Can I use a 50ohm BNC cable as a stand-in for a 1X passive oscilloscope probe in a semi-permanent test fixture? Are there any traps in terms of termination or impedance I need to look out for? This is for a low-speed application, <20MHz, for measuring PSU pk-pk ripple.
I was asking a similar question recently, you should be ok, what you need to check is the capacitance of the cable as the frequency goes up the reactance goes down and you can overload your device under test but being a power supply you will probably be ok.
putting a resistor in series with the "tip" helps, it depends on how accurate you want to be, you will end up with potential distortion.
https://www.eevblog.com/forum/testgear/rigol-dg4162-amplitude-inaccuracies/
Shows how a 50 ohm coax is not a 1x scope probe.
I remember watching the episode Dave did in response that thread, I'll rewatch it, but I think at low frequencies it shouldn't make a difference, right? Plus, the cable capacitance should be swamped in my application, because our testing spec requires 10uF and 0.1uF caps at the point of test.
you will probably be ok, but if you go measuring a high frequency from a high impedence source like that expect distortion and inaccuracies
Can I use a 50ohm BNC cable as a stand-in for a 1X passive oscilloscope probe in a semi-permanent test fixture? Are there any traps in terms of termination or impedance I need to look out for? This is for a low-speed application, <20MHz, for measuring PSU pk-pk ripple.
You can probably get away with it provided the source impedance is less than the cable capacitance's impedance at the frequency of interest.
Try it and see, verifying your measurements by other means!
I remember watching the episode Dave did in response that thread, I'll rewatch it, but I think at low frequencies it shouldn't make a difference, right? Plus, the cable capacitance should be swamped in my application, because our testing spec requires 10uF and 0.1uF caps at the point of test.
At a low enough frequency, it should be OK. But what's low enough? You mentioned 20 MHz. That's a wavelength of 15 meters. A quarter wavelength is 3.75 meters. A coax with improper termination acts like an impedance transformer, and a quarter wavelength of coax will make an open circuit at one end look like a short circuit at the other end. If your cable is a significant fraction of that 3.75 meter length, the signal you're probing will see a very different (read: much lower) impedance than the scope's high impedance input. Is your cable going to be longer than a few cm? Will it be a couple of meters long? Then this is something you'll want to consider if you care about accurate measurements and/or if you care about the load you're putting on the circuit under test.
Thanks for the advice! The cable will be about 1m long, and our main frequencies of interest are DC to ~10MHz, with most of the action happening in the 20KHz-1MHz range. We're using a Tek TDS 3014C with the 20MHz bandwidth limit on, or else the actual PSU ripple gets swamped by high frequency crap.
Testing power supply ripple? Sure. Terminate the cable, and don't forget ample common mode rejection (lots of ferrite beads?). The source impedance (with the traditional 10uF ceramic at the output, or whatever) should be low enough not to mind.
"with most of the action happening in the 20kHz-1MHz range"
Yeah, sure it is...
I've seen more than enough power supplies on "extended bandwidth" to know what that means...
Tim
you will probably be ok, but if you go measuring a high frequency from a high impedence source like that expect distortion and inaccuracies
Its not just high impedance, even low impedance sources can have distortions appear when using a BNC + cliplead adapter. Just saw this at work the other day. For some reason, my coworkers like to use coax + clip lead adatpers on scopes (fair enough I guess, they are laying around for on power supplies etc.. ) But with ~ 1m of coax and a 5 MHz RC filtered signal, a bunch of ripple on the rising edge showed up. Why? The source impedance from the high speed buffer was too low (way below 50 ohms) so the reflections off the 1M input impedance of the scope hit the low impedance source, and bounce back up to the scope again. But the same behavior isnt observed with a signal generator. The sig gen has a 50 ohm series termination, so any signals that bounce back get absorbed by that. With a proper 10x probe, or with a 50 ohm tee'd on the original (crap 1074Z's dont have internal 50 ohm?
) Everything looked fine.
50 ohm series resistor for the probe, 50 ohm termination, if you can set your scope up manually to have a 2x probe, then the readings will be correct too. Guessing a power supply wont complain about a 100 ohm load on it
I remember watching the episode Dave did in response that thread, I'll rewatch it, but I think at low frequencies it shouldn't make a difference, right? Plus, the cable capacitance should be swamped in my application, because our testing spec requires 10uF and 0.1uF caps at the point of test.
When doing power supply testing like this, I have gotten the best results by either using a coaxial connection to a x1 or x10 passive probe or doing what you are suggesting but using a terminated coaxial cable with either a 50 ohm (x2 attenuation) or 450 ohm (x10 attenuation) series resistor. The problem with the coaxial cable and 50 ohm termination is that AC coupling becomes difficult if you want to keep the low frequency response and you have to be careful not to apply large DC voltages to the 50 ohm oscilloscope termination.
Using just the coaxial cable connected directly to the oscilloscope without termination has always resulted in too much pulse distortion and peaking for me. x1 passive probes solve this by using a lossy center conductor to provide critical dampening so they have a clean frequency and phase response. The tip to BNC resistance on a x1 probe is in the hundreds of ohms. The ones I have handy at the moment measure 295 ohms.
One neat little trick that I like to use especially around switching power supplies because of leaking magnetic flux is to solder a short RG-316 coaxial pigtail to the test point and then use a coaxial BNC to probe tip adapter to attach a standard x1 or x10 passive probe. This adds to the input capacitance of the probe but not by a lot and the short length of the pigtail avoids serious reflections.
@David, is this about what you described?
I figured I should allow about 5 ohms for connectors and cable. The coax we have measures about 0.5-0.7 DC ohms per meter, and a couple ohms for the BNCs.
Why not just get a cheapy probe and cut it up and bodge it to the connector you want?
Why not just get a cheapy probe and cut it up and bodge it to the connector you want?
Because modifying purchased equipment (even probes) scares the crap out of management, while building new equipment is a much easier sell.
Why not just get a cheapy probe and cut it up and bodge it to the connector you want?
This is what I am going to try. I've got 4 of the cheap eBay probes coming and I'm going to cut one off and see how well I can make it work with 2" pig tail leads hanging off of it.
@David, is this about what you described?
Those capacitors are going to low pass filter in the input which is probably not what you want. They should be in series with the input resistance.
I figured I should allow about 5 ohms for connectors and cable. The coax we have measures about 0.5-0.7 DC ohms per meter, and a couple ohms for the BNCs.
With 450 ohms of input resistance, you could have +/- 5 ohms while staying within 1%. The 50 ohm termination resistance might be that far off anyway.
I know the caps cause a low pass filter, but they're required to be there parallel to the measured signal by the Intel testing specs. I'm not sure why they were there in the first place though. They cause slight filtering at the test point; a 10uF medium ESR electrolytic, and a 0.1uF ceramic. The ceramic acts super-low-ESR for high frequency noise, the 10uF generally takes the edge off some of the lower frequency stuff. Maybe the Intel engineers figured they'd use them to remove EMI picked up by the cables? But that seems like a sloppy solution.
Hm, how much difference would there be between putting the capacitors in front of the 450 ohm series resistor or behind it?
Hm, how much difference would there be between putting the capacitors in front of the 450 ohm series resistor or behind it?
A huge difference and that is where they are required to be by the specification. What the thing plugs into will have capacitance like that (and more) so they are there to make the environment the power supply is tested in more realistic.
Hm, how much difference would there be between putting the capacitors in front of the 450 ohm series resistor or behind it?
A huge difference and that is where they are required to be by the specification. What the thing plugs into will have capacitance like that (and more) so they are there to make the environment the power supply is tested in more realistic.
The decoupling and bulk capacitance is on the other side of the input resistors across the load. As shown, the high frequency cutoff is 320 Hz. When the capacitors are on the load side, then they are driven by the much lower output impedance produced by the regulator and load raising the high frequency cutoff considerably.
If you are only interested in line and load regulation, then the schematic shows a good way to measure it in a noisy environment without a x1 or x10 probe but if you want to see transient response and noise, then the capacitors need to be moved to the other side of the input resistor or placed in series with it for AC coupling.
I looked through a couple of Intel documents on testing their VRM designs but did not find anything directly applicable other than their requirements for bandwidth to distinguish high frequency noise from load response.
Why not just get a cheapy probe and cut it up and bodge it to the connector you want?
This is what I am going to try. I've got 4 of the cheap eBay probes coming and I'm going to cut one off and see how well I can make it work with 2" pig tail leads hanging off of it.
What? Have you seen the center conductor of scope probes?
Why not just get a cheapy probe and cut it up and bodge it to the connector you want?
This is what I am going to try. I've got 4 of the cheap eBay probes coming and I'm going to cut one off and see how well I can make it work with 2" pig tail leads hanging off of it.
What? Have you seen the center conductor of scope probes?
Besides being thin, the center conductor is some nickel alloy making it difficult to solder to. You can either crimp it or plate the end with something like copper.