V_be at 1uA

Thinking of using discrete darlingtons a bit like comparators, where -in is effectively 2x Vbe. I'd need to get as much accuracy/consistency as possible in this case - which I don't expect to be good. At turned on the situation will be: I_b 1uA. I_c 1mA. V_ce saturated (low). P_diss

Reply to
tabbypurr
Loading thread data ...

Ebbers and Moll. Vbe drops by 60mV per decade of current at room temperature.

One uA isn't all that low. The sugbect is discussed in great detail by people who talk about logarithmic amplifiers.

Keeping the voltage on the collector much the same as voltage being applied to the base seems to be helpful if you want the logarithmic relationship between base-emitter voltage and emitter current to go down nanoamps.

Phil Hobbs has produced what sound like well-informed rants on the subject.

You are stuck with the extended Gummel-Poon model of the bipolar transistor.

If you are on friendly terms with semi-conductor manufacturer you may be alble to get the parameter for the rather superior VB

Reply to
bill.sloman

Ebbers and Moll is probably all you need at 1uA. This is three orders of magnitude below 1mA where the data sheet Vbe is often given, which is to say the Vbe would be about 180mV below the data sheet range (at room temperature).

"The Art of Electronics" spells it out - on page 91 of the 3rd edition.

"Logging" circuits - which exploit this to produce a voltage proportional to emitter current - seem to be okay down to the nanoamp level, if done carefully.

LTSpice mostly offers the Gummel-Poon transistor model. It will also run the rather better VBIC model, but transistor manufacturers treat the VBIC parameters for their transistors as trade secrets.

Apparently VBIC models transistors operating in the inverted mode a whole lot better than Gummel-Poon, and might be able to simulate the "squegging" in bipolar realisations of the Baxandall Class-D inverter, but I've not been able to test this.

--
Bill Sloman, Sydney
Reply to
bill.sloman

National Semiconductor application note AN-29 includes a logging amplifier, and AN-30 talks about nothing else.

--
Bill Sloman, Sydney
Reply to
bill.sloman

May not be relevant, but I use diode connected (single) transistors at 0.1 uA to 1 mA. Vbe variation is +/-5 mV (at a given current) at room temp. I've had one of two (out of a few hundred) that have been outside this range.. so maybe 1%.

George H. (no idea about darlingtons.)

Reply to
George Herold

A pnp emitter follower ahead of an npn has roughly zero offset and zero temperature coefficient, still two transistors.

--
John Larkin         Highland Technology, Inc 

lunatic fringe electronics
 Click to see the full signature
Reply to
John Larkin

And won't oscillate at 1 uA collector current. ;)

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
 Click to see the full signature
Reply to
Phil Hobbs

in is effectively 2x Vbe. I'd need to get as much accuracy/consistency as p ossible in this case - which I don't expect to be good. At turned on the si tuation will be:

on? The graphs I've looked at just don't go down that low. Spice doesn't co ver this territory. Before I get a pile & measure them, has anyone already done likewise?

I tend to do small stuff pushing as much functionality from it as possible , and LTspice is really not set up for that. Most small circuits I don't sp ice at all.

e threshold some, but also raise it. This isn't ideal.

Yes, but Rin is around 1Mohm, and with V_in near the 0v rail there's pretty much zero base current in. If I start adding bias etc R_in falls, I want t o avoid that.

Tempco of a darlington is terrible but it always runs stone cold. Not even a milliwatt. Sziklai would be go lower but life will be easier with darling ton, I won't bore you with why. The tempco & offset are tolerable.

1M? isn't much cop as a Q killer. Ferret bees won't do the buzziness w ith no current flow. There are several ways I could do it, discrete darling ton, prepacked darlington, sziklai, pnp emitter follower, and other approac hes like comparators etc. I'm trying to keep it rock bottom but squeeze tol erable accuracy & high R_in from it. All options are compromises one way or another - no news there.

I've been looking at datasheets. Some graphs do go that low (fwiw) and ther e doesn't seem to be much Vbe difference between 100mA & 500mA parts at 1uA & 30uA I_b, so that seems not to be an issue after all. I'll get about 1.2 v compound Vbe from a discrete darlington, and that's reasonable for the jo b.

Pre-packed darlingtons have less current gain + current loss at the 2nd tr base, resulting in much worse R_in, and are harder to source so I won't use those.

Thanks everyone.

NT

Reply to
tabbypurr

Very useful, thank you George.

NT

Reply to
tabbypurr

that's not going to work, +in as 2 V_be will work, but badly.

--
  When I tried casting out nines I made a hash of it.
Reply to
Jasen Betts

I've always said there's a place for worse opamps & comparators, and there is.

NT

Reply to
tabbypurr

Let us see the final total circuit when you're finished with all these bits and pieces of zero parts count stuff...

Reply to
bloggs.fredbloggs.fred

n is effectively 2x Vbe. I'd need to get as much accuracy/consistency as po ssible in this case - which I don't expect to be good. At turned on the sit uation will be:

n? The graphs I've looked at just don't go down that low. Spice doesn't cov er this territory. Before I get a pile & measure them, has anyone already d one likewise?

I tend to do small stuff pushing as much functionality from it as possible, and LTspice is really not set up for that. Most small circuits I don't spi ce at all.

threshold some, but also raise it. This isn't ideal.

ts and pieces of zero parts count stuff...

It's for a bargraph display. Supplies of LM3914, 1549 etc aren't reliable. The 'comparator', which only compares with a fixed voltage, is just a darli ngton with 1mA collector output, supplied with 1uA in, relying on a beta of at least 33 in saturation, with less beta being ok if/when the other tr ha s more. It has 1megohm on the base to make R_in high, so it begins to turn on at 1.2v in, and is conducting 1mA at or before 2v in. Varying element br ightness this way gives some extra resolution at zero cost. The trs are any npn jellybean.

Comparators etc work but parts where the part number doesn't matter are pre ferred. LED current is controlled by resistors, no fancy variable constant current stuff. Sziklais have advantages but batches of trs are mainly NPN, and darlingtons are good enough in an unheated case.

Parts used are whatever's available/cheapest & works. Unknown part numbers can be used if they pass. Mouser, Element14 etc are too expensive.

NT

Reply to
tabbypurr

How many LED bars? You can get an LM339 quad comparator for 10 cents. A little triangle wave modulation would make the LEDs fade smoothly into one another.

We buy resistor quad packs for 0.6 cents each.

--
John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  
 Click to see the full signature
Reply to
John Larkin

e:

-in is effectively 2x Vbe. I'd need to get as much accuracy/consistency as possible in this case - which I don't expect to be good. At turned on the situation will be:

tion? The graphs I've looked at just don't go down that low. Spice doesn't cover this territory. Before I get a pile & measure them, has anyone alread y done likewise?

d. I tend to do small stuff pushing as much functionality from it as possib le, and LTspice is really not set up for that. Most small circuits I don't spice at all.

the threshold some, but also raise it. This isn't ideal.

s.

bits and pieces of zero parts count stuff...

e. The 'comparator', which only compares with a fixed voltage, is just a da rlington with 1mA collector output, supplied with 1uA in, relying on a beta of at least 33 in saturation, with less beta being ok if/when the other tr has more. It has 1megohm on the base to make R_in high, so it begins to tu rn on at 1.2v in, and is conducting 1mA at or before 2v in. Varying element brightness this way gives some extra resolution at zero cost. The trs are any npn jellybean.

preferred. LED current is controlled by resistors, no fancy variable consta nt current stuff. Sziklais have advantages but batches of trs are mainly NP N, and darlingtons are good enough in an unheated case.

rs can be used if they pass. Mouser, Element14 etc are too expensive.

One's 9 segment, another's 6. There may be another later with 20.

10 cents for a 339 is far too much. Triangle modulation means more parts, m ore labour, more pcb, more cost. Every component is under a cent a piece. T his stuff has to sell in the developing world.

When I worked for a very profitable large corporation I realised the busine ss would fail. That sort of corporate culture builds inefficiency, then bui lds more on top of it. I knew that the first skilful startup would wipe the m out by undercutting them, and they wouldn't have the time or the business /management skill to know how to recover from it. I'm sure they would have laughed if I'd said anything - I was a newbie in a lucrative business with a long history. The business no longer exists. That was where I began to wo rk out the many ways costs could be cut while still maintaining quality.

You're in a very different market and obviously need to source from compani es where everything is traceable back to the manufacturer. For small consum er devices & low cost testgear this works.

NT

Reply to
tabbypurr

ote:

re -in is effectively 2x Vbe. I'd need to get as much accuracy/consistency as possible in this case - which I don't expect to be good. At turned on th e situation will be:

uation? The graphs I've looked at just don't go down that low. Spice doesn' t cover this territory. Before I get a pile & measure them, has anyone alre ady done likewise?

eed. I tend to do small stuff pushing as much functionality from it as poss ible, and LTspice is really not set up for that. Most small circuits I don' t spice at all.

p the threshold some, but also raise it. This isn't ideal.

ons.

se bits and pieces of zero parts count stuff...

ble. The 'comparator', which only compares with a fixed voltage, is just a darlington with 1mA collector output, supplied with 1uA in, relying on a be ta of at least 33 in saturation, with less beta being ok if/when the other tr has more. It has 1megohm on the base to make R_in high, so it begins to turn on at 1.2v in, and is conducting 1mA at or before 2v in. Varying eleme nt brightness this way gives some extra resolution at zero cost. The trs ar e any npn jellybean.

e preferred. LED current is controlled by resistors, no fancy variable cons tant current stuff. Sziklais have advantages but batches of trs are mainly NPN, and darlingtons are good enough in an unheated case.

bers can be used if they pass. Mouser, Element14 etc are too expensive.

more labour, more pcb, more cost. Every component is under a cent a piece. This stuff has to sell in the developing world.

ness would fail. That sort of corporate culture builds inefficiency, then b uilds more on top of it. I knew that the first skilful startup would wipe t hem out by undercutting them, and they wouldn't have the time or the busine ss/management skill to know how to recover from it. I'm sure they would hav e laughed if I'd said anything - I was a newbie in a lucrative business wit h a long history. The business no longer exists. That was where I began to work out the many ways costs could be cut while still maintaining quality.

Clive Sinclair seems to have adopted a similar approach to design - except that no supplier was dodgy enough to fall below his quality demands if thei r parts were cheap enough.

nies where everything is traceable back to the manufacturer. For small cons umer devices & low cost testgear this works.

Thurlby-Thandar did seem to be able to sell cheap test gear that worked, bu t they eased Clive Sinclair out of it after they'd exploited his bright ide as but before his enthusiasm for buying parts which had clearly come from m anufacturers reject bins could wreck their production yields.

--
Bill Sloman, Sydney
Reply to
bill.sloman

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.