High-side and over-the-top current sensing

On Saturday, January 2, 2016 at 12:52:56 AM UTC-5, snipped-for-privacy@yahoo.com wro te:

rote:

worse (that's 1000-10,000 x) than a monolithic part.

r. Offset voltage is

uld have bias

Keeping a monolithic

inda hefty. The cut-in voltage on Q1A is therefore a big unknown. The Vbe c ancelation on Q3-Q4 is a crude attempt, there's no bandwidth to his feedbac k with that set-up and Q1A being driven common emitter.

it,

re

) of

tched

mical

is pretty cheap and available from Mouser. The 10V shutdown is a little slo ppy but should fit the bill. This circuit can be adapted for way higher vol tages than you're dealing with. The output drive biasing is based on scratc hpad calcs using nominal Vbe's, diode drops and current gains. It should be run through SPICE environmentals for best compromise or even correct opera tion. Then you might want to add some terminal protection or your controlle r in the event something blows through.

ts

ng

L,

nt

I like that ZXCT1109 with its ultra-low current, those other parts don't co me close and that LM7301 is absurd. Your zeners are current starved as shown, just a guess but you won't see mo re than around 85% nominal Vz out of those...

rote:

:

e worse (that's 1000-10,000 x) than a monolithic part.

fer. Offset voltage is

would have bias

t. Keeping a monolithic

kinda hefty. The cut-in voltage on Q1A is therefore a big unknown. The Vbe cancelation on Q3-Q4 is a crude attempt, there's no bandwidth to his feedb ack with that set-up and Q1A being driven common emitter.

cuit,

e

more

,

yp) of

matched

nomical

w is pretty cheap and available from Mouser. The 10V shutdown is a little s loppy but should fit the bill. This circuit can be adapted for way higher v oltages than you're dealing with. The output drive biasing is based on scra tchpad calcs using nominal Vbe's, diode drops and current gains. It should be run through SPICE environmentals for best compromise or even correct ope ration. Then you might want to add some terminal protection or your control ler in the event something blows through.

arts

ting

rs

0CL,

Figure 5.

---------

gant

d

come close and that LM7301 is absurd.

more than around 85% nominal Vz out of those...

The Fig. 2 zener is starved, true. That's unavoidable, to keep Iq low. The others can shoot a mA or so, which is plenty.

I don't think starving matters for a 10V zener--knees are pretty sharp. 10v and 4.7v are not the same animals. See, for example, Fig. 5 in this document:

or Fig. 9 in this one:

Zener voltage for the 10v-range parts barely changes from 10uA to 10mA.

Meanwhile, John's prompt got me looking at older jellybean JFET op amps.

TL062 $.22 @ 100 Mouser (Iq=200uA per ampl)

That's only a nickel more than my matched pair.

The TL062 and TL082 (faster, Iq=1.4mA/ampl) advertise Vcm(max) = V+, but don't guarantee parameters above (V+) - 4V. That's too bad, otherwise we could do this:

Figure 6.

--------- 12-17V --//------+--[Rs]--+---------------//- | | TL062 | | |\ [R1] '---|-\ Q2 | | >-+-[R4]--. .--> +--------+---|+/ | 1.5k V / | | |/U1b| ----- TL062 | | | | |\ | '--[R3]--' |

+5 >--|+\ |/ ,-' D1 | >----| Q1 ^ 9.1V .--|-/ |>. 4R 2Q 1R 1U = 8 parts | | |/ | === | U1a | '------------+ | [R2] Iq ~=500uA | ===

U1a removes Q1's Vbe drift from the current sink. The Q2 output stage reduces output voltage drift to mostly that of zener diode D1.

Figure 7:

---------

Alternate output stage:

V+ >--//------+--[Rs]--+---------------//- V+ | | TL062 --- | | |\ | [R1] '---|-\ |/Q2 | | >-+-[R4]---+--| +--------+---|+/ | | |>. | | |/ | | '--> | | | ,-' |\ | '--[R3]--' ^ 10V

+5 >--|+\ |/ | | >----| Q1 === .--|-/ |>. 4R 2Q 1R 1U = 8 parts | |/ | | | '------------+ | [R2] Iq ~=500uA | ===

Cheers, James Arthur

You hardly need to close the loop on Q1. But if half an amp is free, why not?

--

John Larkin         Highland Technology, Inc 

lunatic fringe electronics

ote:

tude worse (that's 1000-10,000 x) than a monolithic part.

wafer. Offset voltage is

al would have bias

rent. Keeping a monolithic

mV, kinda hefty. The cut-in voltage on Q1A is therefore a big unknown. The Vbe cancelation on Q3-Q4 is a crude attempt, there's no bandwidth to his fe edback with that set-up and Q1A being driven common emitter.

circuit,

Vbe

nd more

ped,

he

ed

A(typ) of

a matched

economical

.

elow is pretty cheap and available from Mouser. The 10V shutdown is a littl e sloppy but should fit the bill. This circuit can be adapted for way highe r voltages than you're dealing with. The output drive biasing is based on s cratchpad calcs using nominal Vbe's, diode drops and current gains. It shou ld be run through SPICE environmentals for best compromise or even correct operation. Then you might want to add some terminal protection or your cont roller in the event something blows through.

1 parts

v)

setting

rrors

CS30CL,

d,

.

elegant

and

't come close and that LM7301 is absurd.

ee more than around 85% nominal Vz out of those...

10v

Yup. Saves ~four or five percent Itrip drift over temp. Trivial, but free . (I was waiting for complaints I should've used a MOSFET to avoid d(Hfe)/dt! )

The TL062 is actually cheaper than the one-amp TL061. ST's version even specs. Icc=250uA max. total, vs. TI's 250uA max. *per amp.*

ST's datasheet has a nice diagram of the input stage. ISTM it should work up to V+ if the JFETs' Vgs(th) > 1V, which it almost certainly always is.

Anyway, I'm not the one to guarantee it, so sadly, the TL062 is out.

Cheers, James Arthur