0
\$\begingroup\$

I am planning to use Crystek CCSS-945X-25-100.000 sine wave oscillator to drive the reference CLKin pins of TI PLL IC LMK04832 .

Now in datasheet of oscillator, output power mentioned is 5dBm into 50 Ohms.

But lets say if load is not 50 Ohms but a high impedance load, then what will be the output power of oscillator?

Is sine wave oscillator like this, only required a 50 Ohms load for stable oscillations and if load other than 50 Ohms is connected to output of oscillator it will not oscillate at all?

EDIT:

Actually this sine wave oscillator will be on a separate PCB which we are calling as "CLOCK DISTRIBUTION BOARD".

Sine wave clock from "CLOCK DISTRIBUTION BOARD" will be connected to main board which has LMK04832 PLL through a SMA cable.

I have checked the CMOS version of the oscillator : CVHD-950.

In datasheet of CMOS oscillator ,they mentioned the load as 15pF.

Now I need to know how much load in capacitance, this sine wave oscillator version can drive because this SMA cable will be having some capacitance and also LMK04832 refclock inputs will be having some capacitance.

Is there any way to know the value of load capacitance, this sine wave oscillator can drive ,from the available oscillator specification in datasheet like 5dBm into 50 Ohms?

\$\endgroup\$
3
  • \$\begingroup\$ 5dBm is about 1.8 milliwatts, for a 50 ohm load it means about 0.3 volts. In theory, power to high impedance would be zero, but most likely you mean voltage, and that is not shown in the crystal datasheet. Do you really need to connect these parts together, sounds like the PLL chip would be happier with logic level or differential clock. \$\endgroup\$ – Justme Dec 1 '19 at 17:08
  • \$\begingroup\$ One can only guess at Crystek's internals. Output likely comes from a linear buffer stage having some power gain when loaded with 50 ohms. DC current of 30 mA suggests more than one stage. If a buffer is used, it is unlikely that oscillations would cease when unloaded. But other specs (like sine waveform) could be corrupted...Crystek doesn't say in their data sheet. \$\endgroup\$ – glen_geek Dec 1 '19 at 17:17
  • \$\begingroup\$ SMA is a connector, not a cable. Presumably, though, if you've got SMA connectors you're using a 50\$\Omega\$ coax. Why not just terminate it at the main board? Are you sending the clock to more than one place? \$\endgroup\$ – TimWescott Dec 3 '19 at 21:01
0
\$\begingroup\$

Is there any way to know the value of load capacitance, this sine wave oscillator can drive ,from the available oscillator specification in datasheet like 5dBm into 50 Ohms i.e, how to convert 5dBm into 50 Ohms to 5dBm into ... pF?

Not really, no. These 50 Ω are a specific impedance for maximum power transfer, i.e. impedance matching.

So, it'd be expected that you match your capacitive load (using some sort of matching network) to the real 50 Ω impedance to achieve maximum power sinking.

If you don't do that matching, you'll get an impedance mismatch, which means not all power will be brought to your sinking device. Look for the reflection and transmission coefficient formula for impedance mismatches. It's easy to deduct how much power you'll deliver to that load.

But, really: it's not clear why you'd even want to know whether your sine oscillator can sink 5 dBm into a 15 pF load – that's not a problem you're facing! Remember, the sine is a harmonic oscillation – so a capacitive load will do nothing but shift the phase.

When you look at the LMK datasheet p. 92, you'll notice that you'll need to terminate your differential transmission line with the nominal transmission line impedance.

50Ω single-ended sine wave clock source termination

Your datasheet tells you that yes, you can drive a 50 Ω load with 5 dBm; calculate whether those 5 dBm mean enough voltage across the 50 Ω termination resistor.

If it's not solidly enough, you will need to add a buffer on your clock distribution board (sounds like a good idea, anyway, since you're calling it distribution board, you probably need some fanout!).

Make that buffer a clock buffer that converts the single-ended to a differential signal (since you seem to be concerned with cleanliness of the oscillation rather than exact frequency, this is desirable), and go for fig 23 from p. 91

100Ω diff sine wave clock source termination

\$\endgroup\$
2
  • \$\begingroup\$ Muller : I was asking to convert 5dBm into 50 ohms to ..pf because previously my colleagues used CMOS crystal oscillator which has load specified as 15pF and output voltage as 3.3v. And when they connected CMOS output to LMK input through SMA cable,the voltage reduces significantly from 3.3v. My colleagues said that was happening because of the additional capacitance offered by SMA cable. But you are saying that since sine is a harmonic oscillation, additional capacitance will only shift the phase and will not drop the voltage level like it was dropping in case of CMOS oscillator? \$\endgroup\$ – Alan Dec 3 '19 at 19:12
  • \$\begingroup\$ There's a lot to unpack here. So, clearly, capacitive loading can and will load an output of an oscillator. But: As the datasheet clearly specifies, the sine oscillator is designed to feed a 50Ω impedance – and an (ideal) 50Ω SMA cable is lossless and hence will not lead to any voltage drop. Again, not buffering the oscillator's output on the same PCB is probably a design mistake to begin with. But your colleague might still be pretty wrong. \$\endgroup\$ – Marcus Müller Dec 3 '19 at 19:14

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.