Need Op Amp for design

On 2017-11-20 11:20, Leslie Rhorer wrote:
>> Paul isn't making it complicated, he is on the right track. You can
>> use FETs to replace the Reed switches you have now. If your stack
>> remains below 25V you could pick some that have a Vgs breakdown
>> rating of 30V or higher. If using N-channels (more available and
>> usually cheaper) the top FET would have to be bootstrap-driven but
>> that's not difficult.
>>
>> This way you can use any old opamp and it just gets turned off
>> along with all the FETs when the system is shut down.
>
> Yes, but what is that saving me?  Not cost.  The FETs run about $.50
> each, and I will need at least 6, so that is $3.10 per light, vs.
> less than $.20.


I have never paid 50c for a BSS84 or something similar. For hobbyists I 
have seen them as low as $2 for 50 of them with free shipping on Ebay. I 
would not use such offers commercially, of course, but there you have 
distributors with even lower prices for mass production.


   They certainly would not save me space, as 6
> discrete FETs plus one Op Amp will take more space than 2 quad Op
> Amps.
>

The two quad opamps seem not to work for you, which I understand is the 
reason you posted the question here.

As for space you can also get many FETs as arrays of two or more. Also 
in SC75 package if space is tight.


>
>> What I don't understand is where you mention that cost matters
>> while you only plan on producung 20 units.
>
> I don't know about you, but I am not made of money, and since I will
> not be selling these units, cost matters very much.


Ok, we have a different philosophy. When I design something that runs in 
the million units per year I turn over every penny. But trying to save 
five bucks for a run of 20 units while spending your valuable time on 
trying to find some cheaper solution does not make sense to me. What is 
your time worth per minute?


>                                               ... Yes, I surely
> could afford $3, but why spend the money if it does not provide a
> superior solution?


It does provide what you call a superior situation unless you can find a 
mystery opamp whose parasitic diodes do not come on when it is powered down.


  In what way would discrete FETs provide a
> superior solution?


It works in your situations. It seems your opamps don't.


>               ... (In terms of operational superiority - I realize
> their current draw is a tiny fraction of the Op Amp solution.)
>

Exactly.

-- 
Regards, Joerg

http://www.analogconsultants.com/

"Leslie Rhorer"  wrote in message 
news:b8da0074-51cf-4c2c-82b1-b31d87be209b@googlegroups.com...

> Oh, by the way, you are referring to the outputs of the Op Amps.  They are 
> not relevant to the discussion, because we know they will perform as 
> required, which is to say they will fall below 1V WRT ground whenever 
> their power is removed.  The only issue here is the non-inverting inputs, 
> at least 2 of which will be subjected to potentials higher than 15V when 
> Vcc is removed.

I found "over the top" op-amps such as LT1639 that draws only 16 uA with 
power removed and 20 volts applied. Its output is some 690 mV under such 
conditions, and costs about $6 each.

http://cds.linear.com/docs/en/datasheet/16389fg.pdf

There is also the LT1496 and LT1694 which draws only 235 pA under the same 
conditions. It costs about $7 each.

http://cds.linear.com/docs/en/datasheet/149456fb.pdf

Also look at LT1079, and perhaps FET input op-amps like the TL084.

You could easily make a test circuit and see how the LM324 and other op-amps 
behave. Simulations sometimes don't match real world results, especially in 
unusual conditions like this. For example the LT1674 shows an output of -301 
mV under such conditions, with input current of 235 pA.

You don't show how your voltage monitor works. That may dictate the best 
approach to the power-off problem. It seems that you are off-handedly 
dismissing most of the ideas presented without fully understanding and 
analyzing the cost/performance benefits. You seem to want us to provide free 
advice to save you a few dollars on your flashlights, while our engineering 
efforts have taken maybe an hour from each of us, who would normally charge 
$50-$100 per hour. Your attitude is not very conducive to additional help.

Paul 

On Monday, November 20, 2017 at 5:11:06 PM UTC-5, P E Schoen wrote:
> "Leslie Rhorer"  wrote in message 
> news:b8da0074-51cf-4c2c-82b1-b31d87be209b@googlegroups.com...
> 
> > Oh, by the way, you are referring to the outputs of the Op Amps.  They are 
> > not relevant to the discussion, because we know they will perform as 
> > required, which is to say they will fall below 1V WRT ground whenever 
> > their power is removed.  The only issue here is the non-inverting inputs, 
> > at least 2 of which will be subjected to potentials higher than 15V when 
> > Vcc is removed.
> 
> I found "over the top" op-amps such as LT1639 that draws only 16 uA with 
> power removed and 20 volts applied. Its output is some 690 mV under such 
> conditions, and costs about $6 each.
> 
> http://cds.linear.com/docs/en/datasheet/16389fg.pdf
> 
> There is also the LT1496 and LT1694 which draws only 235 pA under the same 
> conditions. It costs about $7 each.
> 
> http://cds.linear.com/docs/en/datasheet/149456fb.pdf
> 
> Also look at LT1079, and perhaps FET input op-amps like the TL084.
> 
> You could easily make a test circuit and see how the LM324 and other op-amps 
> behave. Simulations sometimes don't match real world results, especially in 
> unusual conditions like this. For example the LT1674 shows an output of -301 
> mV under such conditions, with input current of 235 pA.

Right, wire up some LM324's and let us know*.  A dip over copper clad
with rosin solder (Kester '44 or similar) is good to a nA...at least. 
Well that's what I'd do.  

George H.
*I was thinking JL and others may be in the know.  
 
> 
> You don't show how your voltage monitor works. That may dictate the best 
> approach to the power-off problem. It seems that you are off-handedly 
> dismissing most of the ideas presented without fully understanding and 
> analyzing the cost/performance benefits. You seem to want us to provide free 
> advice to save you a few dollars on your flashlights, while our engineering 
> efforts have taken maybe an hour from each of us, who would normally charge 
> $50-$100 per hour. Your attitude is not very conducive to additional help.
> 
> Paul

> I have never paid 50c for a BSS84 or something similar. For hobbyists I 
> have seen them as low as $2 for 50 of them with free shipping on Ebay. I 
> would not use such offers commercially, of course, but there you have 
> distributors with even lower prices for mass production.

Those are the prices that came up from Newark, Mouser, and DigiKey.

> The two quad opamps seem not to work for you, which I understand is the 
> reason you posted the question here.

No, I did not say they would not work, I asked if they would, and if not, for some alternates.  I know that some Op Amps will be destroyed if either input exceeds the Vcc rail.  Some can be destroyed even if Vcc is floating.  Some survive intact one or both ways.

> As for space you can also get many FETs as arrays of two or more. Also 

I found some dual FETs.  I did not find any with 4 or more.

> Ok, we have a different philosophy. When I design something that runs in 
> the million units per year I turn over every penny. But trying to save 
> five bucks for a run of 20 units while spending your valuable time on 
> trying to find some cheaper solution does not make sense to me. What is 
> your time worth per minute?

In business, a fair bit.  This is not a business venture, and I am not a professional EE.  I am doing this for the joy of creating something unique

> It does provide what you call a superior situation unless you can find a 
> mystery opamp whose parasitic diodes do not come on when it is powered down.

Any Op Amp with PNP or FET inputs will not, as long as the breakdown voltage or greater is applied to the input. The question is not, "Hw much current will it draw?", but rather, "Will it survive the experience without damage.  Yes, of course if one does exceed the breakdown potential, then the unit will draw a lot of current, but the point is moot.

>   In what way would discrete FETs provide a
> > superior solution?
> 
> 
> It works in your situations. It seems your opamps don't.


> 
> 
> >               ... (In terms of operational superiority - I realize
> > their current draw is a tiny fraction of the Op Amp solution.)
> >
> 
> Exactly.
> 
> -- 
> Regards, Joerg
> 
> http://www.analogconsultants.com/

On 2017-11-20 23:27, Leslie Rhorer wrote:
>> I have never paid 50c for a BSS84 or something similar. For
>> hobbyists I have seen them as low as $2 for 50 of them with free
>> shipping on Ebay. I would not use such offers commercially, of
>> course, but there you have distributors with even lower prices for
>> mass production.
>
> Those are the prices that came up from Newark, Mouser, and DigiKey.
>

The prices that come up on my computer are different:

https://www.mouser.com/search/ProductDetail.aspx?R=0virtualkey0virtualkeyBSS84

As I mentioned, for hobby purposes you can get them for a few pennies 
via EBay.


>> The two quad opamps seem not to work for you, which I understand is
>> the reason you posted the question here.
>
> No, I did not say they would not work, I asked if they would, and if
> not, for some alternates.  I know that some Op Amps will be destroyed
> if either input exceeds the Vcc rail.  Some can be destroyed even if
> Vcc is floating.  Some survive intact one or both ways.
>

Which ones would survive and not pull excessive input current when shut 
down? I could imagine there being some sort of boutique amp that does 
this but those tend to be expensive.


>> As for space you can also get many FETs as arrays of two or more.
>> Also
>
> I found some dual FETs.  I did not find any with 4 or more.
>
>> Ok, we have a different philosophy. When I design something that
>> runs in the million units per year I turn over every penny. But
>> trying to save five bucks for a run of 20 units while spending your
>> valuable time on trying to find some cheaper solution does not make
>> sense to me. What is your time worth per minute?
>
> In business, a fair bit.  This is not a business venture, and I am
> not a professional EE.  I am doing this for the joy of creating
> something unique
>
>> It does provide what you call a superior situation unless you can
>> find a mystery opamp whose parasitic diodes do not come on when it
>> is powered down.
>
> Any Op Amp with PNP or FET inputs will not, as long as the breakdown
> voltage or greater is applied to the input. The question is not, "Hw
> much current will it draw?", but rather, "Will it survive the
> experience without damage.  Yes, of course if one does exceed the
> breakdown potential, then the unit will draw a lot of current, but
> the point is moot.
>

The datasheets usually answer that in the fine print such as here on page 4:

http://www.st.com/content/ccc/resource/technical/document/datasheet/16/b5/cf/d3/34/29/4b/09/CD00000489.pdf/files/CD00000489.pdf/jcr:content/translations/en.CD00000489.pdf

Quote "The magnitude of the input voltage must never exceed the 
magnitude of the supply voltage or 15 volts, whichever is less". The 
"whichever is less" would be key in your case.

While a CMOS opamp where a parasitic diode path is coming on or a BJT 
opamp where a BE junction begins to zener (they usually do above 5-7V) 
such as the LM324 may not be destroyed if you provide a large enough 
resistor up front the current that will flow can throw your idea out the 
window. Unless there is some current that is acceptable to you and still 
provides sufficient measuring precision (that can be tough). Nanoamps, 
not really going to happen. One question to ask here though is why 
nanoamps. Is the battery so small that its self discharge is not orders 
of magnitude higher?

Here is a totally different idea but it might take the fun out of the 
project: How about a ready-to-roll Li-Ion stack supervisor IC? Those 
usually can be set into a very low power "ship mode", a mode that is 
used when battery packs with these devices connected are spending a long 
time in containers, in warehouses or on store shelves. Example:

http://www.ti.com/lit/ds/symlink/bq76930.pdf

[...]

-- 
Regards, Joerg

http://www.analogconsultants.com/

John Larkin wrote...
>
> The ancient LM324 has PNP input transistors, and the
> safe input voltages can go to +32 irrespective of V+.

 I also thought of PNP-input single-supply op-amps as
 a good solution, but the O.P. seems not to notice.


-- 
 Thanks,
    - Win

On 2017-11-21 11:35, Winfield Hill wrote:
> John Larkin wrote...
>>
>> The ancient LM324 has PNP input transistors, and the
>> safe input voltages can go to +32 irrespective of V+.
>

The datasheet says "Don't do that" though. Looking at the innards on 
page 4 it seems you'd hit a diode path to V+:

http://www.st.com/content/ccc/resource/technical/document/datasheet/bd/fc/46/43/26/8f/40/7f/CD00001046.pdf/files/CD00001046.pdf/jcr:content/translations/en.CD00001046.pdf

Quote page 4, footnote (1) "Neither of the input voltages must exceed 
the magnitude of (VCC+) or (VCC-)".


>   I also thought of PNP-input single-supply op-amps as
>   a good solution, but the O.P. seems not to notice.
>


Even if there wasn't a diode path, wouldn't the then reversed BE 
junction zener somewhere around 6V?

-- 
Regards, Joerg

http://www.analogconsultants.com/

Winfield Hill <hill@rowland.harvard.edu> wrote:

> John Larkin wrote...

>> The ancient LM324 has PNP input transistors, and the safe input
>> voltages can go to +32 irrespective of V+. 
 
>  I also thought of PNP-input single-supply op-amps as
>  a good solution, but the O.P. seems not to notice.
 
That's ok, Win. These people are probably too young to know who you are, or 
to have read AOE. 

Not to mention the Radiotron Designer's Handbook, the Radiation Lab Series, 
Terman, AT&T, and so on.

This will probably happen more often as time goes on. Never mind. Just keep 
posting your gems as always. Some will listen.

For those who are interested, here are some links:

  Electrons And Holes In Semiconductors
  by William Shockley 1950
  http://preview.tinyurl.com/y77jxe6q

  Radiotron Designers Handbook 1954 Edition
  http://preview.tinyurl.com/hmnpj2r

  Radiation Lab Series
  http://www.febo.com/pages/docs/RadLab/

  Reference Data For Engineers
  http://preview.tinyurl.com/ycjcp6cp

  Terman, 1943
  https://preview.tinyurl.com/ya5ho5dt

- url's verified on Tue, Nov 21, 2017
- all links show a preview so you can see where they will take you

Joerg wrote...
>
>On 2017-11-21 11:35, Winfield Hill wrote:
>> John Larkin wrote...
>>>
>>> The ancient LM324 has PNP input transistors, and the
>>> safe input voltages can go to +32 irrespective of V+.
>>
>
> The datasheet says "Don't do that" though. Looking at
> the innards on page 4 it seems you'd hit a diode path to V+:

 That's ST's LM324A, their non-A part doesn't have the diodes.

>>   I also thought of PNP-input single-supply op-amps as
>>   a good solution, but the O.P. seems not to notice.
>
> Even if there wasn't a diode path, wouldn't the
> then reversed BE junction zener somewhere around 6V?

 These old silicon IC pnp trannies don't suffer from a
 low Veb breakdown.  The abs max spec of most company's
 LM324 says you can go 28/ 32V above the neg rail: e.g.,
 NSC, ON Semi, NXP, UTC, TI, Fairchild, Bay Linear, etc.


-- 
 Thanks,
    - Win

On Tue, 21 Nov 2017 12:05:59 -0800, Joerg <news@analogconsultants.com>
wrote:

>On 2017-11-21 11:35, Winfield Hill wrote:
>> John Larkin wrote...
>>>
>>> The ancient LM324 has PNP input transistors, and the
>>> safe input voltages can go to +32 irrespective of V+.
>>
>
>The datasheet says "Don't do that" though. Looking at the innards on 
>page 4 it seems you'd hit a diode path to V+:
>
>http://www.st.com/content/ccc/resource/technical/document/datasheet/bd/fc/46/43/26/8f/40/7f/CD00001046.pdf/files/CD00001046.pdf/jcr:content/translations/en.CD00001046.pdf
>
>Quote page 4, footnote (1) "Neither of the input voltages must exceed 
>the magnitude of (VCC+) or (VCC-)".
>
>

The original National data sheet specifically allows +32 irrespective
of V+. Can't trust those eye-talian chips. [1]

But DO NOT allow an input to go below V-, operating. All sorts of
weird stuff will happen.

>>   I also thought of PNP-input single-supply op-amps as
>>   a good solution, but the O.P. seems not to notice.
>>
>
>
>Even if there wasn't a diode path, wouldn't the then reversed BE 
>junction zener somewhere around 6V?

PNPs tend to have higher zener voltage. The 324 must have some
special, old-fashioned diffusion.

[1] And why do we sell them good wheat for a few dollars a bushel, and
let them add water, take out the water, and sell it back to us for $8
a pound?

-- 

John Larkin   Highland Technology, Inc   trk

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