On a sunny day (Sat, 12 Aug 2023 07:39:21 -0700) it happened John Larkin
<jlarkin@highlandSNIPMEtechnology.com> wrote in
<cs5fdidecu7mkgtuopp0f8egg2mk8pk8pk@4ax.com>:
>On Sat, 12 Aug 2023 05:40:32 GMT, Jan Panteltje <alien@comet.invalid>
>wrote:
>
>>On a sunny day (Fri, 11 Aug 2023 08:00:42 -0700) it happened John Larkin
>><jlarkin@highlandSNIPMEtechnology.com> wrote in
>><vuicdi5u8vfimu5cgkh9mvl0fkih5sfprr@4ax.com>:
>>
>>>We have a mosfet closed-loop programmable current sink that works fine
>>>most of the time but behaves badly at startup from zero current or at
>>>very low currents. The fix is to drool in a bit of keepalive current
>>>so that the loop never goes open and rails low.
>>>
>>>https://www.dropbox.com/scl/fi/3ql0ew6a2elh562wbsx96/Keepalive.jpg?rlkey=fujknponcgtf31mi3rf072x5e&dl=0
>>>
>>>The 400 volt MMBD5004 would be better for high-voltage versions.
>>
>>This has among other things, my MOSFET heater, driven by filtered PWM from a PIC micro
>> https://panteltje.nl/panteltje/tri_pic/tritium_decay_experiment_black_box_circuit_diagram_IMG_3883.GIF
>>The feedback is via the temperature sensor using the Vbe change in a BUX86 mounted on the same hotplate as the MOSFET
>>Worked OK for many years, switched of now.
>> Tritium decay experiment
>> https://panteltje.nl/panteltje/tri_pic/
>> control loop is in software.
>>
>
>Mosfets make great heaters and transistors make pretty good
>temperature sensors.
>
>One of my guys did an experiment yesterday. A drain-tab-up D2PAK
>mosfet is being sat upon by a G199 CPU cooler, with grease.
>
>A LabJack controls the gate of the fet to dissipate power. Once in a
I had to look up 'LabJack'
https://labjack.com/pages/comparison
seems some 24 bit ones are very expensive, 12 bits $260
I'd use a PIC and maybe a Raspberry Pi, relais?
>while it switches off the drain power supply and back-biases the fet
>substrate diode to quickly measure junction temperature.
I never tried that, interesting, will measure some.
Yes, that should work, I have used a simple forward biased diode in big industrial equipment as temperature sensor.
But a LM35 gives you real temperature.
Thermopcouples work OK too.
This uses a LM35 for cold-side measurement:
https://panteltje.nl/panteltje/pic/th_pic/
here measuring some cold stuff:
panteltje.nl/pub/super_filter/cryocooler/minus_40C_90Vpp_60Hz_img_2617.jpg
Reply by John Larkin●August 12, 20232023-08-12
On Sat, 12 Aug 2023 05:40:32 GMT, Jan Panteltje <alien@comet.invalid>
wrote:
>On a sunny day (Fri, 11 Aug 2023 08:00:42 -0700) it happened John Larkin
><jlarkin@highlandSNIPMEtechnology.com> wrote in
><vuicdi5u8vfimu5cgkh9mvl0fkih5sfprr@4ax.com>:
>
>>We have a mosfet closed-loop programmable current sink that works fine
>>most of the time but behaves badly at startup from zero current or at
>>very low currents. The fix is to drool in a bit of keepalive current
>>so that the loop never goes open and rails low.
>>
>>https://www.dropbox.com/scl/fi/3ql0ew6a2elh562wbsx96/Keepalive.jpg?rlkey=fujknponcgtf31mi3rf072x5e&dl=0
>>
>>The 400 volt MMBD5004 would be better for high-voltage versions.
>
>This has among other things, my MOSFET heater, driven by filtered PWM from a PIC micro
> https://panteltje.nl/panteltje/tri_pic/tritium_decay_experiment_black_box_circuit_diagram_IMG_3883.GIF
>The feedback is via the temperature sensor using the Vbe change in a BUX86 mounted on the same hotplate as the MOSFET
>Worked OK for many years, switched of now.
> Tritium decay experiment
> https://panteltje.nl/panteltje/tri_pic/
> control loop is in software.
>
Mosfets make great heaters and transistors make pretty good
temperature sensors.
One of my guys did an experiment yesterday. A drain-tab-up D2PAK
mosfet is being sat upon by a G199 CPU cooler, with grease.
A LabJack controls the gate of the fet to dissipate power. Once in a
while it switches off the drain power supply and back-biases the fet
substrate diode to quickly measure junction temperature.
In this box, the giant optics oven is heated by mosfets.
https://www.dropbox.com/scl/fi/d8zn6yca7bkun0jkj3riq/Man_Top_1.jpg?rlkey=352n236tashqclwpcn6dd2qf7&dl=0
The temp sensor is a thermistor Wheatstone bridge for high sensitivity
at the target temperature, 30C.
Reply by Jan Panteltje●August 12, 20232023-08-12
On a sunny day (Fri, 11 Aug 2023 08:00:42 -0700) it happened John Larkin
<jlarkin@highlandSNIPMEtechnology.com> wrote in
<vuicdi5u8vfimu5cgkh9mvl0fkih5sfprr@4ax.com>:
>We have a mosfet closed-loop programmable current sink that works fine
>most of the time but behaves badly at startup from zero current or at
>very low currents. The fix is to drool in a bit of keepalive current
>so that the loop never goes open and rails low.
>
>https://www.dropbox.com/scl/fi/3ql0ew6a2elh562wbsx96/Keepalive.jpg?rlkey=fujknponcgtf31mi3rf072x5e&dl=0
>
>The 400 volt MMBD5004 would be better for high-voltage versions.
On Fri, 11 Aug 2023 13:49:22 -0700 (PDT), Fred Bloggs
<bloggs.fredbloggs.fred@gmail.com> wrote:
>On Friday, August 11, 2023 at 12:12:18?PM UTC-4, John Larkin wrote:
>> On Fri, 11 Aug 2023 08:51:43 -0700 (PDT), Fred Bloggs
>> <bloggs.fred...@gmail.com> wrote:
>> >On Friday, August 11, 2023 at 11:01:02?AM UTC-4, John Larkin wrote:
>> >> We have a mosfet closed-loop programmable current sink that works fine
>> >> most of the time but behaves badly at startup from zero current or at
>> >> very low currents. The fix is to drool in a bit of keepalive current
>> >> so that the loop never goes open and rails low.
>> >>
>> >> https://www.dropbox.com/scl/fi/3ql0ew6a2elh562wbsx96/Keepalive.jpg?rlkey=fujknponcgtf31mi3rf072x5e&dl=0
>> >>
>> >> The 400 volt MMBD5004 would be better for high-voltage versions.
>> >
>> >Isn't 0.12V on that drain rather scanty?
>> The 10 ohm resistor can always be tweaked, but this works in
>> simulation. The opamp that closes the loop has a 3 uV max offset and
>> the fet goes to milli-ohms. My project engineer can make it work; I
>> just donated the idea.
>>
>> Of course dual diodes are usually separate chips and don't exactly
>> match. But they do have exponential i-v curves and some ohms, not
>> brick walls.
>>
>> If we specify a minimum input voltage, 2.5 maybe, the keepalive shuts
>> itself off when the voltage is applied so contributes no error.
>
>Is this a current source or variable resistor?
Current sink or resistor. It will be an 8-channel isolated dummy load
module. It's easy to have both programmable resistance and
constant-current modes, as most electronic load boxes do.
Some even have constant-voltage, essentially zener diode, mode, but I
don't think we'll do that. They do want open and short states.
In resistance mode, our customer wants to apply +28 volts to one side
and a PWM mosfet to ground. Or connect us to a full-bridge bipolar
switcher, like a torque motor or stepper driver. The dynamic issues
are horrendous.
The best way to simulate resistors is with resistors, but that had
problems too.
Reply by Phil Hobbs●August 11, 20232023-08-11
John Larkin <jlarkin@highland_atwork_technology.com> wrote:
> On Fri, 11 Aug 2023 21:22:10 +0100, piglet <erichpwagner@hotmail.com>
> wrote:
>
>> On 11/08/2023 20:38, John Larkin wrote:
>>> On Fri, 11 Aug 2023 19:19:14 +0100, piglet <erichpwagner@hotmail.com>
>>> wrote:
>>>
>>>> On 11/08/2023 16:00, John Larkin wrote:
>>>>> We have a mosfet closed-loop programmable current sink that works fine
>>>>> most of the time but behaves badly at startup from zero current or at
>>>>> very low currents. The fix is to drool in a bit of keepalive current
>>>>> so that the loop never goes open and rails low.
>>>>>
>>>>> https://www.dropbox.com/scl/fi/3ql0ew6a2elh562wbsx96/Keepalive.jpg?rlkey=fujknponcgtf31mi3rf072x5e&dl=0
>>>>>
>>>>> The 400 volt MMBD5004 would be better for high-voltage versions.
>>>>>
>>>> Does the sink device have to be a FET - a BJT or Darlington might be
>>>> better in startup?
>>>>
>>>> piglet
>>>
>>> The fet is sure nice. It goes to milliohms and is easy to drive from
>>> an opamp.
>>>
>>> The thing about startup, with any transistor, is to not let the servo
>>> opamp see zero-zero inputs and wind up to some rail. In fact we want
>>> the fet to be biased slightly ON and be ready to go. Audio amps have
>>> the same issue, wanting to keep the output transistors always on a
>>> bit.
>>>
>>
>> Yes, I am hoping with a BJT the base current should guarantee the opamp
>> is always servoing and not railed.
>>
>> piglet
>
> If both the opamp inputs are zero volts, and it has any offset, it
> will rail high or low. The keepalive, and a tiny DAC offset, makes the
> opamp regulate a small constant mosfet current, and that's nice.
>
> This is a textbook opamp+mosfet controlled current sink, which can go
> bananas at zero current.
>
> Right, a BJT and some minimum demand offset will keep the loop closed
> even with nothing available on the collector.
>
> A gate-source resistor on the mosfet might have a similar effect. Kill
> its infinite beta!
>
>
>
The FET avoids the latchup state where the BJT saturates and the loop
regulates off the (greatly increased ) base current.
--
Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC /
Hobbs ElectroOptics Optics, Electro-optics, Photonics, Analog Electronics
Reply by John Larkin●August 11, 20232023-08-11
On Fri, 11 Aug 2023 21:22:10 +0100, piglet <erichpwagner@hotmail.com>
wrote:
>On 11/08/2023 20:38, John Larkin wrote:
>> On Fri, 11 Aug 2023 19:19:14 +0100, piglet <erichpwagner@hotmail.com>
>> wrote:
>>
>>> On 11/08/2023 16:00, John Larkin wrote:
>>>> We have a mosfet closed-loop programmable current sink that works fine
>>>> most of the time but behaves badly at startup from zero current or at
>>>> very low currents. The fix is to drool in a bit of keepalive current
>>>> so that the loop never goes open and rails low.
>>>>
>>>> https://www.dropbox.com/scl/fi/3ql0ew6a2elh562wbsx96/Keepalive.jpg?rlkey=fujknponcgtf31mi3rf072x5e&dl=0
>>>>
>>>> The 400 volt MMBD5004 would be better for high-voltage versions.
>>>>
>>> Does the sink device have to be a FET - a BJT or Darlington might be
>>> better in startup?
>>>
>>> piglet
>>
>> The fet is sure nice. It goes to milliohms and is easy to drive from
>> an opamp.
>>
>> The thing about startup, with any transistor, is to not let the servo
>> opamp see zero-zero inputs and wind up to some rail. In fact we want
>> the fet to be biased slightly ON and be ready to go. Audio amps have
>> the same issue, wanting to keep the output transistors always on a
>> bit.
>>
>
>Yes, I am hoping with a BJT the base current should guarantee the opamp
>is always servoing and not railed.
>
>piglet
If both the opamp inputs are zero volts, and it has any offset, it
will rail high or low. The keepalive, and a tiny DAC offset, makes the
opamp regulate a small constant mosfet current, and that's nice.
This is a textbook opamp+mosfet controlled current sink, which can go
bananas at zero current.
Right, a BJT and some minimum demand offset will keep the loop closed
even with nothing available on the collector.
A gate-source resistor on the mosfet might have a similar effect. Kill
its infinite beta!
Reply by Fred Bloggs●August 11, 20232023-08-11
On Friday, August 11, 2023 at 12:12:18 PM UTC-4, John Larkin wrote:
> On Fri, 11 Aug 2023 08:51:43 -0700 (PDT), Fred Bloggs
> <bloggs.fred...@gmail.com> wrote:
> >On Friday, August 11, 2023 at 11:01:02?AM UTC-4, John Larkin wrote:
> >> We have a mosfet closed-loop programmable current sink that works fine
> >> most of the time but behaves badly at startup from zero current or at
> >> very low currents. The fix is to drool in a bit of keepalive current
> >> so that the loop never goes open and rails low.
> >>
> >> https://www.dropbox.com/scl/fi/3ql0ew6a2elh562wbsx96/Keepalive.jpg?rlkey=fujknponcgtf31mi3rf072x5e&dl=0
> >>
> >> The 400 volt MMBD5004 would be better for high-voltage versions.
> >
> >Isn't 0.12V on that drain rather scanty?
> The 10 ohm resistor can always be tweaked, but this works in
> simulation. The opamp that closes the loop has a 3 uV max offset and
> the fet goes to milli-ohms. My project engineer can make it work; I
> just donated the idea.
>
> Of course dual diodes are usually separate chips and don't exactly
> match. But they do have exponential i-v curves and some ohms, not
> brick walls.
>
> If we specify a minimum input voltage, 2.5 maybe, the keepalive shuts
> itself off when the voltage is applied so contributes no error.
Is this a current source or variable resistor?
Reply by piglet●August 11, 20232023-08-11
On 11/08/2023 20:38, John Larkin wrote:
> On Fri, 11 Aug 2023 19:19:14 +0100, piglet <erichpwagner@hotmail.com>
> wrote:
>
>> On 11/08/2023 16:00, John Larkin wrote:
>>> We have a mosfet closed-loop programmable current sink that works fine
>>> most of the time but behaves badly at startup from zero current or at
>>> very low currents. The fix is to drool in a bit of keepalive current
>>> so that the loop never goes open and rails low.
>>>
>>> https://www.dropbox.com/scl/fi/3ql0ew6a2elh562wbsx96/Keepalive.jpg?rlkey=fujknponcgtf31mi3rf072x5e&dl=0
>>>
>>> The 400 volt MMBD5004 would be better for high-voltage versions.
>>>
>> Does the sink device have to be a FET - a BJT or Darlington might be
>> better in startup?
>>
>> piglet
>
> The fet is sure nice. It goes to milliohms and is easy to drive from
> an opamp.
>
> The thing about startup, with any transistor, is to not let the servo
> opamp see zero-zero inputs and wind up to some rail. In fact we want
> the fet to be biased slightly ON and be ready to go. Audio amps have
> the same issue, wanting to keep the output transistors always on a
> bit.
>
Yes, I am hoping with a BJT the base current should guarantee the opamp
is always servoing and not railed.
piglet
Reply by John Larkin●August 11, 20232023-08-11
On Fri, 11 Aug 2023 19:19:14 +0100, piglet <erichpwagner@hotmail.com>
wrote:
>On 11/08/2023 16:00, John Larkin wrote:
>> We have a mosfet closed-loop programmable current sink that works fine
>> most of the time but behaves badly at startup from zero current or at
>> very low currents. The fix is to drool in a bit of keepalive current
>> so that the loop never goes open and rails low.
>>
>> https://www.dropbox.com/scl/fi/3ql0ew6a2elh562wbsx96/Keepalive.jpg?rlkey=fujknponcgtf31mi3rf072x5e&dl=0
>>
>> The 400 volt MMBD5004 would be better for high-voltage versions.
>>
>Does the sink device have to be a FET - a BJT or Darlington might be
>better in startup?
>
>piglet
The fet is sure nice. It goes to milliohms and is easy to drive from
an opamp.
The thing about startup, with any transistor, is to not let the servo
opamp see zero-zero inputs and wind up to some rail. In fact we want
the fet to be biased slightly ON and be ready to go. Audio amps have
the same issue, wanting to keep the output transistors always on a
bit.
Reply by John Larkin●August 11, 20232023-08-11
On Fri, 11 Aug 2023 19:09:34 +0100, piglet <erichpwagner@hotmail.com>
wrote:
>On 11/08/2023 16:00, John Larkin wrote:
>> We have a mosfet closed-loop programmable current sink that works fine
>> most of the time but behaves badly at startup from zero current or at
>> very low currents. The fix is to drool in a bit of keepalive current
>> so that the loop never goes open and rails low.
>>
>> https://www.dropbox.com/scl/fi/3ql0ew6a2elh562wbsx96/Keepalive.jpg?rlkey=fujknponcgtf31mi3rf072x5e&dl=0
>>
>> The 400 volt MMBD5004 would be better for high-voltage versions.
>>
>Ingenious! At very low terminal voltages the user will see it sourcing a
>low current - don't know if that will be a problem in their application?
>
>piglet
This will be an 8-channel dummy load module and we'll specify some
minimum operating voltage, 2.5 or 3 maybe. The launch customer will
almost always use 28 volts DC, the nominal aircraft DC supply voltage.
Once they apply their 3 volts, the keepalive diode back-biases and
contributes no error. But this is a dummy load to simulate solenoids
and motors, not a precision instrument.