[time-nuts] Performance of upgraded TS-2100

[Esa Heikkinen]

Hello all

I've been running the TS2100 with MTI OCXO and HEOL DESIGN N024 GPS 
upgrades for a few monts now. I want to share some test results of the 
upgrades and how much they actually improve the performance of TS2100. 
Both ot these upgrades should be familar for the followers of this forum.

In my tests I was comparing the TS2100 PPS output against Trimble 
Thunderbolt PPS output in the normal state and simulated "loss of GPS 
coverage" state. This way it was possible to test the performance of the 
new N024 GPS board and MTI OCXO.

Thunderbolt and TS2100 was sharing same GPS antenna. Antenna cable delay 
compensation was set to 100 ns for Thunderbolt. In TS2100 the cable 
delay can be also set in 100 ns steps from engineering menu with telnet. 
After the N024 GPS upgrade the initial offset value is -1. Thus, in my 
setup with long antenna cable with 100 ns. delay the correct offset 
value for TS2100 is -2:

1 ? eng tim offset
Current offset : -2 at offset unit (100ns)

With original GPS I kept this always at zero. Because TS2100 PPS output 
was quite inaccurate with default oscillator, there was no notable 
difference with values 0 and -2 for example. TS2100 manual defines that 
PPS accuracy with original setup is less than 2 microseconds. In my own 
measurement it was just about less than 1 microsecond and changed in 
every second if examined nanosecond level. So the PPS output of TS2100 
was quite "noisy" as default - but good enough for NTP use only.

After upgrading MTI OCXO and new GPS board for HEOL DESIGN (N024) things 
really changed. Now the TS2100 PPS output stays within +-50 ns from 
Thunderbolt PPS. Here's an oscilloscope measurement result over 6 hours 
test period with persist mode selected for traces:

http://www.amigazone.fi/files/ts2100/ts2100.png

Yellow trace is Thunderbolt PPS and blue trace is TS2100 PPS. And here's 
the PPS error on same time period reported by Thunderbolt, logged with 
Lady Heather:

http://www.amigazone.fi/files/ts2100/tb.png

MTI OCXO seems to be quite sensitive to rapid environment temperature 
changes. Normally there's no problem at all, but if there's any kind of 
rapid variations it may interference TS2100 PPS. For example, if TS2100 
top coverplate is removed "on the fly" the PPS is not within +-50 ns 
anymore, it might be even several hunders nanoseconds off soon after 
removing the cover.

Then the freerun performance...

If there's a temporary loss of GPS reception - or antenna is removed 
temporarily, how long does the TS2100 keep the time? This puts the 
oscillator under tough test.

For the NTP use only the maximum acceptaple timing error is about one 
millisecond. TS2100 manual says typical NTP accuracy is 1...10 
milliseconds, so anything less than 1 millisecond should be "kind of" 
acceptable here.

With original oscillator this is reached even less than 90 minutes. 
After OCXO upgrade it was not reached at all in four hours test period. 
Four hours was selected as maximum test time because any loss of GPS 
reception or possible antenna replacement etc. should be take less than 
four hours. Test was terminated if PPS error was more than 1,5 
milliseconds, because the TS2100 was running as NTP server at same time.

In the freerun tests the TS2100 antenna was disconnected, while keeping 
Thunderbolt connected normally. Then the PPS of the TS2100 and 
Thunderbolt was compared with oscilloscope and the difference was 
manually logged after 30, 60, 90, 120, 150 and 180 minutes. After 
upgrades the PPS error was also measured before removing the antenna and 
after 240 minutes of freerun.

Each test was repeated three times. Positive result means that TS2100 
PPS was late and negative means it was early (same way as oscilloscope 
gives these readings). After each test the TS2100 was kept in normal 
state with antenna connected at 48 hours minimum to let it recover.

With standard oscillator and original GPS:

Freerun          TEST1     TEST2     TEST3
time
30 min           -157 us    -20 us   -78 us
60 min           -513 us    -109 us  -202 us
90 min           -1,026 ms  -296 us  -378 us
120 min          -1,680 ms  -530 us  -629 us
150 min          (ended)    -661 us  -943 us
180 min          (ended)    -760 us  -1,310 ms

With MTI OCXO and HEOL DESIGN N024:

Freerun          TEST1     TEST2     TEST3
time
0 min            +25 ns    +38 ns    +31 ns
30 min           +14 ns    +332 ns   -74 ns
60 min           +500 ns   +1,17 us  +238 ns
90 min           +1,58 us  +2,42 us  +900 ns
120 min          +3,16 us  +3,82 us  +1,77 us
150 min          +5,88 us  +5,06 us  +2,88 us
180 min          +11,5 us  +6,36 us  +4,04 us
240 min          +38,4 us  +10,0 us  +7,25 us

I hope that this information is useful and interesting for TS2100 owners 
planning to do the upgrades - or already done them.

Best regards,

-- 
73s!
Esa
OH4KJU