Basic TIC measurement
Another newbie question (sorry). I'm using an HP 5315A Universal
Counter and doing some basic experiments to teach myself.
Here is the setup: 100KHz square wave as the "A" input to the 5315A.
Using a BNC-tee to connect a length of RG58A/U from "A" to the "B"
input and another tee with a 50 Ohm terminator. Using the TI Avg mode
and the gate time maxed at 13.4S, I read a TI of 22.5nS.
Assuming I got the numbers correct, light takes 3.3356nS to travel 1M
in a vacuum. 22.5nS / 3.3356nS/M = 6.745M. If I divide the measured
length of the coax (3.825M) by 6.745M, I get a velocity factor of
56.7%. This is quite a bit lower than the nominal 66% I expected.
So, is the difference due to:
The BNC-tee connectors?
My use of a 100KHz square wave instead of a PPS?
The resolution of the 5315A (100nS single-shot, 1nS averaging)?
Some other failing of the 5315A? It was just out of cal in Feb 2009.
Something else entirely?
All of the above?
The reason I'm using a 100KHz square wave is that I haven't yet modded
either of the Z3801A's to bring out a PPS signal where I could use it.
That is on my short list. The 100KHz from a Spectracom 8140 was the
closest I can get right now. Is this type of measurement useless with
anything other than a PPS signal?
An interesting observation - when I remove the 50 Ohm terminator, I
get a much shorter TI. I assume this is due to reflections back down
the coax.
Joe Gray
KA5ZEC
Joseph Gray wrote:
Another newbie question (sorry). I'm using an HP 5315A Universal
Counter and doing some basic experiments to teach myself.
Here is the setup: 100KHz square wave as the "A" input to the 5315A.
Using a BNC-tee to connect a length of RG58A/U from "A" to the "B"
input and another tee with a 50 Ohm terminator. Using the TI Avg mode
and the gate time maxed at 13.4S, I read a TI of 22.5nS.
What about the difference in propagation delay between the 2 counter inputs?
This could easily amount to several ns and account for your results.
You need to measure this delay.
Split the signal in 2 and feed it to both inputs with nominally equal
length cables.
Measure the difference then swap the 2 cables over and measure the
difference.
Average the 2 differences to get the interchannel differential delay.
This fails if the counter cant measure negative delays and a more
complex procedure is required.
Another issue is if the slew rate of the signal at the trigger threshold
is low then trigger threshold offsets between the 2 counter inputs will
affect the effective differential delay between the 2 inputs.
Assuming I got the numbers correct, light takes 3.3356nS to travel 1M
in a vacuum. 22.5nS / 3.3356nS/M = 6.745M. If I divide the measured
length of the coax (3.825M) by 6.745M, I get a velocity factor of
56.7%. This is quite a bit lower than the nominal 66% I expected.
So, is the difference due to:
The BNC-tee connectors?
My use of a 100KHz square wave instead of a PPS?
It shouldn't matter as within the resolution of your instrument the
cable is nondispersive.
However you need to ensure that the reflection coefficient of each 5315
input is small.
The resolution of the 5315A (100nS single-shot, 1nS averaging)?
Some other failing of the 5315A? It was just out of cal in Feb 2009.
Something else entirely?
All of the above?
Just the differential delay between the 5315 inputs.
The reason I'm using a 100KHz square wave is that I haven't yet modded
either of the Z3801A's to bring out a PPS signal where I could use it.
That is on my short list. The 100KHz from a Spectracom 8140 was the
closest I can get right now. Is this type of measurement useless with
anything other than a PPS signal?
No, as long as cable dispersion is insignificant.
An interesting observation - when I remove the 50 Ohm terminator, I
get a much shorter TI. I assume this is due to reflections back down
the coax.
Yes this will alter the signals seen by each input and due to the finite
slope of the input signal and differences in trigger threshold between
the 2 inputs the effective differential delay will change.
Joe Gray
KA5ZEC
Bruce
Bruce, thanks for the very helpful advice. My comments are interspersed below.
You need to measure this delay.
Split the signal in 2 and feed it to both inputs with nominally equal
length cables.
Measure the difference then swap the 2 cables over and measure the
difference.
Average the 2 differences to get the interchannel differential delay.
This fails if the counter cant measure negative delays and a more
complex procedure is required.
I get 0.9nS one way and 1.1nS with cables swapped. This with tees and
terminators on each input. The cables are almost exactly the same
length. So, I have a 1nS difference between the two inputs.
Another issue is if the slew rate of the signal at the trigger threshold
is low then trigger threshold offsets between the 2 counter inputs will
affect the effective differential delay between the 2 inputs.
I tried it both in trigger mode and in level mode, where the trigger
threshold is supposed to be at zero crossing (according to the
manual). The difference between the two is about 0.1nS, which may be
due to my inaccurate setting of the trigger controls.
However you need to ensure that the reflection coefficient of each 5315
input is small.
How do I do this?
Thanks very much,
Joe Gray
KA5ZEC
Joe
Joseph Gray wrote:
Bruce, thanks for the very helpful advice. My comments are interspersed below.
You need to measure this delay.
Split the signal in 2 and feed it to both inputs with nominally equal
length cables.
Measure the difference then swap the 2 cables over and measure the
difference.
Average the 2 differences to get the interchannel differential delay.
This fails if the counter cant measure negative delays and a more
complex procedure is required.
I get 0.9nS one way and 1.1nS with cables swapped. This with tees and
terminators on each input. The cables are almost exactly the same
length. So, I have a 1nS difference between the two inputs.
Did you use a resistive splitter or just a tee to split the signal?
Another issue is if the slew rate of the signal at the trigger threshold
is low then trigger threshold offsets between the 2 counter inputs will
affect the effective differential delay between the 2 inputs.
I tried it both in trigger mode and in level mode, where the trigger
threshold is supposed to be at zero crossing (according to the
manual). The difference between the two is about 0.1nS, which may be
due to my inaccurate setting of the trigger controls.
One way to see the effect of threshold differences between the 2
channels is to measure the delay between the channels as the signal
transistion (rise, fall) times are increased.
However you need to ensure that the reflection coefficient of each 5315
input is small.
How do I do this?
Either estimate it (use a circuit simulator) from the input impedance (R
and C) and the signal transition times or find a TDR (expensive) to
measure it.
You need to know the signal seen by each trigger circuit and also have
some idea of the trigger circuit bandwidth to estimate the significance
of any such reflections.
A circuit simulator can give you some idea of the significance.
Terminating each input in 50 ohms helps but if their is a significant
wiring length between the input connector and the trigger circuit the
reflection coefficient will not be as low as that when the termination
is at the trigger circuit input.
Thanks very much,
Joe Gray
KA5ZEC
Bruce
Did you use a resistive splitter or just a tee to split the signal?
Just a tee. I have tees and terminators at each input of the 5315A as well.
I guess I could try using the multiple outputs on the back of the
Spectracom 8140, but that would introduce another variable -- the
outputs may not be exactly in phase. Also, isn't a pulse or square
wave better for doing TI measurements? What the heck. I may try that
just to see what happens :-)
Either estimate it (use a circuit simulator) from the input impedance (R
and C) and the signal transition times or find a TDR (expensive) to
measure it.
Well, actually I can borrow a TDR. I'll have to read up on how to
measure reflection coefficient with it. The only time I ever used a
TDR was many years ago, in the Air Force.
Joe Gray
KA5ZEC
Joe
Joseph Gray wrote:
Did you use a resistive splitter or just a tee to split the signal?
Just a tee. I have tees and terminators at each input of the 5315A as well.
You need to use a resistive splitter and not a tee to control reflections.
If you don't have one, make one using 3 resistors mounted in ametal box
with a coax connector for each arm/port of the splitter
This will work reasonably well if your signal risetimes aren't too short
ahnd the component leads arent excessively long.
If the 2 cables attached to the tee have the same length you are
attempting to drive a 25 ohm load from the source attached to the other
arm of the the tee.
If the cable delays are mismatched the impedance seen at the tee will be
more complex.
I guess I could try using the multiple outputs on the back of the
Spectracom 8140, but that would introduce another variable -- the
outputs may not be exactly in phase. Also, isn't a pulse or square
wave better for doing TI measurements? What the heck. I may try that
just to see what happens :-)
You may as well just do it to see what happens, but don't expect much
better performance.
Either estimate it (use a circuit simulator) from the input impedance (R
and C) and the signal transition times or find a TDR (expensive) to
measure it.
Well, actually I can borrow a TDR. I'll have to read up on how to
measure reflection coefficient with it. The only time I ever used a
TDR was many years ago, in the Air Force.
Joe Gray
KA5ZEC
It is instructive to vary the TDR transition time and watch the
amplitude of any reflections reduce as the risetime increases.
Bruce
Bruce,
I guess I could try using the multiple outputs on the back of the
Spectracom 8140, but that would introduce another variable -- the
outputs may not be exactly in phase. Also, isn't a pulse or square
wave better for doing TI measurements? What the heck. I may try that
just to see what happens :-)
You may as well just do it to see what happens, but don't expect much
better performance.
Oops. It has been a long time since I read the manual for the
Spectracom 8140, or used it. I forgot that the rear outputs are fixed
at 10MHz and only the single front output changes with the push
buttons. You re-learn something every day :-)
I guess I'll have to find the correct resistors for a splitter. Stay tuned.
Joe Gray
KA5ZEC
Joseph Gray wrote:
Another newbie question (sorry). I'm using an HP 5315A Universal
Counter and doing some basic experiments to teach myself.
Here is the setup: 100KHz square wave as the "A" input to the 5315A.
Using a BNC-tee to connect a length of RG58A/U from "A" to the "B"
input and another tee with a 50 Ohm terminator. Using the TI Avg mode
and the gate time maxed at 13.4S, I read a TI of 22.5nS.
Assuming I got the numbers correct, light takes 3.3356nS to travel 1M
in a vacuum. 22.5nS / 3.3356nS/M = 6.745M. If I divide the measured
length of the coax (3.825M) by 6.745M, I get a velocity factor of
56.7%. This is quite a bit lower than the nominal 66% I expected.
So, is the difference due to:
The BNC-tee connectors?
My use of a 100KHz square wave instead of a PPS?
The resolution of the 5315A (100nS single-shot, 1nS averaging)?
Some other failing of the 5315A? It was just out of cal in Feb 2009.
Something else entirely?
All of the above?
The HP 5315A is using the bare MRC chip without external interpolators,
so that's why you only get the 100 ns singel-shot resolution, i.e. the
bare 10 MHz coarse counters. The MRC chip is used in 5314, 5315, 5334
and 5335 counters to name a few. I gave away my 5315 counter to a friend
who had no counter, but I still have a 5314 counter (and 5334 and 5335)
around.
With averaging and the 100 kHz and 10 MHz not being locked to each
other, the asynchronous clock beating will cause the averaging to have a
meaningfull effect.
The reason I'm using a 100KHz square wave is that I haven't yet modded
either of the Z3801A's to bring out a PPS signal where I could use it.
That is on my short list. The 100KHz from a Spectracom 8140 was the
closest I can get right now. Is this type of measurement useless with
anything other than a PPS signal?
No, rather the opposite. There is an overbeleif in using PPS signals.
PPS is good for one of the signals, but 100 kHz is good for the other.
An interesting observation - when I remove the 50 Ohm terminator, I
get a much shorter TI. I assume this is due to reflections back down
the coax.
No loading means the current can charge the capacitance quicker.
Cheers,
Magnus