Claude,
I am still confused; while I agree that 1N914 can be used for around 5 decades (has 25nA leakage at room temperature and around 0.5ohm series resistence, both according to the data sheet) and I have to admit I never looked into the JPAD diodes (btw, how much does a JPAD1 cost? can't find any reference about prices) it is not true that BJTs are not good enough or are in any way comparable to 1N914. Take a look at this 1964 reference (it's a classic, but sorry, I can't put the file here):
A Circuit with Logarithmic Transfer Response over 9 Decades
Gibbons, J.; Horn, H.
Circuit Theory, IEEE Transactions on
Volume 11, Issue 3, Sep 1964 Page(s): 378 - 384
Summary: The short-circuit collector current of many silicon transistors is proportional to the exponential of the emitter-base voltage over 9 decades of collector current (10^{-11} ato10^{-2} a). Such a transistor may be connected as a three-terminal feedback network in an operational amplifier to obtain accurate logarithmic transfer response over a dynamic range of 9 decades.
Or an application with MAT02 (very conservatively quoted for 6 decades, and intended for optical sensors):
http://www.edn.com/contents/images/336867f1.pdf
And finally a rather exotic application that extends the dynamic range, using BJTs, to 11 decades and dated 1976
http://scitation.aip.org/getabs/ser...00047000007000854000001&idtype=cvips&gifs=yes
It's not necesarry about special transistors, but about providing some sort of thermal compensation and balancing the offset of the opamp that keeps the transistor in the loop at Vcb=0. This is very good explained here:
http://books.google.com/books?id=2k...bFkLcG&sa=X&oi=book_result&ct=result&resnum=8
Funny enough, usually it's not the transistor that limits the dynamic range, but the opamp offset!
I am still confused; while I agree that 1N914 can be used for around 5 decades (has 25nA leakage at room temperature and around 0.5ohm series resistence, both according to the data sheet) and I have to admit I never looked into the JPAD diodes (btw, how much does a JPAD1 cost? can't find any reference about prices) it is not true that BJTs are not good enough or are in any way comparable to 1N914. Take a look at this 1964 reference (it's a classic, but sorry, I can't put the file here):
A Circuit with Logarithmic Transfer Response over 9 Decades
Gibbons, J.; Horn, H.
Circuit Theory, IEEE Transactions on
Volume 11, Issue 3, Sep 1964 Page(s): 378 - 384
Summary: The short-circuit collector current of many silicon transistors is proportional to the exponential of the emitter-base voltage over 9 decades of collector current (10^{-11} ato10^{-2} a). Such a transistor may be connected as a three-terminal feedback network in an operational amplifier to obtain accurate logarithmic transfer response over a dynamic range of 9 decades.
Or an application with MAT02 (very conservatively quoted for 6 decades, and intended for optical sensors):
http://www.edn.com/contents/images/336867f1.pdf
And finally a rather exotic application that extends the dynamic range, using BJTs, to 11 decades and dated 1976
http://scitation.aip.org/getabs/ser...00047000007000854000001&idtype=cvips&gifs=yes
It's not necesarry about special transistors, but about providing some sort of thermal compensation and balancing the offset of the opamp that keeps the transistor in the loop at Vcb=0. This is very good explained here:
http://books.google.com/books?id=2k...bFkLcG&sa=X&oi=book_result&ct=result&resnum=8
Funny enough, usually it's not the transistor that limits the dynamic range, but the opamp offset!
Thanks syn08 you saved me the trouble. The MAT02 will do 7 decades and Vishay as well as Tellabs have made exact TC compensating resistors for more than 30yr. I think the JPAD diodes were originally for input clamping on electrometers, etc. The equivalent leakage can be obtained on some electrometer FET's like the 2N4117/8 (used in smoke detectors). I have a private stash of fA leakage FET pairs that match forward V down to .065V @20fA. Who said Vbe is .6V?
And you remembered the heated diff-pair
And you remembered the heated diff-pair
The 1976 reference refers to a bjt connected *as a diode*. I couldn't obtain a schematic.
Bob Pease claimed 5 diodes for a bjt. Independently I found that 5 decade figure to be exactly correct. I haven't examined every bjt device. I have no doubt that a bjt with more log-linear range can be produced.
Using standard off the shelf cost-effective parts was the quest. Exotic specialty parts were avoided. A 2N2222A, or 2N4401 costs but a penny or two. Likewise for a 1N914B diode. A JPAD at the time was around $0.25 to $0.50, but I believe they're cheaper now.
The V for a given I and the delta V over a given decade of I are also very important. The diode showed a clear advantage. The 9 decade range of the JPAD parts is much more than I needed. In fact the limiting factor for dynamic range using JPADs was not the diode, but the cleanliness of the board. At 0.25V & 1.0 nA, the equivalent resistance is 0.25V/1e-9A = 250 Mohm. To maintain a 1% error, the pcb resistance should be 100 times that, or 25 Gohm. This is hard to do. Even if an op amp has very low offset voltage, maintaing Gohm pcb impedances is quite difficult, and that is what I found to be the limiting factor more so than the logging element.
The dynamic range is limited by pcb moisture/condensation, contamination, etc.
As I've stated, even a 5 decade bjt is more than enough range for my uses. The signal gain, V vs. I, & delta V vs. Idecade was very important to me, more so than the dynamic range. A diode measured better wrt to these parameters.
Also, if the light reaching the photodiode is weak, then the logging element is operating at the bottom of its log-linear range. If a diode or bjt has 5 decades, but the light source is only strong enough to produce a current in the bottom 2 decades of the 5, then 3 decades go unused.
My patent, U.S. no. 5,670,775, addresses this issue with low cost off the shelf parts. The photodiode amp utilizes both positive & negative feedback. The photodiode current, Ipd, gets multiplied by a resistor ratio (1 + (R1+R2)/R3). This boosts the current in the logging diode, Ilog. So, we operate at the top end of the 5 decade span. If you wish to comment, I'd suggest looking up the patent. It has been in use for over a decade ("decade" meaning 10 yrs. of time). It makes optical log measurements easy where they used to be difficult.
Again, dynamic range in decades is only one measure of how effective a logging element is. I am well aware that a bjt, or diode-connected bjt can work well. I never said otherwise. I just stated that when all is considered, the diode has a couple extra features that were well suited for my specific application.
As far as how good the 1N914B is, consider the following. When using it as a rectifier, the large forward voltage drop at a given current value is generally a BAD thing. When rectifying a signal, who wants a diode with a *larger* forward drop? But if a photodiode current, usually small, is passed through a 1N914B used as a logging element, then we want the highest forward voltage drop we can get. Likewise, to obtain high a/d resolution, it is desirable that over a 1 decade span of current, we get the largest delta V we can get.
A 1N914B is great at what I've just described, and costed a penny in the 1990's. Also, the consistency from part to part, batch to batch, was very good.
What more do you want?
Bob Pease claimed 5 diodes for a bjt. Independently I found that 5 decade figure to be exactly correct. I haven't examined every bjt device. I have no doubt that a bjt with more log-linear range can be produced.
Using standard off the shelf cost-effective parts was the quest. Exotic specialty parts were avoided. A 2N2222A, or 2N4401 costs but a penny or two. Likewise for a 1N914B diode. A JPAD at the time was around $0.25 to $0.50, but I believe they're cheaper now.
The V for a given I and the delta V over a given decade of I are also very important. The diode showed a clear advantage. The 9 decade range of the JPAD parts is much more than I needed. In fact the limiting factor for dynamic range using JPADs was not the diode, but the cleanliness of the board. At 0.25V & 1.0 nA, the equivalent resistance is 0.25V/1e-9A = 250 Mohm. To maintain a 1% error, the pcb resistance should be 100 times that, or 25 Gohm. This is hard to do. Even if an op amp has very low offset voltage, maintaing Gohm pcb impedances is quite difficult, and that is what I found to be the limiting factor more so than the logging element.
The dynamic range is limited by pcb moisture/condensation, contamination, etc.
As I've stated, even a 5 decade bjt is more than enough range for my uses. The signal gain, V vs. I, & delta V vs. Idecade was very important to me, more so than the dynamic range. A diode measured better wrt to these parameters.
Also, if the light reaching the photodiode is weak, then the logging element is operating at the bottom of its log-linear range. If a diode or bjt has 5 decades, but the light source is only strong enough to produce a current in the bottom 2 decades of the 5, then 3 decades go unused.
My patent, U.S. no. 5,670,775, addresses this issue with low cost off the shelf parts. The photodiode amp utilizes both positive & negative feedback. The photodiode current, Ipd, gets multiplied by a resistor ratio (1 + (R1+R2)/R3). This boosts the current in the logging diode, Ilog. So, we operate at the top end of the 5 decade span. If you wish to comment, I'd suggest looking up the patent. It has been in use for over a decade ("decade" meaning 10 yrs. of time). It makes optical log measurements easy where they used to be difficult.
Again, dynamic range in decades is only one measure of how effective a logging element is. I am well aware that a bjt, or diode-connected bjt can work well. I never said otherwise. I just stated that when all is considered, the diode has a couple extra features that were well suited for my specific application.
As far as how good the 1N914B is, consider the following. When using it as a rectifier, the large forward voltage drop at a given current value is generally a BAD thing. When rectifying a signal, who wants a diode with a *larger* forward drop? But if a photodiode current, usually small, is passed through a 1N914B used as a logging element, then we want the highest forward voltage drop we can get. Likewise, to obtain high a/d resolution, it is desirable that over a 1 decade span of current, we get the largest delta V we can get.
A 1N914B is great at what I've just described, and costed a penny in the 1990's. Also, the consistency from part to part, batch to batch, was very good.
What more do you want?
Not much. You claimed that a 1N914 diode is generally better than a BJT and I tried to prove this being incorrect. But If you need only 5 decades and add economic constraints, then I agree the diode is the best solution.
BTW, you are incorrect again regarding the PCB impact. Ever heard of "guarding"? This technique makes the PCB impact almost moot, within reasonable (7-9 decades) limits.
scott wurcer said:
And you remembered the heated diff-pair
It was me, actually...
http://www.diyaudio.com/forums/showthread.php?postid=1828135#post1828135
Claude, I am glad for your input on log amps. This is a good example of what real engineers do, and why. Often, design articles like this are spotlighted in 'EDN', 'Electronics Design', and several other publications that working design engineers read on a regular basis. It is the design 'APPROACH' that is more important than brute force results. This is what gives other engineers new ideas, and 'out of the box' thinking. In this case, I noted several differences between common small signal diodes, and normal bipolar transistors, especially diffusion tradeoffs. Several engineers have alreadiy noted this from your input.
This is a major difference between academic design and 'invention'. Academic design tends to approve certain models and approaches and works them to death. 'Invention' tends to look at something and think, why not do it differently?
Think about the movie 'The Dirty Dozen', where former inmates and lowlifes, outsmart the army while conducting war games. Some call it 'Thinking outside the box' others might say that they were never taught to NOT do it that way.
If you sometime examine my patent on a low noise bipolar transistor preamplifier, you will find that some of you will think that it couldn't work at all, because of your pre-conditioning, as to how transistors work.
This is a major difference between academic design and 'invention'. Academic design tends to approve certain models and approaches and works them to death. 'Invention' tends to look at something and think, why not do it differently?
Think about the movie 'The Dirty Dozen', where former inmates and lowlifes, outsmart the army while conducting war games. Some call it 'Thinking outside the box' others might say that they were never taught to NOT do it that way.
If you sometime examine my patent on a low noise bipolar transistor preamplifier, you will find that some of you will think that it couldn't work at all, because of your pre-conditioning, as to how transistors work.
Hi Claude;
I am in love with diodes too.
Here is my unity gain complementary high power opamp (class AB version):
Edit: common point between R1 and R2 trimmers is grounded, of course.
It was an output stage of a hybrid amp, circa 1982
I used diodes because when it goes to self-protection mode (gradually increasing output resistance with decreasing of a load resistance) Vbe breakdown voltage is not enough to survive. Diodes are better.
I am in love with diodes too.
Here is my unity gain complementary high power opamp (class AB version):
Edit: common point between R1 and R2 trimmers is grounded, of course.
It was an output stage of a hybrid amp, circa 1982
I used diodes because when it goes to self-protection mode (gradually increasing output resistance with decreasing of a load resistance) Vbe breakdown voltage is not enough to survive. Diodes are better.
john curl said:
High Claude,
I have really enjoyed your log amp explanations based on your experience as well as engineering priciples, and I agree with John completely. I have learned from this exchange.
I also applaud your patience with the questions that you have been asked, not getting defensive, but rather patiently explaining to the next level. That is how we all learn.
In fact, both sides learn in such an exchange. Sometimes one person will make what others consider too broad a generalization and others will object or ask for justification. Then the caveats come out. We all learn from such exchanges even if it takes a few iterations and sometimes requires a thick skin. Finally, even after such an exchange, everybody need not agree on everything.
Cheers,
Bob
Lumba Ogir said:Wavebourn,
your circuit was way ahead of its time. Hopefully, all diodes were 1N914B.
I would not say so. People were playing already with complementary opamps. I just simplified one a little bit, to fit in my particular need.
Diodes were KD213 (Russian high frequency rectifiers).
john curl said:
Diodes sometimes are preferable to transistors, even on the places traditionally occupied by transistors. You may see on the picture a very powerful unity gain complementary opamp with diodes instead of transistors for an inverting input.
John, it is nothing to do with personalities. It is about solutions of electronic problems using different ways, please relax.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.