Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

[10GBASE-CX4] FW: Correction to your 12/02 CX4 Presentation

Hi all,
 
Please be aware about next update: there is a mistake in December's "Pre-emphasis" presentation on slide 8 - "Correspondence of pre-emphasis definitions" - thanks a lot to Larry for figuring this out. The whole conversation between Larry, Ze'ev and myself might be found below - for those of you who will be flying from West Cost/Europe and anyway have nothing to do in the plane J... The bottom lane - for Pre-emphasis = 1-V_low/V_high definition next correspondence takes place:
         Pre-emphasis= 2b  =2a /(1+  a ) - Correct one     instead of         Pre-emphasis= 1 - 2b  =(1-a )/(1+  a ) - Incorrect!
 
Spreadsheet with correct conversions is attached.
 
I really-really hope that this is the last time we discuss Pre-emphasis terminology/correspondence issue...
 
But really fancy fact is just coming: for currently selected pre-emphasis value (a = 1/3  ) both equations - corrected and wrong one - yield absolutely same 50% Pre-emphasis value... Does that say that this is the right one? :-) 
 
Regards,
    Dimitry 
 
-----Original Message-----
From: Ze'ev Roth
Sent: יום חמישי 13 פברואר 2003 01:12
To: Dimitry Taich; 'larry rennie'
Subject: RE: Correction to your 12/02 CX4 Presentation

Dimitry/Larry,
The results   Pre-emphasis=2b=2a/(1+a)  are correct as well as b=a/(1+a).
If we wrote differently in the presentation - let's issue a correction.
 
I rechecked Peter's spreadsheet for conversion between the different parameters (he sent it out some time ago), and fortunately he used the definitions - so his transformations are correct.
I think I sent out an excel spreadsheet with conversion (see attachment) as well, this time with the correct equations.
 
Larry thanks for pointing this out.
 
Regards,
Ze'ev
 
-----Original Message-----
From: Dimitry Taich
Sent: Thursday, February 13, 2003 1:09 AM
To: 'larry rennie'
Cc: Ze'ev Roth; Dimitry Taich
Subject: RE: Correction to your 12/02 CX4 Presentation

Larry,
 
Finally I got you... Stupid mistake - I used V_low/V_high ratio itself instead of subtract it from 1... My son would say - "And that guy dare to teach me math?!"
 
BTW, methods' 2 results still seem to be OK - b=a/(1+a), 1-2B=(1-a)/(1+a).  But - Pre-emphasis def should be as you state:
 
                Pre-emphasis=2b=2a/(1+a)
 
Unless Ze'ev will find something new tonight, I'll update CX4 group about that correction tomorrow morning.
 
Thanks a lot for your review and contribution.
 
Regards,
    Dimitry
 
P.S. That's really amazing that currently selected value (1/3) yields same pre-emphasis value for both - correct one and wrong one - definitions!  
-----Original Message-----
From: larry rennie [mailto:Larry.Rennie@nsc.com]
Sent: יום רביעי 12 פברואר 2003 14:15
To: Dimitry Taich
Cc: Ze'ev Roth
Subject: Re: Correction to your 12/02 CX4 Presentation

Dimitry,

Your reply has the following equations under Method 1 ( I have italicized them):

Step 3: Using pre-emphasis def with a parameter (Slide 5) we get next values for V_low & V_high:

                V_low = V_peak*(1-a)/(1+a)V_peak        -->
                1-V_low/V_high = (1-a)/(1+a)
                V_high = V_peak                                         -->

but  the first 2 lines above are not correct,

V_low = V_peak*(1-a)/(1+a)V_peak  should be V_low = V_peak*(1-a)/(1+a). You have an extra V_peak term.
and
1-V_low/V_high = (1-a)/(1+a) should be  [1-V_low/V_high ]= [1- (1-a)/(1+a)] =  2a/(1+a)

More serious (unless I am missing something) is that from slide 8 and method 2 below you say,

pre-emphasis = [1-V_low/V_high]  = 1-2beta], I don't think this is correct. I calculate,

pre-emphasis = [1-V_low/V_high]  and from your slide 4,

V_low = (1-2beta)V_peak, and
V_high = V_peak, therefore, substituting we get,

pre-emphasis = [1-V_low/V_high] = [1- (1-2beta)V_peak)/Vpeak]= 2beta.

I am confusing b and beta?

Regards,

Larry
 
 

Dimitry Taich wrote:

Larry,

Probably I miss something, but our version seems me to be the correct one. You pay our attention on contradiction between "1-2Beta  = (1-alpha)/(1+alpha)" and some "earlier equations" - could you specify which equations exactly do you mean?

Anyway, here is how we got "1-2b  = (1-a)/(1+a)" equation:

Method 1:

Step 1: Pre-emphasis is finally defined as "1-V_low/V_high" (Slide 6)

Step 2: Using pre-emphasis def with b parameter (Slide 4) we get next values for V_low & V_high:

                V_low = (1-2b)V_peak    -->
                                                                                        1-V_low/V_high = 1-2b
                V_high = V_peak         -->

Step 3: Using pre-emphasis def with a parameter (Slide 5) we get next values for V_low & V_high:

                V_low = V_peak*(1-a)/(1+a)V_peak        -->
                1-V_low/V_high = (1-a)/(1+a)
                V_high = V_peak                                         -->

Step 4: combining Step 2 & Step 3 results for "1-V_low/V_high " equation, we get above ratio - 1-2b = (1-a)/(1+a)

Method 2:

Let's analyze time-domain equations for pre-emphasis' filter definition - using b parameter (slide 4) and a parameter (slide 5). We know that coefficients correspondent to different samples - x_n and x_n-1 - are independent (non-fraction equalizer!), so we can compare them one by one. For example, let's treat x_n-1 coefficients - should be equal, right?

                b = a/(1+a)  --> 2b = 2a/(1+a)  --> 1-2b = 1 - 2a/(1+a) = (1+a-2a)/(1+a)=(1-a)/(1+a)

Same result might be achieved for X_n coefficients comparison.

Let me know if you still have problem - and thanks for your patient to read so far :-)

Regards,
        Dimitry

-----Original Message-----
From: larry rennie [mailto:Larry.Rennie@nsc.com]
Sent: יום רביעי 12 פברואר 2003 12:15
To: 'dimitryt@mysticom.com'; Zev Roth
Subject: Correction to your 12/02 CX4 Presentation

Dimitri and Zev,

I was looking at your pre-emphasis definition presentation from the Dec
02 CX4 meeting.  I notice that you have a slight error on slide 8,
"Correspondence of pre-empahisis definitions"

you have 1-2Beta  = (1-alpha)/(1+alpha)

Based on your earlier equations it should be:

1-2Beta = 2*alpha/(1+alpha)

By coincidence, for alpha=1/3, (1-alpha)/(1+alpha) = 2*alpha/(1+alpha)

Is this a correct correction?

Thanks,

Larry

pre_emphasis_to_alpha_ conversion.xls