40106 variations in manufacturers? Odd performance issue - Electronic Circuits Projects Diagrams Free

zevon8 · 30th January 2006, 03:26 PM

Hi.. I have a circuit using 2 garden variety oscillators ( 0.1uF from input to ground, 100K feedback, and 1uF and 1M ) that have been used as part of a larger circuit for almost 10 years with no problems. The small variations from one batch to another are not a problem, due to cap tolerances, etc.

What happened the other day was a batch were all oscillating at nearly twice the frequency normal. Both oscilators were doing this. Careful examination and testing could find that the only difference between the previous functional units and the batch running at nearly twice the speed was that the 40106 IC brand used.

The "good" ones used a Fairchild CD40106BCN and the "bad" ones used a Texas CD40106BE. I have used many brands over the years, likely all the big names, and never had this problem before.

The only thing I can figure is there is a different hysterisis window from one brand to the next, intentional or otherwise. Or maybe I happened to get a batch at one extreme of the makers tolerance? The data sheets for both parts don't show anything that I could see to explain the big difference.

Anyone seen anything like this before? Unless I am mistaken, a 40106B is a 40106B, is a 40106B, no ???

Thanks for reading, any insight greatly appreciated!!

Roff · 30th January 2006, 04:39 PM

Look at the datasheet. The thresholds vary with VCC, of course, but as an example, at VCC=5V and Temp=25C, the hysteresis can be as low as 1V, or as high as 3.6V. Obviously, you can't depend on these parts to give you a predictable oscillation frequency.

audioguru · 30th January 2006, 05:17 PM

The frequency of a two-inverter or two-gate (one input used, not Schmitt trigger) Cmos oscillator is much more stable than a Schmitt trigger oscillator with different ICs or supply voltages:

eblc1388 · 30th January 2006, 05:57 PM

Quote:

Originally Posted by zevon8

Anyone seen anything like this before? Unless I am mistaken, a 40106B is a 40106B, is a 40106B, no ???

No. Similar results apply also to CD4046 PLL from different manufacturers.

zevon8 · 30th January 2006, 07:17 PM

Thanks for the replies, let me take this in order of appearance

Ron H: true the window varies with supply and temperature, in this case tho the supply is very clean and the temperature is stable. Good point, but the frequency of this cct is not critical, it merely flashes a light at around 30 per minute. 35 per minute, nobody will get excited, but 50 times a minute, well they tend to notice, LOL.

audioguru: Yes, I agree, much better results. Normally this is what would be done, but this is an existing design that is not worth a new PCB and redesign. It is one of those circuits where the lowest gate/part count possible was a design priority, almost to an extreme unfortunately.

eblc1388: I will keep that in mind, we have one or two ccts with that part.

Anymore ideas? sofar so good, is it just a normal thing I will have to live with?

( edited hz to min )

Roff · 30th January 2006, 08:11 PM

Quote:

Originally Posted by zevon8

Ron H: true the window varies with supply and temperature, in this case tho the supply is very clean and the temperature is stable.

You missed the point (or at least glossed over it). I only mentioned VCC and temp to focus on one area of the datasheet. My point was that hysteresis can vary by over 3:1 from unit to unit. This means the frequency can vary by about the same amount, independent of other contributors.

zevon8 · 30th January 2006, 10:07 PM

Sorry, didn't mean to appear that way. While the variation can be large unit to unit, it has not posed a problem over the years. Typically the variation has been very minor and tolerable, this is the first time out of many hundreds of units that a frequency shift of nearly 2:1 has been seen. This is from batches using atleast 3 major brands, sometimes mixed on the same board. Fairchild and Motorola chips are nearly identical in performance.

It seems the Texas Instrument units have a "very wide" hysterisis window, compared to that of the other samples I tried in the circuit. One good thing, is that the TI chips are all the same, from random samples from about 500 pcs (from different date lots) they all behave the same. Knowing this I can cahnge the R/C values when using this chip, and hopefully "solve" the problem for now.

I can't imagine the problems someone would have in a hi-speed circuit if part changes resulted in what could be wild shifts in timing.

Thanks for the help, seems to be one of those "practical lessons" that come along once in a while. Too bad the guy who designed this device did not have more "practical lessons."

Roff · 30th January 2006, 10:21 PM

The problem results from using the part for a purpose other than what it was designed to do. As a device to clean up noise on digital lines, it was probably specified as loosely as possible while still maintaining functionality. This reduces the cost by increasing yields. It also probably results in different "centers" on the yield curves of different manufacturers, whether from circuit design or process differences, or both.

30th January 2006, 03:26 PM	(permalink)
zevon8 Experienced Member	40106 variations in manufacturers? Odd performance issue Hi.. I have a circuit using 2 garden variety oscillators ( 0.1uF from input to ground, 100K feedback, and 1uF and 1M ) that have been used as part of a larger circuit for almost 10 years with no problems. The small variations from one batch to another are not a problem, due to cap tolerances, etc. What happened the other day was a batch were all oscillating at nearly twice the frequency normal. Both oscilators were doing this. Careful examination and testing could find that the only difference between the previous functional units and the batch running at nearly twice the speed was that the 40106 IC brand used. The "good" ones used a Fairchild CD40106BCN and the "bad" ones used a Texas CD40106BE. I have used many brands over the years, likely all the big names, and never had this problem before. The only thing I can figure is there is a different hysterisis window from one brand to the next, intentional or otherwise. Or maybe I happened to get a batch at one extreme of the makers tolerance? The data sheets for both parts don't show anything that I could see to explain the big difference. Anyone seen anything like this before? Unless I am mistaken, a 40106B is a 40106B, is a 40106B, no ??? Thanks for reading, any insight greatly appreciated!!

30th January 2006, 04:39 PM	(permalink)
Roff Experienced Member	Look at the datasheet. The thresholds vary with VCC, of course, but as an example, at VCC=5V and Temp=25C, the hysteresis can be as low as 1V, or as high as 3.6V. Obviously, you can't depend on these parts to give you a predictable oscillation frequency. __________________ Ron

30th January 2006, 10:21 PM	(permalink)
Roff Experienced Member	The problem results from using the part for a purpose other than what it was designed to do. As a device to clean up noise on digital lines, it was probably specified as loosely as possible while still maintaining functionality. This reduces the cost by increasing yields. It also probably results in different "centers" on the yield curves of different manufacturers, whether from circuit design or process differences, or both. __________________ Ron

Thread Tools
Show Printable Version Email this Page
Display Modes
Linear Mode Switch to Hybrid Mode Switch to Threaded Mode

30th January 2006, 07:17 PM	(permalink)
zevon8 Experienced Member	Thanks for the replies, let me take this in order of appearance Ron H: true the window varies with supply and temperature, in this case tho the supply is very clean and the temperature is stable. Good point, but the frequency of this cct is not critical, it merely flashes a light at around 30 per minute. 35 per minute, nobody will get excited, but 50 times a minute, well they tend to notice, LOL. audioguru: Yes, I agree, much better results. Normally this is what would be done, but this is an existing design that is not worth a new PCB and redesign. It is one of those circuits where the lowest gate/part count possible was a design priority, almost to an extreme unfortunately. eblc1388: I will keep that in mind, we have one or two ccts with that part. Anymore ideas? sofar so good, is it just a normal thing I will have to live with? ( edited hz to min )

30th January 2006, 10:07 PM	(permalink)
zevon8 Experienced Member	Sorry, didn't mean to appear that way. While the variation can be large unit to unit, it has not posed a problem over the years. Typically the variation has been very minor and tolerable, this is the first time out of many hundreds of units that a frequency shift of nearly 2:1 has been seen. This is from batches using atleast 3 major brands, sometimes mixed on the same board. Fairchild and Motorola chips are nearly identical in performance. It seems the Texas Instrument units have a "very wide" hysterisis window, compared to that of the other samples I tried in the circuit. One good thing, is that the TI chips are all the same, from random samples from about 500 pcs (from different date lots) they all behave the same. Knowing this I can cahnge the R/C values when using this chip, and hopefully "solve" the problem for now. I can't imagine the problems someone would have in a hi-speed circuit if part changes resulted in what could be wild shifts in timing. Thanks for the help, seems to be one of those "practical lessons" that come along once in a while. Too bad the guy who designed this device did not have more "practical lessons."