Mike Its well worthwhile estimating the additional jitter due to this effect when using such a circuit to square up the output of an OCXO: If the input signal characteristics are: Frequency 10MHz Amplitude at the gate input: A = 1.4V pk Threshold mismatch Vt = 1V AM noise: Am = -120dBc/Hz Input signal AM noise bandwidth: BW = 1MHz (eg a low Q bandpass filter). Rms Output jitter due to AM noise is given by delta(t) ~ (1/(2*PI*f))*((Vt/A)/(1 + (Vt/A)*(Vt/A)))*(BW*1)^(Am/20)) i.e. delta(t) ~ 0.5*1.6E-8 *(1E-3) sec ~ 8ps rms. Wideband AM noise as high as -120dBc/Hz is somewhat higher than is typical for a good OCXO. Thus in applications such as a PPS divider this effect is probably insignificant. However it may be useful to use a low Q bandpass filter to limit the integrated AM and PM noise seen at the gate input. Bruce Mike Monett wrote: > > Message: 3 > > Date: Fri, 03 Apr 2009 09:04:59 +1300 > > From: Bruce Griffiths <bruce.griffiths@xtra.co.nz> > > Subject: Re: [time-nuts] Frequency Divider > > > Hal Murray wrote: > > >>> A large resistor connected between the input and output would > >>> accommodate threshold variations better. Even better would be a > >>> feedback loop that adjusts the input bias point to maintain the > >>> output duty cycle at 50%. > > >> Isn't that resistor a feedback loop? > > >> I played with that setup in the lab many years ago. It didn't > >> work as well as I was expecting. I didn't figure out why it > >> didn't work better. > > >> Maybe some gain in the feedback path would help. Then we have to > >> consider stability. Ugh. > > > Hal > > > Yes, a resistor connected between the input and output of an > > inverter is a feedback loop but the loop gain is relatively low. > > > With a high amplitude input threshold variations from the nominal > > can cause the input protection diodes to conduct. > > > Once these diodes conduct the output jitter may deteriorate > > significantly (it does for HCMOS inverters). > > > Using a non inverting integrator in the feedback path can > > accurately stabilise the duty cycle. > > > Bruce > > The 74HC and 74AC input threshold tolerance is +/- 30%. This means > the threshold can vary from 1.5V to 3.5V with a Vcc of 5V. > > This limits the maximum input signal to 3V p-p or +13.5dBm, and > leads to a very subtle flaw discovered in some amazing engineering > work by Martein Bakker, PA3AKE. > > If the threshold is not controlled, it can cause AM noise to convert > to PM noise and degrade the jitter. This occurs in the Analog > Devices AD9910 1GHz DDS chip. > > Martein Bakker discovered this in his noise analysis, and Kevin > Wheatly gave a nice entry in his blog on how to fix it: > > http://www.m0khz.com/?p=589 > > Mike > > _______________________________________________ > time-nuts mailing list -- time-nuts@febo.com > To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts > and follow the instructions there. > >

Correction: For those who didn't spot it, the formula should have been: delta(t) ~ (1/(2*PI*f))*((Vt/A)/(1 + (Vt/A)*(Vt/A)))*SQRT(BW)*10^(Am/20)); Bruce Bruce Griffiths wrote: > Mike > > Its well worthwhile estimating the additional jitter due to this effect > when using such a circuit to square up the output of an OCXO: > > If the input signal characteristics are: > > Frequency 10MHz > Amplitude at the gate input: A = 1.4V pk > Threshold mismatch Vt = 1V > AM noise: Am = -120dBc/Hz > Input signal AM noise bandwidth: BW = 1MHz (eg a low Q bandpass filter). > > Rms Output jitter due to AM noise is given by > > delta(t) ~ (1/(2*PI*f))*((Vt/A)/(1 + (Vt/A)*(Vt/A)))*(BW*1)^(Am/20)) > > i.e. > delta(t) ~ 0.5*1.6E-8 *(1E-3) sec > ~ 8ps rms. > > Wideband AM noise as high as -120dBc/Hz is somewhat higher than is > typical for a good OCXO. > > Thus in applications such as a PPS divider this effect is probably > insignificant. > However it may be useful to use a low Q bandpass filter to limit the > integrated AM and PM noise seen at the gate input. > > Bruce > > > Mike Monett wrote: > >> > Message: 3 >> > Date: Fri, 03 Apr 2009 09:04:59 +1300 >> > From: Bruce Griffiths <bruce.griffiths@xtra.co.nz> >> > Subject: Re: [time-nuts] Frequency Divider >> >> > Hal Murray wrote: >> >> >>> A large resistor connected between the input and output would >> >>> accommodate threshold variations better. Even better would be a >> >>> feedback loop that adjusts the input bias point to maintain the >> >>> output duty cycle at 50%. >> >> >> Isn't that resistor a feedback loop? >> >> >> I played with that setup in the lab many years ago. It didn't >> >> work as well as I was expecting. I didn't figure out why it >> >> didn't work better. >> >> >> Maybe some gain in the feedback path would help. Then we have to >> >> consider stability. Ugh. >> >> > Hal >> >> > Yes, a resistor connected between the input and output of an >> > inverter is a feedback loop but the loop gain is relatively low. >> >> > With a high amplitude input threshold variations from the nominal >> > can cause the input protection diodes to conduct. >> >> > Once these diodes conduct the output jitter may deteriorate >> > significantly (it does for HCMOS inverters). >> >> > Using a non inverting integrator in the feedback path can >> > accurately stabilise the duty cycle. >> >> > Bruce >> >> The 74HC and 74AC input threshold tolerance is +/- 30%. This means >> the threshold can vary from 1.5V to 3.5V with a Vcc of 5V. >> >> This limits the maximum input signal to 3V p-p or +13.5dBm, and >> leads to a very subtle flaw discovered in some amazing engineering >> work by Martein Bakker, PA3AKE. >> >> If the threshold is not controlled, it can cause AM noise to convert >> to PM noise and degrade the jitter. This occurs in the Analog >> Devices AD9910 1GHz DDS chip. >> >> Martein Bakker discovered this in his noise analysis, and Kevin >> Wheatly gave a nice entry in his blog on how to fix it: >> >> http://www.m0khz.com/?p=589 >> >> Mike >> >> _______________________________________________ >> time-nuts mailing list -- time-nuts@febo.com >> To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts >> and follow the instructions there. >> >> >> > > > _______________________________________________ > time-nuts mailing list -- time-nuts@febo.com > To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts > and follow the instructions there. > >