Synapses are important components in a spiking neural network (SNN), which is an advanced and bio-physically plausible neural network. The Spike Timing-Dependent Plasticity (STDP) learning rule, whose target object is the synapse’s plasticity, is a common training method for SNN. Most of the realization of synapses in hardware is usually defined in the amplitude of action potentials to work with leaky integrate-and-fire (LIF) neurons, which mainly focus on the voltage of signals and lack of neural dynamic properties. The all-digital resonate-and-fire (ADRAF) neuron was an alternative spiking neuron model that solved the weak features of LIF neurons. The existing methods of using synapses in digital hardware are not suitable for this new model due to the absence of dynamic characteristics, especially in on-chip learning. This study proposed a new definition of synaptic plasticity based on spike-time transmitting. A new learning method based on STDP was developed to apply to the new plasticity. As a result, a digital synapse that can modify its plasticity automatically depending on the activity of the spikes coming from the pre-synapse neurons was designed and simulated to prove the new learning method. The design was implemented on a 28-nm FPGA device without any multipliers or finite-state machines.