Element for producing half and one clock cycle delay in Cadence

[Shishira]

Hi All,
I want to introduce a delay of half and one clock period in an analog circuit. I need an element/circuit that can produce half-clock period in real circuit to be taped out. I cannot use a D flip flop as the output needed is an amplitude varying square signal. (atch 2). Looking forward to ideas..

 

[FvM]

Sounds like your design has a clock. A switched capacitor circuit can do.

 

[Shishira]

Sounds like your design has a clock. A switched capacitor circuit can do.

I already have a SC circuit in feed-forward path and hence I do not want to increase area and power. Isn't there any other simple way?

 

[BradtheRad]

Sample-and-hold is the all-purpose term for what you need.
A few methods:

* Switched capacitor circuit. (Hold sample for entire clock cycle.)

* Convert each pulse amplitude to a binary value (ADC), hold it briefly via latch, then send through a DAC.

* Store pulse amplitude in an inductor (rather, in the flux field of the inductor). Then discharge the energy to an output stage.

* Certain led's give off light briefly after current is shut off.

* A CdS photoresistor takes a brief time to respond to light intensity.

 

[KlausST]

Hi,

A sample and hold will lose information. As long as it holds it can not react on input signal. And since it is 1.5 cycles you miss one edge.

The requirement of "1.5 cycles" is not useful at all. You need to be able to look into the future to know when 1.5 cycles are over. Esoecially when frequency is not constant.

If frequency is constant, then better state your requirement as a fixed delay time. Like "3.34 ms".

Delaying analog signal is difficult.
For delay times in the nanoseconds you may use a cable with the required length.
For delay in the milliseconds or higher I'd use: ADC --> digital FIFO --> DAC.

Klaus

 

[Shishira]

Hi,

A sample and hold will lose information. As long as it holds it can not react on input signal. And since it is 1.5 cycles you miss one edge.

The requirement of "1.5 cycles" is not useful at all. You need to be able to look into the future to know when 1.5 cycles are over. Esoecially when frequency is not constant.

If frequency is constant, then better state your requirement as a fixed delay time. Like "3.34 ms".

Delaying analog signal is difficult.
For delay times in the nanoseconds you may use a cable with the required length.
For delay in the milliseconds or higher I'd use: ADC --> digital FIFO --> DAC.

Klaus

Hi Klaus,
The problem is to get a delay of 0.5 and 1 cycle delay. Yes 1.5 cycles is not useful

 

[BradtheRad]

Simulation portraying basic sample-and-hold. (Timing is performed in the manner which I grasp from the discussion.)

1) Sample incoming level during first half of cycle.

2) Send amplitude on capacitor to output during second half of cycle .

3) Reset capacitor by discharging it briefly.

Incoming voltage was changed during each idle gap.

Obtain 1.5 cycle delay by attaching a second sample-and-hold which samples the first level at an convenient moment.

 

[Shishira]

Simulation portraying basic sample-and-hold. (Timing is performed in the manner which I grasp from the discussion.)

1) Sample incoming level during first half of cycle.

2) Send amplitude on capacitor to output during second half of cycle .

3) Reset capacitor by discharging it briefly.

Incoming voltage was changed during each idle gap.

Obtain 1.5 cycle delay by attaching a second sample-and-hold which samples the first level at an convenient moment.

View attachment 174435

Hey, that sounds promising thank you. But, attaching a second S&H should give a delay of 1 clock cycle isn't it?

 

[KlausST]

Hi,

I´m a bit confused.

Can you please give an almost complete definition of the signal?
* Frequency (range)
* duty cycle (range)
* HIGH voltage range
* LOW voltage range
* are LOW always flat?
* are HIGH always flat?

Klaus

 

[BradtheRad]

Hey, that sounds promising thank you. But, attaching a second S&H should give a delay of 1 clock cycle isn't it?

Yes and seeing your initial diagram I suppose that's the time length you want. Then extended to 1.5 cycles, it appears?

By putting two or more sample-and-holds in a row you create the bucket brigade concept. Capacitors pass along varying volt levels from one to the next. Hundreds of stages are in analog delay IC's (particularly audio) to generate phasing/ chorus/ echo effects. Example, MN3002 and SAD-1024.

To make a bucket brigade for your project requires some design effort, even to make two or three stages.

 

[Shishira]

Hi,

I´m a bit confused.

Can you please give an almost complete definition of the signal?
* Frequency (range)
* duty cycle (range)
* HIGH voltage range
* LOW voltage range
* are LOW always flat?
* are HIGH always flat?

Klaus

The signal is ,
* fixed frequency say 32 MHz
* It's not a clock signal, it's a correction signal extracted for further operation. (It is sampled w.r.t to a clock signal of frequency 32 MHz)
* Vout range -500mV < Vh < 500 mV

This is a discrete signal and can take any value between 500mV < Vh < 500 mV. For a better understanding of my input signal I attach this waveform simulated from macro model.

 

[KlausST]

Hi,

When I see the diagrams of post#11, I don´t think that it´s possible with 1 or two sampe-and-hold circuits.
The problem I see is how to get the correct timing when one can not rely on the signal edges.

Klaus

 

[Shishira]

Yes and seeing your initial diagram I suppose that's the time length you want. Then extended to 1.5 cycles, it appears?

By putting two or more sample-and-holds in a row you create the bucket brigade concept. Capacitors pass along varying volt levels from one to the next. Hundreds of stages are in analog delay IC's (particularly audio) to generate phasing/ chorus/ echo effects. Example, MN3002 and SAD-1024.

To make a bucket brigade for your project requires some design effort, even to make two or three stages.

Okay, that sounds deep thanks. I actually need two output signals,
a) 0.5 (half ) cycle delay
b) 1 (one) cycle delay

--- Updated Feb 16, 2022 ---

Hi,

When I see the diagrams of post#11, I don´t think that it´s possible with 1 or two sampe-and-hold circuits.
The problem I see is how to get the correct timing when one can not rely on the signal edges.

Klaus

Hi,
Yes that's right, so there is no other way to delay the signal? I want to have an algebraic addition of these signals, say x[n] and x[n-1]. Fig is from Simulink model

 

[KlausST]

Hi,

x[n] and x[n-1] often refer to the digital value of samples. On the digital side it´s rather simple.

Klaus

 

[Shishira]

Hi,

x[n] and x[n-1] often refer to the digital value of samples. On the digital side it´s rather simple.

Klaus

Sorry for using the wrong notations, my question is restricted to the analog side. By n-1 , I just meant adding with the previous value (which is noting but delayed signal.)