[time-nuts] 10MHz LTE-Lite

Orin Eman orin.eman at gmail.com
Thu Dec 25 23:14:54 EST 2014

I had PCBs made by OSHPark for the buffer discussed below.  I just built
one up and it's working fine.  3 outputs giving about 10 dBm each.

The board design is shared at:  https://oshpark.com/shared_projects/pCpmILwj

There are links to the schematic and a picture of the mostly complete
board.  If you have SMA end-launch sockets, they may fit the board (I'd
have to scrape some solder mask off the ground plane for the ones I have to
fit) or you can just solder coax directly.  Use the pads on the bottom of
the board for the coax screen.

Feel free to use the design.  (If you sell anything based off it, please
attribute the source.)


On Sat, Dec 6, 2014 at 10:32 PM, Orin Eman <orin.eman at gmail.com> wrote:

> Hi Said,
> It's a little while since you sent this, but I just finished some testing
> with the LTE Lite.
> I already had a Trimble Thunderbolt and also have an HP 5335A with OCXO.
> The 5335A has shown the Trimble O/P 10 MHz +/- 0.03 Hz for the last few
> years (displayed frequency on the 5335A has drifted down by about 0.04 Hz).
> So, I set up the LTE Lite (10MHz TCXO version) with its antenna about 6'
> from that used by the Trimble.  After letting it settle down, it looked
> good.  I was using the high impedance input of the 5335A (aside: if set for
> DC coupling, the 5335A would read 20MHz from the ringing.  I only have
> about 18" of RG-188 after the pigtail supplied with the LTE Lite).
> Today, I made up a buffer using a 74AC04 and LT1763-3.3 LDO regulator
> supplied with 5V from a bench supply.  Two inverters each into 100 ohms
> then 0.1uF DC block as suggested.  I connected this to the LTE Lite instead
> of directly to the 5335A with the 5335A set to 50 ohms and the 5335A read
> 0.5Hz low!  It settled down to the original reading after a while.  I
> looked at the signal from the buffer using a 50 ohm pass-through terminator
> and the signal looked nice and square.
> A few hours later I went back and it still looked good.  I decided to look
> at the input to the buffer on a TDS-210 'scope with a 10X probe.  There is
> now perhaps 6" of coax to the buffer which has 1 Mohm in parallel.  The
> signal is about 5V pk-pk, but the ringing dies down quickly.  BUT, the
> 5335A was displaying a few hundredths of a Hz higher than 10MHz, whereas it
> was a few hundredths below before!  When I removed the probe, it went a few
> hundredths of a Hz below where it was originally and gradually recovered.
> So, in addition to temperature, the LTE-Lite eval board with 10MHz TCXO
> appears to be sensitive to load as well as temperature, such that just a
> 10X oscilloscope probe will affect the output.  Normally, you probably
> wouldn't notice this, but I switched the loads while it was running.
> Certainly, once put in an enclosure, the 74AC04 buffer would be permanently
> connected and I'd assume any effect noticed above would be during the
> warmup of the LTE-Lite and wouldn't really be noticed.
> I'm sure I forgot some detail or other in the above description.  I
> couldn't find a DIP 74AC04 so I made a PCB for an SMT 784AC04 using MG
> Chemicals 1/32" positive-resist PCB and Eagle for the design; 1206 100 ohm
> series resistors and 1206 0.1uF ceramic DC blocking capacitors.  I laid it
> out for SMA sockets that you could wire RG-316 or similar directly instead.
> I used a 6" MMCX to RG-316 pigtail from the LTE-Lite to my buffer.  I
> mounted one SMA socket on the outputs which is what I used to connect to
> the HP 5335A.  The LT1763 has a 1uF 35V tantalum on the input and 10 uF 30V
> tantalum on the output (they are what I had in the parts bin, or I'd have
> used *smaller* ceramic parts).  I tweaked the design for hand soldering
> with no solder mask (i.e. 20 mil clearances and top layer restrictions
> around most components to keep the ground fill away).
> Orin.
> On Sat, Nov 22, 2014 at 9:44 AM, Said Jackson via time-nuts <
> time-nuts at febo.com> wrote:
>> Hi Paul, Jim, David,
>> Let me address all your emails:
>> Glad you got your boards. $50 in overseas additional charges from your
>> post office sucks!
>> Some hints for experimenting from what I have learned:
>> You definitely want to build a 50 Ohms buffer for the 10MHz boards and
>> the synthesized outputs on all boards; on the 20MHz boards on the Tcxo
>> output it's optional.
>> The biggest problem is building a suitable 3.3V or 5V power supply. I
>> built a buffer using two NC7SZ04 chips receiving the input in parallel with
>> a 1M terminating resistor to ground. Then using a 100 Ohms series resistor
>> on both outputs to get ~55 Ohms equivalent impedance, and combining the two
>> R's to drive the 50 Ohms inputs through a 100nF cap for DC blocking. You
>> can use a 74AC04 just as well. I tried a standard high quality 3-pin
>> regulator and got very bad AM noise modulation due to the large noise on
>> the rail. Then I used a very low noise LDO from LT and that solved the
>> problem and the output is now very clean and drives 50 Ohms inputs with
>> ease.
>> you can grab the very low noise 3.0V rail output from the eval board to
>> power the buffers (see the schematics in the user manual) but loading that
>> creates a bit of heat on the LTE-Lite which will affect stability a tiny
>> bit.
>> On the power consumption, you can see we go through a linear regulator to
>> get 3.3V from 5V USB/EXT power. This is very inefficient. For best power
>> consumption you want to use a very high efficiency buck regulator to
>> generate the 3.3V from your battery. This means you loose the USB interface
>> though as that chip runs from 5V and has an internal LDO.
>> On the zero Ohms R2/R3 resistors, check the schematics - these allow you
>> to power the DIP-14 Tcxo from either the digital 3.3V rail, or the
>> low-noise 3.0V rail (default). The software will auto-detect if you attach
>> a 10MHz or 20MHz Tcxo, no configuration is needed.
>> On drafts, yes that is the number one cause of phase drifts. We put the
>> board into an ESD bag, and put some thin ESD padding material on top. That
>> prevents drafts, and following the EFC curve you can see the unit still
>> reacts slowly to the AC or heaters kicking on. That's normal, and that's
>> why we discipline to GPS.. In our setup the units have typically less than
>> 20ns standard deviation on the RF and 1PPS outputs. The raw gps 1PPS output
>> on the header is even better on average, but has the sawooth error on it.
>> The sawtooth error correction value is in the PSTI NMEA message for those
>> that want to use the raw gps 1PPS output and correct the sawtooth
>> externally. This chip has a very high rate internal system frequency that
>> results in very low residual sawtooth error.
>> On the auto survey process - this is disabled when using 3D mobile mode
>> by shorting pins 1 and 3 on the 3-pin header as described in the "read me
>> first". But be aware that changing that header with power applied results
>> in flash memory corruption, and thus a very bad day. That's why we did not
>> solder the header - to avoid any accidents..
>> Bye,
>> Said

More information about the time-nuts mailing list