stonehenge regulator

On 28 Sep 2019 01:57:35 -0700, Winfield Hill <winfieldhill@yahoo.com>
wrote:

John Larkin wrote...

On 27 Sep 2019, bloggs.fredbloggs.fred@gmail.com wrote:
On September 26, 2019, John Larkin wrote:

I have a 48 volt power supply and I want to drive a
lot of relays and uPs and FPGAs and stuff, so I
figured I'd knock it down to 12 volts first.

Isn't that 12V TVS cutting things close?

A 12 volt TVS doesn't conduct at 12 volts. More like 14.

If the reg fails, it would push 48 volts into 7 PC boards
that are all expecting 12. I could include the TVS and
maybe an 0805 series resistor as sacrifial parts, to
localize the flames.

To protect those parts, use an SCR shutdown with a fuse.

Wow, I haven't done a crowbar in decades. It used to be popular. In
theory, a crowbar SCR needs a fancy gate driver chip, not just a zener
into the gate.

I'm assuming that a hunky (600 watt) TVS will fail hard shorted,
enough to take out a resistor in the 48V switcher input. If I use a
crowbar and a fuse, it will have to come home as an RMA anyhow, if the
fuse blows.

(I should test some 1 ohm 0805 resistors for their behavior as fuses.
I can use my exploder rig. May as well blow up some TVSs too... see
how much energy it takes to open them up.)

We've had bad experiences with surface-mount fuses, so if we crowbar
we'd go with an MDL 5mm or 3AG in a socket on the board.

The switcher failure would be a low probability event, so protecting
against over-voltage on the 12 volt rail is purely optional. If a box
fails, it will come back for repair and the control board (where the
48-12 switcher lives) will need to be repaired. The TVS would just
limit damage downstream, especially on the other 6 boards that get the
shared +12.

We don't usually over-voltage protect the smaller switchers, like the
3.3 and 1 volt supplies. We do check their output voltages.



--

John Larkin Highland Technology, Inc

lunatic fringe electronics

 

[Winfield Hill]

jlarkin@highlandsniptechnology.com wrote...

We've had bad experiences with surface-mount fuses, so if we
crowbar we'd go with an MDL 5mm or 3AG in a socket on the board.

The idea of an SCR is to create really high fault currents,
so an appropriate low-value 0805 resistor should blow out.
If your input is fused, ideally the resistor would go first.
The customer could make a repair, if they were brave enough.

> ... The TVS would just limit damage downstream ...

Exactly, my worry is the TVS, with its very high dissipation,
could fail. Maybe it would fail shorted, but maybe not. An
SCR running at 1V is safer. If you really want to be safe,
use the serious fail-safe circuit Paul devised:

https://www.dropbox.com/s/a33pelolwzrdv37/9x.7_crowbar.pdf?dl=1


--
Thanks,
- Win

 

[Winfield Hill]

jlarkin@highlandsniptechnology.com wrote...

On 28 Sep 2019, Winfield Hill wrote:

... The TVS would just limit damage downstream ...

Exactly, my worry is the TVS, with its very high dissipation,
could fail. Maybe it would fail shorted, but maybe not. An
SCR running at 1V is safer. If you really want to be safe,
use the serious fail-safe circuit Paul devised:

https://www.dropbox.com/s/a33pelolwzrdv37/9x.7_crowbar.pdf?dl=1

Way too complex to protect against a low-probability event.

It was protecting $8k of electronic PCBs. I thought
the impact in your case would be pretty dramatic as
well. A portion of the circuit might be sufficient.


--
Thanks,
- Win

 

On 28 Sep 2019 09:24:38 -0700, Winfield Hill <winfieldhill@yahoo.com>
wrote:

jlarkin@highlandsniptechnology.com wrote...

We've had bad experiences with surface-mount fuses, so if we
crowbar we'd go with an MDL 5mm or 3AG in a socket on the board.

The idea of an SCR is to create really high fault currents,
so an appropriate low-value 0805 resistor should blow out.
If your input is fused, ideally the resistor would go first.
The customer could make a repair, if they were brave enough.

... The TVS would just limit damage downstream ...

Exactly, my worry is the TVS, with its very high dissipation,
could fail. Maybe it would fail shorted, but maybe not. An
SCR running at 1V is safer. If you really want to be safe,
use the serious fail-safe circuit Paul devised:

https://www.dropbox.com/s/a33pelolwzrdv37/9x.7_crowbar.pdf?dl=1

Way too complex to protect against a low-probability event.

I was just explaining to the kids that life is risky. It doesn't make
sense to add weeks or months of engineering to reduce risk a tiny
amount. Done is better than perfect. Done can be sold.





--

John Larkin Highland Technology, Inc

lunatic fringe electronics

 

[Steve Wilson]

On Saturday, September 28, 2019 at 1:57:03 PM UTC-4, Winfield Hill wrote:

jlarkin@highlandsniptechnology.com wrote...

On 28 Sep 2019, Winfield Hill wrote:

... The TVS would just limit damage downstream ...

Exactly, my worry is the TVS, with its very high dissipation,
could fail. Maybe it would fail shorted, but maybe not. An
SCR running at 1V is safer. If you really want to be safe,
use the serious fail-safe circuit Paul devised:

https://www.dropbox.com/s/a33pelolwzrdv37/9x.7_crowbar.pdf?dl=1

Way too complex to protect against a low-probability event.

It was protecting $8k of electronic PCBs. I thought
the impact in your case would be pretty dramatic as
well. A portion of the circuit might be sufficient.

Thanks,
- Win

Here's a much simpler method. Replace the switch with a TL431. Watch the wrap.

Version 4
SHEET 1 1608 1700
WIRE 624 -208 -80 -208
WIRE 880 -208 624 -208
WIRE 880 -160 880 -208
WIRE 624 -144 624 -208
WIRE 768 -144 720 -144
WIRE 816 -144 768 -144
WIRE 832 -144 816 -144
WIRE 768 -80 768 -144
WIRE -80 -48 -80 -208
WIRE -32 -48 -80 -48
WIRE 112 -48 -32 -48
WIRE 176 -48 112 -48
WIRE 320 -48 256 -48
WIRE 384 -48 320 -48
WIRE 464 -48 384 -48
WIRE 608 -48 544 -48
WIRE 672 -48 672 -80
WIRE 672 -48 608 -48
WIRE -80 -32 -80 -48
WIRE 112 -32 112 -48
WIRE 880 -32 880 -64
WIRE 960 -32 880 -32
WIRE 976 -32 960 -32
WIRE 880 0 880 -32
WIRE 768 16 768 0
WIRE 320 48 320 -48
WIRE -80 64 -80 48
WIRE 112 64 112 48
WIRE 208 64 112 64
WIRE 272 64 208 64
WIRE 112 96 112 64
WIRE 880 96 880 80
WIRE 272 112 256 112
WIRE 256 128 256 112
WIRE 112 192 112 176
WIRE 320 192 320 128
FLAG -80 64 0
FLAG 320 192 0
FLAG -32 -48 Vin
FLAG 112 192 0
FLAG 608 -48 Q1B
FLAG 768 16 0
FLAG 880 96 0
FLAG 816 -144 M1G
FLAG 960 -32 Vout
FLAG 256 128 0
FLAG 208 64 R1R3
FLAG 384 -48 R2R4
SYMBOL voltage -80 -48 R0
WINDOW 39 0 0 Left 2
WINDOW 3 -54 150 Left 2
SYMATTR Value PULSE(0 50 0 1m 0 0 1m 1)
SYMATTR Value2 AC 1
SYMATTR InstName V1
SYMBOL res 160 -32 R270
WINDOW 0 32 56 VTop 2
WINDOW 3 0 56 VBottom 2
SYMATTR InstName R2
SYMATTR Value 10k
SYMBOL res 96 -48 R0
SYMATTR InstName R1
SYMATTR Value 10k
SYMBOL res 96 80 R0
SYMATTR InstName R3
SYMATTR Value 549
SYMBOL pnp 720 -80 M270
WINDOW 0 -6 42 VLeft 2
WINDOW 3 89 64 VLeft 2
SYMATTR InstName Q1
SYMATTR Value 2N5401
SYMBOL res 560 -64 R90
WINDOW 0 0 56 VBottom 2
WINDOW 3 32 56 VTop 2
SYMATTR InstName R4
SYMATTR Value 10k
SYMBOL pmos 832 -64 M180
SYMATTR InstName M1
SYMATTR Value Si7489DP
SYMBOL res 752 -96 R0
SYMATTR InstName R5
SYMATTR Value 10k
SYMBOL res 864 -16 R0
SYMATTR InstName R6
SYMATTR Value 48
SYMBOL sw 320 144 M180
WINDOW 0 12 104 Left 2
WINDOW 3 17 11 Left 2
SYMATTR InstName S1
SYMATTR Value TL431
TEXT 128 -288 Left 2 !.tran 0 1m 0 1u
TEXT 128 -320 Left 2 ;'Overvoltage Protection
TEXT 72 288 Left 2 !.model 2N5401 PNP(Is=21.48f Xti=3 Eg=1.11 Vaf=100 Bf=132.1 Ne=1.375 \n+Ise=21.48f Ikf=.1848 Xtb=1.5 Br=3.661 Nc=2 Isc=0 Ikr=0 Rc=1.6 Cjc=17.63p \n+Mjc=.5312 Vjc=.75 Fc=.5 Cje=73.39p Mje=.3777 Vje=.75 Tr=1.476n Tf=641.9p \n+Itf=0 Vtf=0 Xtf=0 Rb=10 Vceo=150 Icrating=600m mfg=Fairchild)
TEXT 72 424 Left 2 !.model Si7489DP VDMOS(Rg=3 Rd=31.2m Rs=1m Vto=-2.4 mtriode=2.2 Kp=35 \n+lambda=0.1 Cgdmax=6n Cgdmin=10p A=1 Cgs=4n cjo=200p M=.3 VJ=.9 Is=3.6p \n+Rb=5.5m mfg=Siliconix ksubthres=.1 Vds=-100 Ron=34m Qg=106n pchan)
TEXT 72 248 Left 2 !.model TL431 SW(Ron=1 Roff=1Meg Vt=2.5 Vh=0)

 

On Saturday, September 28, 2019 at 12:24:56 PM UTC-4, Winfield Hill wrote:

jlarkin@highlandsniptechnology.com wrote...

We've had bad experiences with surface-mount fuses, so if we
crowbar we'd go with an MDL 5mm or 3AG in a socket on the board.

The idea of an SCR is to create really high fault currents,
so an appropriate low-value 0805 resistor should blow out.
If your input is fused, ideally the resistor would go first.
The customer could make a repair, if they were brave enough.

... The TVS would just limit damage downstream ...

Exactly, my worry is the TVS, with its very high dissipation,
could fail. Maybe it would fail shorted, but maybe not. An
SCR running at 1V is safer. If you really want to be safe,
use the serious fail-safe circuit Paul devised:

https://www.dropbox.com/s/a33pelolwzrdv37/9x.7_crowbar.pdf?dl=1

It gets even more complicated when you have a bipolar supply and the load is damaged if the supply goes out of balance. Last one I did used the Motorola MC3423 self- contained overvoltage sense and SCR driver, which is about as complicated as your circuit. IIRC the application literature of the time stressed care must be taken to ensure the duration of the gate drive gets past the I2t threshold of activating the protection device. For the mechanical stuff like breakers and fuses, you need something like 100x peak overcurrent or more to get the trip in us-ms time frame, and that can't always happen. If that can't happen, then you need to stretch the t in the I2t via the gate drive. Looks like it's still available from Rochester.
https://www.onsemi.com/pub/Collateral/MC3423-D.PDF
Dunno if they upgraded to a better IC or someone else came in and stole their business.

--
Thanks,
- Win

 

On Sat, 28 Sep 2019 14:53:01 -0700 (PDT),
bloggs.fredbloggs.fred@gmail.com wrote:

On Friday, September 27, 2019 at 5:29:33 PM UTC-4, John Larkin wrote:
On Fri, 27 Sep 2019 13:47:45 -0700 (PDT),
bloggs.fredbloggs.fred@gmail.com wrote:

On Thursday, September 26, 2019 at 5:56:15 PM UTC-4, John Larkin wrote:
I have a 48 volt power supply and I want to drive a lot of relays and
uPs and FPGAs and stuff, so I figured I'd knock it down to 12 volts
first.

We have LM2567HV-ADJ in stock, a 52 KHz Simple Switcher probably
designed by druids during the last ice age. Figured I'd breadboard the
reg just for fun. A boy gets tired of typing all day.

https://www.dropbox.com/sh/ajtqs7c1nswfhey/AABi7r0gchljjrxka4aHZItfa?dl=0


Works pretty well, actually. Nothing gets very warm at 1 amp out.

Yes, that unshielded drum core will spray field everywhere, but then I
am simulating alternators.

Isn't that 12V TVS cutting things close?

A 12 volt TVS doesn't conduct at 12 volts. More like 14.

If the reg fails, it would push 48 volts into 7 PC boards that are all
expecting 12. I could include the TVS and maybe an 0805 series
resistor as sacrifial parts, to localize the flames.

LOL- localize the flames....
I wonder with your hefty relay drive, how the switcher responds to a sudden discontinuation of current loading. Will the series inductor overvoltage for a while conducting through the TVS? I think so. That would be worth running through a sim.

Done that sim, at least enough to demonstrate that the energy in the
inductor could pull the +12 up pretty good. I don't have an LT Spice
model for the LM2567, which is why I did the little breadboard.

Yes, one reason to have the TVS is to clamp overshoots in the case of
sudden unloading of the 12 volt buss. But in this particular case, a
bit of overshoot won't do any harm. The +12 is driving relays and
2nd-level switchers that can stand 32 volts in. So the TVS basically
protects against a hard switcher failure.

When we have bipolar supplies, we add a big schottky to ground so a
supply can't be pulled through to its opposite polarity, which can
cause all sorts of problems.

It looks like this entire system can be all-positive power supplies.



--

John Larkin Highland Technology, Inc

lunatic fringe electronics

 

On Friday, September 27, 2019 at 5:29:33 PM UTC-4, John Larkin wrote:

On Fri, 27 Sep 2019 13:47:45 -0700 (PDT),
bloggs.fredbloggs.fred@gmail.com wrote:

On Thursday, September 26, 2019 at 5:56:15 PM UTC-4, John Larkin wrote:
I have a 48 volt power supply and I want to drive a lot of relays and
uPs and FPGAs and stuff, so I figured I'd knock it down to 12 volts
first.

We have LM2567HV-ADJ in stock, a 52 KHz Simple Switcher probably
designed by druids during the last ice age. Figured I'd breadboard the
reg just for fun. A boy gets tired of typing all day.

https://www.dropbox.com/sh/ajtqs7c1nswfhey/AABi7r0gchljjrxka4aHZItfa?dl=0


Works pretty well, actually. Nothing gets very warm at 1 amp out.

Yes, that unshielded drum core will spray field everywhere, but then I
am simulating alternators.

Isn't that 12V TVS cutting things close?

A 12 volt TVS doesn't conduct at 12 volts. More like 14.

If the reg fails, it would push 48 volts into 7 PC boards that are all
expecting 12. I could include the TVS and maybe an 0805 series
resistor as sacrifial parts, to localize the flames.

LOL- localize the flames....
I wonder with your hefty relay drive, how the switcher responds to a sudden discontinuation of current loading. Will the series inductor overvoltage for a while conducting through the TVS? I think so. That would be worth running through a sim.

 

On Saturday, September 28, 2019 at 6:19:36 PM UTC-4, jla...@highlandsniptechnology.com wrote:

On Sat, 28 Sep 2019 14:53:01 -0700 (PDT),
bloggs.fredbloggs.fred@gmail.com wrote:

On Friday, September 27, 2019 at 5:29:33 PM UTC-4, John Larkin wrote:
On Fri, 27 Sep 2019 13:47:45 -0700 (PDT),
bloggs.fredbloggs.fred@gmail.com wrote:

On Thursday, September 26, 2019 at 5:56:15 PM UTC-4, John Larkin wrote:
I have a 48 volt power supply and I want to drive a lot of relays and
uPs and FPGAs and stuff, so I figured I'd knock it down to 12 volts
first.

We have LM2567HV-ADJ in stock, a 52 KHz Simple Switcher probably
designed by druids during the last ice age. Figured I'd breadboard the
reg just for fun. A boy gets tired of typing all day.

https://www.dropbox.com/sh/ajtqs7c1nswfhey/AABi7r0gchljjrxka4aHZItfa?dl=0


Works pretty well, actually. Nothing gets very warm at 1 amp out.

Yes, that unshielded drum core will spray field everywhere, but then I
am simulating alternators.

Isn't that 12V TVS cutting things close?

A 12 volt TVS doesn't conduct at 12 volts. More like 14.

If the reg fails, it would push 48 volts into 7 PC boards that are all
expecting 12. I could include the TVS and maybe an 0805 series
resistor as sacrifial parts, to localize the flames.

LOL- localize the flames....
I wonder with your hefty relay drive, how the switcher responds to a sudden discontinuation of current loading. Will the series inductor overvoltage for a while conducting through the TVS? I think so. That would be worth running through a sim.


Done that sim, at least enough to demonstrate that the energy in the
inductor could pull the +12 up pretty good. I don't have an LT Spice
model for the LM2567, which is why I did the little breadboard.

Yes, one reason to have the TVS is to clamp overshoots in the case of
sudden unloading of the 12 volt buss. But in this particular case, a
bit of overshoot won't do any harm. The +12 is driving relays and
2nd-level switchers that can stand 32 volts in. So the TVS basically
protects against a hard switcher failure.

Also eliminates any possibility of avalanche breakdown of the coil driver turning off, although almost any output capacitance should buy enough time to prevent that too. There shouldn't be secondary breakdown for a driver turning on.

When we have bipolar supplies, we add a big schottky to ground so a
supply can't be pulled through to its opposite polarity, which can
cause all sorts of problems.

It looks like this entire system can be all-positive power supplies.



--

John Larkin Highland Technology, Inc

lunatic fringe electronics

 

[Steve Wilson]

On Saturday, September 28, 2019 at 4:40:18 PM UTC-4, Steve Wilson wrote:

On Saturday, September 28, 2019 at 1:57:03 PM UTC-4, Winfield Hill wrote:
jlarkin@highlandsniptechnology.com wrote...

On 28 Sep 2019, Winfield Hill wrote:

... The TVS would just limit damage downstream ...

Exactly, my worry is the TVS, with its very high dissipation,
could fail. Maybe it would fail shorted, but maybe not. An
SCR running at 1V is safer. If you really want to be safe,
use the serious fail-safe circuit Paul devised:

https://www.dropbox.com/s/a33pelolwzrdv37/9x.7_crowbar.pdf?dl=1

Way too complex to protect against a low-probability event.

It was protecting $8k of electronic PCBs. I thought
the impact in your case would be pretty dramatic as
well. A portion of the circuit might be sufficient.

Thanks,
- Win

The previous version allowed the input to exceed the 36V maximum of the TL431. Here's the fix. Watch the wrap.

Version 4
SHEET 1 1608 1700
WIRE 144 -208 -80 -208
WIRE 384 -208 144 -208
WIRE 688 -208 384 -208
WIRE 944 -208 688 -208
WIRE 144 -160 144 -208
WIRE 384 -160 384 -208
WIRE 944 -160 944 -208
WIRE 688 -144 688 -208
WIRE 832 -144 784 -144
WIRE 880 -144 832 -144
WIRE 896 -144 880 -144
WIRE 832 -80 832 -144
WIRE -80 -48 -80 -208
WIRE -32 -48 -80 -48
WIRE 16 -48 -32 -48
WIRE 144 -48 144 -80
WIRE 240 -48 144 -48
WIRE 384 -48 384 -80
WIRE 448 -48 384 -48
WIRE 528 -48 448 -48
WIRE 672 -48 608 -48
WIRE 736 -48 736 -80
WIRE 736 -48 672 -48
WIRE -80 -32 -80 -48
WIRE 16 -32 16 -48
WIRE 144 -32 144 -48
WIRE 944 -32 944 -64
WIRE 1024 -32 944 -32
WIRE 1040 -32 1024 -32
WIRE 240 0 240 -48
WIRE 304 0 240 0
WIRE 320 0 304 0
WIRE 944 0 944 -32
WIRE 832 16 832 0
WIRE -80 64 -80 48
WIRE 144 64 144 48
WIRE 384 80 384 48
WIRE 384 96 384 80
WIRE 944 96 944 80
WIRE 16 112 16 48
WIRE 48 112 16 112
WIRE 336 112 48 112
WIRE 16 144 16 112
WIRE 336 160 320 160
WIRE 320 176 320 160
WIRE 384 192 384 176
WIRE 16 240 16 224
FLAG -80 64 0
FLAG 384 192 0
FLAG -32 -48 Vin
FLAG 16 240 0
FLAG 672 -48 Q1B
FLAG 832 16 0
FLAG 944 96 0
FLAG 880 -144 M1G
FLAG 1024 -32 Vout
FLAG 320 176 0
FLAG 48 112 R1R3
FLAG 448 -48 R2R4
FLAG 144 64 0
FLAG 304 0 Q2B
FLAG 384 80 Q2E
SYMBOL voltage -80 -48 R0
WINDOW 39 0 0 Left 2
WINDOW 3 3 -182 Left 2
SYMATTR Value PULSE(0 50 0 1m 0 0 1m 1)
SYMATTR Value2 AC 1
SYMATTR InstName V1
SYMBOL res 368 -176 R0
SYMATTR InstName R2
SYMATTR Value 10k
SYMBOL res 0 -48 R0
SYMATTR InstName R1
SYMATTR Value 10k
SYMBOL res 0 128 R0
SYMATTR InstName R3
SYMATTR Value 549
SYMBOL pnp 784 -80 M270
WINDOW 0 -6 42 VLeft 2
WINDOW 3 89 64 VLeft 2
SYMATTR InstName Q1
SYMATTR Value 2N5401
SYMBOL res 624 -64 R90
WINDOW 0 0 56 VBottom 2
WINDOW 3 32 56 VTop 2
SYMATTR InstName R4
SYMATTR Value 10k
SYMBOL pmos 896 -64 M180
SYMATTR InstName M1
SYMATTR Value Si7489DP
SYMBOL res 816 -96 R0
SYMATTR InstName R5
SYMATTR Value 10k
SYMBOL res 928 -16 R0
SYMATTR InstName R6
SYMATTR Value 48
SYMBOL sw 384 192 M180
WINDOW 0 12 104 Left 2
WINDOW 3 17 11 Left 2
SYMATTR InstName S1
SYMATTR Value TL431
SYMBOL npn 320 -48 R0
SYMATTR InstName Q2
SYMATTR Value 2N3904
SYMBOL res 128 -176 R0
SYMATTR InstName R7
SYMATTR Value 10K
SYMBOL res 128 -48 R0
SYMATTR InstName R8
SYMATTR Value 10k
TEXT 128 -288 Left 2 !.tran 0 1m 0 1u
TEXT 128 -320 Left 2 ;'Overvoltage Protection
TEXT 72 368 Left 2 !.model 2N5401 PNP(Is=21.48f Xti=3 Eg=1.11 Vaf=100 Bf=132.1 Ne=1.375 \n+Ise=21.48f Ikf=.1848 Xtb=1.5 Br=3.661 Nc=2 Isc=0 Ikr=0 Rc=1.6 Cjc=17.63p \n+Mjc=.5312 Vjc=.75 Fc=.5 Cje=73.39p Mje=.3777 Vje=.75 Tr=1.476n Tf=641.9p \n+Itf=0 Vtf=0 Xtf=0 Rb=10 Vceo=150 Icrating=600m mfg=Fairchild)
TEXT 72 504 Left 2 !.model Si7489DP VDMOS(Rg=3 Rd=31.2m Rs=1m Vto=-2.4 mtriode=2.2 Kp=35 \n+lambda=0.1 Cgdmax=6n Cgdmin=10p A=1 Cgs=4n cjo=200p M=.3 VJ=.9 Is=3.6p \n+Rb=5.5m mfg=Siliconix ksubthres=.1 Vds=-100 Ron=34m Qg=106n pchan)
TEXT 72 328 Left 2 !.model TL431 SW(Ron=1 Roff=1Meg Vt=2.5 Vh=0)
TEXT 456 136 Left 2 ;Replace the switch with a TL431

 

[Tim Williams]

So what? Do it without a regulator.
https://www.seventransistorlabs.com/Images/Discrete_Tube_Supply.png

I lie, there's a TL431. So what, it's like a single transistor (making this
a "10"-transistor-lab product, alas; if you ignore the complementary
follower and feedback, the core circuitry uses a mere seven however).

Hmm, don't know if I have any pictures of a finished converter with this.
Have built a few though. Works fine. Sometimes weird modes (subharmonics,
poorly optimized compensation?), eh, it's simple.

Nice thing about low gain devices (which I think includes the ancient
Simpler Switchers, for better and for worse) is they don't make much for
harmonics. I've got one of these on a shortwave radio and another on a
Theremin, just CLCLC output filters is fine even for such sensitive
purposes.

Tim

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

"John Larkin" <jlarkin@highland_atwork_technology.com> wrote in message
news:79cqoehqeqmmiqt5ekuck5pe7a8r65anic@4ax.com...

I have a 48 volt power supply and I want to drive a lot of relays and
uPs and FPGAs and stuff, so I figured I'd knock it down to 12 volts
first.

We have LM2567HV-ADJ in stock, a 52 KHz Simple Switcher probably
designed by druids during the last ice age. Figured I'd breadboard the
reg just for fun. A boy gets tired of typing all day.

https://www.dropbox.com/sh/ajtqs7c1nswfhey/AABi7r0gchljjrxka4aHZItfa?dl=0


Works pretty well, actually. Nothing gets very warm at 1 amp out.

Yes, that unshielded drum core will spray field everywhere, but then I
am simulating alternators.

 

[Winfield Hill]

Steve Wilson wrote...

On September 28, 2019, Steve Wilson wrote:
On September 28, 2019, Winfield Hill wrote:
jlarkin@highlandsniptechnology.com wrote...
On 28 Sep 2019, Winfield Hill wrote:

... The TVS would just limit damage downstream ...

Exactly, my worry is the TVS, with its very high
dissipation, could fail. Maybe it would fail
shorted, but maybe not. An SCR running at 1V is
safer. If you really want to be safe, use the
serious fail-safe circuit Paul devised:

https://www.dropbox.com/s/a33pelolwzrdv37/9x.7_crowbar.pdf?dl

Way too complex to protect against a low-probability event.

It was protecting $8k of electronic PCBs. I thought
the impact in your case would be pretty dramatic as
well. A portion of the circuit might be sufficient.

The previous version allowed the input to exceed the
36V maximum of the TL431. Here's the fix. [ snip ]

Posts using LTspice are hard to see, save file, open, etc.

Some issues with your idea:
(1) It's a voltage clamp, which must be fast, but also
its control loop must be stable at high currents.
(2) The Si7489DP FET's Vgs is limited to 20 volts.
(3) If a fault persists for any length of time, the Si7489DP
will overheat and fail. So you need to add a crowbar.

Paul's circuit handles these issues, and also times actions
carefully to avoid employing the crowbar, if the clamp can
handle the problem and save the day. As a bonus, it turns
off the source's AC power. It would have been nice to add
some performance waveforms to the article.



--
Thanks,
- Win

 

[Steve Wilson]

On Sunday, September 29, 2019 at 3:17:38 PM UTC-4, Winfield Hill wrote:

Steve Wilson wrote...
On September 28, 2019, Steve Wilson wrote:
On September 28, 2019, Winfield Hill wrote:
jlarkin@highlandsniptechnology.com wrote...
On 28 Sep 2019, Winfield Hill wrote:

... The TVS would just limit damage downstream ...

Exactly, my worry is the TVS, with its very high
dissipation, could fail. Maybe it would fail
shorted, but maybe not. An SCR running at 1V is
safer. If you really want to be safe, use the
serious fail-safe circuit Paul devised:

https://www.dropbox.com/s/a33pelolwzrdv37/9x.7_crowbar.pdf?dl

Way too complex to protect against a low-probability event.

It was protecting $8k of electronic PCBs. I thought
the impact in your case would be pretty dramatic as
well. A portion of the circuit might be sufficient.

The previous version allowed the input to exceed the
36V maximum of the TL431. Here's the fix. [ snip ]

Posts using LTspice are hard to see, save file, open, etc.

Posts showing the ASC file are extremely useful. You seem to have managed.

Unfortunately, it is difficult to include the PLT file to show the waveforms. Perhaps the solution is to zip the files and upload them to google drv.

Try a different newsreader. I find XNews to be extremely useful to save any post that contains useful information, such as problems with ICs or other information.

Some issues with your idea:
(1) It's a voltage clamp, which must be fast, but also
its control loop must be stable at high currents.

It is not a clamp. There is no control loop. It simply opens the FET when the input voltage exceeds the limit.

> (2) The Si7489DP FET's Vgs is limited to 20 volts.

I added a 12V zener from gate to src in later versions. Fixed.

(3) If a fault persists for any length of time, the Si7489DP
will overheat and fail. So you need to add a crowbar.

Mosfets are available with milliohms of on resistance. This would take a lot of current to cause overheating. The problem would shift back to the source supply.

A crowbar could make a bad situation worse. A better solution would be to measure the current and shut the FET off if it exceeds some limit. This would need to store the fault and provide a reset. An additional enhancement would be to check for undervoltage and turn off the FET. These additions would require some LEDs to indicate the problem and show the status.

Paul's circuit handles these issues, and also times actions
carefully to avoid employing the crowbar, if the clamp can
handle the problem and save the day. As a bonus, it turns
off the source's AC power. It would have been nice to add
some performance waveforms to the article.

Paul's circuit is way too complex. Crowbars can cause severe damage.

No need to shut off the power. Just shut off the FET.

Run the ASC file and look at whatever waveforms you desire.

Thanks for your reply.

 

On 29 Sep 2019 12:17:17 -0700, Winfield Hill <winfieldhill@yahoo.com>
wrote:

Steve Wilson wrote...
On September 28, 2019, Steve Wilson wrote:
On September 28, 2019, Winfield Hill wrote:
jlarkin@highlandsniptechnology.com wrote...
On 28 Sep 2019, Winfield Hill wrote:

... The TVS would just limit damage downstream ...

Exactly, my worry is the TVS, with its very high
dissipation, could fail. Maybe it would fail
shorted, but maybe not. An SCR running at 1V is
safer. If you really want to be safe, use the
serious fail-safe circuit Paul devised:

https://www.dropbox.com/s/a33pelolwzrdv37/9x.7_crowbar.pdf?dl

Way too complex to protect against a low-probability event.

It was protecting $8k of electronic PCBs. I thought
the impact in your case would be pretty dramatic as
well. A portion of the circuit might be sufficient.

The previous version allowed the input to exceed the
36V maximum of the TL431. Here's the fix. [ snip ]

Posts using LTspice are hard to see, save file, open, etc.

Some issues with your idea:
(1) It's a voltage clamp, which must be fast, but also
its control loop must be stable at high currents.
(2) The Si7489DP FET's Vgs is limited to 20 volts.
(3) If a fault persists for any length of time, the Si7489DP
will overheat and fail. So you need to add a crowbar.

Paul's circuit handles these issues, and also times actions
carefully to avoid employing the crowbar, if the clamp can
handle the problem and save the day. As a bonus, it turns
off the source's AC power. It would have been nice to add
some performance waveforms to the article.

Maybe an updated version of Paul's circuit would be one tiny uP doing
all the thinking.



--

John Larkin Highland Technology, Inc

lunatic fringe electronics

 

[Winfield Hill]

Steve Wilson wrote...

On September 29, 2019, Winfield Hill wrote:

Some issues with your idea:
(1) It's a voltage clamp, ...

It is not a clamp. There is no control loop. It simply
opens the FET when the input voltage exceeds the limit.

Excuse me, you are right. OK, that changes everything.


--
Thanks,
- Win

 

[mpm]

On Friday, September 27, 2019 at 7:33:45 AM UTC-4, mpm wrote:

On Thursday, September 26, 2019 at 5:56:15 PM UTC-4, John Larkin wrote:
I have a 48 volt power supply and I want to drive a lot of relays...

And what's wrong with old tech?

I just approved two designs with the LM2575HVT, another "simple switcher" part....

One of my pet weekend projects is to breadboard the LM2575 (actually, using a PCB) and try about a dozen different values for the coil under various loading. I'd just like to get a good feel for the part, since it is dirt simple to use and is a low-freq switching design.

Of course, I've had the coils for about a year now, and they are still sitting in the DigiKey box.

I also wanted to characterize its EMI profile and see if I could get it to oscillate under some "off-datasheet conditions".

In general, I think this "old tech" gets a bad rap sometimes.
While better designs exist, the old ones are still pretty good!
Although I can't remember the last time I used a 7805.