[ibis-macro] Re: Jitter Bird Graphic, and Backchannel Presentation

  • From: Mike Steinberger <msteinb@xxxxxxxxxx>
  • To: fangyi_rao@xxxxxxxxxxx
  • Date: Sun, 22 May 2011 20:05:32 -0500

Fangyi-

So far, IBIS-AMI has only addressed differential mode behavior, and there's nothing about Rx_Noise_pad that attempts to go beyond that. Thus, in the context of your question, the only equation that is defined is <(Vin_p - Vin_n)^2> = Rx_Noise_Pad^2.

Mike S.

On 05/22/2011 07:54 PM, fangyi_rao@xxxxxxxxxxx wrote:

Hi, Mike;

Thanks for the answers. Noise at the Rx output node, Vout, is determined by noise at the Rx positive input node, Vin_p, noise at the Rx negative input node, Vin_n, transfer function from Vin_p to Vout, XF_p, and transfer function from Vin_n to Vout, XF_n, as

Vout = XF_p*Vin_p + XF_n*Vin_n

Can you further clarify how noise powers of <Vin_p^2> and <Vin_n^2> and correlation function <Vin_p*Vin_n> are defined in terms of Rx_Noise_Pad?

Fangyi

*From:*ibis-macro-bounce@xxxxxxxxxxxxx [mailto:ibis-macro-bounce@xxxxxxxxxxxxx] *On Behalf Of *Mike Steinberger
*Sent:* Thursday, May 12, 2011 11:03 AM
*To:* ibis-macro@xxxxxxxxxxxxx
*Subject:* [ibis-macro] Re: Jitter Bird Graphic, and Backchannel Presentation

Fangyi-

Thanks for the questions. Answers inserted below.

Mike S.

On 05/11/2011 03:20 PM, fangyi_rao@xxxxxxxxxxx <mailto:fangyi_rao@xxxxxxxxxxx> wrote:

Hi, Walter;

Thanks for the graphs. I have two questions:

Noise "Rx_Noise_Pad" is injected into the analog circuit. What is the impedance of the noise source? How is the noise source connected to the circuit? Can you provide a SPICE netlist so that this parameter is unambiguously defined?

The word "injected" is not doing you any favors. Rx_Noise_Pad is a noise voltage _measured_ at the circuit node which is the input to the receive amplifier.

There may be a noise injection process; and if there is, its goal is create conditions which result in a specific noise voltage measured at the input to the receiver amplifier. For example, the noise injection process could be a current source driving current into the input of the receiver. If the goal of that current source is to create a voltage Vn at the receiver input, the input impedance of the receiver is Zr (approximately equal to Z0=50ohms) and the source impedance of the interconnect network is Zs (also approximately equal to Z0) at the input to the receiver, then the current from a current source injecting into that circuit node should be

In = Vn/(Zr||Zs) = Vn/((Zr*Zs)/(Zr+Zs))

The fact of the matter is that in many SPICE simulations, the noise is actually inserted in the transistor models, and Rx_Noise_Pad is the noise voltage that _would_ be injected into the input in order to produce the output noise voltage that was _actually_ produced by the transistor internal to the circuit. This is the whole concept of a noise figure http://en.wikipedia.org/wiki/Noise_figure - very familiar to microwave circuit designers and communications analysts, but otherwise requiring some time and effort to get used to.

Since parameters Rx_DCD, Rx_RJ and Rx_SJ are jitters external to receiver as stated in the BIRD, how do Rx model developers know their values? Similarly, how do Rx model developers know the value of Rx_Noise_Pad which is injected to the Rx analog circuit?

This is actually two different questions with two different answers.

2a. Rx_DCD, Rx_RJ and Rx_SJ represent the effect of processes such as power supply noise that are external to the receiver. As such, the model developer may or may not know their values for a given system or application. For example, the vendor could specify a maximum tolerable level of power supply noise, and the model developer could translate those worst case values into jitter parameters supplied with the model. !!!-> In this case, the documentation supplied with the model should clearly state these worst case conditions. <-!!! It may also be that the model developer does not have any specific information from which to calculate these jitter parameters. In that case, "Better to remain silent and let the World think you a fool than to open your mouth and remove all doubt."

2b. Rx_Noise_Pad is calculated from the receiver noise figure and input impedance. These are values that only the circuit designer is sure to have access to.

Thanks,

Fangyi

*From:*ibis-macro-bounce@xxxxxxxxxxxxx <mailto:ibis-macro-bounce@xxxxxxxxxxxxx> [mailto:ibis-macro-bounce@xxxxxxxxxxxxx] *On Behalf Of *Walter Katz
*Sent:* Tuesday, May 10, 2011 1:13 PM
*To:* IBIS-ATM
*Subject:* [ibis-macro] Jitter Bird Graphic, and Backchannel Presentation

All,

Presentation made to IBIS ATM today.

Walter

Walter Katz

wkatz@xxxxxxxxxx <mailto:wkatz@xxxxxxxxxx>

Phone 303.449-2308

Mobile 720.333-1107

Other related posts: