CMOS analog delay cricuit

Thread Starter

penumbra

Joined Dec 8, 2012
76
Hello everyone,

I am trying to build a circuit that can delay a random signal by about a millisecond. I am using nmos/pmos. The thing is that i have tried the bucket brigade approach, but i can't seem to get it working. Any ideas on how to implement it?


Thanks.
 

KMoffett

Joined Dec 19, 2007
2,770
More details please.
I'm assuming you are talking about delaying an AC signal.
What about the bucket brigade circuit doesn't work.
Can you post the schematic?

Ken
 

ErnieM

Joined Apr 24, 2011
7,992
How often do you want to sample the signal? That depends on how exact you wish the delay, and also how long the signal exists: you want to make sure you will get a sample when the signal is high before it goes away or you will loose it.

Once you know how fast you will be sampling divide the 1 mS by the sampling time (or multiply by the rate. That gives you how many stages you need.

INCORRECT Example: You want to sample every 10 nS: 1mS/10nS = 100 so you need a 100 bucket brigade shirt register.

CORRECTED Example: You want to sample every 10 uS: 1mS/10uS = 100 so you need a 100 bucket brigade shirt register.


The CD4517B is a dual 64 bit shift register, so something like that may work well for you.
 
Last edited:

Thread Starter

penumbra

Joined Dec 8, 2012
76
Thank you for your reply guys.

Well, the signal is not exactly an AC signal, but rather an arbitary signal. The total spike takes 3.5ms to complete, then there is some random signal and then again another spike. Basically, it is an action potential, like shown here:
http://content.answcdn.com/main/content/img/oxford/Oxford_Sports/0199210896.action-potential.1.jpg

This happens about 10-100 times a second, and we will be sampling at 50Khz.

So i need to make an analog memory that can take in the signal and then delay it at the output by about a 1ms. The original signal must be maintained.

Also, i tried putting pmos transistors in steps with capacitors. How big of a capacitor do i need?

I cannot use an existing IC. SO basically i can use bucket brigade, but make my own buckets (i.e. the transistor and capacitors in each stage)

Thanks.
 

Ron H

Joined Apr 14, 2005
7,012
How often do you want to sample the signal? That depends on how exact you wish the delay, and also how long the signal exists: you want to make sure you will get a sample when the signal is high before it goes away or you will loose it.

Once you know how fast you will be sampling divide the 1 mS by the sampling time (or multiply by the rate. That gives you how many stages you need.

Example: You want to sample every 10 nS: 1mS/10nS = 100 so you need a 100 bucket brigade shirt register.

The CD4517B is a dual 64 bit shift register, so something like that may work well for you.
1mS/10nS=100,000, not 100. Perhaps you meant uS instead of nS.
 

ErnieM

Joined Apr 24, 2011
7,992
Oops,yes I did. Post edited WITH comments so you post still makes sense <grin>
 

Thread Starter

penumbra

Joined Dec 8, 2012
76
Okay, so anyone looked at the action potential image i posted earlier?

Thanks.
 

MrChips

Joined Oct 2, 2009
19,269
Another solution is to digitize the waveform and then play it back with the desired delay.

Why do you want to delay the signal by 1ms?
 

Thread Starter

penumbra

Joined Dec 8, 2012
76
I need to delay the signal so it can be recorded later, like a type of a temporary memory.

Cannot use Digital, has to be analog.
 

MrChips

Joined Oct 2, 2009
19,269
Still not clear.

A CCD is a recorder. You are already recording the signal.

Why the restriction that it must be analog?

As pointed out in other threads, it is best to give us the whole picture. What are you trying to do?
 

Thread Starter

penumbra

Joined Dec 8, 2012
76
Hi,

I've already given detailed information. Basically this has to go on a chip, fabricated and hence keeping it all analog, it cannot be DSP. All i am saying is, if i need to implement the bucket-brigade for delay, how should i go about it?

Senor,

if i wanted to use Digital, i would have (it's easy, simply implement a DSP), but i am restricted by my design parameters.

Thanks.
 

ErnieM

Joined Apr 24, 2011
7,992
The only analog recording units I know of look somewhat similar to this:



If I was attempting to capture this signal I'd go digital. Even the inexpensive USB PIC18 based scopes could capture this signal.
 

Thread Starter

penumbra

Joined Dec 8, 2012
76
Ok then, this is my detailed answer.

I already said that i need to delay the signal so it can be recorded externally.

Basically, by the time the action potential occurs and is detected (it has to cross a certain potential difference to make sure it is action potential and not the background noise) it is too late to record the signal, the part before the threshold is obviously lost. So we delay this signal in some sort of an analogue memory to use and analyze it later.

I just want to know if i can implement a buck-brigade or anyother way to delay an analogue signal. The thing is, i have to work with analogue electronics so cannot simply take those IC's that came some time ago and implement them. It has to be a custom design IC.

Thanks.
 

Ron H

Joined Apr 14, 2005
7,012
I already said that i need to delay the signal so it can be recorded externally.

Basically, by the time the action potential occurs and is detected (it has to cross a certain potential difference to make sure it is action potential and not the background noise) it is too late to record the signal, the part before the threshold is obviously lost. So we delay this signal in some sort of an analogue memory to use and analyze it later.
That is not a good reason for not doing it digitally.
You can continuously digitize the signal and store it in a small RAM. When the signal crosses the threshold, record a few more milliseconds, then dump a 5ms (or whatever) snapshot to some sort of static memory or recirculating RAM for later recording or analysis.
 

Thread Starter

penumbra

Joined Dec 8, 2012
76
That is not a good reason for not doing it digitally.
You can continuously digitize the signal and store it in a small RAM. When the signal crosses the threshold, record a few more milliseconds, then dump a 5ms (or whatever) snapshot to some sort of static memory or recirculating RAM for later recording or analysis.
See designing an analogue delay is part of the challenge. I know i can use digital, but i am restricted to not use it as it makes things more complex from programming point of view. Secondly, the design has to be cmos compatible, that is, it should be something that can easily be fabricated on the die, next to say, a differential amplifier and biasing circuitry.

If you guys can help me with regards to bucket brigade more, it would be helpful as i know bucket-brigade concept should work, but how should i go about simulating and implementing it?

Thanks.
 

Ron H

Joined Apr 14, 2005
7,012
See designing an analogue delay is part of the challenge. I know i can use digital, but i am restricted to not use it<snip>
So this appears to be homework. I don't believe you have posted your work on the problem. We don't do homework here, but we will help, after you have shown us you have made your best effort.
 

WBahn

Joined Mar 31, 2012
24,674
I suspect the problem that you are having is that you are trying to implement a CCD-like circuit in a CMOS process. The problem is that the CCD devices use the gate capacitances and then use overlapped gates (usually in three phases) to sweep the charge from gate to gate. But with most CMOS processes, even most dual gate ones, you can't have overlapping gates. You can come pretty close using chains of FET capacitors that you phase the gate voltages on, but the loss from bucket-to-bucket still adds up pretty quickly.
 
Top