DS90UB953-Q1: Operating temperature issue of paired SER - DES at synchronous mode

Hello Torsten,

The characteristics of the channel will change over temperature and it is important to make sure the channel still meets our spec even at these low temps.  Also you can try to run the map tool that is described in the link below but if you are completely losing LOCK then it is not going to be much help.

https://www.ti.com/lit/an/snla301/snla301.pdf

https://www.ti.com/lit/ug/snlu243/snlu243.pdf

The temp sensor value going below the limit just triggers an alarm bit.

Regards,

Nick

Hello Nick,

thanks for your immediate answer.

The docs are quite interesting, but we made a margin analysis at room temperatur that shows excellent results.

I.e., the SER-DES couple is working reliably and stable. We are looking for the reasons, causing a link fail below -20°C.

The predecessor system, a camera based on 953, but working in non-synchronous clocked mode and STP wiring passed down to -40°C.

May be REFCLOCK = 25MHz in synchronous mode could cause a larger jitter?

By the way: The temp sensor value going below: Is there any impact? What is the precise temperature threshold?

Regards,

Torsten

Hi Torsten, 

I agree the results look good at room temp.  Is it possible to run the test at cold?  Would be interesting to see if better AEQ settings could help with this problem.  As for previous systems passing, are you switching from STP to coax?  Im not sure I understand the differences, however STP connection should perform better than a single-ended connection.  This is part of the channel characteristics that need to pass our spec at cold.

Temp threshold can be changed and it doesn't impact the part it is just an alarm for the system to use if needed.

Regards,

Nick

Hi Nick,

we developed a new camera with coax interface in contrast to our former one, that works with STP interface. Because last camera operates reliably down to -40°C at non-synchronous clock mode I switched over the actual coax one to non-synchronous clock mode.

Now the camera seems to be stable at -40°C operating temperature without CSI-2 data errors (figure with 10m coax cable).

Hence our configuration for synchronous mode with 25 MHz oscillator at REFCLK of DES 960 should not be optimal.

Is there a recommendation for 953 - 960 settings in synchronous mode available, could you advise suitabe AEQ settings?

Regards

Torsten    

Hi Torsten,

I think you will always see better performance in an STP system than a single ended one.  As for the Async vs sync mode the only difference would be the backchannel speed.  Were you able to run the MAP program again at -40?  

Regards,

Nick

Hi Nick,

the decision whether STP or coax is strongly determined by the specific customer application, hence we developed both types. May be our settings for operation of PoC with synchronous mode are not optimal especially at very low temperature.

Paired to a 954EVM board we got following margin analysis plots at room temperature  resp. started at -33°C:

 Regards,

Torsten

Hi Torsten,

Room temperature looks really good, but the cold start looks passing as well.  Should be able to set AEQ limits with the data that was provided and get have a working link correct?

Also I understand you want to make STP and coax applications, I was just saying we tend to see slightly better performance with STP however should work in both applications.  

Regards,

Nick

Hi Nick,

I am still tackling the low temperature issue of our PoC camera based on 953. At present I bypass the PoC input network of power supply in order to exclude an impact of frequency response and cool down the camera. By the way: For synchronous mode of SER - DES a BC frequency of 50 Mbps seems to be mandatory. Although synchronous mode is it possible to transmit with only 2 Gbps line rate (and how to)?

Regards,

Torsten

Additional questions:

Register 0xA5 at DES (at 960 as well at 954) delivers REFCLK_FREQ of DES: At both boards our ECU with 960 and the 954EVM I get frequency values that differ from connected oscillator with 25MHz. What is the reason why?

Moreover, RX_FREQ register at DES (0x4f, 0x50) deliver a RX frequency. How does this value correspond to the CLK_OUT for imager at SER?

Hi Torsten,

You can operate the forward channel at 2Gbps, are you using the devices in sync mode or non-sync mode?  

The 0xA5 register uses the internal oscillator to determine the external oscillator frequency and the internal oscillator accuracy is not going to be as reliable as an external oscillator.  Therefore this register may read 24MHz instead of 25 as expected.

Registers 0x4F and 0x50 read the pclk frequency which in a CSI serializer is not very meaningful. 

Regards,

Nick

Hi Nick,

using devices in non-sync clock mode, we are able to arrange a 2Gbps forward channel.

But how to configure SER and DES, if 2Gbps is desired in sync mode?

Regards,

Torsten 

Hi Torsten,

You should be able to change the back channel rate to half the speed in register 0x58, however there is a way on the 953 side but it requires some reserved register writes.  Let me see what I can share.

Regards,

Nick

Hi Nick,

thanks for your information. In recent time we had half back channel rate by setting 0x58 to 0x5d at DES (in non-sync mode). But for sync mode 0x58 to 0x5e seems to be more stable (and is the default value). Can you find out another setting to half?

Regards,

Torsten

Hi Torsten,

I will look into getting you this information.  Can I ask why is it needed though?  Why wont the 4Gbps work in the customers application?

Regards,

Nick

Hi Nick,

using 2 Gbps (if sufficient for transmission content) instead of 4 Gbps leads to a more robust and longer transmission distance.

But I continued testing in order to find out some more hints of the low temperature issue:

I cooled down the operating camera and logged the GENERAL_STATUS register of 953 via I2C and ALP tool.

Register value of 0x52 changed as follows:    0x45 (i.O. with image stream) – 0x47 (+CRC_ERR BC; stream error) – 0x67 (+ Reserved?; stream error) – 0x27 (no DES lock status; stream error) – 0x21 (no Forward Channel High speed lock; no stream) – (sometimes 0x20 (no BC Link detect).

I analysed the DOUT signal of single ended signal at coax cable. Beginning by 0x52:= 0x67 there are some disruptions. If the camera is cooled down completely (< -33°C) there are some stable signal diagrams:

What could be the reason why?

Regards,

Torsten

Hi Torsten,

I am not sure that using 2Gbps will result in any increase in transmission distance since our channel spec is only defined for cables up to 15m and isn't dependent on the speed of the forward channel.  With that said is there anyway to measure the quality of the channel at room temp vs cold temp.  Is it possible to compare s-paramters of the channel at cold vs room temp?  If the only way to recreate the communication issues is too cool everything down to <-33  degrees then I would argue this is a system level issue and the channel characteristics are affected by temperature.

Regards,

Nick