Hi-Speed Layout on ADS5400 EVM

Hi,

The signal traces from the SMA input J1 (and from the optional SMA J2 if a differential input is used and the baluns bypassed) are by necessity single ended controlled impedance because the components themselves (such as the SMA connectors, zer-ohm jumper resistors, baluns or transformers and AC coupling caps) are too bulky to let us tightly couple the traces. But after all the optional components on the input circuitry we are able to tightly couple the trace from R25/26 to form a differential controlled impedance path up to the ADC.  In both cases, however, we notched out the layer2 ground plane under the input circuitry and referenced to a ground plane on layer three so that we *could* make the traces wider.  We did this so that the solder pads for the various components could 'blend in' with the width of the signal trace instead of having a large solder pad relative to the width of the trace.  This was to reduce the lumps of parasitic capacitance to ground at each of the solder points since we have so many solder options available for the input circuit.  We can't do much about the size of the solder pads, but we can size up the trace by playing with the layer stackup to match signal to solder pad.

Regards,

Richard P.

Dear Richard,

thank you very much for your answer - now  I understand why there is a big ground area in *power* layer #3 and the notch in ground plane #2.

Is the differential configuration from R25/26 to the ADC 100Ohm / Differential Grounded Coplanar (I think it's grounded coplanar because the coplanar ground plane is very close to the transmission lines)

Regards,

Michael K.

Hi,

I took a look at the design database with one of our board layout people.  The differential trace in question is a 10-mil wide trace with a 6-mil spacing between the trace.  (And with the ground plane on layer 2 below notched away the actual ground reference below is on layer three, which is a 5mil pre-preg plus 6mil pre-preg away.)  These values were chosen to give a 100 ohm differential impedance.  The ground plane on the top layer on either side of the pair is spaced at 30-mil away.  So i would not call this coplanar.  A general rule of thumb is that something that is 3-W away is not tightly coupled, and with the width W=10mil i can see where the spacing out to the ground pour on the top was set to be 30mil to be 3W.

I also note that the LVDS pairs are to be 100 ohm, and the board stackup calls for them to be 4-mil width with 6-mil spacing and the ground plane below is on layer 2.  The original revision of the EVM also had the tightly coupled portion of the analog input set to use the same trace width and space as the LVDS, but when we notched out the ground plane under the analog inputs then the dimensions of the analog path increased throughout, which was the intent.

Regards,

Richard P.

HI Richard,

thanks for your very (very) useful information. How thick is the FR-4 core for the 1st and 2nd layer? I think it's thickness adds to the two pre-pregs of 5 and 6 mil so that we have a distance of (5mil + 6mil + FR-4) as distance between the traces on top and the ground reference.

Regards,

Michael K.

Dear Forum,

I have another question concerning the input matching of the 100Ohm differential lines on the ADS5400 Eval Board. There are two (why two?) baluns on the Eval board which translate the single ended input signal for the ads into a 100Ohm differential signal. Now I see a 100Ohm matching resistor close to the input pins of the ads5400 - but the datasheet says that a 100Ohm matching resistor is already included on chip.

So, why is there a extrernal 100Ohm resistor? is it a 50 Ohm differential line? I have no idea. Please help me ...

Thanks, Regards, Michael

Hi,

The two back to back baluns (or transformers, when they are used) provide a better single ended to differential conversion, with lower 2nd adn 3rd harmonics than a single balun provides, in many instances.  A single balun will (depending on input frequency and specs of the balun) still leave some 2nd and 3rd harmonic content after the single ended to differential conversion.  A  double balun often offers better balance although although with more loss.  If the signal coming in is already differential then the double balun would not be necessary.

Since the back to back balun presents a 1:1 coupling ratio - the 50 ohm single ended input before the balun represents a 50 ohm differential impedance after the baluns.  And so the termination must be an effective 50 ohms differential.  This is why we have the external 100 ohm in parallel with the internal ~100 ohm termination.  If the input were two 50 ohm differential cables then the differential impedance after the baluns would be 100 ohms and the external resistor would not be appropriate.  (So i suppose for that length of 100 ohm closely coupled trace from the last balun to the termination resistor, we can only make the trace 100 ohm differential or 50 ohm differential, and one would be optimal for single ended input and the other optimal for differntial input, and the trace will be sub optimal for one or the other input choices.)

Regarding your earlier question on the board stackup, I am attaching the fab drawing showing the stackup.

Regards,

Richard P.

Hi Richard, and again, thank you very much.

So I assume from your post from 10 Feb 2010 5:20 PM , the Differential Line between the Baluns and the ADC are 100Ohm Differential Impedance (that's what you wrote in you r post) with a 50Ohm matching resistor? Please excuse my repeatedly questioning but for me that's very uncommon :-) I just want to make sure I understood that right.

Thanks and Regards,

Michael K.

Hi,

I confirmed that the differential trace between the baluns adn the ADC are indeed 100 ohm differential, which would be optimal *if* the EVM is set up for a differential input comprised of two 50 ohm input cables.  Then the termination on the ADS5400 is optimum without the external 100 ohm resistor in parallel.  But if we accept a single 50 ohm input and convert that to differential with the transformers or baluns, then we want to terminate the line with 50 ohm.  And then we would want the differential trace to have been designed for 50 ohm differential.  We cannot lay out the differential portion of the trace to be either 50 ohm or 100 ohm when we want it to be, so we picked one case to be optimal.  Hmm, in hindsight i think the single ended input to the EVM is the more common case and we should have optimized for that.    As you say - the 100 ohm trace terminated into 50 ohm is uncommon - that is a nice way of putting it.   :-)  Trying to allow flexibility for too many things i think.

Regards,

Richard P.

Aaaaah - now i see. And I thought you had indeed already optimized for that - that confused me :-)

So, if I use a single ended RF input (that's what I am going to do for my application) I will use a balun and layout the differential Lines going out of the balun as a 50Ohm differential pair and terminate it with 50Ohms (by an additional external 100Ohm resistor).

The actual EVM setup (with 100Ohm diff Lines and 50 Ohms Termination) is hereby not matched correctly. Can one estimate the lack of performance because of  this non-matched state?

Regards,

Michael