Forums

Current transformer

Started by Tim Williams November 30, 2011
On Sat, 3 Dec 2011 11:48:23 -0500, "P E Schoen" <paul@pstech-inc.com>
wrote:

>"John Larkin" wrote in message >news:vcmgd7117hiv4sjm27tlks6ofat0f648en@4ax.com... > >> Some people make high-frequency CTs > >> Julian Bergoz makes CTs up to 500 MHz. Metglas cores, secret recipes. > >> http://www.bergoz.com/ > >I wonder if these are Hall effect, but there does not seem to be a >requirement for excitation voltage. Or maybe a Rogowski coil with built-in >integrator? But I don't think that would work all the way from 0.5 Hz to 500 >MHz. And up to 20 kA.
I met Julian once, big burly funny not-typically-French guy. He likes to allude to his tricks without giving things away entirely. He did say that some of his core materials have u around a million. He mostly makes real CTs, not Hall things. Halls are slow, I think, and not very stable.
> >> Pearson makes amazing CTs, some up to a couple of hundred MHz, >> and some down to 0.15 Hz > >> http://www.pearsonelectronics.com/ > >Those appear to be Hall Effect. They have a limited current range, and >accuracy depends on the excitation voltage. But they have a wide frequency >range, which includes DC.
I have a Pearson current sensor that works from below 1 Hz to many KHz. It's a CT with an internal burden resistor, and doesn't need power.
> >> There are also Rogowski coils, with no core at all. The induced >> voltage is the derivative of the current being measured. They >> are generally not used with a burden resistor. > >These were used for a long time in the welding industry: >http://www.iernc.com/manufactures/D/Duffers%20Scientific%20Inc/ >and were adapted for use in circuit breaker testing: >http://www.pstech-inc.com/cbtest.htm > >We still use the air-core CTs at www.etiinc.com, for test sets up to 6000 >amps nominal (60kA peak). We make our own, using phenolic core inductors >placed around the high current bus, and the instrumentation uses an RC >integrator to reproduce the current. There seems to be no theoretical limit >to the frequency response or current range, but what we use has an output of >about 100 mV at 1000 A. Since we need to measure current as low as 50 amps, >and the integrator drops the voltage by a factor of about 20:1 at 60 Hz, the >instrumentation needs to measure as low as 250 uV. At the upper end, 60 kA >produces 6 volts. > >Here is more on Rogowski coils: >http://www.dynamp.com/dynamp/LDADocum.nsf/c2270fbdd892ac3e86256e75000ad88a/e710af6d3e0f6255862565d7004b19db/$FILE/Report.pdf >http://www.dynamp.com (in case that link doesn't work) >http://www.latest-science-articles.com/Technology_Science/Study-of-Sensor-Theory-Centered-on-Rogowski-Coil-for-Heavy-Current-Measurement-A-23405.html >http://www.pacw.org/no-cache/issue/autumn_2007_issue/protection_rogowski/rogowski_coil_designs/complete_article/1/print.html >http://customcurrentprobes.com/ >http://www.highcurrenttech.com/products.htm > >> We've made home-made CTs with ferrite cores, risetimes in the >> 10 ns sort of range. We used them for measuring the currents >> into drift step-recovery diodes. > >Here are high frequency CTs for PCB mounting, 20 kHz to 250 kHz. But not in >the 10 nS range. >http://www.triadmagnetics.com/catalog_template.php?productCategoryId=7 > >I have also used some of these for 60 Hz current measurement. They have a >rather good range of current with accuracy if you use the correct burden >resistor: >http://www.talema.net/en/products/ct.shtml >The 100 amp version worked well from less than 1 amp to more than 1000 amps. > >> There are DC current transformers, too, interesting animals. > >> http://www.gmw.com/electric_current/Danfysik/866_867/867.html > >These are definitely Hall Effect.
There are Hall effect CTs, open-loop and zero-flux, but the Danfysic units aren't Halls. They are compound devices using two toroidal cores and both AC and second-harmonic DC feedback. I think they have app notes that explain the theory. Or google dcct . The Danfysik units are amazing... stunning linearity and PPM zero offsets. John
"John Larkin"  wrote in message=20
news:6plkd7ducj2f4ql0k99v03mm4b64abvvob@4ax.com...

> I met Julian once, big burly funny not-typically-French guy. He > likes to allude to his tricks without giving things away entirely. > He did say that some of his core materials have u around a > million. He mostly makes real CTs, not Hall things. Halls are > slow, I think, and not very stable.
Interesting. The link on the GMW site for more information is broken: http://www.gmw.com/technicalnotes/install_5452.html
> I have a Pearson current sensor that works from below 1 Hz > to many KHz. It's a CT with an internal burden resistor, and > doesn't need power.
Probably an iron core of some type?
> There are Hall effect CTs, open-loop and zero-flux, but the > Danfysic units aren't Halls. They are compound devices using > two toroidal cores and both AC and second-harmonic DC > feedback. I think they have app notes that explain the > theory. Or google dcct .
> The Danfysik units are amazing... stunning linearity and > PPM zero offsets.
I think this describes them, known as "integrated parametric current=20 transformers": http://www.bergoz.com/index.php?option=3Dcom_content&view=3Darticle&id=3D= 27&Itemid=3D34 http://www.gmw.com/beam_diagnostics/Bergoz/ipct/pdf/IPCT-07-02%20TechNote= .pdf Interesting range of devices here: http://www.hellotrade.com/gmwassociates/product.html There were lots of simplified explanations, but this seems to be rather=20 detailed (and apparently a rather old scan of a typewritten document): http://lss.fnal.gov/conf/C720919/p674.pdf And this gives details of the instrumentation: http://cdsweb.cern.ch/record/1183400/files/CERN-BE-2009-019.pdf Thanks for the information. Paul John=20
"P E Schoen" <paul@pstech-inc.com> wrote in message 
news:ypsCq.2074$Q83.1500@newsfe17.iad...
> Here are high frequency CTs for PCB mounting, 20 kHz to 250 kHz. But not > in the 10 nS range. > http://www.triadmagnetics.com/catalog_template.php?productCategoryId=7
Yes, I meant to say CST- earlier. These things ring strongly at two nearby frequencies, making a sort of AM-DSB-SC waveform at a MHz or so. Impossible to filter sharply with just RCs, it's not merely HF trash. When they say 250kHz, they *mean* 250kHz, because you aren't getting any useful output beyond that. When you're building a power supply that *starts* at 250kHz, you find you have to spend like gangbusters, or wind custom. Tim -- Deep Friar: a very philosophical monk. Website: http://webpages.charter.net/dawill/tmoranwms
On Sat, 3 Dec 2011 12:57:48 -0600, "Tim Williams"
<tmoranwms@charter.net> wrote:

>"P E Schoen" <paul@pstech-inc.com> wrote in message >news:ypsCq.2074$Q83.1500@newsfe17.iad... >> Here are high frequency CTs for PCB mounting, 20 kHz to 250 kHz. But not >> in the 10 nS range. >> http://www.triadmagnetics.com/catalog_template.php?productCategoryId=7 > >Yes, I meant to say CST- earlier. These things ring strongly at two >nearby frequencies, making a sort of AM-DSB-SC waveform at a MHz or so. >Impossible to filter sharply with just RCs, it's not merely HF trash. >When they say 250kHz, they *mean* 250kHz, because you aren't getting any >useful output beyond that. > >When you're building a power supply that *starts* at 250kHz, you find you >have to spend like gangbusters, or wind custom. > >Tim
As I noted, we just did a few turns around a ferrite toroid. We were driving a DSRD, just a power diode that happens to accidentally have the right diffusion profile. We slammed 48 volts across it in the forward direction until the current built up to 50 amps or so, then reverse biased it from a 400 volt source in series with a bit of inductance. It conducts in the reverse direction for 50 ns or so, then snaps open. Big voltage spike. The CT really helped us see what was going on. Sometimes it's better to use a current shunt, and couple that back down to ground with a transmission-line transformer. Or just a shunt and a TPS2024. John
"John Larkin" <jjlarkin@highNOTlandTHIStechnologyPART.com> wrote in 
message news:pe3ld751do09mv345nj68ptiv6uvn42n62@4ax.com...
> As I noted, we just did a few turns around a ferrite toroid.
See, that'll work -- I've seen tens-of-turns toroids used in RF equipment for things like VSWR meters, matching transformers, plain old inductors and RFCs, etc. Most times, the resonance is sufficiently out of band, or has the wrong impedance or Q to cause any trouble. The trouble comes when you're doing a high ratio, like if you want over 100 turns, evenly distributed around the core, so you aren't burning a hundred watts in the poor burden resistor under whatever load it is you're testing. That needs a progressive wind at least, if not more creative measures.
> We were > driving a DSRD, just a power diode that happens to accidentally have > the right diffusion profile. We slammed 48 volts across it in the > forward direction until the current built up to 50 amps or so, then > reverse biased it from a 400 volt source in series with a bit of > inductance. It conducts in the reverse direction for 50 ns or so, then > snaps open. Big voltage spike. The CT really helped us see what was > going on.
Even so, a couple turns probably had a little ringing when that thing snapped off. Though that'll be as true with the rest of the circuit. Without a perfectly smooth RF signal path, you get reflections off everything. As I recall, the waveform on your website still has some trash on it, but only a few volts bounce, not at all bad when you're making a 20kW pulse into 50 ohms.
> Sometimes it's better to use a current shunt, and couple that back > down to ground with a transmission-line transformer. Or just a shunt > and a TPS2024.
There are, of course, coaxial resistors for that sort of thing. Or whatever those fancy swirls you're etching into manganin are doing. Tim -- Deep Friar: a very philosophical monk. Website: http://webpages.charter.net/dawill/tmoranwms
Tim Williams a &#2013265929;crit :
> "John Larkin" <jjlarkin@highNOTlandTHIStechnologyPART.com> wrote in > message news:pe3ld751do09mv345nj68ptiv6uvn42n62@4ax.com... >> As I noted, we just did a few turns around a ferrite toroid. > > See, that'll work -- I've seen tens-of-turns toroids used in RF equipment > for things like VSWR meters, matching transformers, plain old inductors > and RFCs, etc. Most times, the resonance is sufficiently out of band, or > has the wrong impedance or Q to cause any trouble. > > The trouble comes when you're doing a high ratio, like if you want over > 100 turns, evenly distributed around the core, so you aren't burning a > hundred watts in the poor burden resistor under whatever load it is you're > testing. That needs a progressive wind at least, if not more creative > measures. >
The problem is that if your primary side conductor is quite small WRT your core diameter, the small leakage inductance that you have on the primary side is multiplied by N^2 on the secondary side, which can be quite a lot for high turns ratios. It's that inductance that resonate with the secondary parasitics. The first thing is to reduce as much as possible that empty space.
>> We were >> driving a DSRD, just a power diode that happens to accidentally have >> the right diffusion profile. We slammed 48 volts across it in the >> forward direction until the current built up to 50 amps or so, then >> reverse biased it from a 400 volt source in series with a bit of >> inductance. It conducts in the reverse direction for 50 ns or so, then >> snaps open. Big voltage spike. The CT really helped us see what was >> going on. > > Even so, a couple turns probably had a little ringing when that thing > snapped off. Though that'll be as true with the rest of the circuit. > Without a perfectly smooth RF signal path, you get reflections off > everything. As I recall, the waveform on your website still has some > trash on it, but only a few volts bounce, not at all bad when you're > making a 20kW pulse into 50 ohms. > >> Sometimes it's better to use a current shunt, and couple that back >> down to ground with a transmission-line transformer. Or just a shunt >> and a TPS2024. > > There are, of course, coaxial resistors for that sort of thing. Or > whatever those fancy swirls you're etching into manganin are doing. > > Tim >
-- Thanks, Fred.
On Dec 2, 1:55=A0am, "Tim Williams" <tmoran...@charter.net> wrote:
> "BillSloman" <bill.slo...@ieee.org> wrote in message > > news:d7f1d4b0-c8ee-48d8-840d-d1276c73d4b3@p2g2000vbj.googlegroups.com... > > > You seem to have failed to take into acount the parallel capacitances > > of the windings. No electronic component is pure - in the sense of > > presenting only resistive, capacitative or inductive impedance - and > > inductors/transformers are more imperfect than most. > > Indeed, effective parallel parasitic capacitance is a valuable concept. > Sadly, it's just that, a concept -- the *actual* capacitance from end to > end of, say, a solenoidal coil (i.e., as more advanced modelers call it, =
a
> helical resonator) is dramatically smaller than the turn-to-turn > capacitance. > > Consider, if instead of a helix, you had a stack of rings. =A0It's the sa=
me
> basic structure, except skewed by a turn, so the turns aren't turns, > they're loops. =A0Now ground ALL the rings, except for just the two ends. > What is the capacitance between those two rings? > > The capacitance will not only be small due to distance, but almost > entirely shielded by the turns in between them. =A0When you unground them=
,
> all the intervening turns have their own capacitance, but it still doesn'=
t
> even act as an ideal capacitive divider, because there is finite > propagation delay along the structure (i.e., the speed of light) and > because the interspersed turns have a comparable loading all their own > (self capacitance to free space as well as "mutual capacitance" to > adjacent turns).
I once amused myself by splitting the notional 1pF parallel capacitance of a single layer winding into two 2pF capacitors across two - linked - mutual inductances in series, fudging the two series inductances and the coupling to come out at roughly the original single inductance. It was all very rough and I can't remember the numbers. What I can remember is that I might as well not have bothered. Once you've got deep enough into high frequencies that the propagation delay along the solenoid matters, you are obviously into a different ball-park, but I was a bit surprised how useful the simple lumped component model was in my application. <snip> -- Bill Sloman, Nijmegen
"Bill Sloman" <bill.sloman@ieee.org> wrote in message 
news:7e2e53a7-05a8-43ff-ab5e-23718d179842@g21g2000yqc.googlegroups.com...
> Once you've got deep enough into high frequencies that the propagation > delay along the solenoid matters, you are obviously into a different > ball-park, but I was a bit surprised how useful the simple lumped > component model was in my application.
Oh, if you've got the right model and the system is reasonable (possibly, it must be non-dispersive), it even works all the way up to the first resonance. This is true of 1/4 and 1/2 wave resonators, in coax / stripline form, and still reasonable for microstrip (which is dispersive). Probably because the infinnitessimal circuit is a scaled down model of the same layout. Take a chunk of coax and it looks like series L and parallel C on almost any scale (limited by diameter, when higher modes take over); Z_o = sqrt(L/C) and F_res = 1 / (2*pi*sqrt(L*C)) are always true. I don't think it applies for helical and thin toroidial resonators, because the calculated effective parallel capacitance and inductance vary with frequency (going to extreme values at resonance, of course); and this is going to be a consequence of the infinnitessimal model, which involves mutual inductances and chains of capacitors, producing dispersive modes that are difficult to model in the same way. Tim -- Deep Friar: a very philosophical monk. Website: http://webpages.charter.net/dawill/tmoranwms
> >I appreciate your deep knowledge of the subject, >however maybe I should not read about it, >as it makes something that I always perceived as extremely simple, >sooooo complicated. > >Current transformer: >current_in / turns_ratio = current_out. (for a low load resistor or
short).
>current_out x load_resistor = voltage_out. > >Of course there are some details, >but those can usually be disregarded as the errors caused by those are
very small.
>NORMALLY. > >One lesson: Do Not Leave Out The Load Resistor, as then > Ze Foltage Kan Beacome Ferry High >
Why? I dont understand why one cant simply use the open ckt voltage on the secondary as a measure of the primary current. Heres my argument: B = prim. ct. * I(geometry-dependent stuff) (Biot Law); (I() => integral) flux = B * A; A => area Vsec = d(coupled-flux)/dt For sinusoidals therefore, the open circuit is proportional to the prim. current. If you consider the ct. transformer as a coupled inductor, then the actual calculations are quite simple. Right? vkj --------------------------------------- Posted through http://www.Electronics-Related.com
"vkj" <tranquine@n_o_s_p_a_m.n_o_s_p_a_m.gmail.com> writes:
> Why? I dont understand why one cant simply use the open ckt voltage on the > secondary as a measure of the primary current. Heres my argument:
Because a current transformer is normally in the 1000:1 range. Do you really want to try to measure 120,000 volts? (also: this is why it's important to hook up the load resistor *before* you put the current transformer on the primary wire)