quantized
Member level 2
- Joined
- Jul 6, 2012
- Messages
- 51
- Helped
- 1
- Reputation
- 2
- Reaction score
- 1
- Trophy points
- 1,288
- Activity points
- 1,887
Hi,
I've been reading about and SPICEing a lot of current mode logic (MCML) lately, and I find it really fascinating.
However, all the papers I've come across use a DC tail current.
Is there any work on current mode logic (preferably MOSFET but not necessarily) using an AC tail current? The advantage here would be the ability to do charge recovery. The charge from current sent through the circuit is recycled to the power supply (a huge LC oscillator) rather than being dumped to ground.
With a single-phase supply you can't do much computation when the current is near zero, so it probably makes sense to have at least two phases for alternating stages of logic. Ideally the output voltage from each stage should cross Vth at the same time that the next stage's supply voltage crosses zero; this should give a quasi-adiabatic behavior -- it's not reversible, but it does arrange for gate voltages to change the least when Vds is large and change the most when Vds is small.
An alternative arrangement for polyphase AC would have gates appear in pairs (or triples) with one member of each pair (or triple) on each phase of the supply; the (differential) output of each gate in one stage drives the inputs of all gates in the next stage, so at every point in time at least one gate in each stage has enough current going through it to compute an output. In the two-phase case, when the differential outputs of one gate cross the outputs from the other gate in the pair will be reaching maximum separation, so the sum of the differences in the two differential outputs never falls to zero. This should provide reduced latency and decouple the computational speed from the cycle time of the power supply. On the other hand it has a huge area cost and probably increases the energy per operation quite a bit too.
I've been reading about and SPICEing a lot of current mode logic (MCML) lately, and I find it really fascinating.
However, all the papers I've come across use a DC tail current.
Is there any work on current mode logic (preferably MOSFET but not necessarily) using an AC tail current? The advantage here would be the ability to do charge recovery. The charge from current sent through the circuit is recycled to the power supply (a huge LC oscillator) rather than being dumped to ground.
With a single-phase supply you can't do much computation when the current is near zero, so it probably makes sense to have at least two phases for alternating stages of logic. Ideally the output voltage from each stage should cross Vth at the same time that the next stage's supply voltage crosses zero; this should give a quasi-adiabatic behavior -- it's not reversible, but it does arrange for gate voltages to change the least when Vds is large and change the most when Vds is small.
An alternative arrangement for polyphase AC would have gates appear in pairs (or triples) with one member of each pair (or triple) on each phase of the supply; the (differential) output of each gate in one stage drives the inputs of all gates in the next stage, so at every point in time at least one gate in each stage has enough current going through it to compute an output. In the two-phase case, when the differential outputs of one gate cross the outputs from the other gate in the pair will be reaching maximum separation, so the sum of the differences in the two differential outputs never falls to zero. This should provide reduced latency and decouple the computational speed from the cycle time of the power supply. On the other hand it has a huge area cost and probably increases the energy per operation quite a bit too.
Last edited: