Help w/ Signal Generator (voltage varies w/ freq adjustment)

[GenSoft3d]

I recently acquired an older Wavetek 2500 signal generator (.4 - 1100 MHz) and HP 54100A 1 GHz Digitizing scope for a project and am having a difficult time in getting an accurate voltage reading using anything other than a direct-connect BNC cable, and even that seems to have it's issues.

With the BNC-BNC cable (connected directly between the generator output and the scope's 50 Ohm input) and the generator set to the max output (13dB / 1V), the scope is showing a drop in voltage of roughly 50% over the full range (.4 - 1100 MHz). Is this drop in voltage abnormal or is it to be expected? Here's a list of measurements showing the drop:

.4 MHz = 945mV RMS
10 MHz = 892mV
100 MHz = 757mV
500 MHz = 681mV
1100 MHz = 450mV

Now, when I attempt the same test using two separate BNC-to-clip cables between the generator and scope (and connecting to grnd) I lose the consistent voltage drop over the full range of freq and instead am seeing random spikes in the voltage as I sweep the entire range. The spikes are repeatable, meaning they consistently appear at specific frequencies.

Not being familiar with the equipment and having purchased it used, I was hoping that someone might be knowledgeable enough to know if the hardware is faulty or if I'm doing something wrong in the process. I would guess that the cables are the culprits but what about the 50% voltage drop with the direct BNC-BNC cable? Or is that to be expected as the freq is increased?

Thanks for any help!

 

[sophiecentaur]

It always helps to have a range of other bits of equipment so that you can compare performances and eliminate the dodgy one so you have my sympathy.
I see from the spec that the flatness in level should be within +/- 1.2dB for a new instrument and this is more like 6dB. Perhaps your output stage is a bit tired or an internal connector needs tightening up. That BNC cable could be the culprit; do you have another one to compare (they don't cost too much). The peakiness of the response with the two 'croc' leads could be expected from the discontinuity at the join; it's definitely not 50Ohms over the joining section. Do the peaks vary as you waggle it about?

How does the scope display look using its own calibration waveform?

 

[GenSoft3d]

Thanks a ton for the help... I really appreciate that!

I connected the BNC cable to the scopes output and it's showing a clean square wave w/ freq = 500.005 kHz, Vrms = 289.3mV, Vp-p = 414.0 mV. I'm not noticing any significant signal variation when I wiggle the cables as you suggested, but there is a slight change. I don't have another BNC cable on-hand but I will try to pick one up within a day or so and see if that helps at all.

I did another test with signal generator connected directly to the scope and here are the results from that:

MHz____db____mV

..1...+13...918
..1...-3...153
550...+13...600
550...-3...104
1000...+13...427
1000...-3...91

I have a feeling that the generator is causing the drop - perhaps it's the output stage as you mentioned. I think I might be able to get by with it for my specific project if I take the known drop into account. However, I still need to figure out the spikey thing that's occurring with the other cables so that I can tell the difference between the unwanted spikes and the actual resonance peaks that I'm trying to measure.

Would you mind offering some more details on the discontinuity that you mentioned where the cables are joined? Is there a way that I can rectify it? BTW, I'm using min-grabbers rather than the crocs, though I doubt that matters any.

 

[sophiecentaur]

A section of line that isn't 50 Ohms will have a standing wave in it and the signal passed will not be constant level with frequency. The impedance of the section of twin and not coax can have an impedance of more like 200ohms and can be a significant fraction of a wavelength at the high end of your range. UHF is fussy about this sort of thing and good lines are important up there. Many BNC leads are only intended for video frequencies, too. Wait to see how your new lead behaves.

 

[GenSoft3d]

Thanks for the detailed response... that helps a ton! The generator has a type N connector and I decided to go with a small type N-to-BNC adapter and use standard BNC cables, but maybe that was a bad idea? I wasn't able to locate any type N cables w/ clips and so I just opted for that route but it sounds like that kind of cable doesn't exist for a reason based on what you're saying about the sensitivity of the UHF range. Heck, I couldn't even find a type N to BNC cable that I could connect between the generator and scope so there must be a reason for that as well I assume?

So, I'm now wondering how I'll be able to test my circuit if the two line approach introduces too much impedance into the system? Would it be a reasonable approach to build a test holder of sorts that has BNC and/or type N connectors directly embedded into it and then connect the cables directly to it - bypassing the clips and probe entirely? I'm trying to wrap my head around how to go about testing my circuit under these conditions where the two-line approach seems to break the continuity of the signal. Any ideas?

I'm hoping to pick up a BNC cable today so I will report back when I'm able to test it out.

Thanks again for all your help!

 

[sophiecentaur]

It all depends on your particular application (oh boy, how many times have I written those words?).
Your test equipment need not be 'very much' better than the thing you're testing but when you build stuff to work at hundreds of MHz you need to be careful or the measurements can be seriously compromised. For UHF transistor circuits on PCB, it is normal to use 50Ohm microstrip on the board and appropriate connectors on the board so as to avoid reflections at the transition with the cables in and out. Type N was designed for use at higher frequencies than BNC, although many bits of equipment function well enough with the cheaper connector (just look at the naff TV antenna connectors, which work well enough to bring us all off air TV signals).
If your new lead doesn't help then you could try to get access to another scope or generator (known to be working correctly) to establish which is the faulty unit.
Presumably you don't want to be delving into the sig gen yourself and checking internal DC supply volts, connections and for signs of 'roasting' in there. The manual appears to be available on line for not much money or free if you wanted to have a look. But that sort of thing only suits 'certain types' of person!
BTW, what device do you hope to be testing / measuring?

 

[GenSoft3d]

Well, I called it a "circuit" but there's no PCB or state of the art circuit involved. It basically consists of a custom air-core copper coil that's coupled to a xtal oscillator. I purchased the high frequency generator and scope specifically for tuning the small coil in the circuit after I came across these videos that show some methods for measuring the SRF of a coil:

coil SRF method 2

coil SRF method 3

The coils in these videos are tuned to a much lower frequency than mine so I'm not sure that I can use the same methods for my application given that my coil has a much higher SRF of 1 GHz (based on my calculations).

 

[sophiecentaur]

I found http://www.rac.ca/tca/RF_Coil_Design.html, recently, for calculating the performance of coils. Could be useful for you if you don't already have the equivalent.
It suggests that the SRF of your coil would be more in the region of 100 to 200Mhz, rather than 1GHz. Slightly easier frequencies for you to deal with.

If you are just looking for resonances then actual levels are not too important because a good sharp peak or dip will show itself over any slight slope from the generator. The resonance of the coil will be a parallel resonance - going high impedance at the SRF. Operating at frequencies more than 100MHz, the other stray capacitances can be relevant compared with the self capacitance of around 1pF and the way the coil is mounted can have more effect on the resonant frequency than the actual coil construction. (RF is magic and a nightmare)
Those videos show measurements at much more 'reasonable' frequencies - not surprisingly.

Why have you chosen to use such a low inductance coil? You could establish the principles at much lower and manageable frequencies to start with.

You should be able to find the resonant frequency of the crystal easily (flying leads are fine for that) in the way described in the video. If you just connect the crystal across the coil and look for a resonance then all that particular coil will do is to reduce the volts across the crystal at resonance because of its low shunt reactance. The resonance you will see in the crystal will be a series resonance which will go low reactance at resonance.

If you don't like the sums, you might try a simulation to see what effect to expect. An interesting thing, with the real components would be to see what happens when the coil is made to resonate with an appropriate parallel (tuning) capacitor at the same frequency as the crystal - or at an overtone of the crystal. Your two resonant circuits would affect each other. The coil should exhibit a broad peak of resonance and, at the crystal series resonance, there would be a sharp dip near that broad peak - plus some interesting other response at the parallel resonance of the crystal, perhaps.

 

[Bob S]

Are you accounting for the skin effect attenuation of RG-58? How long is your coax? See http://www.w4rp.com/ref/coax.html

 

[GenSoft3d]

Thanks for the link Bob... I didn't consider that! The cable is almost 6 ft in length and at 1 GHz it looks like bout 1.3 attenuation. That's definitely accounting for a good sum of what I'm seeing. I'll try to pick up a much shorter cable today and see how much it improves.

SC, thanks for the link to the calculator... I haven't come across that one in my searching. However I'm still getting something around 1 GHz for some reason. Can't see how I'm off though. I'm using the following dimensions:

coil diam: 9.32 mm
coil length: 11.399 mm
wire diam: 1.0236 mm
coil pitch: 2.28 mm
N - turns: 5

Using these figures I'm getting 1141.39 MHz. Have you seen this calc that treats the coil as a helix waveguide?

http://hamwaves.com/antennas/inductance.html

It claims to be more accurate and it also includes David Knight's empirical formula and some others for higher accuracy. I'd like to know what you think of it if you get a chance to check it out. Using the same figures above I am getting an SRF of 742.85 MHz which is supposed to be λ/4 (parallel) self-resonant frequency of n=0 sheath helix mode. It states that it's the first self-resonant frequency of the coil which is also the frequency for which the coil appears as a quarter-wave resonator.

To complicate matters further, what I have is actually an inverse conical coil so the standard calcs for helical coils aren't going to suffice. I've not been satisfied with the formulas I've found so far for conical coils. Deriving the inductance based on the two helical and flat spiral formulas seems too vague to me. Do you happen to know of a good conical formula by chance?

So that understand what you are saying; did you mean that the coil would not effect the resonant freq of the xtal but would only reduce the volts across it? Or am I misunderstanding? Also, are you referring to the shunt reactance of the coil or the xtal?

I will most definitely start with a much larger coil I think and then work my way down so that I can get an idea of what I'm looking at, etc... Do you think that the methods shown in the video would be applicable for my coil SRF testing (swapping caps and using the bulbs as indicators)?

Thanks again for all the help!

 

[sophiecentaur]

That link on coil inductance is interesting. There is no complete solution to any of these calculations and they all use different approximations. What operating frequency were you putting into the calc? I can agree that it gives around 740MHz SRF. I wonder whether the coil is a bit on the short side for that treatment - the actual length of the wire used is only around 150mm, which is only a fraction of the wavelength (0.4) at 750MHz. The applications he shows in the photo (a TWT and a helical antenna) use a much longer coil.

But if you try with a higher inductance you may find life easier in general.
RG58 loss should be only 0.5dB/m at 500MHz, I think.

 

[GenSoft3d]

Giving this more thought, I think I might be going about this the wrong way. From what I understand the coil will lose it's inductance qualities at SRF and this may be something I want to avoid considering that I want it to serve as an inductor in the xtal oscillating circuit. In RF design it's desirable to have the operating freq of the coil be well below it's SRF to avoid this issue and so I may need to re-think my approach entirely and perhaps I won't need to bother with the higher-freq measurements required for SRF after all.

I think I got off-track by the idea that perhaps the xtal freq would settle on a higher overtone at resonance when coupled with a higher freq coil like I'm working with. However, I'm now thinking that it might be more realistic to say that the main contributing factor in determining the resonant freq of the circuit would be the capacitance value of the xtal coupled with the inductance value of the coil. If so then the resulting resonance may (should) be much lower than the coil's SRF I'm thinking.

What would you expect to see happen if the xtal is operating at it's fundamental freq (say 1 MHz) and then a coil of such low inductance as I mentioned (+/-.2 uH) were attached to it? Would it be reasonable to expect only a slight shift in the resonant freq of the xtal, or would it be a major shift resulting in a resonant freq that was far from the xtal's fundamental? Care to speculate?

At this point, I think I will pick up a function generator that covers the lower range and start with finding the xtal's fundamental freq and then go from there. Once I establish that then I will try adding the coil to the circuit and see how it effects it. That might be the best approach I think.