bandgap reference current accuracy and osc frequency

[lhlbluesky]

i have two questions need your help.
1, i have designed a bandgap reference, the reference voltage is 1.25V, i want to use this voltage (almost PVT independent) to generate reference current, as the picture shows. the reference voltage and R are both positive temp coef, and designed to get a almost zero temp coef reference current Iref, however, the R varies very much with corner(ss or ff), and the variation can be more than 20% percent, so Iref varies also. i know, i can use a external resistor Rext replacing the internal R, but my design doesn't allow this Rext, so R is internal resistor. i want to know, how to make Iref unchanged with corner variation? besides, i want to do a trimming for internal R with a register, so i can calibrate the process deviation of R for +/-100%, but i have a question, i don't know how to monitor the process change of R, that is, i can't judge R is smller or larger than expected. about this, is there some method or advice? besides, is there any other way to generate a process independent reference current? thanks.

2, about the OSC frequency. i have to design a RC OSC based on cap charging and discharging, and the frequency of OSC output is 4MHz, tolerence is about +/-15% ~ +/-20%, i want to know, if i don't use calibration (such as current calibration), can i realize the frequency accuracy of +/-15%? as i konw, capacitor can have a process deviation more than 15%, am i right? or is there any method to ensure the 15% frequency accuracy without calibration (that is, to decrease the influence of process deviation)? i need your help.
thanks all.

 

[aryajur]

Hi,
THe bandgap reference circuit you have would require trimming also. To trim that you would have to measure it and do the trimming. So why not measure the current and trim the resistor also?
For the oscillator if you have a trimmed bandgap reference it may be possible to self adjust a oscillator a little bit by converting the output to 50% duty cycle (using /2 logic) and level translating it to the bandgap or other regulated voltage pass it through a differentiator and filter it out, which would give you a voltage proportional to the frequency. You can compare that to the Bandgap and then adjust your capacitive or resistor trim to get the desired frequency.

 

[lhlbluesky]

hi, aryajur. 'So why not measure the current and trim the resistor also?'
but how to measure the current? i don't have extra pin for the current measurement for my design. is there some other method? pls help me, thanks.

 

[lhlbluesky]

about 'For the oscillator if you have a trimmed bandgap reference it may be possible to self adjust a oscillator a little bit by converting the output to 50% duty cycle (using /2 logic) and level translating it to the bandgap or other regulated voltage pass it through a differentiator and filter it out, which would give you a voltage proportional to the frequency. '
i don't understand very clearly, 'by converting the output to 50% duty cycle ', whose output, bandgap or osc? if OSC, it is square wave also with 50% duty cycle. and how to ' translating it to the bandgap '? and what is the meaning of 'pass it through a differentiator and filter it out'?
pls explain it more clearly for me, thanks very much.
besides, i want a simple realization, which would be area and power optimized and minimized.
and i want some other solutions, thanks.

 

[aryajur]

If you don't have extra pins then you may have to multiplex pins and create test modes where he functionality of a pin is changed specifically for testing purposes.
In the oscillator I meant the output of the oscillator changed to 50% duty cycle. Then it can be converted to a ground to bandgap rail clock simply by running an inverter off the buffered bandgap voltage. If you differentiate the square wave by a simple C in series with the signal and then put a low pass filter you would get a voltage proportional to your oscillation frequency and you can use that to measure and self callibrate your oscillator.

 

[lhlbluesky]

hi, aryajur, ' If you differentiate the square wave by a simple C in series with the signal ', then, i get a sharp pulse, am i right? but how to 'get a voltage proportional to your oscillation frequency' using the sharp pulse (may be -vref to vref)? and what is the role of the low pass filter?
can you give me a simple diagram or waveform about the generated signal after the square wave? thanks.

and any other ideas, i want it would be area and power optimized and minimized, not very complicated. thanks.

 

[watersky]

how is the frequency of your osc calculated? for example: the frequency is changed with VCC. if the frequncy is fixed, a good bandgap voltage is normal, the current is calculated by vbg and R. the frequency is found this current and C. three variable parameter are vbg, R and C. vbg is good. R and C are variable, so the frequency is variable.through trimming current that is observed by vbg and R, needful frequency is assured by trim pad amount.

 

[lhlbluesky]

if i only consider the variation of C, assuming vbg and R is fixed, then, how to self calibrate the osc frequency without calibration circuit? thanks.

 

[lhlbluesky]

can anyone help me again?

 

[aryajur]

Sorry I do not have access to my computer right now so I could give you some schematic. But as you said after differentiation you would get sharp pulses. The height of the pulses would be proportional to the height of the square wave supply so if the square wave is running off a regulated rail from the bandgap your height of the pulses is defined well. Then if you pass the sharp pulses through a low pass filter a simple R-C filter the voltage on the C would be constant since it will remove the high frequency component of the pulses. The filter cut off should be much lower than your oscillator frequency. Now this gives you a voltage proportional to the frequency. So you can have a feedback loop that tries to make this voltage equal to the voltage that would give you the desired frequency by adjusting the capacitor of your oscillator. Now this feedback loop will be complicated. Depending on whether you have varactors or not you may not be able to get good continuous trim for your oscillator so all you would be able to do is to have multiple comparators on this voltage that turn ON/OFF certain capacitor banks based on the voltage