The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Design >> Mixed-Signal Design >> Code Dependent kick on Ladder
https://designers-guide.org/forum/YaBB.pl?num=1323896085

Message started by jefkat on Dec 14^th, 2011, 12:54pm

Title: Code Dependent kick on Ladder
Post by jefkat on Dec 14^th, 2011, 12:54pm

Hi folks,
Is there a clever way to minimize the code dependent kick on the ladder. I have a differential signal and I need to quantize it at 5 bit resolution and at 8MHz. I can not spend much current . So any hints, references etc , most welcome..
Thanks
shaf

Title: Re: Code Dependent kick on Ladder
Post by raja.cedt on Dec 14^th, 2011, 12:56pm

hello,
could be please elaborate more, what do you mean by ladder.

Thanks,
raj.

Title: Re: Code Dependent kick on Ladder
Post by jefkat on Dec 14^th, 2011, 11:03pm

Resistive Ladder I meant.

Title: Re: Code Dependent kick on Ladder
Post by raja.cedt on Dec 15^th, 2011, 1:31am

hello,

I understand now, you are talking about flash ADC reference disturbance because of comparators switching. So better make reference less sensitive to sudden change other make coupling less to reference. Use per-amplifier at the comparator input otherwise use good DE-coupling cap at each tap of the resistive divider (this is called bowing ).

Thanks,
Raj.

Title: Re: Code Dependent kick on Ladder
Post by jefkat on Dec 15^th, 2011, 9:24am

Thanks Raja for the comments. Using pre-amplifiers is out of question as I have no current left . Not so sure about using decoupling caps. Decoupling caps will slow the ladder down further.

raja.cedt wrote on Dec 15^th, 2011, 1:31am:

Title: Re: Code Dependent kick on Ladder
Post by raja.cedt on Dec 15^th, 2011, 10:36am

hello,
you are wrong, decoupling caps doesn't slow down your ckt's. It is just bias ckts for comparator. This is very imp when you have switch cap comparator. Any how thing about resistors-string bowing

Thanks,
Raj.

Title: Re: Code Dependent kick on Ladder
Post by Sanjay Rajasekhar on Dec 16^th, 2011, 1:41am

There is probably a nicer way to deal with this. If you have a good parasitic extractor, you could probably do some simulations with parasitics to check the strength of the kickback. Once you have a fairly good idea, you could choose a cap value which when used between the output of the comparator and the ladder tap point would result in the same strength of kickback. Now all you have to do is to connect one end of this cap to the ladder tap point, and the other end to the inverted comparator output. So when the comparator is pushing some charge onto the ladder, now there is this new cap which is absorbing it, leaving the ladder tap point voltage largely unchanged. I'm pretty sure that this cap value wont be very high.

However beware of creating inadvertent negative feedback paths which could result in multiple triggers on your comparators.

Regards,
Sanjay

Title: Re: Code Dependent kick on Ladder
Post by RobG on Dec 16^th, 2011, 8:13am

jefkat wrote on Dec 15^th, 2011, 9:24am:

Thanks Raja for the comments. Using pre-amplifiers is out of question as I have no current left . Not so sure about using decoupling caps. Decoupling caps will slow the ladder down further.

It is a power/speed/accuracy trade-off. The caps will reduce the glitch size, but the smaller glitches will take longer to settle to zero. For example, increasing the filtering might take a 100mV glitch that takes 100nS to settle and change it into a 1mV glitch that takes 10uS to settle. If your error tolerance is much more than 1mV you don't care how long it takes to settle so using caps might make sense in that case.

The down side to filtering is that the taps will never settle to their exact value. You need to check the average current coming out of the taps and make sure the IR drop isn't too large.

Code dependent currents can cause problems, but I wouldn't think it would be a problem at 5 bits, 8MHz. Check your average current for different codes and see how it will affect the average voltage of the taps of your ladder.

Title: Re: Code Dependent kick on Ladder
Post by loose-electron on Dec 18^th, 2011, 2:25pm

Consider a sample and hold circuit which is in "hold" mode while
all the transient switching is going on.

Title: Re: Code Dependent kick on Ladder
Post by vivkr on Dec 19^th, 2011, 5:37am

jefkat wrote on Dec 14^th, 2011, 12:54pm:

Your signal is differential. Are you also using a differential threshold for the comparators? If not, then you should as you can cut down the effective impedance presented by the ladder to any differential kickback by a factor of 2. I have seen many people use a single-ended threshold where they take one end off the ladder and the other is grounded or tied to a common mode voltage.

Vivek

Title: Re: Code Dependent kick on Ladder
Post by raja.cedt on Dec 19^th, 2011, 6:23am

hello vivek,
if you use pure differential reference then only differential impedance offered by ladder will be half, where as if you only single ended reference and and by applying one side reference and one side common mode voltage to comparator then i guess you don't see reduction in differential impedance, am i correct? correct me if am wrong.

Thanks,
Raj.

Title: Re: Code Dependent kick on Ladder
Post by loose-electron on Dec 21^st, 2011, 10:59am

OH, this is on the bias reference chain to a flash ADC?

1. Resistance in series with the comparator input
that connects to the bias reference chain.

2. If needed, filter capacitors on each end of that resistor.

the Pi section RC filter created is a LPF in both directions

try it, you will like it!

:)

Title: Re: Code Dependent kick on Ladder
Post by weber8722 on Jan 2^nd, 2012, 1:49am

jefkat wrote on Dec 14^th, 2011, 12:54pm:

Hi,

beside adding caps to the reference ladder (which is usually very good). Another good way might be to use a comparator with kick-back cancellation capacitance :). This would have a similar effect than a pre-amp, but without more current. Usually dummy mosfets are inserted in cross-coupled manner. This way the charge injection to ref ladder is canceled (at least to e.g. 90%). The slow-down of the comp should be small (like 15%).

Bye Stephan

Title: Re: Code Dependent kick on Ladder
Post by raja.cedt on Jan 2^nd, 2012, 4:57am

hello,
do you mean neutralization capacitors?

Thanks,
Raj.