Continue to Site

Welcome to EDAboard.com

Welcome to our site! EDAboard.com is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.

Can equalization be used to counter package/die parasitic

Status
Not open for further replies.

Dydeb

Newbie
Joined
May 25, 2020
Messages
2
Helped
0
Reputation
0
Reaction score
0
Trophy points
1
Activity points
74
Hi all,

I've recently been working on high-speed digital communications (multi-GHz) and wasn't able to figure something out about equalization and all the cases where it is used.

From my understanding, equalization is used only for countering the effect of channel attenuation: if you have a channel with a 20 dB loss than you will use an EQ with approximately 20 dB gain (or less) in the pass-band to flatten out the channel frequency response up to your frequency of interest.

My question is, could you use equalization to counter the behavior from package/die parasitics as well? Since a parasitic capacitor will also act as an attenuator at higher frequencies, its frequency response would be similar to that of a channel. In this case, could a receiver with large package/die parasitics be improved through EQ as well?

Thank you all for your replies
 

From my limited reading I make it that the EQ on
modern transmission line interfaces is meant to
idealize the far-end match, to maximally open
up the eye. Particularly as the physical plant
outside the IC, is unknown at time of chip
design / build / ship.

The effects of bond wires and leadframe are
more knowable and could be embedded to the
product electrical / physical design. This is the
norm for many RFIC functions, padding the
input wirebond with a little excess C to get
flat(-ish) 50 ohms Zpin in the advertised
working band.

Those parasitics are probably a nit, in the big
(end-to-end) picture, and might not see the EQ
step resolution to null them - but they are also
a minor fraction of the total, and don't deserve
nulling on their own (only as part of the whole)?
 

Hi dick_freebird,

Thanks for your reply. I agree with what you said which can be applied in most cases, and perhaps I should have been clearer with where my question came from.

In my case it's a scenario where the IC die parasitic capacitance dominates the system with a low frequency pole (couple of pF with a 50 Ohm termination impedance -> below 2 GHz cutoff) which could be as you said improved through design (through inductive peaking for example). In this scenario, I was wondering if perhaps equalization could also be applied (i.e: if your equalization filter applied a transfer function similar to j*w*L? would it have the same effect?)
 

Typically in a SerDes, the output spectrum contains harmonics as you are trying to send out square waves (bits). Any filter designed to tune out the parasitics will have a narrow band. As a result, if you place one of those filters, you will wreak havoc with the sharp edges of your square waves and thereby impacting your eye diagram significantly.
 

A typical Equalization Circuit consists an All Pass Filter with a certain slope so overall Gain Roll-off is compensated by this filter.
Using a EQ circuit in a small package is not possible but it can be used externally to compensate the Gain slope.
 

My question is, could you use equalization to counter the behavior from package/die parasitics as well? Since a parasitic capacitor will also act as an attenuator at higher frequencies, its frequency response would be similar to that of a channel.

The issue solved with your equalizer example is signal pass loss.

The issue with package parasitics is mismatch, i.e. reflection. So even if you equalize the transmission, by making matching equally bad at all frequencies, you will have trouble due to multiple reflections along the signal path.

But indeed, some compensation works. If you start from a given shunt parasitic, you can turn that into a low pass (series L, two shunt C) to improve matching in the low frequency range.
 

sure, there are all sorts of equalizers out there. Group delay equalizers are actually much more common than gain slope equalizers.

as far as compensating for package parasitics….not sure what that means. Most parasitics add a lowpass response (due to shunt capacitance), so a gain slope equalizer that has less loss at higher frequency would work.

at some point, though, physics steps in. You can not passively match out a reactance any more than Fano's limit allows, so keep those parasitic reactances as low as you can
 

Status
Not open for further replies.

Part and Inventory Search

Welcome to EDABoard.com

Sponsor

Back
Top