Post by thread:[EE] Measuring characteristic impedance of coax ca

> I have a customer who wants to measure the characteristic impedance of a > length of coax. Actually, we know what it is supposed to be according to > the > manufacturer, but they want to verify it. > > Test equipment available: high accuracy multimeters (5-1/2 digit?), cheap > analog oscilloscope, relatively low frequency signal generator (sine, > square, triangle) (top end is maybe 1Mhz), frequency counter, bench > supply. > > Willing to buy high-precision R's or C's if needed. > > Accuracy required: hopefully a few percent, 10% is some help. > > I am really a digital guy, so this is a bit out of what I do. > > Any ideas? How about this? Drive the coax with a square wave through a resistor that is the same as the suspected characteristic impedance (the driving resistance is not critical here). Watch the coax input with a scope. Vary the resistance at the far end of the coax until the scope shows a square wave. With anything other than the characteristic impedance terminating the coax, you should see the reflection. If the driving impedance is the same as the characteristic impedance, you should see only one reflection. In that case, the square wave will either have a step up or down before settling at the proper voltage depending on whether the terminating impedance is low or high. Good luck! Harold -- FCC Rules Updated Daily at http://www.hallikainen.com - Advertising opportunities available! Not sent from an iPhone

> How about this? > > Drive the coax with a square wave through a resistor that is the same as the > suspected characteristic impedance (the driving resistance is not critical here). > Watch the coax input with a scope. Vary the resistance at the far end of the coax > until the scope shows a square wave. With anything other than the characteristic > impedance terminating the coax, you should see the reflection. If the driving > impedance is the same as the characteristic impedance, you should see only one > reflection. In that case, the square wave will either have a step up or down before > settling at the proper voltage depending on whether the terminating impedance is low > or high. Flyback of a 'scope sweep waveform can often be a good source of fast edge. You will also need to make sure the total source impedance presented to the cable is the nominal cable impedance. Assuming that the cable already has connectors on it then standard BNC type termination (either end or feedthough) will do for the far end. I take it they are just wanting to verify that it is within manufacturers specification (rather than exact to the last 0.1 ohm etc). -- Scanned by iCritical.

Harold Hallikainen wrote: {Quote hidden}>> I have a customer who wants to measure the characteristic impedance of a >> length of coax. Actually, we know what it is supposed to be according to >> the >> manufacturer, but they want to verify it. >> >> Test equipment available: high accuracy multimeters (5-1/2 digit?), cheap >> analog oscilloscope, relatively low frequency signal generator (sine, >> square, triangle) (top end is maybe 1Mhz), frequency counter, bench >> supply. >> >> Willing to buy high-precision R's or C's if needed. >> >> Accuracy required: hopefully a few percent, 10% is some help. >> >> I am really a digital guy, so this is a bit out of what I do. >> >> Any ideas? >> > > How about this? > > Drive the coax with a square wave through a resistor that is the same as > the suspected characteristic impedance (the driving resistance is not > critical here). Watch the coax input with a scope. Vary the resistance at > the far end of the coax until the scope shows a square wave. With anything > other than the characteristic impedance terminating the coax, you should > see the reflection. If the driving impedance is the same as the > characteristic impedance, you should see only one reflection. In that > case, the square wave will either have a step up or down before settling > at the proper voltage depending on whether the terminating impedance is > low or high. > > Good luck! > > Harold > > > > Also, the longer the cable being tested, the larger the reflection for a given rise time / mismatch. Kerry

To get the sort of accuracy you're asking for, you really need a network analyser of some sort. The impedance will vary with frequency - especially from audio frequencies through to VHF etc. The main component of this is skin effect, where the inductance changes slightly as the centre conductor moves into its RF region. The outer conductor (screen) impedance is also likely to change somewhat. There is also an effect known as "structural return loss" which is due to manufacturing variations in the cable. Where these variations correspond to a 1/2 wavelength the cable impedance can change markedly as the cable starts to operate as a filter. Obviously manufacturers try to minimise this but it can be s real problem with the cheaper grades of coax. For an overall appraisal the TDR method is probably going to be the easiest to implement with the cable terminated in it's nominal impedance. But you may get different results when using it at RF. What sort of frequency rang is involved? - is it used for data (i.e a TDR method may be more appropriate anyway) or RF, in which case some sort of network analyser is really needed. If it's to be used over a narrow frequency band you may be able to get an idea with a bridge and a signal generator but it's not going to be easy. RP On 30 November 2011 05:22, Kerry Wentworth <.....kwentworthKILLspam@spam@skunkworksnh.com> wrote: {Quote hidden}> Harold Hallikainen wrote: >>> I have a customer who wants to measure the characteristic impedance of a >>> length of coax. Actually, we know what it is supposed to be according to >>> the >>> manufacturer, but they want to verify it. >>> >>> Test equipment available: high accuracy multimeters (5-1/2 digit?), cheap >>> analog oscilloscope, relatively low frequency signal generator (sine, >>> square, triangle) (top end is maybe 1Mhz), frequency counter, bench >>> supply. >>> >>> Willing to buy high-precision R's or C's if needed. >>> >>> Accuracy required: hopefully a few percent, 10% is some help. >>> >>> I am really a digital guy, so this is a bit out of what I do. >>> >>> Any ideas? >>> >> >> How about this? >> >> Drive the coax with a square wave through a resistor that is the same as >> the suspected characteristic impedance (the driving resistance is not >> critical here). Watch the coax input with a scope. Vary the resistance at >> the far end of the coax until the scope shows a square wave. With anything >> other than the characteristic impedance terminating the coax, you should >> see the reflection. If the driving impedance is the same as the >> characteristic impedance, you should see only one reflection. In that >> case, the square wave will either have a step up or down before settling >> at the proper voltage depending on whether the terminating impedance is >> low or high. >> >> Good luck! >> >> Harold >> >> >> >> > Also, the longer the cable being tested, the larger the reflection for a > given rise time / mismatch. > > Kerry > > >

I just tried my method of measuring L and C and indeed it does not work very well (at least not when measuring with audio frequencies). I have two 4.5 foot lengths of cable, one RG-58C/U (50 ohms) and the other RG-59B/U (75 ohms), both terminated at both ends with BNC connectors. For the 50 ohm cable, I measured C=145.2pF and L=1.05uH at 10kHz, using a bench top high quality digital LCR meter (Danbridge CT-20). The C indicated a high Q while the L had such a poor Q that the meter auto-detected it as a resistance with parasitic series inductance. The far end was shorted with a shorted female BNC connector for the L measurement. For the 75 ohm cable, I measured C=96.8pF and L=1.3uH. sqrt(L/C)= 85 ohms for the 50 ohm cable and 116 ohms for the 75 ohm cable. Correct direction (75 ohm is larger), but not even the correct ratio of impedances. The problem is the L measurement. The theoretical C per foot for RG-58C/U is 30.8 pF/ft, which would be 138.6pF for this 4.5 foot length. My measurement is only 5% off. For the RG-59B/U, the value should be 20.5pF/ft or 92.3pF for this cable. I was once again about 5% high. This is no doubt partially due to the BNC connectors and parasitics in the leads from the test jig to the BNC connector. I tried to obtain a value for the inductance external to the cable by connecting my BNC "short" directly to the leads going to the test jig and I got 0.6uH, which gives you an idea of the magnitude of the parasitics involved here. Subtracting 0.6uH from each of the two above L values and re-computing Zo for each gives 55 ohms for the 50 ohm cable and 85 ohms for the 75 ohm cable. Much closer but still very hand-wavy when I have only one (partially) significant digit in my L measurement after subtracting-out the stray inductance. I bet that I could do this much more accurately with my 1MHz HP4271A LCR meter, but I have not yet gotten around to making a test jig for it so I am not even sure if it works properly :( Sean On Tue, Nov 29, 2011 at 4:43 PM, Richard Prosser <rhprosserKILLspamgmail.com> wrote: {Quote hidden}> To get the sort of accuracy you're asking for, you really need a > network analyser of some sort. The impedance will vary with frequency > - especially from audio frequencies through to VHF etc. The main > component of this is skin effect, where the inductance changes > slightly as the centre conductor moves into its RF region. The outer > conductor (screen) impedance is also likely to change somewhat. There > is also an effect known as "structural return loss" > which is due to manufacturing variations in the cable. Where these > variations correspond to a 1/2 wavelength the cable impedance can > change markedly as the cable starts to operate as a filter. Obviously > manufacturers try to minimise this but it can be s real problem with > the cheaper grades of coax. > > For an overall appraisal the TDR method is probably going to be the > easiest to implement with the cable terminated in it's nominal > impedance. But you may get different results when using it at RF. > > What sort of frequency rang is involved? - is it used for data (i.e a > TDR method may be more appropriate anyway) or RF, in which case some > sort of network analyser is really needed. If it's to be used over a > narrow frequency band you may be able to get an idea with a bridge > and a signal generator but it's not going to be easy. > > RP > > > > On 30 November 2011 05:22, Kerry Wentworth <.....kwentworthKILLspam.....skunkworksnh.com> wrote: >> Harold Hallikainen wrote: >>>> I have a customer who wants to measure the characteristic impedance of a >>>> length of coax. Actually, we know what it is supposed to be according to >>>> the >>>> manufacturer, but they want to verify it. >>>> >>>> Test equipment available: high accuracy multimeters (5-1/2 digit?), cheap >>>> analog oscilloscope, relatively low frequency signal generator (sine, >>>> square, triangle) (top end is maybe 1Mhz), frequency counter, bench >>>> supply. >>>> >>>> Willing to buy high-precision R's or C's if needed. >>>> >>>> Accuracy required: hopefully a few percent, 10% is some help. >>>> >>>> I am really a digital guy, so this is a bit out of what I do. >>>> >>>> Any ideas? >>>> >>> >>> How about this? >>> >>> Drive the coax with a square wave through a resistor that is the same as >>> the suspected characteristic impedance (the driving resistance is not >>> critical here). Watch the coax input with a scope. Vary the resistance at >>> the far end of the coax until the scope shows a square wave. With anything >>> other than the characteristic impedance terminating the coax, you should >>> see the reflection. If the driving impedance is the same as the >>> characteristic impedance, you should see only one reflection. In that >>> case, the square wave will either have a step up or down before settling >>> at the proper voltage depending on whether the terminating impedance is >>> low or high. >>> >>> Good luck! >>> >>> Harold >>> >>> >>> >>> >> Also, the longer the cable being tested, the larger the reflection for a >> given rise time / mismatch. >> >> Kerry >> >> >> -

> How about this? > > Drive the coax with a square wave through a resistor that is the same as > the suspected characteristic impedance (the driving resistance is not > critical here). Watch the coax input with a scope. Vary the resistance at > the far end of the coax until the scope shows a square wave. With anything > other than the characteristic impedance terminating the coax, you should > see the reflection. If the driving impedance is the same as the > characteristic impedance, you should see only one reflection. In that > case, the square wave will either have a step up or down before settling > at the proper voltage depending on whether the terminating impedance is > low or high. One issue with this method is that to do it with cheap equipment you need a long peice of cable. IIRC wave propogation speed in coax is of the order of 2x10^8 meters per second. So if your cable is a meter long your signal generator and scope will need to resolve a reflection arround 10ns after the original transition. OTOH if you have 100 meters of cable to test then your signal generator and scope will only need to resolve a reflection arround 1us after the original transition

On 1 December 2011 15:30, peter green <plugwashspam_OUTp10link.net> wrote: > Justin Richards wrote: >> Is there an implied frequency when discussing characteristic impedance. >> When we need to analyze cables with respect to their RF >> characteristics we end up with a large table S parameters that have >> been recorded over a range of frequencies. >> >> So when a cable is quoted at 75 ohms, that must be for a given >> frequency. If so what is it. >> > An ideal coaxial cable (made up of perfect conductors and perfect > dielectrics in a perfect geometery) will have a constant characteristic > impedance across all frequencies. I'm not sure on the details but my > understanding is that if the characterstic impedance is changing > significantly with frequency you are probablly above the cable's usable > frequency range. > Justin, Cable characteristics are normally referred to at a standard frequency, depending on cable likely application. 200MHz, 400MHz and 1GHz being typical but impedance should be pretty stable above a few MHz. Mathematically the impedance is give by Z=sqrt((R+jWL)/(G+jWC)) Since G is normally close to zero in most cases this reduces to sqrt(L/C) at high frequencies - where WL is >> R. (W = 2 * PI * f) .. The inductance is not constant although it only changes slightly as skin effect moves the current to the outside of the centre conductor. The R term is also not constant with frequency, it increases due to skin effect (with sqrt(frequency)) and linearly with frequency as dielectric loss starts to dominate. There is also a loss component proportional to frequency involving the screen, but this is normally not significant. Insulators like PVC have a very frequency dependent dielectric constant also (This adds to the G term IIRC), which is why PVC is not used for RF cable insulation - although the effect is present with other materials as well. Richard P

> Justin Richards wrote: > > Is there an implied frequency when discussing characteristic impedance. > > When we need to analyze cables with respect to their RF > > characteristics we end up with a large table S parameters that have > > been recorded over a range of frequencies. > > > > So when a cable is quoted at 75 ohms, that must be for a given > > frequency. If so what is it. > > > An ideal coaxial cable (made up of perfect conductors and perfect dielectrics in a > perfect geometery) will have a constant characteristic impedance across all > frequencies. I'm not sure on the details but my understanding is that if the > characterstic impedance is changing significantly with frequency you are probablly > above the cable's usable frequency range. The other option is that the source and load are imperfectly matched, and large changes in effective impedance at the end of the cable can be seen due to the length being close to an odd integer multiple of a 1/4 wavelength at the frequency of measurement. -- Scanned by iCritical.

I am replying a bit late on this and couldn't read the other replies completely but I would suggest using a time-domain-reflectometer. This precisely fits your requirements, using a high speed capacitor with a PIC having CTMU, people have reported that they reach a resolution level of "less than 1ns". I am currently trying to achieve better, you might want to look into AN1375 and en542792.pdf. Good luck. On Tue, Nov 29, 2011 at 3:35 PM, Bob Ammerman <@spam@picramKILLspamroadrunner.com> wrote: {Quote hidden}> I have a customer who wants to measure the characteristic impedance of a > length of coax. Actually, we know what it is supposed to be according to the > manufacturer, but they want to verify it. > > Test equipment available: high accuracy multimeters (5-1/2 digit?), cheap > analog oscilloscope, relatively low frequency signal generator (sine, > square, triangle) (top end is maybe 1Mhz), frequency counter, bench supply. > > Willing to buy high-precision R's or C's if needed. > > Accuracy required: hopefully a few percent, 10% is some help. > > I am really a digital guy, so this is a bit out of what I do. > > Any ideas? > > -- Bob Ammerman > RAm Systems > > >