Q: When should I divide the VCO by 2? A: Choosing whether or not to divide the VCO output by 2 depends on several factors. Do you need a 50% duty cycle? It's difficult to generate a 50% duty-cycle if you use a non-differential VCO. Even with a differential VCO, it's hard to keep duty-cycle within +/- 3% over all PVT. Dividing by 2 is an easy way to get a good duty-cycle. If you need your VCO to cover a large frequency range, then selectively dividing by 2 is an easy way to extend the usable range of your VCO. Typically, VCO performance is poor when the control voltage hovers close to the threshold voltage of the transistors in the Voltage-to-Current (V2I) converter and is also poor at high control voltages when the V2I current sources are pulledout of saturation. So, it may be beneficial to divide the VCO by 2 at low frequencies and divide by 1 at high frequencies to keep the control voltage in the VCO's "sweet spot". So, what's the downside? If you need a 2.5GHz clock for PCI Express and you are not able to generate a 5GHz VCO in your technology, div-by-2 is not an option. In addition, you need to budget for PVT variations. VCO gain/speed may vary by as much as 2-3X over all process/power-supply/temperature corners. Also, you're adding delay when you divide-by-2. The more delay in your path, the more jitter you'll pick up from power-supply noise. This is likely a minor concern. If the signals which selects between div-by-1 and div-by-2 clocks is not properly synchronized, then you can create clock glitches which may corrupt your state machines or create tester indeterminacy. An alternative to dividing the VCO to extend its range is to add VCO range controls. e.g. programmable current sources to modify the VCO gain (common) or programmable number of VCO stages (uncommon).