Level Shifter 3.3V to 28V

[Robert Latest]

Only dozens I guess? But Jeorg is always complaining about integrated
solutions being so expensive! But I guess that's at a thousand dozen,
rather than just a dozen. :^)

In fact I'm doing this circuit (just the schematic) as a favor for a friend
who wants to build it on a breadboard. I'm not going to be able to sell him
your design!

If you like the 555,

Heavens, no!

Tim (loves discrete)

I noticed. I like 'em, too.

robert

 

[Robert Latest]

It's the wide range
CCS (from about 5nA -- leakage current -- to 15mA full bias) and JFET
follower (pA leakage?) that allow my circuit to cover such a wide range.

Actually I don't quite understand how your circuit works. As I can see you
have a "charge" CCS (the top left Qs) and a "discharge" CCS (the bottom
current mirror), and the diodes decide if your cap gets charged (anode of
right diode "high") or discharged (anode low, discharge current is
difference between charge and discharge CCS).

What I don't get:

1. Why the paralleled Qs?

2. How is I_dis = 2*I_ch ensured?

3. Shouldn't the right diode be connected to the other leg of the
comparator? Let's say the voltage on the cap is rising until the left Q of
the diff pair starts conducting, tripping the comparator. Then the 4403
conducts, pulling the anode of the right diode up. How is the cap ever going
to discharge? Shouldn't it be the other way round?

I wish you'd number your components ;-)

robert

 

[John Larkin]

I wonder how someone can use an FPGA but be stumped by the simplest of data
sheets. Is this what EE is all about nowadays?

Heck, I wouldn't touch an FPGA with a ten-foot pole but I sure know my
basics.

robert

If you program an fpga in vhdl or verilog and treat it as a pure logic
engine, you don't have to know much about electricity.

John

 

[John Larkin]

OK, here's something about _me_: I built a few drivers for "normal clocks"
yars ago and used LM358s for the 24V H-bridges. Back then I was stupid and
forgot the freewheel diodes for the inductive load, but heck, all these
units have been putting out their minute-pulses flawlessly for some 12 years
now.

robert

That would be an interesting product: a dds synthesizer and stepper
motor kit to retrofit old mechanical clocks, like the ones on
buildings. Some toothed belts and pulleys could get you into the old
hand drive, and the dds would let you match any ratios.

John

 

[John Larkin]

Well, just yesterday I designed-in a 555 (a 7555, actually) for the first
time in m,y life. I need an oscillator whose frequency can be changed over a
range of at least 200:1 with a 100k 10-turn pot. You need something with
some oomph to whack that timing cap around with R < 500 ohms.

Anything better for this than the 555? 100k is the limit for the pot because
they get pricey for higher values.

robert

LM5112 maybe.

John

 

[John Larkin]

I suggested that opamps might be useful as level translators. That's
not an argument over details.



Sure, but that's only once, and that voltage may be free anyhow, like
from an led pilot light or something. Are you this annoying in real
life?

---


---
Sounds like lots details to me...

Built in current limiting?

Geez, John, I think most of us aren't worried about that except, as
it relates to this case, that we keep the current through the sink
to less than that which will result in a logic '1'.

Maybe the loads can short? We don't know. I do know that an opamp can
be a lot of stuff in a small, very cheap package.

Got an example less expensive than an LM339/LM393 and a few
resistors?

Quad opamp and no resistors. But don't use an LM324 for this; the
sections can interact when used open-loop.

A lot of it is routine, as engineering tends to be. But then, we do a
lot of stuff, and some of it is cool. And my life is fairly cool
because I designed it to be that way; most anybody can do that if they
want to.

And, since when is 'me' not part of 'we'?

So you want me to only talk about electronics in the abstract, things
I have no personal connection to, stuff out of peer-reviewed journals?
Drop dead.

John

 

[Spehro Pefhany]

Quad opamp and no resistors. But don't use an LM324 for this; the
sections can interact when used open-loop.

Interact to the point of not giving the correct state? I've seen
interaction in the 10's of uV, IIRC, but nothing too crazy.

One Japanese design from the eighties used half a cheap dual op-amp as
a very low level amplifier, and the other half as a comparator. They
used resistors to feed back the (presumably minus of the measured
average value of the) interaction so they could save a few pennies
parts cost. Clever.

Best regards,
Spehro Pefhany

 

[John Larkin]

Interact to the point of not giving the correct state? I've seen
interaction in the 10's of uV, IIRC, but nothing too crazy.

I've seen some outputs, running closed-loop, spike several volts when
another channel railed. It has to do with multi-collector shared
current sources saturating, I think. I'm pretty sure it was the
National parts. I recall Jim saying that some peoples' 324's do it,
and some don't. Anyway, it's another interesting hazard. The 324 is a
really terrible opamp; massive crossover distortion and ghastly
substrate diode issues, too.

One Japanese design from the eighties used half a cheap dual op-amp as
a very low level amplifier, and the other half as a comparator. They
used resistors to feed back the (presumably minus of the measured
average value of the) interaction so they could save a few pennies
parts cost. Clever.

Might be the same problem. I used 324's as comparators (single-slope
adc's in an electric meter) and had terrible interactions. Dropped in
an LF347 and everything worked.

John

 

[Jim Thompson]

I've seen some outputs, running closed-loop, spike several volts when
another channel railed. It has to do with multi-collector shared
current sources saturating,
Correct.

I think. I'm pretty sure it was the
National parts. I recall Jim saying that some peoples' 324's do it,
and some don't.

Yep. That's part of what provoked me, at GenRad, to de-list Motorola
as a vendor for 324's.

Anyway, it's another interesting hazard. The 324 is a
really terrible opamp; massive crossover distortion

Terrible distortion. My cure, for active filters, was to load the
output to class-A.

and ghastly
substrate diode issues, too.

Not sure what you're referring to here. Pulling an input way below
ground??

Might be the same problem. I used 324's as comparators (single-slope
adc's in an electric meter) and had terrible interactions. Dropped in
an LF347 and everything worked.

John

...Jim Thompson

 

[Robert Latest]

On 29 Nov 2007 08:29:22 GMT, Robert Latest <[email protected]>
wrote:

LM5112 maybe.

Nice! 7 amp cap discharge would make for some serious frequency range.

robert

 

[John Larkin]

Yep. That's part of what provoked me, at GenRad, to de-list Motorola
as a vendor for 324's.


Terrible distortion. My cure, for active filters, was to load the
output to class-A.


Not sure what you're referring to here. Pulling an input way below
ground??

Not way below, more like -0.3 volts! If you do that to a single input,
*all* the amp stages can go bonkers. It's really amazing.

The old National linear book, printed on dirty brown tissue paper, had
a tiny, barely-readable footnote on the '324 data sheet, in about
3-point type, that warned of the -0.3 volt thing.

Hey, here it is:

http://s2.supload.com/free/LM324.JPG/view/


John

 

[[email protected]]

Yeah, but the f range is way too limited. Tim's circuit is great, but I need
dozens of those so I'd prefer an integrated solution.

robert

You could perhaps use a PIC675, it has a 10bits AD IIRC, internal osc.
If you grab reference and voltage divider-pot from the same supply,
and the required output freq is not too high, then this can be very
stable.
The internal osc is 4MHz IIRC, and can be software calibrated.
If you need more accuracy, use a xtal, a bunch of those on one xtal
should
work.
Advantage: No capacitor needed, only the PIC and the pot.
Several kHz should be possible, if not more.
And pins left for other fun things, micro-power too.
Low supply, hows' that?

 

[Robert Latest]

Not way below, more like -0.3 volts! If you do that to a single input,
*all* the amp stages can go bonkers. It's really amazing.

The old National linear book, printed on dirty brown tissue paper, had
a tiny, barely-readable footnote on the '324 data sheet, in about
3-point type, that warned of the -0.3 volt thing.

Hey, here it is:

http://s2.supload.com/free/LM324.JPG/view/

Actually it's in the current datasheets of the LM393 as well! I got bitten
by this just yesterday in a circuit where the 393 goes haywire when the
input goes negative, fed through a 50k resistor, despite my having added
clamp diodes. I put in a Schottky diode but haven't tested this yet.

Very annoying. I don't have any resonable negative rail to work with (just
-40V), there's little space on the board etc. I can yank up the input series
resistor and pray to get away with the Schottky. This is a one-off.

0.3V is weird. I always figured I'd pretty safe with a diode drop. Not so.

robert

 

[Jim Thompson]

On Thu, 29 Nov 2007 08:16:26 -0800, John Larkin
[snip]

and ghastly
substrate diode issues, too.

Not sure what you're referring to here. Pulling an input way below
ground??

Not way below, more like -0.3 volts! If you do that to a single input,
*all* the amp stages can go bonkers. It's really amazing.

The old National linear book, printed on dirty brown tissue paper, had
a tiny, barely-readable footnote on the '324 data sheet, in about
3-point type, that warned of the -0.3 volt thing.

Hey, here it is:

http://s2.supload.com/free/LM324.JPG/view/


John

Hmmmm! -0.3V was a CYA thing, it took a vBE below ground to
disrupt... and I've only seen issues with a single OpAmp, not the
whole package... UNLESS: OpAmp railing tripped the PNP mirrors on
certain brands ??

...Jim Thompson

 

[Jan Panteltje]

You could perhaps use a PIC675, it has a 10bits AD IIRC, internal osc.
If you grab reference and voltage divider-pot from the same supply,
and the required output freq is not too high, then this can be very
stable.
The internal osc is 4MHz IIRC, and can be software calibrated.
If you need more accuracy, use a xtal, a bunch of those on one xtal
should
work.
Advantage: No capacitor needed, only the PIC and the pot.
Several kHz should be possible, if not more.
And pins left for other fun things, micro-power too.
Low supply, hows' that?

PS, you could have 3 in one PIC too.
+, GND, 3 x pot in, 3 x out = 8 pins.
The ADC has a mux.

 

[John Larkin]

On Thu, 29 Nov 2007 08:16:26 -0800, John Larkin
[snip]

and ghastly
substrate diode issues, too.

Not sure what you're referring to here. Pulling an input way below
ground??

Not way below, more like -0.3 volts! If you do that to a single input,
*all* the amp stages can go bonkers. It's really amazing.

The old National linear book, printed on dirty brown tissue paper, had
a tiny, barely-readable footnote on the '324 data sheet, in about
3-point type, that warned of the -0.3 volt thing.

Hey, here it is:

http://s2.supload.com/free/LM324.JPG/view/


John

Hmmmm! -0.3V was a CYA thing, it took a vBE below ground to
disrupt... and I've only seen issues with a single OpAmp, not the
whole package... UNLESS: OpAmp railing tripped the PNP mirrors on
certain brands ??

...Jim Thompson

It was not far from 0.3 on the National parts, and all four opamps
would rail, one polarity at first, then they'd rail in the other
direction as the clamping current was increased. I wouldn't joke about
a thing like this.

John

 

[Robert Latest]

You could perhaps use a PIC675, it has a 10bits AD IIRC, internal osc.
If you grab reference and voltage divider-pot from the same supply,
and the required output freq is not too high, then this can be very
stable.

Of course I thought about that, especially since I could just read out, say,
10 pots via a multiplexer and control 10 timing channels that way.

Problem is: I've never gotten my head around any embedded design, so I'm
staying in the analog domain.

BTW, the gadget in question is a "sequencer" that actuates some relays after
individually programmable delays. It's going to be used for special effects
in a film studio.

robert

 

[John Larkin]

Actually it's in the current datasheets of the LM393 as well! I got bitten
by this just yesterday in a circuit where the 393 goes haywire when the
input goes negative, fed through a 50k resistor, despite my having added
clamp diodes. I put in a Schottky diode but haven't tested this yet.

A *power* schottky should work.

Very annoying. I don't have any resonable negative rail to work with (just
-40V), there's little space on the board etc. I can yank up the input series
resistor and pray to get away with the Schottky. This is a one-off.

0.3V is weird. I always figured I'd pretty safe with a diode drop. Not so.

robert

The LM35 family of temperature sensors is similarly delicate. The
thing about them is that National *tells* you to pull the outputs
below ground to read negative temperatures. My advice is to not do it.
If you want to read below 0C, use the Fahrenheit version and don't
pull it below ground.

John

 

[John Fields]

I suggested that opamps might be useful as level translators. That's
not an argument over details.

---
That's not what I was talking about.

Read down to where I wrote: "Sounds like lots details to me..." for
a clue
---

Sure, but that's only once, and that voltage may be free anyhow, like
from an led pilot light or something.

---
Yeah, that's a good idea. With the spec's for the LED's Vf all over
the place and its tempco causing that Vf to drift, it's going to be
really likely that it's going to fit in just right between the
logic's Voh an Vol. Oh, well, though, you've got it covered with
that: "or something."
---

Are you this annoying in real
life?

---
If you're getting annoyed, then this _is_ real life.
---

Maybe the loads can short? We don't know.

---
Youre grasping at straws.
---

I do know that an opamp can
be a lot of stuff in a small, very cheap package.

---
Really???

Geez, thanks, John! That's good to know.
---

Quad opamp and no resistors. But don't use an LM324 for this; the
sections can interact when used open-loop.

---
Got a part number that'll work and be less expensive than my way?
---

A lot of it is routine, as engineering tends to be. But then, we do a
lot of stuff, and some of it is cool. And my life is fairly cool
because I designed it to be that way; most anybody can do that if they
want to.


So you want me to only talk about electronics in the abstract, things
I have no personal connection to, stuff out of peer-reviewed journals?

---
Certainly not, but that's not what I was talking about. What I
_was_ talking about was that since you said "we" in reference to the
design, then you must have been a part of it, which is in conflict
with your statement: "I didn't design this particular level
shifter,"

 

[Tim Williams]

BTW, the gadget in question is a "sequencer" that actuates some relays after
individually programmable delays. It's going to be used for special effects
in a film studio.

As in one-shots? Ah, then you may want this instead:
http://webpages.charter.net/dawill/Images/One-shot Timer.gif

Some day I should redraw that with a two-transistor flip-flop.

Operation: rising edge causes Qbar to fall. Capacitor charges by about 3V,
then the feedback junk flops and the capacitor discharges. And of course,
other values may get better performance (like for minimum flop current from
R_T).

Heck, a monostable can be made with two transistors, around four resistors
and a capacitor. This is kind of a verbose version of that.

Tim