Reply by Ricky C May 6, 20202020-05-06
I've looked into a CMOS FF and it is a nice choice instead of the two NFETs and two resistors.  The NFETs would have worked just fine, but I can get two FFs in a single package and use the other one for something else.  Also it is pretty easy to assure the FF starts in the right state. 

I even rigged up a serviceable emulation of the FF using the LTspice digital library.  I was ready to run the simulation and I got an error that wasn't clear.  Turns out the library I got for the TI regulator is encrypted!!!  WTF!!!  Why does TI have such a hard on to interfere with using LTspice.  I mean really!  TINA has lost the war.  The world is an LTspice domain and that's a natural fact.  

The TI wasn't a perfect fit anyway.  It is only specified down to 2.2 volts and I want the design to work down to 2.  I spent a couple of hours looking but did not find one part at Digikey that matched all my requirements.  I found a couple that were really close.  One had a disturbing overshoot when enabled.  I am using the LDO to prevent the 5 volts on the supercap from damaging the sounder which is rated for 2 to 4 volts.  If it overshoots to 4.6 volts, that defeats the point!  

The device also needs to have very low drop out at low currents, 4 mA and low double digit uA of supply (ground) current when operating.  The main thing though is to maintain the drop out voltage down to 2.0 volts or even lower.  

I found one that did everything perfectly, but it is only available in fixed voltages.  Mouser will order the 4.0 volt part and say you can buy qty 1.  I don't know the lead time and I'm not sure I won't have to buy the reel. 

I'll try again Thursday when I have time to work on this again. 

-- 

  Rick C.

  --- Get 1,000 miles of free Supercharging
  --- Tesla referral code - https://ts.la/richard11209
Reply by Ricky C May 5, 20202020-05-05
On Tuesday, May 5, 2020 at 12:43:22 PM UTC-4, bitrex wrote:
> On 5/5/2020 2:14 AM, Ricky C wrote: > > On Monday, May 4, 2020 at 7:27:25 PM UTC-4, bitrex wrote: > >> On 5/4/2020 12:09 PM, Ricky C wrote: > >>> On Monday, May 4, 2020 at 7:25:05 AM UTC-4, piglet wrote: > >>>> On 04/05/2020 10:41 am, Ricky C wrote: > >>>>> Now I need to find a logic family that works from 4 volts down to under 2 volts and doesn't draw excess current. I did a quick look at LVC and with 2 volt inputs it's starting to draw excess current. I'll check the old CD4000 family to see if that will work below 2 volts Vdd. > >>>>> > >>>> > >>>> CD4000 might work that low but be very feeble, think Rdson values in > >>>> tens of kilo-ohm. 30+ year old 74HC (not HCT) parts are spec'd for Vdd > >>>> 2V and will work much lower (with worsening performance). > >>>> > >>>> I had 74HC163 counters that retained internal states right down to Vdd > >>>> of 50mV (yes - millivolts) > >>>> > >>>> piglet > >>> > >>> I realized the required logic is even simpler than I thought. Once I made a state diagram I realized one of the two FFs controlling the indicators can actually be removed as the output was essentially the same as an input. That leaves one FF to control the other indicator which can be muted. > >>> > >>> Taking advantage of wire and diode connections allows the transistor based FF to be set and reset on the same net. Then the mute control drives a transistor which acts as an AND gate combining the mute signal with the alarm signal to drive the indicator. > >>> > >>> So as long as the functionality doesn't change, this is a pretty simple solution. > >>> > >> > >> To get flip-flops with guaranteed operation down below 2 volts you'll > >> have to use something a little exotic like these Fairchild/ON > >> "TinyLogic" parts with guaranteed down to 1.65: > >> > >> <https://www.mouser.com/datasheet/2/308/NC7SZ74-D-1812305.pdf> > >> > >> They are very nice high-performance flops though the US8 package is a > >> little tedious to hand-solder to a surfboard for prototyping, use a > >> microscope if you have one and are gonna do that. > > > > Thanks for the info. "Below 2 volts" is just that I don't want the thing going crazy. The sounder is rated for 2 to 4 volts, so I am basing my power life on that range, charging the supercap up to 4 volts and running it down as long as it works. > > > > I actually simplified the logic a bit. I had been looking at two FFs to control the alarm state as well as the mute for the audio alarm. But once I tried implementing it I found a number of nice simplifications so the circuit is just one FF (implemented with two transistors and a transistor to act as an AND gate to combine the alarm with the mute signal for the sounder. The run time with 2.5 F is 20 minutes... just meeting the requirement. If I want to extend the time I can pulse the alarm by some duty cycle which might make it more apparent. Then I need something additional to create the pulsing. A 555 timer seems overkill, but maybe it is perfect kill? I believe there is a CMOS version that runs at very low currents. Maybe not all that low. I'm finding 100 uA at 5 volts. I'd like to at least find low double digits uA. > > > > I need to ask more about the alarm requirements, but I think they want me to define them. > > > > Flops and logic made from discrete BJTs doesn't sound particularly > low-power to me.
NFETs with 1 Mohm resistors are adequately low power. I had them on the board anyway so why not use them for a FF? I like keeping the BOM short.
> There is a CMOS 555 but as you say it's not > particularly cheap or low-current, me being me I'd think about using the > low-power ATTiny variant that will run down to 1.8 volts, when it's not > doing anything it can be put in the deep sleep and draw only nA.
Yeah, I hate to use an MCU when it's not really required. So far the circuit has gotten simpler rather than more complex. Realizing I didn't need a three state FSM was a help. A two state mute circuit is simpler. I'm looking at implementing that with a cap, but that requires nA currents and I have little confidence since a fingerprint in the wrong place can cause a failure.
> There's a circuit for a one-shot/timer on page 21 of the Lm13700 > datasheet that draws (almost) zero stand-by current and triggers on 2 > volts. The 13700 requires about 3-ish volt supply to operate properly > but looking at the circuit and the internal schematic of the 13700 I > suspect it would work OK with a lower supply rail if the output was > bootstrapped to the positive rail to boost its own supply as the output > is positive-going.
Doesn't look so good to me. I can kill power to the device entirely since it has to be run from its own regulator. It's the active current that is important. Ten or twenty uA is not bad since the load it is controlling is 3 mA. I'm thinking it can be pulsed on for a second and off for up to 9 seconds and still be effective, but that's a question that I'll need to get some feedback on. I know I am over thinking this design, but I don't really have many requirements. I think I'm expected to figure them out myself. For grins I took a look at FFs and found an 74HCS72 single dual with negative edge trigger which is required unless I add an inverter or use a cap to couple a signal transition to a clear input. I'm looking at that and it looks pretty good. The second FF could be used with a pair of resistors, capacitors and diodes to make an oscillator, no? Every input to the device has a Schmitt trigger so a slow transition won't cause noise! -- Rick C. ++ Get 1,000 miles of free Supercharging ++ Tesla referral code - https://ts.la/richard11209
Reply by bitrex May 5, 20202020-05-05
On 5/5/2020 2:14 AM, Ricky C wrote:
> On Monday, May 4, 2020 at 7:27:25 PM UTC-4, bitrex wrote: >> On 5/4/2020 12:09 PM, Ricky C wrote: >>> On Monday, May 4, 2020 at 7:25:05 AM UTC-4, piglet wrote: >>>> On 04/05/2020 10:41 am, Ricky C wrote: >>>>> Now I need to find a logic family that works from 4 volts down to under 2 volts and doesn't draw excess current. I did a quick look at LVC and with 2 volt inputs it's starting to draw excess current. I'll check the old CD4000 family to see if that will work below 2 volts Vdd. >>>>> >>>> >>>> CD4000 might work that low but be very feeble, think Rdson values in >>>> tens of kilo-ohm. 30+ year old 74HC (not HCT) parts are spec'd for Vdd >>>> 2V and will work much lower (with worsening performance). >>>> >>>> I had 74HC163 counters that retained internal states right down to Vdd >>>> of 50mV (yes - millivolts) >>>> >>>> piglet >>> >>> I realized the required logic is even simpler than I thought. Once I made a state diagram I realized one of the two FFs controlling the indicators can actually be removed as the output was essentially the same as an input. That leaves one FF to control the other indicator which can be muted. >>> >>> Taking advantage of wire and diode connections allows the transistor based FF to be set and reset on the same net. Then the mute control drives a transistor which acts as an AND gate combining the mute signal with the alarm signal to drive the indicator. >>> >>> So as long as the functionality doesn't change, this is a pretty simple solution. >>> >> >> To get flip-flops with guaranteed operation down below 2 volts you'll >> have to use something a little exotic like these Fairchild/ON >> "TinyLogic" parts with guaranteed down to 1.65: >> >> <https://www.mouser.com/datasheet/2/308/NC7SZ74-D-1812305.pdf> >> >> They are very nice high-performance flops though the US8 package is a >> little tedious to hand-solder to a surfboard for prototyping, use a >> microscope if you have one and are gonna do that. > > Thanks for the info. "Below 2 volts" is just that I don't want the thing going crazy. The sounder is rated for 2 to 4 volts, so I am basing my power life on that range, charging the supercap up to 4 volts and running it down as long as it works. > > I actually simplified the logic a bit. I had been looking at two FFs to control the alarm state as well as the mute for the audio alarm. But once I tried implementing it I found a number of nice simplifications so the circuit is just one FF (implemented with two transistors and a transistor to act as an AND gate to combine the alarm with the mute signal for the sounder. The run time with 2.5 F is 20 minutes... just meeting the requirement. If I want to extend the time I can pulse the alarm by some duty cycle which might make it more apparent. Then I need something additional to create the pulsing. A 555 timer seems overkill, but maybe it is perfect kill? I believe there is a CMOS version that runs at very low currents. Maybe not all that low. I'm finding 100 uA at 5 volts. I'd like to at least find low double digits uA. > > I need to ask more about the alarm requirements, but I think they want me to define them. >
Flops and logic made from discrete BJTs doesn't sound particularly low-power to me. There is a CMOS 555 but as you say it's not particularly cheap or low-current, me being me I'd think about using the low-power ATTiny variant that will run down to 1.8 volts, when it's not doing anything it can be put in the deep sleep and draw only nA. There's a circuit for a one-shot/timer on page 21 of the Lm13700 datasheet that draws (almost) zero stand-by current and triggers on 2 volts. The 13700 requires about 3-ish volt supply to operate properly but looking at the circuit and the internal schematic of the 13700 I suspect it would work OK with a lower supply rail if the output was bootstrapped to the positive rail to boost its own supply as the output is positive-going.
Reply by Ricky C May 5, 20202020-05-05
On Monday, May 4, 2020 at 7:27:25 PM UTC-4, bitrex wrote:
> On 5/4/2020 12:09 PM, Ricky C wrote: > > On Monday, May 4, 2020 at 7:25:05 AM UTC-4, piglet wrote: > >> On 04/05/2020 10:41 am, Ricky C wrote: > >>> Now I need to find a logic family that works from 4 volts down to under 2 volts and doesn't draw excess current. I did a quick look at LVC and with 2 volt inputs it's starting to draw excess current. I'll check the old CD4000 family to see if that will work below 2 volts Vdd. > >>> > >> > >> CD4000 might work that low but be very feeble, think Rdson values in > >> tens of kilo-ohm. 30+ year old 74HC (not HCT) parts are spec'd for Vdd > >> 2V and will work much lower (with worsening performance). > >> > >> I had 74HC163 counters that retained internal states right down to Vdd > >> of 50mV (yes - millivolts) > >> > >> piglet > > > > I realized the required logic is even simpler than I thought. Once I made a state diagram I realized one of the two FFs controlling the indicators can actually be removed as the output was essentially the same as an input. That leaves one FF to control the other indicator which can be muted. > > > > Taking advantage of wire and diode connections allows the transistor based FF to be set and reset on the same net. Then the mute control drives a transistor which acts as an AND gate combining the mute signal with the alarm signal to drive the indicator. > > > > So as long as the functionality doesn't change, this is a pretty simple solution. > > > > To get flip-flops with guaranteed operation down below 2 volts you'll > have to use something a little exotic like these Fairchild/ON > "TinyLogic" parts with guaranteed down to 1.65: > > <https://www.mouser.com/datasheet/2/308/NC7SZ74-D-1812305.pdf> > > They are very nice high-performance flops though the US8 package is a > little tedious to hand-solder to a surfboard for prototyping, use a > microscope if you have one and are gonna do that.
Thanks for the info. "Below 2 volts" is just that I don't want the thing going crazy. The sounder is rated for 2 to 4 volts, so I am basing my power life on that range, charging the supercap up to 4 volts and running it down as long as it works. I actually simplified the logic a bit. I had been looking at two FFs to control the alarm state as well as the mute for the audio alarm. But once I tried implementing it I found a number of nice simplifications so the circuit is just one FF (implemented with two transistors and a transistor to act as an AND gate to combine the alarm with the mute signal for the sounder. The run time with 2.5 F is 20 minutes... just meeting the requirement. If I want to extend the time I can pulse the alarm by some duty cycle which might make it more apparent. Then I need something additional to create the pulsing. A 555 timer seems overkill, but maybe it is perfect kill? I believe there is a CMOS version that runs at very low currents. Maybe not all that low. I'm finding 100 uA at 5 volts. I'd like to at least find low double digits uA. I need to ask more about the alarm requirements, but I think they want me to define them. -- Rick C. +- Get 1,000 miles of free Supercharging +- Tesla referral code - https://ts.la/richard11209
Reply by bitrex May 4, 20202020-05-04
On 5/4/2020 12:09 PM, Ricky C wrote:
> On Monday, May 4, 2020 at 7:25:05 AM UTC-4, piglet wrote: >> On 04/05/2020 10:41 am, Ricky C wrote: >>> Now I need to find a logic family that works from 4 volts down to under 2 volts and doesn't draw excess current. I did a quick look at LVC and with 2 volt inputs it's starting to draw excess current. I'll check the old CD4000 family to see if that will work below 2 volts Vdd. >>> >> >> CD4000 might work that low but be very feeble, think Rdson values in >> tens of kilo-ohm. 30+ year old 74HC (not HCT) parts are spec'd for Vdd >> 2V and will work much lower (with worsening performance). >> >> I had 74HC163 counters that retained internal states right down to Vdd >> of 50mV (yes - millivolts) >> >> piglet > > I realized the required logic is even simpler than I thought. Once I made a state diagram I realized one of the two FFs controlling the indicators can actually be removed as the output was essentially the same as an input. That leaves one FF to control the other indicator which can be muted. > > Taking advantage of wire and diode connections allows the transistor based FF to be set and reset on the same net. Then the mute control drives a transistor which acts as an AND gate combining the mute signal with the alarm signal to drive the indicator. > > So as long as the functionality doesn't change, this is a pretty simple solution. >
To get flip-flops with guaranteed operation down below 2 volts you'll have to use something a little exotic like these Fairchild/ON "TinyLogic" parts with guaranteed down to 1.65: <https://www.mouser.com/datasheet/2/308/NC7SZ74-D-1812305.pdf> They are very nice high-performance flops though the US8 package is a little tedious to hand-solder to a surfboard for prototyping, use a microscope if you have one and are gonna do that.
Reply by Ricky C May 4, 20202020-05-04
On Monday, May 4, 2020 at 6:03:59 AM UTC-4, piglet wrote:
> On 04/05/2020 9:08 am, Ricky C wrote: > > LTspice is a nice program... well, it's ok in many respects and great in others. But it can be a PITA to find models for. > > > > I'm looking at a number of possible parts to use in a new design but many of them have no spice models. I'm not so worried about finding an accurate LED model. But I'm using a Schottky diode to prevent back feeding of power through a regulator and I need a good match to the Vf and Ir. > > > > I tried going through the available LEDs in LTspice and few of them are current parts. I don't want to do the same search for a Schottky diode as they have a lot more of them and many are likely not useful for a number of reasons. A part I picked from Digikey is PMEG40T20ER. I thought they had no model, but I got a reply from support and they told me how to find it. They list it under documentation. Why not? Now to figure out if I can connect it to my schematic. > > > > I considered using the current limiting part Win is using, the FPF2125 which would do a great job of running current into the supercap and includes the back current block. With a comparator this should be all that is needed to charge the supercap. But no model. I'd have to cobble something up that *should* work the same, but who knows? > > > > Dialog Semi has programmable devices for this like the SLG46116. I don't see any indication it includes a block of the back current and doesn't limit the current, but otherwise it might include enough configurable logic and analog to do nearly the whole thing in one chip! No model for LTspice, but they have their own simulator. Not sure how the tools work and what other parts of the system can be included. Their web site is all marketing heavy and not so much info on the details. > > > > Maybe I'll download the Dialog tools and see how well they work. > > > > Someone posted an inrush limiting circuit that gave me an idea to just construct my own regulator by adding a PNP to the PFET with a sense resistor. That works ok, but there's just not enough head room to keep the current up once the voltage on the cap approaches 4V compared to the input of 5V. That's no big deal since the cap is mostly charged by that point. At the end there is still about half a volt across the drain/source of the PFET with almost 4V across the source/gate. I'm using a BSS84 in the simulation, in no small part because that was in the library. Is there another part that would have a lower drain/source voltage while passing 100mA with 4 volts on the gate? > > > > The inrush limiter I posted in response to Win's post is not really a > regulator, instead it uses a resistor to limit inrush and then shorts > the resistor with the p-fet switch once the drop across the resistor > falls. That creates a second inrush surge but huge gate capacitance can > help partly tame that. > > You could try FDC638P which is specified for 2.5Vgs and is in the > default LT Spice IV library (I haven't checked if it is in LTSVII).
Yes, that part seems to work much better in the simulation with the current holding at 80 mA until the full 4 volts is achieved. Thanks. The data sheet provides good thermal info... or I thought so until I noticed the example copper area layout doesn't fit this device. It would short most of the pins together. I guess they just use it for "illustrative" purposes. I'm surprised how simple this constant current design is. I was worried putting the back current Schottky in the regulating leg with the sense resistor might result in a sneak path for current through the 2N2907, but the BE junction is lower reverse current than the Schottky. Add one half of an LT6700-3 to the circuit and the current is a constant 80 mA until the supercap reaches 4 volts from a 5 volt supply. Even with a 4.7 volt source it holds 80 mA. I see the voltage across the PFET drop to 125 mV at the point the supercap is 4 volts. Nice. I was still exploring the device Win used which has no sim model. On support replied and they suggested a similar current limiting device, FPF2195 which does have a spice model. The current regulation setting is very loose though. Set it for 100 mA (the minimum) and you can get anywhere between 75 and 125 mA. The concern is the power from that current at the full 5 volts supply with zero volts on the supercap. It takes some time to put volts on these things so the device will heat up. Using the transistor design at 80 mA, the peak power is 390 mW in the PFET. 60 seconds later it is still 240 mW. I think if I pour a square inch of 2 oz copper on each side of the board connected by vias it should stand up fine. The data sheet says with a 2 oz square in on one side of the board the theta JA is 78 &deg;C/W giving a rise of 30 &deg;C worse case. Thanks for the tip. I'm using LTspice XVII, btw. It has a few UI quirks that rear heads when using the cntl-tab to switch windows. I don't know what that is about. Seems the cntl key changes the function of the cursor in the waveform display that mucks it up and I have to reset to view all and zoom in again if I mess up. -- Rick C. -+ Get 1,000 miles of free Supercharging -+ Tesla referral code - https://ts.la/richard11209
Reply by Ricky C May 4, 20202020-05-04
On Monday, May 4, 2020 at 7:25:05 AM UTC-4, piglet wrote:
> On 04/05/2020 10:41 am, Ricky C wrote: > > Now I need to find a logic family that works from 4 volts down to under 2 volts and doesn't draw excess current. I did a quick look at LVC and with 2 volt inputs it's starting to draw excess current. I'll check the old CD4000 family to see if that will work below 2 volts Vdd. > > > > CD4000 might work that low but be very feeble, think Rdson values in > tens of kilo-ohm. 30+ year old 74HC (not HCT) parts are spec'd for Vdd > 2V and will work much lower (with worsening performance). > > I had 74HC163 counters that retained internal states right down to Vdd > of 50mV (yes - millivolts) > > piglet
I realized the required logic is even simpler than I thought. Once I made a state diagram I realized one of the two FFs controlling the indicators can actually be removed as the output was essentially the same as an input. That leaves one FF to control the other indicator which can be muted. Taking advantage of wire and diode connections allows the transistor based FF to be set and reset on the same net. Then the mute control drives a transistor which acts as an AND gate combining the mute signal with the alarm signal to drive the indicator. So as long as the functionality doesn't change, this is a pretty simple solution. -- Rick C. -- Get 1,000 miles of free Supercharging -- Tesla referral code - https://ts.la/richard11209
Reply by piglet May 4, 20202020-05-04
On 04/05/2020 10:41 am, Ricky C wrote:
> Now I need to find a logic family that works from 4 volts down to under 2 volts and doesn't draw excess current. I did a quick look at LVC and with 2 volt inputs it's starting to draw excess current. I'll check the old CD4000 family to see if that will work below 2 volts Vdd. >
CD4000 might work that low but be very feeble, think Rdson values in tens of kilo-ohm. 30+ year old 74HC (not HCT) parts are spec'd for Vdd 2V and will work much lower (with worsening performance). I had 74HC163 counters that retained internal states right down to Vdd of 50mV (yes - millivolts) piglet
Reply by piglet May 4, 20202020-05-04
On 04/05/2020 9:08 am, Ricky C wrote:
> LTspice is a nice program... well, it's ok in many respects and great in others. But it can be a PITA to find models for. > > I'm looking at a number of possible parts to use in a new design but many of them have no spice models. I'm not so worried about finding an accurate LED model. But I'm using a Schottky diode to prevent back feeding of power through a regulator and I need a good match to the Vf and Ir. > > I tried going through the available LEDs in LTspice and few of them are current parts. I don't want to do the same search for a Schottky diode as they have a lot more of them and many are likely not useful for a number of reasons. A part I picked from Digikey is PMEG40T20ER. I thought they had no model, but I got a reply from support and they told me how to find it. They list it under documentation. Why not? Now to figure out if I can connect it to my schematic. > > I considered using the current limiting part Win is using, the FPF2125 which would do a great job of running current into the supercap and includes the back current block. With a comparator this should be all that is needed to charge the supercap. But no model. I'd have to cobble something up that *should* work the same, but who knows? > > Dialog Semi has programmable devices for this like the SLG46116. I don't see any indication it includes a block of the back current and doesn't limit the current, but otherwise it might include enough configurable logic and analog to do nearly the whole thing in one chip! No model for LTspice, but they have their own simulator. Not sure how the tools work and what other parts of the system can be included. Their web site is all marketing heavy and not so much info on the details. > > Maybe I'll download the Dialog tools and see how well they work. > > Someone posted an inrush limiting circuit that gave me an idea to just construct my own regulator by adding a PNP to the PFET with a sense resistor. That works ok, but there's just not enough head room to keep the current up once the voltage on the cap approaches 4V compared to the input of 5V. That's no big deal since the cap is mostly charged by that point. At the end there is still about half a volt across the drain/source of the PFET with almost 4V across the source/gate. I'm using a BSS84 in the simulation, in no small part because that was in the library. Is there another part that would have a lower drain/source voltage while passing 100mA with 4 volts on the gate? >
The inrush limiter I posted in response to Win's post is not really a regulator, instead it uses a resistor to limit inrush and then shorts the resistor with the p-fet switch once the drop across the resistor falls. That creates a second inrush surge but huge gate capacitance can help partly tame that. You could try FDC638P which is specified for 2.5Vgs and is in the default LT Spice IV library (I haven't checked if it is in LTSVII). piglet
Reply by Ricky C May 4, 20202020-05-04
On Monday, May 4, 2020 at 4:41:37 AM UTC-4, bitrex wrote:
> On 5/4/2020 4:08 AM, Ricky C wrote: > > LTspice is a nice program... well, it's ok in many respects and great in others. But it can be a PITA to find models for. > > > > I'm looking at a number of possible parts to use in a new design but many of them have no spice models. I'm not so worried about finding an accurate LED model. But I'm using a Schottky diode to prevent back feeding of power through a regulator and I need a good match to the Vf and Ir. > > > > I tried going through the available LEDs in LTspice and few of them are current parts. I don't want to do the same search for a Schottky diode as they have a lot more of them and many are likely not useful for a number of reasons. A part I picked from Digikey is PMEG40T20ER. I thought they had no model, but I got a reply from support and they told me how to find it. They list it under documentation. Why not? Now to figure out if I can connect it to my schematic. > > > > I considered using the current limiting part Win is using, the FPF2125 which would do a great job of running current into the supercap and includes the back current block. With a comparator this should be all that is needed to charge the supercap. But no model. I'd have to cobble something up that *should* work the same, but who knows? > > > > Dialog Semi has programmable devices for this like the SLG46116. I don't see any indication it includes a block of the back current and doesn't limit the current, but otherwise it might include enough configurable logic and analog to do nearly the whole thing in one chip! No model for LTspice, but they have their own simulator. Not sure how the tools work and what other parts of the system can be included. Their web site is all marketing heavy and not so much info on the details. > > > > Maybe I'll download the Dialog tools and see how well they work. > > > > Someone posted an inrush limiting circuit that gave me an idea to just construct my own regulator by adding a PNP to the PFET with a sense resistor. That works ok, but there's just not enough head room to keep the current up once the voltage on the cap approaches 4V compared to the input of 5V. That's no big deal since the cap is mostly charged by that point. At the end there is still about half a volt across the drain/source of the PFET with almost 4V across the source/gate. I'm using a BSS84 in the simulation, in no small part because that was in the library. Is there another part that would have a lower drain/source voltage while passing 100mA with 4 volts on the gate? > > > > Adding e.g. diode models to LTSpice is easy, you can use standard SPICE > models if you have one. just go to to program files > LTC > LTSpicexxx > > lib > cmp and click at top right of the file explorer there should be an > icon that shows LTSpice as the associated program, paste the .model in > at the end of the list with the others and save the file, it will be > available to select next time you run the program. > > If you can find a PSpice model for something you should have a working > LTSpice model also, just open the PSpice model in LTSpice, highlight the > name of the part immediately after the ".model" declaration, right click > and select "auto-generate symbol" and a symbol will be created for the > model and you can find it under "AutoGenerated" in the component browser > window and just drop in like any other component from the included library. > > I did that for the LM317 and LM337 on that circuit earlier, works fine.
Thanks for the info. I wasn't quite getting it but a little google-fu and I got a new symbol (a rectangle with the numbers 1 and 2) connected to the model file. Info on using LTspice is a lot more available than it used to be. Turns out this part number is a bit better in reverse current, but a bit worse in forward voltage. So I'm keeping the one I found in LTspice, it's a couple pennies cheaper at Digikey and they've got a lot more of them in stock. I'm pretty pleased with this part of the design. It is working well with nothing marginal that I can see. The rest of the circuit is crap. I don't know how complicated the logic needs to be. I started out thinking simply and used a pair of FETs to "remember" the state. I think the state machine needs to be a bit more complex than is reasonable with a hand full of transistors. Now I need to find a logic family that works from 4 volts down to under 2 volts and doesn't draw excess current. I did a quick look at LVC and with 2 volt inputs it's starting to draw excess current. I'll check the old CD4000 family to see if that will work below 2 volts Vdd. -- Rick C. + Get 1,000 miles of free Supercharging + Tesla referral code - https://ts.la/richard11209