LMH0318: Jitter performance of re-clocked output

Hi Nasser,

Thank you for your response. I have tried setting register 0x99 with the value 0x01, but observed no difference in jitter on the Phabrix unit. Below is the pseudo code of setting up the LMH0318 with the added line to change the loop bandwidth register. Can you confirm that this is how register 0x99 would be configured?

--LMH0318 Initialization with loop BW change

SPIWrite (0xFF, 0x04)      --Enable channel registers

SPIWrite (0x16, 0x25)     --Enable full temperature range

SPIWrite (0x3E, 0x00)     --Initialise CDR state machine control

SPIWrite (0x55, 0x02)     --Initialise CDR state machine control

SPIWrite (0x6A, 0x00)     --Initialise CDR state machine control

SPIWrite (0x03, 0x00)     --CTLE setting. Use 00 as track is very short (EQ to Driver)

SPIWrite (0x99, 0x01)     -- **Line add to change loop BW. Default is 0x04, suggestion is 0x01

SPIWrite (0x0A, 0x0C)     --Assert CDR reset

SPIWrite (0x0A, 0x00)     --De-assert CDR reset

Also, the following code is where I add two lines to override and select raw data for out0. As mentioned, this improves jitter by using the raw data.

--LMH0318 Initialization with raw data selected for out0

SPIWrite (0xFF, 0x04)      --Enable channel register

SPIWrite (0x16, 0x25)     --Enable full temperature range

SPIWrite (0x3E, 0x00)     --Initialise CDR state machine control

SPIWrite (0x55, 0x02)     --Initialise CDR state machine control

SPIWrite (0x6A, 0x00)     --Initialise CDR state machine control

SPIWrite (0x03, 0x00)     --CTLE setting. Use 00 as track is very short (EQ to Driver)

SPIWrite (0x09, 0x20)     --Enable over-ride           ** line added to select raw data

SPIWrite (0x1C, 0x40)     --Out0 Raw Data               ** line added to select raw data

SPIWrite (0x0A, 0x0C)     --Assert CDR reset

SPIWrite (0x0A, 0x00)     --De-assert CDR reset

Many thanks ,

Alan.

Hi Nasser,

Again, many thanks for your reply.

That actual problem I am addressing is that the development of the new system was performed using a calibrated Phabrix SXe and as you stated the measured jitter of 0.14UI is within spec for an SDI. So no problems and the unit passed our development tests for production release.

However the Production test department uses a calibrated Tektronix WFM8000 unit and this unit is showing 0.39UI of jitter on the same system. In Production the system is failing the SD spec. Other example systems have been tried and all behaviour the same where they pass on the Phabrix but fail on the Tektronix. Please note this is for SD only, HD and 3G streams are fine on both.

I wanted to see if the LMH0318 could be configured to improve jitter so that both measurement systems (the Phabrix and Tektronix) would show results within the specified range. With the reclocker set to raw data both the Phabrix and the Tektronix showed exceptional jitter performance <0.08UI.

I was amazed to see that raw data through the LMH0318 gave such a clean eye and good jitter performance for both the Phabrix and the Tektronix. In fact the Tektronix was marginally better. But reclocked data on the Tektronix is significantly worse.

Please note I am only using very short cable of 0.5 metres from a video test pattern generator to the Unit under test. And also 0.5 metres from the Unit under test to the Phabrix or Tek measurement unit.

I have tried your recent posted changes without any effect. I don’t fully understand the changes, because I thought register 03 (eq index value) had no effect if register 2D[3] is cleared for manual mode.

I know the control mechanism for the LMH0318 across the SPI interface works, as I can do things like adjust the OUT0_VOD or of course override the settings to produce raw data.

It’s a strange problem. I would have thought I would see some difference by changing eq parameters in the LMH0318 even it makes things worse. Perhaps the use of short cables is preventing seeing any significant differences.

I really appreciate your time and help.

Thanks

Alan.

Hi Alan,

I am assuming for your test you are using color bar as well.

In our lab, we use both Pahbrix and also Tek WFM8300. With Tek WFM8300 we have measured under 0.1UI of alignment jitter for SD rate.

Do you see a difference if you change reg 0x99 from a value of 0x04 to 0x01?

Regards,, Nasser

Hi Nasser,

I am using a 100% color bars test pattern. And it is 625 so SD rates.

Thank you for checking with the Phabrix and WFM8300.

I don't see any difference at all in jitter or eye diagram when changing register 0x99 to 0x01.

I am thinking that because raw data shows 0.08UI on the Phabrix and 0.07UI on the WFM then the return loss network is OK in the design.

I see there is a TI eval module for the LMH0318. If this eval board still exists, then perhaps it would be a good approach to acquire one and perform similar tests on it. I have contacted Thileepaan (UK TI FAE) concerning the eval board. But if you have any more thoughts or ideas, then it would be very much appreciated. 

Thanks

Alan.