FeelTech FY6600 60MHz 2-Ch VCO Function Arbitrary Waveform Signal Generator

[rhb]

A 50 ohm source connected to a 50 ohm cable connected to a 1 meg ohm input is going to have a reflection coefficient of 0.9999.  With most of the output reflected back into  the source, it will be severely distorted.

A 50 ohm terminator on a tee will work OK up into the middle HF region, but as you go higher the reflection from the stub going into the scope will become an issue.  To avoid that, buy some 50 ohm thru terminations.  The Chinese BNC ones I bought are actually quite decent and good to over 1 GHz.

If my 20 GHz SD-24 sampling head arrives today and works properly I'll add some scope traces to this post later.  I'll be setting up to test SMA-F to N-F adapters, so little effort to substitute a BNC tee & terminator and a thru.

[maxwell3e10]

All of this transmission-line considerations of cable impedance and load impedance matching apply only if the cable is long enough to be a transmission line, which means its length is comparable to the wavelength of the wave. With light speed on the order of 1 ft/nsec one doesn't need to worry about impedance matching for a 3-ft cable and 100 nsec pulses.

[radiolistener]

Here is difference with different cable length, source is 10 dBm:

Blue - through 0.15 m cable
Yellow - through 1 m cable




The same cables, the same signal, but with 50 Ohm input:



Here is direct connection through BNC-BNC adapter (cable length 0.01 m):

[maxwell3e10]

Thanks. Try the same for a sine with f=10 MHz, the difference will be smaller. Also would be interesting to look at square pulses. There will be some ringing on the transitions, but the flat regions will not be affected.

[radiolistener]

Thanks. Try the same for a sine with f=10 MHz, the difference will be smaller. Also would be interesting to look at square pulses. There will be some ringing on the transitions, but the flat regions will not be affected.

Here we are, the same cables, square wave, rise time 1 ns.

Blue - 0.15 meters
Yellow - 1 meter



the same with 50 Ohm input:



And through BNC-BNC adapter (0.01 m):



As you can see, the difference between 0.15 meters and 1 meter cable is 7.96 ns delay. It is pretty clear visible.

Also, as you can see, even with 50 Ohm input there is some small impedance mismatch. This is because this square wave has too wide bandwidth (more than 1 GHz or something like that) and my oscilloscope input has some impedance mismatch at about 1 GHz and above... (I'm using oscilloscope with 100 MHz bandwidth).

Square wave is too hard for my scope 

[pantelei4]

Also, as you can see, even with 50 Ohm input there is some small impedance mismatch.

This is not related to the number and level of steps. The level of quantization remains the same, and if it can be seen, this is enough.

[radiolistener]

This is not related to the number and level of steps. The level of quantization remains the same, and if it can be seen, this is enough.

DAC works at 250 MHz, so there is about 1-2 ns rise time transition between two levels. Are you sure that you can measure voltage difference on this transition with no proper termination at the end of cable?

[pantelei4]

DAC works at 250 MHz, so there is about 1-2 ns rise time transition between two levels. Are you sure that you can measure voltage difference on this transition with no proper termination at the end of cable?

In order to see, a low frequency is selected, for example 1 kHz or even lower.
The quantization level is a constant voltage signal, the RF bandwidth does not matter.

[radiolistener]

pantelei4, but these quantization levels are changed with rise time about 1-2 ns. It doesn't depends what frequency you selected, because DAC sample rate is fixed at 250 MHz. With any output frequency, transition between two quantization levels will occurs with fixed interval 4 ns and with fast rise time.

With low frequency you will get longer interval between changes. But how you know if this interval is one step 100 ns or 4 steps 25 ns? Since your cable doesn't have proper termination, short intervals may be overlapped with ringing in the cable and you will not be able to recognize short steps.

[pantelei4]

Since your cable doesn't have proper termination, short intervals may be overlapped with ringing in the cable and you will not be able to recognize short steps.

The steps will be long.

[radiolistener]

pantelei4, just reviewed your screenshot, ok. I can agree, it looks like this is not the case when short pulse is hidden due to cable ringing. But who knows... Measurement with impedance match is more reliable 

So, FY6600 has 12 bit sine wave?

[pantelei4]

So, FY6600 has 12 bit sine wave?

Yes.

[rhb]

All of this transmission-line considerations of cable impedance and load impedance matching apply only if the cable is long enough to be a transmission line, which means its length is comparable to the wavelength of the wave. With light speed on the order of 1 ft/nsec one doesn't need to worry about impedance matching for a 3-ft cable and 100 nsec pulses.

The above is completely wrong.

Here is a 36 ps rise time, 10 MHz square wave displayed on a 200 MHz Instek MSO-2204EA with a 50 ohm source feeding a 1 M ohm input.  This is one of the excellent units sold by Leo Bodnar.  There is *no* cable, just a BNC male connector mounted on the board.

Aside from the appearance of the square wave I call your attention to the measured rise time shown in the lower left corner.  Nominally BW = 0.35 divided by rise time.

 

The second  image is the same source and same scope settings except that the source is now connected using a 50 ohm thru termination 1.5" long and the gain adjusted.  Again, note the measured rise time in the lower left.

 


Just for good measure here is the same setup but using a 100 ps pulse instead.  The scope is in dot mode with infinite persistence.  The narrower impulse is with the terminator, the wider is without.  I'd like to point out that the pulse is only 1/10th the I GSa/s sample rate of the DSO for the benefit of those who claim that you can miss a short pulse if your sample rate is not fast enough.

[maxwell3e10]

rhb, in all your work in the oil business, have you ever worked out solutions to Maxwell's equations in a cylindrical waveguide with a center conductor?

What you are showing is not really relevant, you are looking at rise times of very fast pulses with a bandwidth way beyond that of the equipment used. So, the rise time and wiggles is just a function of the oscilloscope input stage ringing. It can depend on gain setting, quality of BNC connector, even how it is rotated. Plus, one shouldn't always trust automatic scope measurements, the difference in those rise times is not a factor of 2. And to compare pulse width, they should be made the same height.

[rhb]

I suggest you find a good primer on the use of oscilloscopes and read it.  In particular the section on input impedance and bandwidth.

[rhb]

rhb, in all your work in the oil business, have you ever worked out solutions to Maxwell's equations in a cylindrical waveguide with a center conductor?

Just for the record, the last time I solved that was 40 years ago when I went back to school to get an MS in geology after taking a BA in English lit.

The class average on the Physics II (EM) final was 45.  I had the highest score, 89.  I derived the solution to a complex capacitor problem from first principles during the exam.

That's the only grade I ever got I truly cared about.  And that was because some twit sitting next to me in class, after hearing I had a BA in English lit,  remarked I'd never pass because  "This is where they weed out the engineers."   

[maxwell3e10]

I suggest you find a good primer on the use of oscilloscopes and read it.  In particular the section on input impedance and bandwidth.

Good for you. But people who have PhD like to understand things at a deeper level than an instruction manual.

[rhb]

I suggest you find a good primer on the use of oscilloscopes and read it.  In particular the section on input impedance and bandwidth.

Good for you. But people who have PhD like to understand things at a deeper level than an instruction manual.

ROFL!!!!  I was merely suggesting you learn to crawl before you try to walk.

I was paid to sit as judge and jury on Stanford professors and their students for several years.  I worked as a contract consultant and my client paid around $35K/yr for consortium membership.  I was sent to the annual meetings to make sure the work they did was worth the money.  There were others, but you probably have not heard of Mines or the others.

Clearly not a world you're familiar with.   I know you have a BS, but it's obviously not a Bachelor of Science degree, even allowing for the horrible degradation of academic standards.

Attached are photos taken of the same 36 ps rise time pulser on a Tek 485.  It needs a full cal run, but it's still very good, especially compared to the horrible step responses of DSOs such as the Keysight MSOX3104T I returned  or the R&S RTM3104 I had on demo. It just isn't up to my (or Tek's) rather stiff demands for a proper scope step response.  A full cal on one of these is a full day's work which is why it's not in full cal.  Adjusting the front end attenuators takes a couple of hours.

For the benefit of the youngsters, the 485 was *the* premier portable analog scope for many years and has a 350 MHz BW.  It also has both 50 and 1 M ohm input selectable by switch.  The sweep is set at 2 ns/div for both photos.

Rather than spoil the fun, I shall allow our "PhD" friend here to explain which is 50 ohm and which is 1 M ohm and why the 36 ps rise time square wave looks so different.  In this case, aside from gain adjustments, the only change is the input impedance setting.  My apologies to the old guys who realize I botched the gain settings relative to the 0% & 100% lines.

BTW  For the benefit of the more intelligent readers of this farce, Leo Bodnar supplies with each of his pulsers a plot from a CSA803A with either an SD-30 or SD-32 head.  The SD-30 is a 40 GHz,  8.75 ps rise time head.  The SD-32 is a 50 GHz, 7 ps rise time head.  I'm not sure which he used for my plot.  Leo only claims < 40 ps in his listings to allow for individual unit variations.  The laser driver he uses has a 21-23 ps rise time IIRC, so he also offers 3.5 mm and 2.48 mm versions with faster rise times, albeit at higher cost as the connectors are *much* more expensive.  A 1 MHz square wave version of his BNC pulser will give exceptional TDR capability for very low cost with whatever scope you have.  Leo very graciously will provide that instead of 10 MHz if you ask. I have 1 MHz & 10 MHz square wave units and a 100 ps impulse unit.  They are fantastic!

[BU508A]

I suggest you find a good primer on the use of oscilloscopes and read it.  In particular the section on input impedance and bandwidth.

Good for you. But people who have PhD like to understand things at a deeper level than an instruction manual.

Arrogance will not help you understanding why proper termination is necessary.
But maybe if you will have a look at this, then perhaps you will understand.
And if you want to discuss this topic with someone on your level, I can recommend Shariar from The Signal Path.
He is a RF expert and has a PhD.

First: introducing s-parameters, the basics:
https://www.edn.com/design/test-and-measurement/4437010/S-parameters-basics

A paper from CERN:
https://cds.cern.ch/record/1415639/files/p67.pdf

a more general article from Wikipedia:
https://en.wikipedia.org/wiki/Transmission_line

a nice usecase of not correctly terminated transmission lines, example by Rohde & Schwarz
https://www.rohde-schwarz.com/de/file/Distance-to-Fault_Measurements.pdf

And here is the link to The Signal Path if you want to discuss this topic with a PhD and not some mediocre EEs.
http://thesignalpath.com/blogs

And just a last remark:
Every, really every transmission line MUST have a proper termination, otherwise it will fail.
This is more true, the higher the frequencies are. High frequencies means: fast square signals.
All the high frequencies are in the edges covered. The related topic is "Fourier-Analysis", explained here:
https://en.wikipedia.org/wiki/Fourier_analysis

HTH,

BU508A

[maxwell3e10]

No one argues that a transmission line needs termination. The question is what constitutes a transmission line. The general rule is that the length of the line needs to be comparable to the wavelength of the signal transmitted. That is why impedance matching only becomes an issue at higher frequencies (shorter wavelength).

I perfectly stand by the original statement:

All of this transmission-line considerations of cable impedance and load impedance matching apply only if the cable is long enough to be a transmission line, which means its length is comparable to the wavelength of the wave. With light speed on the order of 1 ft/nsec one doesn't need to worry about impedance matching for a 3-ft cable and 100 nsec pulses.

It has already been shown by radiolistener that changing impedance matching only affects the ringing on pulse transitions (which necessarily have much higher frequencies or shorter wavelengths), it does not affect the DC value of 100 nsec -long steps (apart from trivial factor of 2 attenuation). So if one only cares about those 100 nsec flat steps, there is no need for termination. 

Now as for rhb tests, it doesn't even involve a cable, just a direct signal connection as far as I understand. As was discussed on another thread (https://www.eevblog.com/forum/testgear/oscilloscope-input-noise-comparison/), oscilloscopes often have different input circuits for 1 MΩ and 50 Ω signal paths, so the rise time difference is not surprising at all. I don't know about the details of Tek 485 input, only that the bandwidth is quoted as 350 MHz (50 Ω) / 250 MHz (1 MΩ) at http://w140.com/tekwiki/wiki/485.

I am sorry for tweaking rhb, he just likes to brag about his credentials so much that I couldn't resist.

[tinhead]

I couldn't resist.

https://en.wikipedia.org/wiki/Idiocracy

[BU508A]

No one argues that a transmission line needs termination. The question is what constitutes a transmission line. The general rule is that the length of the line needs to be comparable to the wavelength of the signal transmitted. That is why impedance matching only becomes an issue at higher frequencies (shorter wavelength).

Nope. The general rule is, what the source "sees" when it is going to leave a signal. If you put a wire into thin air,
then it will see the impedance of free space, around 377Ohm. If the Impedance of the source is different from that,
which is mostly the case, then it will come to distortions caused by reflections. If you put a piece of cable with the
impedance of 50 Ohm to the source, no matter how long it is, then the source will "see" these 50 Ohms at first.
If the impedance of the source differs from these 50 Ohms, you'll have reflections, distortions etc.
That is why the impedance of the source should match the impedance of the transmission line, which is seen
by the source at first.
Assuming the transmission line is homogen all over and it hits the end, then the same will happen as at
the beginning.
Therefore: no matter, how your transmission line is looking, if you want to avoid reflection and distortion, then
you have to follow the rule:
impedance of the source = impedance of the transmission line = impedance of the target.
This is independet of the length.

For this reason we have antennas. They adapt the impedance of the source to the impedance of the target.
In ideal cases you have a SWR of 1:1 which means that nearly all of the energy coming from the source will
reach the target. No reflection. No distortion.

[maxwell3e10]

This should really go to a different thread. But two questions come to mind. a) What is the impedance of your audio headphones "transmission line"? and b) If you design an RF circuit (below 1 GHz), does every node on it sees 50 Ohm?
   

[radiolistener]

It has already been shown by radiolistener that changing impedance matching only affects the ringing on pulse transitions

my example is not the worst case for impedance mismatch. It's just random mismatch. So it cannot be used as example that the difference is not so significant.

For clarification here is impedance measurement of my oscilloscope 1M input:
- at 10 MHz = 57 - j884 Ohms
- at 100 MHz = 43.9 - j96.5 Ohms

As you can see, my oscilloscope input impedance at 10 MHz is not 1 MOhm, but 57-j884 Ohm. So, you're needs to take this into account.

And don't expect that your oscilloscope will give you the same impedance mismatch, because your oscilloscope may have different input impedance at 1 MOhm mode.

[BU508A]

The rule is the same, but in this frequency range the distortion is very low. The main reason for this is,
that the source has usually a very low impedance and the target in respect a high impedance. If some energy is
reflected, it will be absorbed by the source.
Most of the audio equipement which uses unbalanced connections have a source and target impedance of 47kOhm.
But as I said, in this frequency range the "wrong" impedance of the cable doesn't really matter.

Regarding the RF circuit: no one said, that it must have an impedance of 50 Ohm. The only condition is:
source Z = transmission line Z = target Z. If the signal leaves the circuit one have to make sure,
that the impedances are adapted, for example by using a Pi-network.