the transistors output.

The AtMega328 can generate the 8 PWM output, easily filtered and drive a mirrored small NPN to ground, driving the base of the upside BDxxx power transistors.  The voltage feedback can also be read by 8x ADCs of the 328 and adjust the output to reach perfectly the voltage you want at the output.  This will work *EXACTLY* like a digital controlled power supply with a upside PNP driver.

 

The nice thing of this solution is using just one pin of the AtMega1284 to "talk" with the Atmega328 every 100ms or so.

Even that the 328 has 6x PWM native, I would use 8x PWM output generated by software (assembly is recommended), very very easy to do it.

Just a 8 bits incremental counter (0~255) every 0.1ms, comparing the counter value to the 8 other registers having the value 0~255 for the individual PWMs, when the counter value is higher than the individual PWM registers, zero the relative bit in an "output" port, when the counter reach value 255 just flip up all 8 bits of the output port to the transistors... voilá, a 10kHz PWM 8 channels, with less than 20 assembly instructions.

 

You can even try to use such PWM to drive directly the BDxxx power PNPs via a small NPN emitter grounded, collector resistor to base of BDxxx, having PWM at the voltage output, needing filtering, avoiding to make it linear, not needing any heatsink.

 

I use to project power supplies, both linear and switching, and I see your application just 8 individual power supplies controlled by an AVR.

 

Wagner Lipnharski
Orlando Florida USA