11
\$\begingroup\$

As a research for the problem I described here I found this circuit by Maxim:

enter image description here

This is clock doubler, and must be a really good fit in my case as input frequency is very well defined.

However looking through the datasheets I found that MAX9010 outputs TTL levels, while 74VHC86 accepts CMOS levels (0.7 * Vcc). In general I can not find high speed comparator with CMOS outputs operating @ 5V.

Should I pay special attention to this issue - what are the conditions when circuit may fail producing proper clock?

Can you give feedback on the circuit in general? My assessment that it should work properly doubling 21.47727 MHz to 42.95454 MHz with R1=1k and C1=15pF (however for sure will need prototyping and adjustment in real life).

P.S. Last days I reviewed a lot of designs for managing clocks, and my feeling is that they in high degree a kind of "marketing articles" and not appropriate for the direct application - articles talk a lot about circuits' pros, but almost none state the cons (arising from propagation delays, frequency ranges etc) thus it is really bad idea to implement what is said directly without modelling and proper simulation for target conditions.

Update: as I suspected this circuit is an ideal design designed to work in ideal conditions. When built in real life, it does not function properly without investment into the following areas:

  1. power must be maximally clean. Due to noise in the power rails voltage divider will have level fluctuating, causing spikes at the output of the comparator and false positives;
  2. comparator may (will) sink some current from the voltage divider (reference voltage) at its positive input at the time of switching. It may also change reference point slightly;
  3. RC with such small capacitance is very subject to be influenced by other capacitances around and EMI, changing tuned duty cycle (at best) or making x2 multiplication stage malfunctioning.

In addition, I have built this circuit using MAX999, but its LTSpice model is faulty. It is confirmed by the Maxim support, hopefully they will fix it.

I am going to drop this design, considering ICS501 instead.

\$\endgroup\$
8
\$\begingroup\$

However looking through the datasheets I found that MAX9010 outputs TTL levels, while 74VHC86 accepts CMOS levels (0.7 * Vcc).

That's a good spot and I agree with you - maybe you should inform Maxim on their dodgy circuit. Shame on them.

Should I pay special attention to this issue - what are the conditions when circuit may fail producing proper clock?

Yes, you can't use those two chips together without lowering the power rail on the 74 series. Maybe try a MAX999 - it's slightly faster on propagation time (4.5 ns) but, importantly, hits the rails on the output hence will drive the 74 chip.

\$\endgroup\$
  • \$\begingroup\$ Will try MAX999. Datasheet also says about 3.5 mV hysteresis, and that's good for this application. Faster chip timing is not that important I guess, it can be adjusted by the input RC filter properties. Seems uneasy to source 1 gate VHC86 (best solution because choosing 4-gate chip will waste 3 gates as I do not see the place to use spare ones). \$\endgroup\$ – Anonymous Jul 1 at 19:07
  • 1
    \$\begingroup\$ This is likely one of those times where the design will work just fine in practice, despite being slightly outside of what the datasheet promises will work. Just because the device is guaranteed to register a logical high at V >= 0.7*Vcc doesn't mean it's guaranteed to fail to register a logical high at V = 0.66*Vcc \$\endgroup\$ – nitro2k01 Jul 2 at 1:08
  • 3
    \$\begingroup\$ @nitro2k01 - that's a recipe for things to go not-so-hilariously wrong down the line. \$\endgroup\$ – TLW Jul 2 at 2:02
  • \$\begingroup\$ Going to use MAX999+LVC1G86 to build doubler, and LVC2G74+LVC1G08 to build clock /3 circuit. All powered from 5V. \$\endgroup\$ – Anonymous Jul 2 at 10:52
3
\$\begingroup\$

The delay you require is only 25nS. I would consider simplifying your circuit to use two or all three of the other gates in the 74HC86 package to provide the delay, their nominal Tpd is 11nS at 5v into 15pF. Without extra capacitive loading, their delay might be a bit less. Their delay will be strongly affected by the rail voltage, so only use this method if the rail is well regulated.

\$\endgroup\$
  • \$\begingroup\$ I saw these circuits. Prefer timing to be more controllable - to my experience different manufacturers may have large spread in properties of their devices, and if another manufacturer is chosen, circuit may misbehave. \$\endgroup\$ – Anonymous Jul 1 at 19:02

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.