XHRA-2HPA USB power detection

Technical discussions around xCORE processors (e.g. General Purpose (L/G), xCORE-USB, xCORE-Analog, xCORE-XA).
New User
Posts: 3
Joined: Fri Oct 30, 2015 2:51 pm

XHRA-2HPA USB power detection

Postby mars2 » Sat Jul 08, 2017 11:38 am

I have a few questions regarding the power detection of the XHRA-2HPA.

Since I want to isolate my I2S board from USB, I use digital isolators and an independent power supply.
Pin 23, the USB_VBUS USB power detection is isolated and driven over an optocoupler circuit.
I don`t need any 5 volts, except for pin 23, the USB_VBUS Power Detect Pin.

Instead of using an extra 5V regulator (min. a zener diode), my question now is:

1. Is it possible to feed this pin with only 3.3V, resp. does the detection work with this voltage?

According to the data sheet, this is an I/O pin, and the supply voltage for the remaining pins should be around this level.
For this specifically pin, I cann`t find this, or any further information (e.g. min./max. voltage). Also the functional context with pin 15, the SELF_PWR pin as "USB power option" is missing.

2. I think I have to pull up this pin up for "self power", with my independent supply, is this correct?

For my understanding, the USB Power Detection Pin 23 serves shutting down some internal circuits, when there`s no USB connection available.

3. In this case, can I shut down the 24Mhz clock for some kind of standby mode (have also SPDIF inputs onboard) and is a reset necessary for waking the chip up again?

Thanks for any help!
User avatar
XCore Expert
Posts: 743
Joined: Thu Jun 10, 2010 11:43 am

Postby mon2 » Sat Jul 08, 2017 1:58 pm

Hi. A few comments:

1) be aware that the external flash required for this device must be pre-programmed out of circuit AND be in QSPI mode (bit inside the flash must be "1" for QSPI mode) before you can use this design. Assorted discussions on this topic on this forum. If required, send me a private email and can send the dump of the flash -> simply program into a blank flash device AND enable the QSPI bit and solder into your design to use. After this programmed flash device is installed only then you can then move to use the USB interface to alter the configuration of this device or perform firmware updates, etc.

There is no support inside the ROM code of this XMOS CPU to program a blank flash device using USB!

2) alternate option is to tri-state the CPU so that the QSPI interface pins are floating -> then assert the same target i/o lines to program a blank flash device using an external tool

3) no need for an opto-coupler for the VBUS line. You could use a single gate buffer that is 5 volt tolerant yet power the single gate buffer @ 3v3 and then the output of this 3v3 Vbus signal will feed into your digital isolator

For example, Diodes Inc. p/n 74LVC125.

"The device is designed for operation with a power supply range of 1.65V to 5.5V. The inputs are tolerant to 5.5V allowing this device to be used in a mixed voltage environment."

https://www.diodes.com/products/logic/s ... 74LVC1G125

4) VBUS is a nasty signal that produce large spikes and damage downstream parts. Many developers have witnessed such events where the non-isolated CPU pins were damaged. Seriously consider ESD protection on the USB interface pins. Must consider VBUS rail protection for overshoot voltage and transients. In addition, apply low cost in-rush protection devices. This will lead to a design that just keeps on working when out in the field.

Check the post from bowerymarc who charted the scope traces that offer proof on the Vbus spikes with respect to the cap values on the same line. Scary...

* see his last post on the first webpage for the spikes
* XMOS has a documented a few passive parts to be used to suppress these VBUS transients but there are better options

For example, SP3010 is a solid device for ESD protection but does not assist on VBUS protection. However, beefy enough to offer upto IEC 61000-4-5 protection. An external Vbus diode like Fairchild FESD05P30Z will be suitable for Vbus line protection. Finally, AP2331 is a great choice for the in-rush protection.

After this front end conditioning, feed the protected Vbus line into your digital isolator (Silabs has some great parts we have used for years without issues). The isolator can remain to operate @ 3v3. Be sure to pay attention to close by cap requirements + FR4 isolation widths under the isolator.

Have you selected a suitable isolated DC-DC power supply? The spec of the isolated power supply should meet or exceed the isolator you select - the design will be as weak as the weakest component...

5) The VBUS pin on the XMOS CPU has to be be 5 volt tolerant based on the fact that it can be direct connected to USB but ok to electrically treat as a 3v3 pin. The logic high threshold can be met using a 3v3 swing.

6) Do not recommend the shutdown of the 24 Mhz clock which is the clock used to clock the USB state machine. The XMOS ref design uses a pre-programmed PLL from Silabs. This part is available through Digikey and emits 3 desirable clock values. However, saw some designs on Aliexpress where fixed clock oscillators were used in the design. If you are concerned about current draw, consider to use MEMS / pure silicon based oscillators which are apparently much lower in current consumption and will also survive harsh falls of the end product without damage.

7) Stay away from electrolytics, tantalum caps. Consider ceramics only but do read up on how ceramic caps can derate from their marked value:

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

Electrolytics can die over time; Tantalums act like a short circuit upon death.

The above are thoughts we would implement for our version of the design. Make a better mouse trap and the customers will flock to your version of the design...

Suggest to rough up the schematic and post if practical for a final review before moving to PCB layout, production.

Disclaimer: Not an audio developer but have high volumes of isolated designs out in the field in critical sectors.

A few comments:

Ok, more than a few :)
User avatar
Active Member
Posts: 42
Joined: Wed Apr 06, 2011 8:02 pm

Postby data » Sat Jul 08, 2017 10:06 pm

Regarding capacitors -- I too try to use ceramic wherever possible, and I have not designed tantalum caps into anything for many years; I am not aware of any compelling advantages they offer over modern multilayer ceramics, at least for what I do (audio mostly).

Unfortunately, in audio signal paths, multilayer ceramic caps have a serious disadvantage: they tend to be microphonic, particularly at high values. This all but rules out their use for AC coupling. 'Lytics still rule that application unfortunately.

The use of lytics gives designs a finite life, before recapping becomes necessary -- typically 15-25 years. If there is ever a viable alternative to them, I will switch right away.
New User
Posts: 3
Joined: Fri Oct 30, 2015 2:51 pm

Postby mars2 » Sun Jul 09, 2017 12:29 am

Thanks a lot, mon2 for your few :-), and very detailed comments!

1) After flashing my first device (Spansion S25FL1, as recommended) in QSPI mode with the XMOS supplied firmware, I had to recognize, that I bought the wrong flash type, resp. the one with quad mode
disabled. Since I couldn`t set this status register bit, or write into the status register with my programmer, I finally ordered the version with the quad bit enabled.

2) Programming the flash "out of system" works just fine for me.

3) This is a very good idea, but sadly I don`t have a spare pin on my isolator. I have to feed this signal straight over the optocoupler into the USB_VBUS pin.

4) When I have my optocoupler, supplied with a stable, regulated voltage in between the VBUS connection, I shouldn`t have any spike, nor overshooting problems.
The USB-interface itself has a TPD2E2U06 TVS diode protection device in front of the isolator.

5) That`s very good news! ...and further simplifies my design.

6) I see...my design uses a fixed clock oscillator and yes, my concerns are about current draw, furthermore lifetime and unnecessary radiation when not in use (when I use the SPDIF inputs).
Yes, a MEMS-based silicon oscillator could be the better choice.

7) I completely agree with you mon2 and data - I never used tantalums.
Whenever possible (except higher voltages), I stay away from electrolytics and use polymer caps.
They are pricier but offer longer lifetime, have no dry-out behavior, lower ESR and a higher ripple current!
From the ceramic types, I use multilayer X7R and COG`s for lower capacitances.
I followed your link mon2 and this article really makes my thinking. Seems that bigger values and smaller sizes of X7R and X5R ceramic caps vary a lot with the voltage supplied (lesser with temperature)
- a fact that in this dimension, is absolutely new to me.

I`m only aiming at building a mouse trap after my taste and knowledge, there are certainly several ways to build a reliable and fine one

...but I really hope to catch a splendid specimen...:-))
Experienced Member
Posts: 94
Joined: Thu Nov 26, 2015 11:47 pm

Postby akp » Tue Jul 11, 2017 3:22 pm

I have had some problems with class 2 ceramics (e.g. X7R) where a vendor device we use charged and discharged them periodically -- making them sing at the frequency they were charged/discharged at. Not good when placed close to a very sensitive hydrophone. So I tend to prefer class 1 ceramics (NP0, C0G) even though they are bigger if there is any chance the applied voltage could have any AC components on it. When class 1 ceramics become unacceptably large and I don't think I can use a class 2 (e.g. an unregulated DC input will have lots of AC noise on it) -- then I use tantalum polymer. They are more reliable than tantalum, supposedly don't suffer from ignition, and have less strict derating.

Film are also thought to be good for audio paths but they are pretty big and (more importantly) quite expensive.

Return to “Processors”

Who is online

Users browsing this forum: No registered users and 11 guests