Author Topic: Have I gone mad?  (Read 17008 times)

0 Members and 1 Guest are viewing this topic.

Offline king.osloTopic starter

  • Frequent Contributor
  • **
  • Posts: 432
  • Country: no
Have I gone mad?
« on: April 29, 2012, 01:10:56 am »
Hello there,

Until now I was certain this was true:

P-channel MOSFET conducts when the voltage at the gate is more negative than the threshold voltage relative to the source.
N-channel MOSFET conducts when the voltage at the gate is more positive than the threshold voltage relative to the source.

Is this true?

I am sorry to ask such a fundamental question, but LTspice seems to be suggesting I am wrong (see the attachment). If the above was true, I would expected the green voltage to switch between 172V and 0V @ 25% duty cycle.

Thanks.M


EDIT: What I actually want is the 172V at 75% duty cycle!
« Last Edit: April 29, 2012, 11:31:12 am by king.oslo »
 

Online IanB

  • Super Contributor
  • ***
  • Posts: 11885
  • Country: us
Re: Have I gone mad?
« Reply #1 on: April 29, 2012, 01:15:06 am »
I have only a sketchy knowledge of FETs, but I wonder, shouldn't we consider source in one case and drain in the other?
 

Offline amspire

  • Super Contributor
  • ***
  • Posts: 3802
  • Country: au
Re: Have I gone mad?
« Reply #2 on: April 29, 2012, 01:44:37 am »
You have two source followers which are non-inverting and a bit less then unity gain,  it is working as it should.

My thinking is a relay may be the easiest solution particularly if you can arrange for it to switch off when the regulator falls below 14V, or the voltage before the regulator drops to less then 17V. The reason I like a relay is the relay can short out the voltage across the divider, so there is zero volts - a second contact can switch off the 172V as well.

An alternative suggestion is to use a 3 position 2 pole make-before-break switch for an off-standby-on switch.  In the off mode,the input power is off and the divider is shorted. In the standby mode, the input mains is on, bit the divider is still shorted.  In the on mode, then mains is on, and the divider is unshorted. Becuae you have to switch through the standby position, it ensures the power up-power down sequence is correct.

Richard.

Edit: what I think you intended was the P-channel mosfet with a 10K resistor gate to source. Then a N-channel mosfet or NPN transistor with a 180K resistor between the drain/collector and the gate of the p-channel mosfet. Source/emitter to 0v. You then add a simple zener circuit to turn the n-channel/npn on when the voltage of the regulator is over 12V.
« Last Edit: April 29, 2012, 02:23:38 am by amspire »
 

Offline Kremmen

  • Super Contributor
  • ***
  • Posts: 1289
  • Country: fi
Re: Have I gone mad?
« Reply #3 on: April 29, 2012, 07:14:02 am »
Hello there,

Until now I was certain this was true:

P-channel MOSFET conducts when the voltage at the gate is more negative than the threshold voltage relative to the source.
N-channel MOSFET conducts when the voltage at the gate is more positive than the threshold voltage relative to the source.

Is this true?

I am sorry to ask such a fundamental question, but LTspice seems to be suggesting I am wrong (see the attachment). If the above was true, I would expected the green voltage to switch between 172V and 0V @ 25% duty cycle.

Thanks.M
For starters, there are N channel and P channel devices and both come in 2 kinds: the enhancement mode and the depletion mode transfer characteristic. So you have 4 fundamentally different kinds of MOSFET. I took the liberty of attaching a picture from the 3rd ed. of Physics of Semiconductor Devices by Sze and Kwok, hopefully fair use covers this. If not i will of course remove it. Perhaps that picture clarifies the differences.

Then, your circuit simply has the problem that halfway through you switch the common reference potential from source - to source +. This can be done but it has pitfalls and your circuit falls into one of them. M1 contols the gate of M2 in phase with the clock pulse, switching it between - and circa 11 volts or so. Actually you don't really need the fet because the clock source already does the same thing. Now for M2 the reference (node where the source pin is connected) is the + pole of the voltage source. So nominally the source voltage is 0 and the drain voltage is -172 V. This would be the way to orient a P channel fet voltages. Now the idea of this circuit is that M1 produces voltages alternating between 0V and some -15 V to the gate of M2, in M2 reference orientation above. Unfortunately this is not what happens in practice. M1 only swings some 11V above its own reference which is the -172 V rail in M2 reference frame. In M2 reference it is between -161 and -172 volts. So the "output" (i.e. M2 source) swings just enough to maintain the fet in a conducting state, the voltage keeping a constant difference to that applied on the gate.
You can confirm this by checking the M2 gate voltage in LTspice. You will see that it keeps a constant voltage difference against M2 source. Every time the pulse source and M1 switch, there will be a current spike in M2 gate circuit as the charge rebalances due to the change in source voltage.
This circuit can easily be modified to do what you want. Just take M2 gate voltage from M1 drain and switch the resistor from M1 source to its drain. The output phase of course changes around but that sould not be a major issue.
« Last Edit: April 29, 2012, 11:04:14 am by Kremmen »
Nothing sings like a kilovolt.
Dr W. Bishop
 

Offline Psi

  • Super Contributor
  • ***
  • Posts: 9946
  • Country: nz
Re: Have I gone mad?
« Reply #4 on: April 29, 2012, 08:36:30 am »
Hm.. This is something i wasn't aware of, and probably should have been.

Learn something everyday i guess.
Greek letter 'Psi' (not Pounds per Square Inch)
 

Online ejeffrey

  • Super Contributor
  • ***
  • Posts: 3718
  • Country: us
Re: Have I gone mad?
« Reply #5 on: April 29, 2012, 10:00:37 am »
You will probably never see a depletion mode MOSFET.  All jfets are depletion mode because an enhancement mode jfet would require forward biasing the junction.  Of course, discrete jfets are not that common these days either.
 

Offline king.osloTopic starter

  • Frequent Contributor
  • **
  • Posts: 432
  • Country: no
Re: Have I gone mad?
« Reply #6 on: April 29, 2012, 11:07:32 am »
You will probably never see a depletion mode MOSFET.

Why? Are they not useful?M
 

Offline king.osloTopic starter

  • Frequent Contributor
  • **
  • Posts: 432
  • Country: no
Re: Have I gone mad?
« Reply #7 on: April 29, 2012, 11:53:30 am »
Richard, I had no success with your suggestion :( I attached a screenshot to show what I did.

PS Richard: I thought I would use this circuit to instantly empty the caps on the 100V reference whenever the supply for the chopper opamp is off, so that I get no voltage @ power on and no latched up. I have had a few 170V shocks whilst prototyping because I forgot the stored energy in the caps. This will sort me out, right?


Kremmen, your suggestion worked as I expected, thanks, however, I made the mistake to write that I want the 172V square wave to have duty cycle 25%. What I really want is the square wave at 75%.

That can be done with a single n-ch depletion transistor, or two n-ch enhancement trasistors, right?

I am trying to find a depletion mode nmos on farnell. How do I search for it? Is that an nmos with a negative VGS?

And thanks for the diagram of the fets. Excellent stuff.M
« Last Edit: April 29, 2012, 12:00:19 pm by king.oslo »
 

Offline amspire

  • Super Contributor
  • ***
  • Posts: 3802
  • Country: au
Re: Have I gone mad?
« Reply #8 on: April 29, 2012, 12:07:00 pm »
Those spice results are perfect. It is working, except for the fact you have shorted the p-channel to 0v, instead of connecting it to a 22K resistor. I assumed you wanted the p-channel to be a switch before the 22K resistor. Perhaps your idea was something different.

Richard.

 

Offline SeanB

  • Super Contributor
  • ***
  • Posts: 16283
  • Country: za
Re: Have I gone mad?
« Reply #9 on: April 29, 2012, 12:11:48 pm »
Too complex. Lose the resistors and M2, and if the pulse polarity is wrong then use a single CMOS gate as an inverter, driven from the 15v supply.

If all you want is to discharge the capacitors what is wrong with a bleed resistor? If you want fast discharge use a power rail of lower voltage and place a relay across it, with the bleed resistor connected to the NC contact.

Depletion mode transistors are both rare and low voltage RF devices. Add to that expensive and easy to blow up.
 

Offline king.osloTopic starter

  • Frequent Contributor
  • **
  • Posts: 432
  • Country: no
Re: Have I gone mad?
« Reply #10 on: April 29, 2012, 12:16:43 pm »
Perhaps your idea was something different.

Yes, I probably have to work on my communication. As months have passed, I have learnt to trust you knowledge of electronics, Richard. Thanks for taking the time to help me :)



Too complex. Lose the resistors and M2, and if the pulse polarity is wrong then use a single CMOS gate as an inverter, driven from the 15v supply.

If all you want is to discharge the capacitors what is wrong with a bleed resistor? If you want fast discharge use a power rail of lower voltage and place a relay across it, with the bleed resistor connected to the NC contact.

Thanks SeanB. Cannot do the bleed resistors. It will screw up the circuit. Perhaps I can do as you suggested or use two NFETs.
 

Offline SeanB

  • Super Contributor
  • ***
  • Posts: 16283
  • Country: za
Re: Have I gone mad?
« Reply #11 on: April 29, 2012, 12:31:20 pm »
A relay would be easiest, i am looking at a tiny EBD26/2-28 relay in a board here, a tiny ( size of a 8 pin cerdip on edge) relay rated for 115V0.5A with DPDT contacts, and a 28V coil. Sits in a sandwich made from 5 PCB's bolted together.
 

Offline king.osloTopic starter

  • Frequent Contributor
  • **
  • Posts: 432
  • Country: no
Re: Have I gone mad?
« Reply #12 on: April 29, 2012, 12:39:55 pm »
A relay would be easiest, i am looking at a tiny EBD26/2-28 relay in a board here, a tiny ( size of a 8 pin cerdip on edge) relay rated for 115V0.5A with DPDT contacts, and a 28V coil. Sits in a sandwich made from 5 PCB's bolted together.

Sounds tiny Sean, but the problem is that I have solid state fetish  :-\

Check out the attachment. For the sake of learning: As you can see, the voltage drops to 3V at drain of M2, not 0V. Is this because because the M2-resistance is too high?
« Last Edit: April 29, 2012, 12:49:38 pm by king.oslo »
 

Offline SeanB

  • Super Contributor
  • ***
  • Posts: 16283
  • Country: za
Re: Have I gone mad?
« Reply #13 on: April 29, 2012, 12:59:47 pm »
Move R1 to the other side of R2 and rerun the sim.........
 

Offline king.osloTopic starter

  • Frequent Contributor
  • **
  • Posts: 432
  • Country: no
Re: Have I gone mad?
« Reply #14 on: April 29, 2012, 01:10:46 pm »
SeanB, you are right.

But dang, this doesn't work. I am loosing current through R1 when the 172V rail is high. This will screw up accuracy.M
 

Offline SeanB

  • Super Contributor
  • ***
  • Posts: 16283
  • Country: za
Re: Have I gone mad?
« Reply #15 on: April 29, 2012, 01:21:48 pm »
Make R1 1M or more, and add a resistor of 10% of that value across the drive transistor. With a Mosfet you can go very high in value, provided you use a device so that the gate capacitance is not too much of a low pass filter at the frequency that you are switching, and you are prepared to live with the slow switching ( relative to the nanosecond you achieve with a low impedance drive) and slightly increased dissipation. As you will probably be using a high voltage device select one with a low maximum current ( reduces size of die and thus capacitance) and the highest voltage you can get easily in a TO220 housing ( will easily handle the switching power without a heatsink needing to be added other than either mounting vertically or bolting to a pad on the board). That will reduce bleed current, and not make the edges too slow.
 

Offline king.osloTopic starter

  • Frequent Contributor
  • **
  • Posts: 432
  • Country: no
Re: Have I gone mad?
« Reply #16 on: April 29, 2012, 01:44:28 pm »
Make R1 1M or more, and add a resistor of 10% of that value across the drive transistor. With a Mosfet you can go very high in value, provided you use a device so that the gate capacitance is not too much of a low pass filter at the frequency that you are switching, and you are prepared to live with the slow switching ( relative to the nanosecond you achieve with a low impedance drive) and slightly increased dissipation. As you will probably be using a high voltage device select one with a low maximum current ( reduces size of die and thus capacitance) and the highest voltage you can get easily in a TO220 housing ( will easily handle the switching power without a heatsink needing to be added other than either mounting vertically or bolting to a pad on the board). That will reduce bleed current, and not make the edges too slow.

The problem with this circuit is that it draws some current through the resistors whilst the power on.M
 

Offline SeanB

  • Super Contributor
  • ***
  • Posts: 16283
  • Country: za
Re: Have I gone mad?
« Reply #17 on: April 29, 2012, 01:53:26 pm »
Then the power for the first unit will have to come from another rail, with a voltage of between 10 and 20V using a 10k or so resistor. 15V will ensure the mosfet turns on fully, but a lower voltage will work if the device is above the threshold voltage at that gate voltage, at the expense of having a higher on resistance and a temperature dependance.
 

Offline amspire

  • Super Contributor
  • ***
  • Posts: 3802
  • Country: au
Re: Have I gone mad?
« Reply #18 on: April 29, 2012, 02:26:46 pm »
I just went back and looked at the circuit that need the changes to stop the latchup:
https://www.eevblog.com/forum/projects-designs-and-technical-stuff/100vdc-reference-circuit/msg96357/#msg96357

The problem with any electronic switch to short out the output capacitor is that the capacitor needs to be shorted before there is any voltage - and as the voltage first starts to rise, there is no voltage to turn this switch on.

So my solution is to use the circuit I suggested as a switch before the 22K resistor, and I would remove the output capacitor - it should work totally fine without it, and removing the capacitor means that when the ouput voltage is switched off, then there is no energy strage device left past the switch, so nothing to cause latchup.

So I would take this circuit:

https://www.eevblog.com/forum/beginners/have-i-gone-mad/?action=dlattach;attach=23764

I would open up the drain of M2 and remove the 1K R2. I would then use this circuit as a switch before the 22K resistor with the M2 source going to 172V and the drain going to the 22K.  This now means the 172V has to be switched ON rather then a short switched off.

A side reason why I would remove the output capacitor is that the 1K resistor was added for stability - to add a zero on the output to guarantee a high stability. The trouble is that with the 1K, any leakage in a capacitor directly on the output will cause an error, and since it is not needed, it may as well go.

Richard.
 

Offline king.osloTopic starter

  • Frequent Contributor
  • **
  • Posts: 432
  • Country: no
Re: Have I gone mad?
« Reply #19 on: April 29, 2012, 04:02:47 pm »
The problem with any electronic switch to short out the output capacitor is that the capacitor needs to be shorted before there is any voltage - and as the voltage first starts to rise, there is no voltage to turn this switch on.

I do not think this will be a problem. The 170V rail comes from a seperate transformer than the +-15V rails.


I would open up the drain of M2 and remove the 1K R2. I would then use this circuit as a switch before the 22K resistor with the M2 source going to 172V and the drain going to the 22K.  This now means the 172V has to be switched ON rather then a short switched off.


You are advising me to:

- lose the output cap. OK.
- Not to drain the capacitors when the power switches off, rather blocking the output from the caps so that the voltage cannot reach the opamp? Did I understand correctly?

M
 

Offline amspire

  • Super Contributor
  • ***
  • Posts: 3802
  • Country: au
Re: Have I gone mad?
« Reply #20 on: April 29, 2012, 10:27:29 pm »
Yes, you have understood. Add a 1 M ohm resistor across the 172 v to discharge the capacitors when the power is off.

Richard.
 

Offline free_electron

  • Super Contributor
  • ***
  • Posts: 8517
  • Country: us
    • SiliconValleyGarage
Re: Have I gone mad?
« Reply #21 on: April 29, 2012, 11:06:40 pm »
Here is another piece of information about mosfets that few people realize..

a N channel conducts of the Gate is more positive than the channel.
a P channel conducts of the Gate is more negative than the channel.

There is no such thing as 'drain and source'. How the current flows to through the channel is irellevant. we use the name of drain to denote the pin the current flows into and source for the pin the current flows out of. but they can switch places ! the mosfet does not care.

in chip design both are just a terminal.
in classic three pin devices the source is the pin where the substrate ( the bulk)  is connected ( a mosfet is a 4 terminal device. ) this bulk connection gives you a free diode ( called the intrinsic diode) across drain and source.

In harddisks we use mosfets in this fashion to create the drive voltage for the dsa. Instead of using a simple boost converter with an inductor and forward diode we put a mosfet. whenever the diode would conduct we actually turn on the mosfet. so we don't suffer 0.6 volts loss across the bulk diode. ( in step downs this is called synchronous rectification , in boost converters you can do the same). Once the coil is discharged we turn off the mosfet and the bulk diode is reverse plarized so it blocks the higher voltage from returning. During a controlled powerdown we actually turn on the mosfet again so the current can now flow the other way and is correctly discharged. The mosfets don't care at all.


Professional Electron Wrangler.
Any comments, or points of view expressed, are my own and not endorsed , induced or compensated by my employer(s).
 

Offline amspire

  • Super Contributor
  • ***
  • Posts: 3802
  • Country: au
Re: Have I gone mad?
« Reply #22 on: April 30, 2012, 12:17:57 am »
There is no such thing as 'drain and source'. How the current flows to through the channel is irellevant. we use the name of drain to denote the pin the current flows into and source for the pin the current flows out of. but they can switch places ! the mosfet does not care.

What you say is true except that mosfets with separate substrate connections are now so rare, they have to be treated as a special exception. This mean the idea of drain and source is useful in that the source on almost every mosfet is internally shorted to the substrate. A pity really, as you can do some interesting things with a separate substrate  connection.

You are right though in that it is really worth remembering that when the mosfet turns on, it conduct currents in both directions. For less then 0.6V source-drain voltage, you can more or less swap the source and drain. There are advantages in turning the mosfet on to conduct reverse currents instead of relying on the substrate diode - lower voltage drop, and no stored charge in the diode junction to slow the mosfet switching speed down.

Richard.
 

Offline vxp036000

  • Regular Contributor
  • *
  • Posts: 167
Re: Have I gone mad?
« Reply #23 on: April 30, 2012, 12:22:32 am »
Only true in CMOS, and even then not always. This is generally not true at all for discrete FETs.  Reverse the source and drain on an RF power FET and watch what happens to the gain (if the device doesn't smoke, that is).  Substantially different doping profiles are used for the source and drain of a FET to maximize performance.  Reversing source and drain on a FET is just as bad as reversing the collector and emitter on a BJT.

Here is another piece of information about mosfets that few people realize..

a N channel conducts of the Gate is more positive than the channel.
a P channel conducts of the Gate is more negative than the channel.

There is no such thing as 'drain and source'. How the current flows to through the channel is irellevant. we use the name of drain to denote the pin the current flows into and source for the pin the current flows out of. but they can switch places ! the mosfet does not care.

in chip design both are just a terminal.
in classic three pin devices the source is the pin where the substrate ( the bulk)  is connected ( a mosfet is a 4 terminal device. ) this bulk connection gives you a free diode ( called the intrinsic diode) across drain and source.

In harddisks we use mosfets in this fashion to create the drive voltage for the dsa. Instead of using a simple boost converter with an inductor and forward diode we put a mosfet. whenever the diode would conduct we actually turn on the mosfet. so we don't suffer 0.6 volts loss across the bulk diode. ( in step downs this is called synchronous rectification , in boost converters you can do the same). Once the coil is discharged we turn off the mosfet and the bulk diode is reverse plarized so it blocks the higher voltage from returning. During a controlled powerdown we actually turn on the mosfet again so the current can now flow the other way and is correctly discharged. The mosfets don't care at all.
« Last Edit: April 30, 2012, 12:28:38 am by vxp036000 »
 

Offline SeanB

  • Super Contributor
  • ***
  • Posts: 16283
  • Country: za
Re: Have I gone mad?
« Reply #24 on: April 30, 2012, 06:27:21 pm »
The only commonly available device with separate source is the humble 4007 6 FET unit, where you get 4 devices with separate source and drain connections for playing around with. Of course limited to 15V and 5mA like most CMOS gates.
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf