XTAG-3 connection problems

Technical questions regarding the xTIMEcomposer, xSOFTip Explorer and Programming with XMOS.
mcs5
Junior Member
Posts: 5
Joined: Mon Jan 25, 2016 12:22 am

XTAG-3 connection problems

Postby mcs5 » Mon Sep 04, 2017 6:27 pm

I am trying to connect to a board using the XU216-512-TQ128 with the XTAG-3 adapter - but without much luck...

The JTAG pins are connected to the xSYS header, as described in the datasheet. Power and clock is provided to the chip. Is anything else needed for the connection to work?

This is what I get, when I run xrun -l:

C:\Documents and Settings\Bruger>xrun -l

Available XMOS Devices
----------------------

ID Name Adapter ID Devices
-- ---- ---------- -------
0 XMOS XTAG-3 Ib857mvh None
User avatar
mon2
XCore Expert
Posts: 697
Joined: Thu Jun 10, 2010 11:43 am
Contact:

Postby mon2 » Tue Sep 05, 2017 12:47 am

Hi.

1) Try different USB ports? USB 2.0 ports may have better results than USB 3.0 ports but do test all the USB ports on your box.

2) Which OS is being used for this testing?

3) Assuming the board under test is custom? Can you share the relevant parts of the schematic? Such as power supply, power supply sequencing, Clock, JTAG connection details.
mcs5
Junior Member
Posts: 5
Joined: Mon Jan 25, 2016 12:22 am

Postby mcs5 » Tue Sep 05, 2017 1:18 am

mon2 wrote:Hi.

1) Try different USB ports? USB 2.0 ports may have better results than USB 3.0 ports but do test all the USB ports on your box.

The computer I'm using only has USB 2.0 ports - but I will try the other two as well.

2) Which OS is being used for this testing?

XP - I can try using Win 7 at work in a day or two. I'm using software version 14.2.4

3) Assuming the board under test is custom? Can you share the relevant parts of the schematic? Such as power supply, power supply sequencing, Clock, JTAG connection details.

It is custom - based mostly on a demoboard schematic. I could have screwed some part of it up of course :)

The power input on the schematic is +7.
Attachments
XMOS_part_gray.png
(85.04 KiB) Not downloaded yet
XMOS_part_gray.png
(85.04 KiB) Not downloaded yet
User avatar
mon2
XCore Expert
Posts: 697
Joined: Thu Jun 10, 2010 11:43 am
Contact:

Postby mon2 » Tue Sep 05, 2017 2:08 am

Hi. Thanks for the schematic post. Will review in more details when less tired but for starters,

U28 = 74ACH1G08 (wrong # with 73ACH1G08 in the schematic), is a push-pull gate. This is not suitable for this portion of the design. That is, this push-pull gate will provide for a logic high or logic low so it cannot be logically OR'd with the XTAG interface.

What you need here is an open drain version of this AND gate. Otherwise, you will have contention between this gate output and the XSYS reset line and who knows who will win.

You can consider to replace this gate with the 74AHC1G09 which is open drain output. Respectively, being open drain, the output does NOT output a HIGH level. So you must compensate by applying an external pull-up resistor to +3v3. Consider to add a 10k pull-up here for the #RESET line. With this combination, either leg of the inputs can pull the #RESET line LOW and no one will be challenging the other lines for the high since it is only the pull-up resistor that will perform this task.

Summary, open drain output = HI-Z when HIGH; LOW = LOW. For this reason, you need an external pull-up resistor on such outputs.

====

Some additional feedback...

1) ADM1085 is an open drain output. For this reason, you must apply an external pull-up resistor onto pin 4 of this component to define the weak high. Apply a 10k to +3v3 here. XMOS uses the same value in their ref designs. Without this pull-up, this output pin of the voltage supervisor will be high-z.

2) IL717 isolators are great but more expensive than their Silabs drop in replacements. We have used both in high volumes and Silabs is a valid replacement with cost savings. Please note that we are not using in audio field so best to review the results under your test conditions.

3) Recom is a great company for the isolated DC-DC module.

4) Be sure X4 is clocking out ok @ 24 Mhz. Most clock oscillators offer pin 1 to be the EN (Enable if HIGH) function. To be sure the clock is enabled, consider to strap pin 1 to +3v3. We have seen an oscillator mfr who missed the required internal pull-up on pin 1 and the clock did not enable so best to cover yourself here.

From a quick review, rest looks to be solid but may have more to say when rested. However, the above raised issues will be show stoppers and can explain why you are not seeing the positive results. Hope this helps.
Last edited by mon2 on Tue Sep 05, 2017 2:35 am, edited 1 time in total.
mcs5
Junior Member
Posts: 5
Joined: Mon Jan 25, 2016 12:22 am

Postby mcs5 » Tue Sep 05, 2017 2:34 am

The XSYS reset line controls one input of the AND gate - so I don't think there should be any fighting?

And yes, it is a 74AHC1G08. 73 is my code for connected to the +3 supply rather than VCC.
User avatar
mon2
XCore Expert
Posts: 697
Joined: Thu Jun 10, 2010 11:43 am
Contact:

Postby mon2 » Tue Sep 05, 2017 2:37 am

Ah yes. Knew I was tired. My bad. Thought this output was paired with the XSYS #RST. This is not the issue but the other raised ones should be reviewed.
User avatar
mon2
XCore Expert
Posts: 697
Joined: Thu Jun 10, 2010 11:43 am
Contact:

Postby mon2 » Tue Sep 05, 2017 2:44 am

Assuming U45 is being powered ?
User avatar
mon2
XCore Expert
Posts: 697
Joined: Thu Jun 10, 2010 11:43 am
Contact:

Postby mon2 » Tue Sep 05, 2017 11:52 am

More...

1) The XMOS devices can consume some heavy amounts of current. Respectively, be sure your power rails are supporting of this requirement. Concerned about the power source feeding your circuit and also if the 500mA is ample (NCP5501DT50G) which is then fed into the Recom isolated 5 vout DC-DC module @ 200mA. It appears that this 200mA rate power supply is then feeding into the 1v0 switcher by ST. According to the XMOS datasheet, you should offer support of up to 600mA on the 1v0 rail. The 3v3 rail may be ok with the 150mA but best to check. The ST switcher is fine but limited to the max current support upstream.

So the Recom @ 200 mA (max) is being used to power the +3v3 and +1v0 rails of the XMOS CPU which may be too weak but is dependent on your XMOS IP and CPU usage.

However, for initial board testing and the observed no-detect issue, this is not the root cause.

2) Please confirm if the grounds on the left side of the NVE IL717E is the same as the right ride of these isolators. If yes, then the key feature of these isolators is lost and these parts will become costly level shifters. For best results, the grounds must be kept separate. Silabs Si8641 series are drop in parts and available at varying speeds and isolation values at lower costs. Each vendor claims to use different isolation methods so again, best to explore for your needs.

3) Consider to add a few despikes and bulk caps onto each Vdd rail of the digital isolators. Recall that Silabs recommends the use at least 1uf onto each power rail. Ceramics are fine but do quickly review this note on such parts:

https://www.maximintegrated.com/en/app- ... vp/id/5527

http://pdfserv.maximintegrated.com/en/an/TUT5527.pdf

4) Assuming that QSPI flash U22 is being powered.
mcs5
Junior Member
Posts: 5
Joined: Mon Jan 25, 2016 12:22 am

Postby mcs5 » Tue Sep 05, 2017 2:26 pm

mon2 wrote:1) ADM1085 is an open drain output. For this reason, you must apply an external pull-up resistor onto pin 4 of this component to define the weak high. Apply a 10k to +3v3 here. XMOS uses the same value in their ref designs. Without this pull-up, this output pin of the voltage supervisor will be high-z.

That's exactly what I thought about before going to bed last night. Is the output of the ADM1085 open drain without a pull-up? And yes it is. That could explain something...

2) IL717 isolators are great but more expensive than their Silabs drop in replacements. We have used both in high volumes and Silabs is a valid replacement with cost savings. Please note that we are not using in audio field so best to review the results under your test conditions.

As far as I remember if was the jitter specs that made us choose the IL-types.

4) Be sure X4 is clocking out ok @ 24 Mhz. Most clock oscillators offer pin 1 to be the EN (Enable if HIGH) function. To be sure the clock is enabled, consider to strap pin 1 to +3v3. We have seen an oscillator mfr who missed the required internal pull-up on pin 1 and the clock did not enable so best to cover yourself here.

It should have an internal pull-up, and we haven't had problems so far. The oscillators are from Fox.

mon2 wrote:1) The XMOS devices can consume some heavy amounts of current. Respectively, be sure your power rails are supporting of this requirement. Concerned about the power source feeding your circuit and also if the 500mA is ample (NCP5501DT50G) which is then fed into the Recom isolated 5 vout DC-DC module @ 200mA. It appears that this 200mA rate power supply is then feeding into the 1v0 switcher by ST. According to the XMOS datasheet, you should offer support of up to 600mA on the 1v0 rail. The 3v3 rail may be ok with the 150mA but best to check. The ST switcher is fine but limited to the max current support upstream.

So the Recom @ 200 mA (max) is being used to power the +3v3 and +1v0 rails of the XMOS CPU which may be too weak but is dependent on your XMOS IP and CPU usage.

The DC-DC module fitted is actually the 2W type with the same footprint, so I think the 400mA should be enough. I will check the 3V3 rail also.

2) Please confirm if the grounds on the left side of the NVE IL717E is the same as the right ride of these isolators.

It's not. There are two separate ground planes on the board.

3) Consider to add a few despikes and bulk caps onto each Vdd rail of the digital isolators. Recall that Silabs recommends the use at least 1uf onto each power rail.

I use the 100n caps recommended for the IL717.

Ceramics are fine but do quickly review this note on such parts:

https://www.maximintegrated.com/en/app- ... vp/id/5527

I read that some time ago. I use physically large X7R caps to try to avoid that problem...

mon2 wrote:Assuming U45 is being powered ?
4) Assuming that QSPI flash U22 is being powered.

Yes, they are both powered. All pins that are declared as power pins are hidden by OrCAD... The power pins on the XMOS are declared as input pins, so they show up on the schematic.

Hope this helps.

It does! Thanks a lot!
User avatar
mon2
XCore Expert
Posts: 697
Joined: Thu Jun 10, 2010 11:43 am
Contact:

Postby mon2 » Tue Sep 05, 2017 4:26 pm

It should have an internal pull-up, and we haven't had problems so far. The oscillators are from Fox.


Ironically, we were sourcing from the supplier in Taiwan (Hosonic) who claimed to build for Fox but that was about 10-12 years ago. Met them @ Taitronics and bought quite a bit of volume from them till we had this issue. They did not want to take back the 3-5 pieces we found to be defective with pin 1 not supporting the OE at their shipping expense from our office in Canada for the root cause analysis. That was a turn off (pun intended) for us and we switched suppliers. Have used China Fronter in China and recently (past few years) started to source from WTL International. Both are always at the Electronic fairs in HK and their quality has held up with fair pricing. For example, the 24 Mhz oscillator (3225 footprint) you have on board should be around $ 0.35 - $ 0.40 USD in single T&R qty. Crystals are $ 0.12 USD to $ 0.07 USD varying with the property of the part.

Due this reason, all of our designs now hard strap the pin # 1 of the SMD oscillator to Vcc to guarantee the enabling of the oscillator.

Return to “Development Tools and Programming”

Who is online

Users browsing this forum: No registered users and 8 guests