# Confused about the difference between frequency synchronization and phase synchronization

I am not sure if this is the correct place to ask this question but it looked like the most fitting SE site.

I study software engineering and I am taking a course about networking, where synchronization came up.

I am not really capable of understanding the difference between these two synchronization types, it seems to me that phase sync. is "stronger" then frequency sync. in some sense but I am not really sure.

Could anyone clear this up for me?

• Different clocks at the Tx and Rx can make the carrier frequencies have an offset (called CFO). Delays in the propagation can cause the carrier to have a carrier phase offset (called CPO). In some channels like satellite or underwater acoustic channels the speed of movement can also cause a CFO due to Doppler. Commented Jul 17, 2021 at 18:33
• Please post a reference to define your terms. There are many different types of applications that use "synchronization" and they don't all mean the same thing. Commented Jul 17, 2021 at 19:06
• @Hilmar Here is what was said in class: Frequency synchronization : at any given time the skew is 0; Phase synchronization : at any given time the offset between the clocks is an integer number of phases
– EL_9
Commented Jul 17, 2021 at 20:32

## 3 Answers

Some simple visual examples:

Two Sinusoids w/ frequency and phase lock

Two Sinusoids w/ frequency lock but a phase offset Note that this might also be considered phase locked depending on the application, if the receiver can estimate for this fixed phase offset and compensate for it elsewhere

Two sinusoids with phase lock but with frequencies that are integer multiples of eachother Note how about every 65 samples the phase aligns again. This is something you commonly see as an option in a PLL when taking in a reference clock and generating an output clock that has a frequency an integer factor higher but maintains a fixed phase relationship to the input.

Two sinusoids where one is a very slightly different frequency then the other In some applications, this may be what "frequency locked without phase lock" actually looks like, since the estimated frequency may have some minor variations w.r.t. the true frequency. Note that clearly there is no fixed phase relationship between the two like I showed in the second example - the phase relationship between the two could be fairly random over time. It's worth pointing out however that even if you had phase lock, it would not be perfect either! If you compared the true vs the estimated phase, it too would have minor variations but the average over time should be close to the true phase and should not continuously wander.

In the physical layer of a network, data (a stream of $$0$$'s and $$1$$'s) often travels serially from one device to another (from one node in the network to another node in the network) as evidenced by names such as USB (Universal Serial Bus). There are various clocks in the physical layer that need to be synchronized in both frequency and phase, and the mechanisms for achieving such synchronization is of intense interest to many readers of dsp.SE, but I am going to ignore those synchronizations and assume that for the purpose of software engineering and networking, the physical layer can be replaced by a pipe through which flows a stream of bits. However, logically, in the higher levels (as per 7-layer OSI model) of the network, this serial bit stream is organized into bytes which are grouped together into words, and at even higher level into blocks or frames or packets etc. There are lots of different clock signals in the system, with a bit clock signal governing the individual bits traveling along the serial bus, and a separate byte clock (hopefully synchronized with the bit clock) that tells the receiving node how to divide the received stream of bits into bytes. The byte clock signals the start of a byte, and the next byte clock signal occurs after eight bit clock signals and signifies the start of the next byte. That is, the bit clock and the byte clock are synchronized in frequency.

With the above in mind, consider the received data stream $$\cdots ~0~1~1~0~0~1~0~1~1~1~1~0~0~1~0~0~0~0~1~1~1~0~1~1~1~0~\cdots$$ which the transmitter node had obtained from successive bytes $$01100101$$,$$11100100$$, $$00111011$$, $$10\cdots$$. The byte clock at the receiver node might not be synchronized in phase with the byte clock at the transmitter node; and so it might well divide up the received bit stream into bytes as $$\cdots01$$, $$10010111$$ $$10010000$$, $$11101110$$ instead. That is, the byte clock at the receiver is synchronized to the byte clock at the transmitter in frequency but not in phase. There are eight possible ways for the two byte clocks to be synchronized in frequency and in only one of these ways are the byte clocks synchronized in phase as well. When the two byte clocks are synchronized in both frequency and phase, the receiver node correctly divides the bit stream into $$01100101$$,$$11100100$$, $$00111011$$, $$10\cdots$$ matching what the transmitter node intended to send to the receiver node. Similar considerations apply to word clocks if logically, the data stream needs to be blocked into $$16$$ or $$32$$-bit or $$64$$-bit words.

make a example here: frequency synchronization: you have a watch and i have a watch <> now my watch is 19:00:01 yours is 19:00:02, we keep the same tick speed as 1 second per ideal second, so after 1 hour, my watch is 20:00:01 and yours 20:00:02. this is frequency synchronization. if you want to have the phase synchronization: we need to always has the same time, not only the same tick period.

frequency synchronization: only need to same periods during same time. phase offset: the start time need to align at first, and frequency need to synchornized