Hi All,
I decided to built an astable oscillator on 555 basis. Required frequency is 1050 Hz, (in other experiment it will be 2050 Hz). Ambient temperature varies from 0 to +40 ° C.
Power supply from voltage regulator LM7812. The datasheet shows 12V at output with line regulation accuracy of 120mV , and load regulation accuracy of 240mV.
Timing capacitor will be polystyrene type, 0,1uF, with TC of 80 ppm and timing resistors with TC of 50 ppm.
Is it real to obtain frequency with drift of 1 % or better with those components?

 

The National Semiconductor datasheet that I looked at says that the initial accuracy is ±2.5%. Drift should be about ±0.1% with the numbers you mentioned.

Who made the '555 that you are planning to use?

 

Don't like to use large circuits, special IC or modules

 

? Where did that come from?

 

Timing capacitor will be polystyrene type, 0,1uF, with TC of 80 ppm and timing resistors with TC of 50 ppm.

This is a potential problem.

Assuming your timing resistor actually is a fixed resistor and a trim pot in series, their temperature coefficients and that of the timing capacitor should be equal and opposite. That is, if the capacitor tempco is +80 ppm, then the two resistors should combine for a tempco of -80 ppm. In this way, the time constant will be, well, constant over your temperature range.

Separate from that, 80 ppm and 50 ppm over a 40 degree temperature range combine for an error of just over 0.8%. This requires that the temperature tracking of the ratios of the 555's internal component errors is way less than 1%. It is not. Building and testing the circuit is a pretty low-effort task. I recommend you build one and see how it behaves, but it does not look good on paper.

If it doesn't work, and alternative is to replicate the circuit function of a 555 with a dual, low-error, low-drift comparator, a low-drift voltage reference (something missing in the 555), and some 0.1% resistors.

ak

 

Just a note: I suspect the nominal values of the components are assumed to be such at or about 25°C, not at either freezing or 40°

 

In fact, capacitor's TC is -80ppm, resistor's TC is +/-50 ppm.
If 555 is replaced by several IC, the whole circuit is significantly complicated. By the way, what comparators and voltage references may be recomended?
I guess at begining, accuracy of 1% is not a very high requirement.
It seems to me, a better way is to use a clock quartz and a divider , CD4060 or so. In other hand, the oscillator on CMOS utilizes high value resistors , 22 megohm or so. It seems , such a circuit is sensitive to PCB leakages and to RFI interferences and need to be shielded

 

In fact, capacitor's TC is -80ppm, resistor's TC is +/-50 ppm.
If 555 is replaced by several IC, the whole circuit is significantly complicated. By the way, what comparators and voltage references may be recomended?
I guess at begining, accuracy of 1% is not a very high requirement.
It seems to me, a better way is to use a clock quartz and a divider , CD4060 or so. In other hand, the oscillator on CMOS utilizes high value resistors , 22 megohm or so. It seems , such a circuit is sensitive to PCB leakages and to RFI interferences and need to be shielded

Even a quartz crystal would be temperature dependent. I haven’t done the math, but temperature variation shifts in the crystal will introduce some error in this solution.

However don’t despair. There is a way to get an extremely accurate frequency from a crystal. It’s called an Oven Controlled Crystal Oscillator. The reason this works is that you use a heater (ie oven built from a couple of resistors) to maintain the crystal’s operating temperature independent of ambient temperature. Roman Black has a very good article describing how to build your own (as he notes, you could buy one, but it might be less expensive and more fun to build your own). Click on the link above to read the article.

 

It seems to me, a better way is to use a clock quartz and a divider , CD4060 or so.

I agree, but that is not the title of your thread.

In other hand, the oscillator on CMOS utilizes high value resistors , 22 megohm or so.

A problem with the 4060 in your application is that I don't think there is a standard crystal that is a binary multiple of your desired output frequencies. The R-C oscillator configuration is based on the very common, hysteretic two-gate design, which is dependent on the input stage transition voltage level. The problem is that that voltage is not stable with temperature, or repeatable from one part to another. Again, cheap to build and test.

ak

 

I agree, but that is not the title of your thread.



A problem with the 4060 in your application is that I don't think there is a standard crystal that is a binary multiple of your desired output frequencies. The R-C oscillator configuration is based on the very common, hysteretic two-gate design, which is dependent on the input stage transition voltage level. The problem is that that voltage is not stable with temperature, or repeatable from one part to another. Again, cheap to build and test.

ak

1024Hz and 2048Hz also acceptable. The rest part of the device will be retuned.
Most important is that frequency stays stable at different temperatures

 

Why do You need such high accuracy, and such a wide Temp range ?
What are You building ?
.
.
.

 

1024Hz and 2048Hz also acceptable. The rest part of the device will be retuned.
Most important is that frequency stays stable at different temperatures

Binary-multiple frequencies change everything. You will get the most stability with the least effort by buying a crystal oscillator device with whatever stability requirement you need, at Digi-Key, Mouser, etc., and running the output through a simple ripple divider (CD4020, 4024, etc.). Zero oscillator design, simple confirmation testing. A 1.0485 MHz, 20 ppm part is $1.51 in 1's. Less at higher multiples.

https://www.digikey.com/en/products/detail/sitime/SIT8008BI-11-33E-1-048576/12316409

These types of parts use either 3.3 V or 5 V. What other power sources are available in your design / system?

ak

 

You can get 16.8MHz crystal, divided by 16 to give 1.050MHz, divided by 1000 for 1050Hz.
16.384MHz divided by 8000 = 2048Hz.

 

I had no idea that a multiple of 1050 Hz would be a standard part, but OK.

Still, four IC's plus a high-frequency oscillator circuit seems like a lot for a single output frequency square wave, given that 1024 Hz is acceptable.

ak

 

(Some text removed for clarity)
In other hand, the oscillator on CMOS utilizes high value resistors , 22 megohm or so. It seems , such a circuit is sensitive to PCB leakages and to RFI interferences and need to be shielded

The 22 Meg resistor you mention is probably a bias resistor and neither it nor leakage currents affect the frequency of the crystal.

 

It is a bias resistor, as shown here on page 8: https://assets.nexperia.com/documents/data-sheet/HEF4060B.pdf

In the National Semi datasheet, it is shown as 10 M.

ak