Transistor And gate

Thread Starter

HarrisonG

Joined Aug 1, 2016
73
Screenshot_2016-08-12-23-01-52.png
This bipolar transistor AND gate configuration is with the purpose to be independent of Hfe right? And one more question. When do we need to take care of varying Hfe under? I think we don't need to worry about it if we want to light up a simple.thing like a led with a transistor, but we sure need to add an emitter resistor in amplifiers where varying with high temperatures beta can cause noise.
 
Last edited:

#12

Joined Nov 30, 2010
18,224
This bipolar transistor AND gate configuration is with the purpose to be independent of Hfe right?
All good designs are independent of Hfe.
Hfe is different for each transistor in a batch and each temperature. A good designer does not bet the quality of the design on the gain of a transistor.
 

WBahn

Joined Mar 31, 2012
29,978
This is essentially RTL (resistor-transistor logic).

What you need to do in a design such as this is determine reasonable operating specifications. For instance, how many other gates does the output need to be able to drive (known as the fanout)? You also need to decide what your logic levels are, both for inputs and outputs. Then you need to decide what operating temperature you want your logic to work over. Given all of that (and a few other specs involving speed and power consumption) you determine what the minimum and maximum values of beta (and other parameters) or for all the devices. Then you can either choose extreme cases for all of the parameters that push each of the performance specifications as hard as possible and choose components that meet spec under all of those cases. However, this becomes prohibitively complex and time consuming almost immediately. So instead what you do is run thousands (perhaps hundreds of thousands) of simulations in which the parameters are randomly varied over their possible range and you test each combination to see if it meets all of the performance specs or not. This is called a Monte Carlo simulation. You decide what fraction of your circuits are allowed to fail to meet spec and you then adjust the design and/or the component nominal values until you satisfy this goal.

Monte Carlo simulations are done for both analog and digital circuits.
 

hp1729

Joined Nov 23, 2015
2,304
View attachment 110388
This bipolar transistor AND gate configuration is with the purpose to be independent of Hfe right? And one more question. When do we need to take care of varying Hfe under? I think we don't need to worry about it if we want to light up a simple.thing like a led with a transistor, but we sure need to add an emitter resistor in amplifiers where varying with high temperatures beta can cause noise.
How high must the input voltage be for a four or eight input AND gate? VCE sat x 7 plus VBE? Still sounds like it would work.
 

#12

Joined Nov 30, 2010
18,224
What you need to do in a design such as this is determine reasonable operating specifications.
That was quite an education. Designing for an acceptable percentage of failures in a batch. Perfectly valid if you test the parts and don't ship the parts that fail to meet specs under all conditions of time and temperature.
 

WBahn

Joined Mar 31, 2012
29,978
That was quite an education. Designing for an acceptable percentage of failures in a batch. Perfectly valid if you test the parts and don't ship the parts that fail to meet specs under all conditions of time and temperature.
While it is quite common to test 100% of all parts, it is almost unheard of to test them over all conditions because this is simply prohibitive (for commercial or consumer-scale quantities). Instead, a narrower set of test specs is developed that is performed at one or, at most, a very small number of conditions. If the parts meet the narrower specs under the test conditions, then it is assumed that they will meet spec under the full range of test conditions (with an acceptably high expected pass rate). This results in an extremely low number of parts that ship that don't meet the full spec. The overwhelming majority of those parts will still be perfectly usable in the actual application they are put in and so only a tiny fraction of a very tiny fraction will cause problems and the tests are intended to make that fraction acceptable. For applications that are critical, more extensive tests are usually developed and often these are done either at the receiving point of the components or at some intermediate part of the assembly process, although it may be contracted back to the component manufacturer. This is quite expensive, so it is only done when needed and only on the actual parts that are involved (as opposed to the entire production run).
 

#12

Joined Nov 30, 2010
18,224
The overwhelming majority of those parts will still be perfectly usable in the actual application they are put in
The only time I did production runs, we found a 3% failure rate of new chips within 30 days.
You can decide if that's an underwhelming failure rate.
 

AnalogKid

Joined Aug 1, 2013
10,986
When do we need to take care of varying Hfe under?
In practical terms, you don't. Digital logic circuits such as yours are based on all transistors acting as saturated switches. There are no emitter resistors because emitter degeneration is exactly what you do not want. Emitter degeneration effectively reduces the gain of a transistor stage, and you want all transistors operating at maximum gain. As long as the collector current for each transistor is only 10 to 20 times greater than the base current, the transistors will be hard-saturated and Hfe variations due to temperature, aging, production variables, etc. never will be apparent in the circuit's operation.

ak
 

WBahn

Joined Mar 31, 2012
29,978
The only time I did production runs, we found a 3% failure rate of new chips within 30 days.
You can decide if that's an underwhelming failure rate.
For some things that might be pretty good and for others it might be abysmal. I can't decide based on the information I have whether it was underwhelming or not.

I also can't tell if the 3% was for chips that had passed initial testing, nor whether the 30 days was part of the testing plan or not to capture burn-in and infancy effects.
 

#12

Joined Nov 30, 2010
18,224
The 3% was for LM723 chips that worked in the first place. They passed the final inspection QC tests and the products were shipped to the customer. 3% of the products came back within 30 days with a dead chip.
 

WBahn

Joined Mar 31, 2012
29,978
The 3% was for LM723 chips that worked in the first place. They passed the final inspection QC tests and the products were shipped to the customer. 3% of the products came back within 30 days with a dead chip.
I would think that that would have been an unacceptable failure rate (and mode). Did they take steps to investigate and correct it?
 

#12

Joined Nov 30, 2010
18,224
Did they take steps to investigate and correct it?
No. I quit and went for the Laser Repair Depot job at a 30% pay increase, and the boss sold the company to some "real" engineers that couldn't specify a transformer core large enough to hold all the windings. That was a good time to get out of Dodge!
 
Top