Using TCXO as clock source

Technical discussions around xCORE processors (e.g. xcore-200 & xcore.ai).
Post Reply
Hagrid
Active Member
Posts: 44
Joined: Mon Jul 29, 2013 4:33 am

Using TCXO as clock source

Post by Hagrid »

I am looking at using a temperature compensated oscillator chip to drive an XUF208 to improve frequency stability for an instrumentation application.

The output specs on the TCXO state:
Clipped sine wave output voltage: 0.8Vp-p
(http://www.farnell.com/datasheets/1701313.pdf)

I can't see anything in the TXCO datasheet about the midpoint voltage, so presumably it would be Vdd/2.

Do I need to buffer this to full CMOS 3.3V logic levels to use it? Or can it drive the XUF208 as is?


henk
Respected Member
Posts: 347
Joined: Wed Jan 27, 2016 5:21 pm

Post by henk »

Hi,

To me it looks like you will need something extra between the TCXO and the xcore. On the xcore side V(IH) = 2V and V(IL) = 0.7V, which would suggest that you need at least a 1.3V p-p swing; more than the 0.8V offered by the TCXO?
Hagrid
Active Member
Posts: 44
Joined: Mon Jul 29, 2013 4:33 am

Post by Hagrid »

Thanks.

There is a buffer needed it seems for the TCXO. Apparently other oscillators have it inbuilt but TCXO's don't because it produces heat which just adds to the stability problem.

Seems like a simple enough problem to solve - it is just bumping up the parts count.
Hagrid
Active Member
Posts: 44
Joined: Mon Jul 29, 2013 4:33 am

Post by Hagrid »

Before I blow something up... My prototype PCBs should be on hand sometime this week, but perhaps they are wrong.

For reference, the chip I am using is XUF208-256-TQ64.

The issue of concern is the maximum voltage for the oscillator driving the chip.

Another thread mentioned 1.8V in connection with a 200-series, so I have just gone digging.

xCORE-200 USB sliceKIT Hardware Reference Manual contributes this to the subject:
4.5 Device Clock
It should be noted that if an external clock signal is used the maximum signal level
should be 1.8V. The output from the 1.8V DCDC regulator on the device can be
used to provide an appropriate supply (VDD1V8 pins).
This does not apply if a crystal is used to provide a clock to the device as the clock
signal is generated internally from the 1.8V supply.
[I didn't notice this the first time around]

However, this seems to be contradicted by the schematics.

Figure 2 "Simplified Reference Schematic" shows a 24MHz oscillator powered by 3.3V.

Main schematic shows the oscillator being buffered by U26, a 3.3V buffer, with a 33ohm output resistor.

So far, I have found no reference in either the datasheet for my target chip nor the datasheet for the chip used on the USB sliceKIT to 1.8V.

Right now, my design uses a 3.3V buffer for the TCXO, but I don't have a series resistor in the output. Trace length from buffer to chip is 6mm.

So, apart from the one comment in the hardware reference manual, I am not seeing anything else that supports this 1.8V maximum on the clock pin.

Advice is welcomed...
User avatar
mon2
XCore Legend
Posts: 1913
Joined: Thu Jun 10, 2010 11:43 am
Contact:

Post by mon2 »

Hello Hagrid. Although no direct experience in using TXCO, recommending for you to consider a single TXCO with 3V3 output such as:

http://www.digikey.com/product-detail/e ... ND/3907426

Some other comments:

1) the datasheet for your controller:

http://www.xmos.com/download/private/XU ... 1.9%29.pdf

does not mention any relationship to requiring a 1V8 CLK oscillator

2) the example schematic in the same datasheet is powering the CLK oscillator @ 3V3.

3) the supporting SliceKit for this controller has a reference clock with a 3V3 swing voltage.

Image

4) there is no 1V8 power rail on the XUF208 controller being used

For the above reasons, suspecting that the reference to requiring a CLK source of 1V8 is a typo in the related documentation.

Only XMOS can confirm and then revise this documentation for accuracy. If you wish to proceed with caution, consider to source a 1V8 TCXO and also a 3V3 TCXO. Start with the 1V8 in your design for testing while waiting for an official response from XMOS. However, fairly confident that you will be fine with the 3V3 clock oscillator since XMOS is doing the same on their SliceKits for this controller.

On the other hand, the XS1-U6A-64-FB96 Datasheet does indeed highlight the need for a 1V8 clock source throughout the referenced documentation. The XS1-U6A does feature a 1V8 power rail so concluding that the documentation was incorrectly cut & pasted for the XUF208.

Image

Summary - the clock source is 3V3 for the XCORE-200 devices and the documentation referencing the 1V8 clock for your XUF208 controller is a typo. This same paragraph is referenced inside the:

U16 sliceKIT Core Board Hardware Manual

http://www.farnell.com/datasheets/1825477.pdf
* section 4.6

on which a different controller is deployed (XS1-U16A-128-FB217) which does indeed require a 1V8 clock source.

Your XUF208 does not have this requirement.
Hagrid
Active Member
Posts: 44
Joined: Mon Jul 29, 2013 4:33 am

Post by Hagrid »

Thanks, mon2.

It really does look like a copy/paste error in the XMOS documentation. The convincing arguments are: (1) the XUF doesn't have a 1.8V supply anywhere, (2) nor does it have a crystal oscillator that might use one as an alternative to the clock input, (3) the datasheet is silent on the subject and the requirement for a max 1.8V clock input would be a critical design parameter, (4) reference design uses 3.3V.
henk
Respected Member
Posts: 347
Joined: Wed Jan 27, 2016 5:21 pm

Post by henk »

Hi Hagrid and mon2,

Many thanks for reporting this - it is definitely an error. On XCORE200 devices CLK is sampled against VDDIO (or VDDIOL if there is a separate left supply), so power your oscillator from VDDIO(L).

For future reference, if you find something like this, a link would be very helpful to make the document reference unambiguous.

It will be fixed in the next release.

Cheers,
Henk

PS, the general purpose XCORE200 slicekit hardware documentation has the same error.
Post Reply