TL783 - are there any newer high-voltage linear regulators?

[John Nagle]

Is there anything later than the TL783 in a high-voltage (125V or so)
linear regulator? That's all DigiKey lists, and the TL783 is from 1981.

John Nagle

[John Larkin]

On Thu, 31 Aug 2017 11:33:14 -0700, John Nagle <na...@animats.com>
wrote:

> Is there anything later than the TL783 in a high-voltage (125V or so)
>linear regulator? That's all DigiKey lists, and the TL783 is from 1981.
>
> John Nagle

There are a few 60V parts around. And there are tricks to help out a
lower voltage reg.

Or make your own around a depletion fet. What are your numbers?




--

John Larkin Highland Technology, Inc
picosecond timing precision measurement

jlarkin att highlandtechnology dott com
http://www.highlandtechnology.com

[Phil Hobbs]

On the plus side, if you can handle the dropout voltage, it's a pretty
good chip.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC
Optics, Electro-optics, Photonics, Analog Electronics

160 North State Road #203
Briarcliff Manor NY 10510

hobbs at electrooptical dot net
http://electrooptical.net

[Tim Williams]

https://www.digikey.com/short/qc79rt ???

Tim

--
Seven Transistor Labs, LLC
Electrical Engineering Consultation and Contract Design
Website: http://seventransistorlabs.com

"John Nagle" <na...@animats.com> wrote in message
news:oo9kh4$keo$1...@dont-email.me...

[TTman]

A rare name.... I worked with a guy called Nagle in Goppingen ( Germany)
in 1975..... any relation to you ????

---
This email has been checked for viruses by Avast antivirus software.
https://www.avast.com/antivirus

[Tim Williams]

"TTman" <pcw1...@ntlworld.com> wrote in message
news:ooa148$toa$1...@dont-email.me...

> A rare name.... I worked with a guy called Nagle in Goppingen ( Germany)
> in 1975..... any relation to you ????

*Shrug*, I know a Daigle. The Simpsons have a Lindsey Naegle character.
Can't be too rare.

(but I'm not one to judge, there are hundreds of Tim Williamses) :)

[John Nagle]

On 08/31/2017 11:56 AM, Phil Hobbs wrote:
> On 08/31/2017 02:33 PM, John Nagle wrote:
>> Is there anything later than the TL783 in a high-voltage (125V or so)
>> linear regulator? That's all DigiKey lists, and the TL783 is from 1981.
>>
>> John Nagle
>
> On the plus side, if you can handle the dropout voltage, it's a pretty
> good chip.
>
> Cheers
>
> Phil Hobbs

It's the huge dropout voltage that's the problem. There has to
be about 15V between IN and OUT for it to work.

Linear Technology has some "high voltage" parts, but they top
out at 80V.

On Semi has a NCP781 150 V, 100 mA Very High Voltage Linear
Regulator, but the output voltage cannot exceed 15V.

Some other vendors have 60V parts. The TL783 is the
only part I can find that can regulate up to 120V.

There are some parts good up to 450V, but they require big
input/output differentials. They're for getting a little bit of
+5 directly from the AC line.

(The application is a driver for Teletype selector solenoids.
The upstream circuitry charges up a 2uf cap to 120V to provide
pull-in against the huge inductance, then dumps that into the
output. After discharge, a sustain supply takes over. I have
all that working for 220 ohm 5.2 henry selector magnets.

The problem is that there are several different magnets,
with different resistances and inductances. All of them need
60mA, and the big jolt from the cap will overcurrent the low
resistance solenoids. The higher resistance ones don't need
this protection. So I want to put current limiting to 60mA on
the output. It has to work over the voltage range 12-120VDC.)

John Nagle

[Tim Williams]

"John Nagle" <na...@animats.com> wrote in message

news:oob02n$a71$1...@dont-email.me...

> The problem is that there are several different magnets,
> with different resistances and inductances. All of them need
> 60mA, and the big jolt from the cap will overcurrent the low
> resistance solenoids. The higher resistance ones don't need
> this protection. So I want to put current limiting to 60mA on
> the output. It has to work over the voltage range 12-120VDC.)

Well pffbt, put a DMOS on there and be done with it!

[pcdh...@gmail.com]

>Well pffbt, put a DMOS on there and be done with it!

Gotta be careful with the protection circuitry though, especially driving inductors. Ask George. ;)


Cheers

Phil Hobbs

[John Nagle]

On 09/01/2017 12:10 AM, Tim Williams wrote:
> "John Nagle" <na...@animats.com> wrote in message

>> So I want to put current limiting to 60mA on
>> the output. It has to work over the voltage range 12-120VDC.)
>
> Well pffbt, put a DMOS on there and be done with it!

Thanks. Looking at

http://www.mouser.com/ds/2/205/98704-13273.pdf

for starters.

John Nagle

[whit3rd]

On Thursday, August 31, 2017 at 11:56:39 PM UTC-7, John Nagle wrote:

> (The application is a driver for Teletype selector solenoids.
> The upstream circuitry charges up a 2uf cap to 120V to provide
> pull-in against the huge inductance, then dumps that into the
> output. After discharge, a sustain supply takes over. I have
> all that working for 220 ohm 5.2 henry selector magnets.
>
> The problem is that there are several different magnets,
> with different resistances and inductances. All of them need

> 60mA,...I want to put current limiting to 60mA on

> the output. It has to work over the voltage range 12-120VDC.)

But, a voltage regulator has LOW output impedance, and to drive an
inductor fast, you want HIGH output impedance. Like, return the negative
side of the solenoid through an NPN collector, with base grounded and
emitter through 200 ohm limit resistor to -12V.

[pcdh...@gmail.com]

>But, a voltage regulator has LOW output impedance, and to drive an
>inductor fast, you want HIGH output impedance.  Like, return the negative
>side of the solenoid through an NPN collector, with base grounded and
>emitter through 200 ohm limit resistor to -12V.

You need high voltage to get a big dI/dt. Low impedance doesn't hurt and may help.

Cheers

Phil Hobbs

[George Herold]

How important is cost? LTC makes some nice HV opamps.
LTC6090 140V rail to rail output. ~$4 for 100.

George H.

[bloggs.fred...@gmail.com]

How much power are you talking about? It's a simple matter to make your own with discretes, and your 125V does not suggest that state-of-the-art efficiency is a requirement.

[John Nagle]

> On 09/01/2017 12:10 AM, Tim Williams wrote:
> > Well pffbt, put a DMOS on there and be done with it!
>

That's a promising idea, but the two-terminal approach
shown in Fig. 3 of this tech note

http://ww1.microchip.com/downloads/en/AppNotes/AN-D66.pdf

puts too much resistance in the circuit. To get 2V at
the gate requires a sense resistor of about 33 ohms.
The load is only 55 ohms. I'm going to waste a third of
the energy in the sense resistor.

Is there some part that contains one of those
DMOS FETs and a gate amp to make it a regulator?

John Nagle

[whit3rd]

A voltage regulator (series pass type) just steals some of the high voltage, which (unless something's
odd in the coil's insulation resistance) is a needless addition to the circuit. Current regulation
is more appropriate than voltage regulation.

To really drive a solenoid fast, a second inductor and current source to bias it is
a good power supply, better than a filter capacitor and voltage source. The constant
power drain from the current source, though, is an energy hog.

[Tim Williams]

"John Nagle" <na...@animats.com> wrote in message

news:ooceoh$tp2$1...@dont-email.me...

> puts too much resistance in the circuit. To get 2V at
> the gate requires a sense resistor of about 33 ohms.
> The load is only 55 ohms. I'm going to waste a third of
> the energy in the sense resistor.

Well, hopefully not just a third, but up to a factor of a hundred: you gave
a 10:1 spread in voltage, and a fixed current limit spec.

Is that not what you were expecting?

Or to put it another way, if you can give a _complete_ description of the
required operation, you'll get more sensible ideas. :-)

[Phil Hobbs]

The idea is to use a PMOS wrapped around the regulator--drain to +HV,
gate to REG_INPUT, source to REG_OUTPUT, with a resistor from drain to
gate to supply bias. You pick a MOSFET with a high enough threshold
that the regulator's dropout is not a problem. It's been done with PNPs
wrapped round IC regulators from the very early days of ICs.

You'll need to protect the regulator from various fault scenarios.

[Phil Hobbs]

On 09/01/2017 05:52 PM, Phil Hobbs wrote:
> On 09/01/2017 04:13 PM, John Nagle wrote:
>> > On 09/01/2017 12:10 AM, Tim Williams wrote:
>> > > Well pffbt, put a DMOS on there and be done with it!
>> >
>> That's a promising idea, but the two-terminal approach
>> shown in Fig. 3 of this tech note
>>
>> http://ww1.microchip.com/downloads/en/AppNotes/AN-D66.pdf
>>
>> puts too much resistance in the circuit. To get 2V at
>> the gate requires a sense resistor of about 33 ohms.
>> The load is only 55 ohms. I'm going to waste a third of
>> the energy in the sense resistor.
>>
>> Is there some part that contains one of those
>> DMOS FETs and a gate amp to make it a regulator?
>>
>> John Nagle
>>
>
> The idea is to use a PMOS wrapped around the regulator--drain to +HV,
> gate to REG_INPUT, source to REG_OUTPUT, with a resistor from drain to
> gate to supply bias. You pick a MOSFET with a high enough threshold
> that the regulator's dropout is not a problem. It's been done with PNPs
> wrapped round IC regulators from the very early days of ICs.
>
> You'll need to protect the regulator from various fault scenarios.

Belay that--I have way too much going on. In a low voltage version,
it's source to +V, gate to REG_INPUT, drain to REG_OUTPUT, resistor from
source to gate. The IC regulator only passes enough current to bias the
FET by way of the resistor.

In the HV version it's probably simpler to use an op amp.

(I need a vacation really really badly ATM.)

[John Nagle]

On 09/01/2017 01:39 PM, Tim Williams wrote:
> "John Nagle" <na...@animats.com> wrote in message
news:ooceoh$tp2$1...@dont-email.me...
>> puts too much resistance in the circuit. To get 2V at
>> the gate requires a sense resistor of about 33 ohms.
>> The load is only 55 ohms. I'm going to waste a third of
>> the energy in the sense resistor.
>
> Well, hopefully not just a third, but up to a factor of a hundred:
you gave a 10:1 spread in voltage, and a fixed current limit spec.
>
> Is that not what you were expecting?
>
> Or to put it another way, if you can give a _complete_ description of
the required operation, you'll get more sensible ideas. :-)

Complete description:

https://github.com/John-Nagle/ttyloopdriver

Full design in KiCAD, including the board layout and
SPICE model.

This works fine for a 220 ohm 5 Henry solenoid as the load,
with about 60mA being maintained through the cycle. The
circuit charges up a 2uf cap to 120V, and dumps that into
the solenoid. Then a 12V sustain supply takes over and holds
the solenoid in. This happens on each bit time, which is
at least 22ms.

All this is powered from a USB port, which is why it's a bit
complicated. It draws about 300mA at 5V.

Now that I have all that working, I just want to
throttle it back for a different selector magnet
which is 55 ohms with a smaller inductance, but still wants
a 60mA signal. So I'd like to put a 60mA current limiter
on the output. There are a few different selector magnets
on various old Teletypes, all designed for 60mA constant current.
The output will start at 120VDC and drop to as low as 4V as the inductor
charges. But the current is supposed to be constant.

(The traditional Teletype power supply is an isolated 120VDC supply
with a 2K 10W resistor in series, with over 90% of the energy
going into heating up the ballast resistor.)

John Nagle

[Tim Williams]

"John Nagle" <na...@animats.com> wrote in message

news:ood80i$vf6$1...@dont-email.me...

> Complete description:
>
> https://github.com/John-Nagle/ttyloopdriver
>
> Full design in KiCAD, including the board layout and
> SPICE model.
>
> This works fine for a 220 ohm 5 Henry solenoid as the load,
> with about 60mA being maintained through the cycle. The
> circuit charges up a 2uf cap to 120V, and dumps that into
> the solenoid. Then a 12V sustain supply takes over and holds
> the solenoid in. This happens on each bit time, which is
> at least 22ms.

Hmm, so, no specs on the coils. At least, I didn't see any more info on the
main page (I'm not going to go digging).

So you want a high voltage slug of charge to overcome the inductance, then a
minimum 12V when on? And the slug should be limited, so it stops charging
the inductance beyond 60mA?

What's wrong with this? :)
https://www.seventransistorlabs.com/Images/Fast_Solenoid_Driver.png

Q2 and D2 don't need to be 100V, but I presume you'll want similar turn-off
speed.

Or if you don't want a hard 12V minimum, but always with current limiting
(say for short circuit protection purposes), Q2 or preferably D1 can be
enhanced with a CCS circuit.

DN3545 might not handle the energy needed (which is, hmm, 5H at 60mA is
~10mJ, that's not too bad), but there's TO-220 versions out there. I just
threw down what was handy in my library...

You don't even need C1 if the DC-DC is designed properly. A good old
fashioned UC3843 current mode controller will limit short circuit output
current to a fixed value. You could make it boost on top of +12V, so D1 is
actually the rectifier diode as well, C1 is really small (0.1uF or less?),
and the switcher itself takes the place of Q1-R1. Q2-D2 would still be as
shown.

[John Nagle]

On 9/2/2017 1:38 AM, Tim Williams wrote:

> What's wrong with this? :)
> https://www.seventransistorlabs.com/Images/Fast_Solenoid_Driver.png

The DMOS approach looks good. I tried the current limiter in LTSpice,
and it works well:

https://image.ibb.co/emzQwv/currentsource06.png

I'm not clear on why R3 works at 16 ohms. Vgs for that DMOS part
is -1.5V, and that's what the SPICE model has. Supposedly,
you should set R to Vgs/I, where I is the desired current
limit. Or so says "Current sources and voltage references",
by Harrison, fig 7.9. So R = 1.5V / 0.060A = 25 ohms.
But in Spice, that's too big; 16 ohms works.

Here's the entire cap charger in LTSpice.

https://image.ibb.co/fWkepF/capcharger01.png

This is a driver for 60mA current loop devices, powered
from a USB port. There's a USB to serial controller, not
shown. I've built and used this, but without the current
limiter. Without it, it only works for 220 ohm 5.5H selector
magnets. There are some other devices in the Teletype world
that want 60mA but have different solenoids, so this thing
needs current limiting to be more general.

John Nagle
www.aetherltd.com

[Jim Thompson]

On Mon, 4 Sep 2017 12:13:33 -0700, John Nagle <na...@animats.com>
wrote:

>On 9/2/2017 1:38 AM, Tim Williams wrote:
>
>> What's wrong with this? :)
>> https://www.seventransistorlabs.com/Images/Fast_Solenoid_Driver.png
>
>The DMOS approach looks good. I tried the current limiter in LTSpice,
>and it works well:
>
>https://image.ibb.co/emzQwv/currentsource06.png
>
>I'm not clear on why R3 works at 16 ohms. Vgs for that DMOS part
>is -1.5V, and that's what the SPICE model has. Supposedly,
>you should set R to Vgs/I, where I is the desired current
>limit. Or so says "Current sources and voltage references",
>by Harrison, fig 7.9. So R = 1.5V / 0.060A = 25 ohms.
>But in Spice, that's too big; 16 ohms works.

The "Vgs" spec is the voltage at which there is *no* (or minimal)
current.

Nothing is quite as abrupt as LTspice "idealisms" might imply >:-}

To see what I mean, "curve trace" the device: Sweep VDS with a small
increment, say 10mV, while doing a STEP VGS in 100mV steps.

>
>Here's the entire cap charger in LTSpice.
>
>https://image.ibb.co/fWkepF/capcharger01.png
>
>This is a driver for 60mA current loop devices, powered
>from a USB port. There's a USB to serial controller, not
>shown. I've built and used this, but without the current
>limiter. Without it, it only works for 220 ohm 5.5H selector
>magnets. There are some other devices in the Teletype world
>that want 60mA but have different solenoids, so this thing
>needs current limiting to be more general.
>
> John Nagle
> www.aetherltd.com
>

...Jim Thompson
--
| James E.Thompson | mens |
| Analog Innovations | et |
| Analog/Mixed-Signal ASIC's and Discrete Systems | manus |
| STV, Queen Creek, AZ 85142 Skype: skypeanalog | |
| Voice:(480)460-2350 Fax: Available upon request | Brass Rat |
| E-mail Icon at http://www.analog-innovations.com | 1962 |

I'm looking for work... see my website.

Thinking outside the box...producing elegant & economic solutions.

[Tim Williams]

"John Nagle" <na...@animats.com> wrote in message

news:ook8ck$q8q$1...@dont-email.me...

> I'm not clear on why R3 works at 16 ohms. Vgs for that DMOS part
> is -1.5V, and that's what the SPICE model has. Supposedly,
> you should set R to Vgs/I, where I is the desired current
> limit. Or so says "Current sources and voltage references",
> by Harrison, fig 7.9. So R = 1.5V / 0.060A = 25 ohms.
> But in Spice, that's too big; 16 ohms works.

Harrison is pretty wrong then...

It's a couple of reasons:
1. Source-referred impedance is already 1/Gm. And, roughly speaking, Gm ~=
Idss / Vgs(th).
2. But Gm isn't constant, it varies. In the large signal range (mA+), it's
quadratic (Gm proportional to Id, i.e., Id goes as Vgs^2). So the expected
source resistor differs a lot from Harrison's example near Idss (after all,
0 ohms gives you Idss only, not infinity!).
3. But it's not just quadratic, it's exponential at low currents. So you
expect Gm to drop very quickly in that range (but not suddenly to zero, as
SPICE models sometimes do, as JT alluded to).

Vgs(th) is measured somewhere along this curve (between the exponential and
quadratic ranges), but how much depends on device size, and measurement
condition. So you can't tell where you're starting on that curve.

In short: you can solve for the value, but you need to use a model as you've
done above, or use the datasheet -- assuming they have reasonably complete
DC curves, and assuming the DC characteristics are stable with temperature,
or if not, using worst case conditions.

This was a lot easier back in the toob days, because the DC curves are
stable, and the transfer curve is less aggressive, closer to linear
(Vg^(3/2) or so).

>
> Here's the entire cap charger in LTSpice.
>
> https://image.ibb.co/fWkepF/capcharger01.png
>

Ah yeah, if you move L2 from GND to V2+, D1 will take the job of D6. And
then reduce C1-C2 as much as you can tolerate, and see how the cap charger
limits current (if it doesn't limit peak current to a fixed value, you can
use something like a '3843 to do it more simply). Then you have automatic
built in current limiting, and don't need M1, and don't need S1 except for
improved turn-off speed.

And the cap charger would be supplied from the +15V, not the 5V, unless you
have enough power available from 5 to do it anyway, in which case you can
nix the 15 entirely. (Speaking of, why run the cap from 5V and not all the
power stuff from 15? That's really odd.)

[John Nagle]

Thanks.

>>
>> Here's the entire cap charger in LTSpice.
>>
>> https://image.ibb.co/fWkepF/capcharger01.png
>>
>
> Ah yeah, if you move L2 from GND to V2+, D1 will take the job of D6.
> And then reduce C1-C2 as much as you can tolerate, and see how the cap
> charger limits current (if it doesn't limit peak current to a fixed
> value, you can use something like a '3843 to do it more simply). Then
> you have automatic built in current limiting, and don't need M1, and
> don't need S1 except for improved turn-off speed.
>
> And the cap charger would be supplied from the +15V, not the 5V, unless
> you have enough power available from 5 to do it anyway, in which case
> you can nix the 15 entirely. (Speaking of, why run the cap from 5V and
> not all the power stuff from 15? That's really odd.)

It's set up that way because the power supply is 5V, from
a USB port. The 15V supply comes from an 1W integrated DC-DC
converter (XP Power IE0515S) powered from the +5 input.
I'm not sure that thing could take the spikes it would face
if it were at the bottom of L2.

The cap charger is charging the caps up to a fixed voltage.
This would work fine if we knew the properties of L4 in advance.
I have that working without output current regulation.
But L4 can be 5.4H, 220ohms, or 1.2H, 55 ohms, or some other
possibilities, depending on what model Teletype is plugged in.
They're all 60mA devices, so a current limit at 60mA makes the
device work for all such loads. I had a version with a jumper
and a ballast resistor, but that's so 1980s. Has to be automatic
today.

All this current limiting does essentially nothing with the
5.4H 220 ohm load. It's there for lower-resistance loads.

S1 is actually a CPC1510G optoisolator. Turn-off time
does matter; if the solenoid L4 doesn't release fast, the machine
will mis-type. That's why the snubbing is a bit strange. The goal
is to get the current down as fast as possible. The usual reverse
diode thing results in too long a current flow after turn-off,
because that huge inductance takes a while to discharge.
The present snubbing arrangement gets a fast release from the solenoid.

Anyway, the new boards just came in from Shenzhen, so I'll assemble
this soon and see how it works.

John Nagle

[George Gonzalez]

Very long ago I found a AM radio by the side of the road in a pile of junk. It was a very cheap GE transformerless radio.
From it I salvaged the audio output transistor, a TO-66 style case. This transistor was fed right from the 120 VAC rectified power line so it was capable of handling 160 volts. I used it to switch the selector magnets on my Model 15 TTY, I forget if I wired it up as 20 or 60 mA. Anyway, that transistor worked just fine for years. Oh and for current limiting I had a resistor capable of dropping about 120 volts from the 160 volt DC power supply, so it must have been around 2K or 6K or so. So anyway, you don't need a constant-current driver so much, a resistor should work just fine.