High Voltage Forum

Tesla coils => Dual Resonant Solid State Tesla coils (DRSSTC) => Topic started by: Hydron on January 11, 2020, 12:04:59 AM

Title: CT performance investigations (Pearson and DIY)
Post by: Hydron on January 11, 2020, 12:04:59 AM
Note: I originally posted this a few years back on 4hv, but unfortunately attachments are broken on that forum and they're kinda necessary to view this properly, so I've reposted this here. "Recently" isn't really accurate anymore, and the Pearson family has grown to include a couple of baby SMA-connector model 2877 CTs. The dodgy BNC on the 7800 has also been fixed as described here: https://highvoltageforum.net/index.php?topic=435.0 (https://highvoltageforum.net/index.php?topic=435.0). I've also since had a play with some other ways of constructing DIY CTs that might be less susceptible to noise, but they haven't really been tested in anger yet - if I have any luck then I'll make a new topic.

I have recently acquired a trio of Pearson wide band current transformers/monitors, so I though that I'd do a bit of work investigating their behavior and that of a home-made DRSSTC CT (constructed for primary current measurement and ZCS bridge switching).

- 7800, 3525 and 4100
- 0.01, 0.1 and 1V/A respectively into high-Z input, half that into 50R
- 2MHz, 15MHz and 35MHz high frequency 3dB point respectively
- All rated for <6deg phase shift at <10% of high 3dB freq
- Performance measurements on all 3 indicate that they meet datasheet specifications (tested to 10MHz)
- At frequencies >10% of high 3dB freq phase shift is strongly affected by position of wire in central hole
- If wire is well centered phase shift seems significantly lower than Pearson specifies

- Cascaded CT wound on mystery ferrites, probably EMI suppression ones
- Bottom right holds the first winding, and this is where the wire with current to be measured passes through
- Top (blue) CTs are for bridge ZCS switching and OCD
- Bottom left CT is for measurement, and is terminated with 10R burden and 39R termination resistors inside the BNC
- CTs are ~1000:1 (I think 31:1 x 32:1) overall, giving 0.01V/A output on the BNC (into high-Z input, half that into 50R)

The first problem I ran into was the huge and heavy Pearson 3525 being too heavy to lug from the UK (where I bought it and now live) to NZ (where my DRSSTC lives).

This means that I only had the 7800 and 4100 to compare against the DIY CT, and the only one that can be trusted for low phase shift at frequencies over 200kHz is the 4100.

The 4100 at 1V/A isn't ideal for TC measurements even with a 20dB (10x) attenuator, as it's only rated for 500Apk/5Arms, so I had to get creative:
- At the end of this paper http://mdk2001.web.cern.ch/mdk2001/Proceedings/SessionPoster/Cordingley.pdf, linked in this thread: http://4hv.org/e107_plugins/forum/forum_viewtopic.php?175696 there is a suggestion of cascading a simple CT with a wide band CT for measuring high currents at high frequencies (while giving up low freq perfomance). I had a go at this with a 10:1 CT on a decent ferrite ring, seen below:

- I compared the 4100 cascaded with a 10:1 DIY CT against the 3525, giving the following result:

(sorry - I couldn't find the original of this so it's only the thumbnail left from 4hv!)

(Note the scales change between captures, and should be checked carefully - e.g. the phase shift above looks big but is under 10deg. Also note that the Pearson 4100 on it's own is only specified for <6deg phase shift between 1.4kHz and 3.5MHz, and the 3525 only up to 1.5MHz)

- This looks really good for frequencies >50kHz - almost as good as the same comparison without the cascaded 10:1 CT!

When I got to NZ I ran a couple of comparisons of the 4100 Pearson (with and without the 10:1 CT on it's input) against the 7800 using the primary circuit of my DRSSTC.

Comparing the 4100 (with 10:1 CT on input and 20dB atten) with the 7800 with the coil running at 850Apk shows no difference other than slightly lower gain on the 7800:

If the 4100 (without the 10:1 CT on the input, but with a 20dB attenuator) is directly compared with the 7800, it has significantly lower susceptibility to noise than the 7800, but it must be remembered that the signal at the CT output is 100x larger than that of the 7800, which is also possibly at a disadvantage being a clamp-on model:

(7800 red, 4100 blue)

Finally I compared the Pearson 4100 (with 10:1 input CT and 20dB attenuator = 0.01V/A) with the 0.01V/A DIY CT, first measuring gain/phase:

Looks rather good, easily usable to over 1MHz!
Seems the cascaded CT approach used for most DRSSTCs is actually a very good one, and not just because of the low number of windings needed vs a single stage 1000:1 CT

When tested in a DRSSTC, we see that the main advantage of the Pearson in this application is noise immunity (it's shielded, whereas the DIY CT is not). Bridge switching noise, especially on the first couple of transitions where ZCS isn't perfect, clearly shows up:

(Red is Pearson, blue is DIY CT)

In conclusion:

The cascaded home-wound approach of CT construction used for DRSSTCs seems to provide extremely good performance for the application. They are not as noise immune as the shielded wide-band commercial models, but this is mostly seen on the initial low-current transitions, seems to clear up as the bridge switches higher current (where accuracy is most needed anyway!). Low frequency performance is likely poor (and certainly won't get anywhere near the 5Hz of the 7800), but it irrelevant in this application.

It can also be seen that the cascaded CT approach can be used to allow a low-current wide-band CT to measure at least an order of magnitude higher current than otherwise possible, while maintaining good noise immunity. Again, this will be high-freq only, but still fine for DRSSTC use. This could be really handy if a 1V/A unit is available cheaply (mine was under 50 USD, albeit with the BNC in bad condition) and higher current capability or a larger center hole is needed.

Something that could be interesting for someone to try in the future is to attempt to shield a DIY CT in a similar manner as done for the commercial CTs (small gap in shield, sometimes even overlapped, to avoid a shorted turn). Might be a good way to get decent performance for TC measurements on a very low budget, without having to luck out on ebay!
Title: Re: CT performance investigations (Pearson and DIY)
Post by: T3sl4co1l on January 11, 2020, 04:39:55 AM
FYI, the biggest limitation is the helicotoroidial waveguide modes.  Er, that is to say... consider the winding as a spiral-walled waveguide, which as it has a loop continuity condition (when the winding is distributed evenly over the core, so the start and end leads are adjacent), means you have integer numbered standing waves.

Compare to the 1/4-wave stub, which can be wound into a helical shape, a helical resonator; a Tesla coil secondary being a fine example.  The fields at the two ends don't need to match, indeed they are complementary at 1/4 wave.  Hence the high voltage at the top, and the high current at the bottom (assuming it's grounded to a nice large plane, of course).

The wave speed is rather slow, because it's loaded by the core.  This also raises Zo, and dampens it some, which helps, or can help anyway.  This makes lossy ferrites desirable (or most of the Pearson CTs I think use nanocrystalline cut cores).  Conductive (powdered iron, stripwound?) materials tend to short out the resonant mode, acting more like a ground plane so you get a more ordinary transmission line stub with a lower impedance (Zo ~ 100 ohms, instead of ~kohms) and higher velocity.

How much does it matter?  Well... Triad CST206-1A says 200kHz range... and they mean it. ::) There's an intense resonance at 300kHz or so.  And because Q and Zo are high, you can't practically dampen it at the terminals -- you'd need to insert an R||L in series with the burden resistor (or something like that) that will gobble up a huge amount of flux, ruining its current capacity and LF cutoff.  On the order of, well, 1kohm and whatever L is the same reactance at 300kHz, so, 530uH or thereabouts?  So yeah, a lot of wasted impedance just to try and dampen a parasitic mode that shouldn't be there by design.

So what can you do about it?  Well, if you use a current loop to couple the CT, you'll notice that when it's positioned perpendicular to the coil terminals, the resonance is at a minima.  Indeed, the resonating wave has two components, with nodes aligned parallel and perpendicular to the coil terminals -- if we position the driving loop at right angles, the null lands on the terminals and we don't see it in circuit (or, as much).

We can also think about shorting out the modes.  Say we use two driving loops in parallel, positioned at right angles.  This forces the resonant modes to butt heads, at least reducing their impedance, but potentially also forcing the minimum resonant mode up to 2 or 4 wavelengths.  Which puts a lot more of the wave into the core, where it can experience higher losses.

The limiting case being a primary winding that's completely surrounding the core, which is what the Pearson CTs manage -- the case is solid, brass I think, except for a slit around the inside.  It doesn't much matter where the actual source current is (to first order anyway; there's still some effect, as you've noted!), as the field seen by the core and secondary is shaped by the housing.

The other thing you can do, is use less overall winding length.  The resonance is proportional to the speed of light times the winding length (and a modest geometry factor).  Smaller cores with shorter windings are preferable, which also prioritizes higher µ_r and Bsat.  So ferrite and nanocrystalline are excellent choices.

Of course, few turns means a poor ratio, so the cascaded design can be quite effective.  The frequency response of each CT stacks, but this is an arithmetic rather than geometric difference.  (That is to say, two N-turn CTs cascaded has a ratio of N^2 and might have a bandwidth half that of one alone, but the bandwidth of a single CT with N^2 turns is about 1/N times...)

And of course of course, preferably the cascaded CTs should be made the same way, i.e. with a slitted shield to suppress resonant modes.

Cool fact: particle physicists make their CTs this way.  Consider a beamline, a metal tube through which bunches of charged particles are traveling, electrons or protons or whatever.*  Suppose we cut the beamline so there's a slot all around, then we put a ferrite core over the slot, and put a larger pipe section over that.  The larger pipe is bonded to the main pipe with end caps.  So, effectively, the pipe diameter increases momentarily, the extra space being filled with a core.  Suppose further, we connect a bunch of coax cables across the slot, spaced evenly.  Now whenever a bunch passes by, the field emitted by them appears across the gap in the beam tube.  And since these bunches can be traveling extremely fast, and are very tightly focused (maybe mm total length, and a cross section of, Idunno, fractional mm to µm I think?), the measured current sense pulse is extremely short -- picoseconds, easily!  The current is quite feeble as well (fractional mA for the more intense accelerators, AFAIK), so you can appreciate a good bit of work is needed to make use of this signal. :)  (The timing signals are routed throughout the facility, for synchronizing experiments, beam deflectors, samplers, cameras, detectors, etc.)

*Particles end up grouped into tight packets or "bunches", rather than a continuous beam, when electromagnetic acceleration is used -- LINACs and synchrotrons.  The particles ride a traveling wave, the ones on the leading edge get pushed forward (drawing energy from the field), while those on the trailing edge get pulled backwards (delivering energy to the field).  These waves are typically held in superconducting (typically electropolished niobium at 2.2K) resonators, with Q factors comparable to quartz crystals!

This has been your daily dose of physics trivia... ;D ;D

Title: Re: CT performance investigations (Pearson and DIY)
Post by: Hydron on January 22, 2020, 11:30:26 PM
I definitely appreciated the daily physics trivia, thanks!

It took a little longer than planned, but I have done a quick check of the frequency response of a newer DIY CT, using a combination of a hand-wound toroid and a commercial 50:1 high(ish) frequency CT, both mounted on a PCB as shown in the photos below:

The shown CT board is missing it's connector and a common mode choke I put a footprint on the board for (but may not use/need), and is wound for 1000:1 (20:1 on the ferrite toroid, 50:1 on the Murata CT) for use in my 160mm coil, with the burden resistor located on the driver/controller board. Note that the holes bolt to the grounded heatsink, and the board is mounted on the coil for conveniently passing the bridge output wire through as shown below:

In an attempt to reduce noise coupling from the bridge output into the CT output I used a short piece of co-ax to pass the signal through the second of the cascaded CTs, with the shield (of ONE end only) connected to RF ground via the mounting holes on the PCB, and the central core carrying the current to be sensed. This can be seen on the later photos too.

The Murata CT I used is a 56050C (https://www.murata-ps.com/datasheet?/data/magnetics/kmp_5600.pdf), rated at 20-200kHz, 10A primary rating and a terminating resistance sized for 1V/turn. Using this in a DRSSTC even with a 10 or 20:1 prior step-down, it will definitely see more than 10A peak through the primary, but it's also being used at a higher than 20kHz frequency and with a lower burden resistor, so hopefully I avoid any saturation (heating shouldn't be an issue at DRSSTC duty cycles).

After your post and comment on the Triad 200kHz limit, I thought I'd measure it's frequency response like in my first post. I did this test using a CT wound for 500:1 (10 turns on the first stage), a 47ohm burden resistor (nearest match to a 50R system I could quickly find in my through hole resistor box) and in comparison to a Pearson 2788 CT plus a 20dB attenuator to roughly match the nominal gain between the two to ~0.05V/A driving a 50 ohm load (47R burden means the DIY one is a little over 0.5dB down from 0.05V/A).
The test setup is shown below (with a Pearson 4100, which was changed to the higher frequency capable 2788 once I saw how far it could go!):

Frequency response results (below) were significantly better than expected, nearly getting to a few MHz before everything started going pear shaped (I noticed the output waveform getting non-sinusoidal over about 4MHz, which obviously isn't obvious in a gain/phase plot and wasn't really what I expected from something I assumed would be linear!). I suspect that using the lowest turn-count part in the series (1/6th of the highest 300 turn option) helped out, as the bandwidth rating of 200kHz (the same as the triad one which you said went to pot above that) applies to all of them regardless of turn count.

P.S. If anyone is interested in the PCB files then I'm happy to share, but be warned that they'll be in Altium or something easily exportable from Altium and are obviously customised for my application (e.g. with connection via a S/STP patch cable to the control/driver board - a generic board should probably use screw terminals and ditch some of the other junk I have on there).
Title: Re: CT performance investigations (Pearson and DIY)
Post by: T3sl4co1l on January 22, 2020, 11:45:26 PM

Title: Re: CT performance investigations (Pearson and DIY)
Post by: Hydron on January 23, 2020, 12:33:44 AM
I took another look at the output waveform and it starts to look distorted above 4-5MHz for some reason (not quite sure how - was expecting things to be linear!), so I've updated my earlier post from "near 10MHz" to "a few MHz". An example of the distortion is attached below:
 [ Invalid Attachment ]
(orange is reference from pearson CT)

Not an issue when I'm planning to use it at <1MHz, but worth bearing in mind. Also it picks up a LOT of noise when not band limited to 20MHz, whereas the pearson picks up very little.
Title: Re: CT performance investigations (Pearson and DIY)
Post by: T3sl4co1l on January 23, 2020, 10:03:17 PM
Probably there's a peak up there, and it's not so much distortion as tuning in to a harmonic that your signal source has, which is normally invisible (-20dB or less compared to fundamental?) but which becomes visible due to that gain (tuning).  This should show up on the sweep, of course... :unsure:

High frequencies are also asymmetrical around the core, so, couple to ambient fields.  Which again, is another way of saying that a [slotted] shield around the secondary is of great help. :)

Title: Re: CT performance investigations (Pearson and DIY)
Post by: Hydron on January 23, 2020, 11:50:22 PM
Yeah invisible harmonics were the only rational explanation I could think of, but I certainly couldn't see them in the range I tested (or obviously in the reference signal).

As for noise, I'm less worried about general pickup from ambient sources as I am about stuff from the high dv/dt switching events - my testing was done at 10mA/div whereas it would be used at much higher currents.
Title: Re: CT performance investigations (Pearson and DIY)
Post by: T3sl4co1l on January 25, 2020, 06:44:21 AM
Of course, much stronger signals come with much louder ambients (ambiance? ;D ).  The usual effect is to get some of commutation (either the switching loop, so, a spike up or down during switching; or a fraction of the load, so, a squarewave superimposed on the reading; and of course whatever ringing from either source) as an error signal.  And yeah, the transient stuff you can filter out, but it brings into question: how representative is the signal really, and how much do you need to filter to clean it up, and since you're filtering, how does that impact your response time (for fault protection, or phase lock, or whatever).

Which for most power applications, should be pretty alright.  (Those damn CST-206s though?... ::) )

Title: Re: CT performance investigations (Pearson and DIY)
Post by: Uspring on February 04, 2020, 12:10:49 PM
Thank you for the intro to the "helicotoroidial waveguide modes". I rather liked the analogy to a Tesla coil secondary winding, which in this context appears as a transmission line. The resonances than would be standing waves of a circular transmission line, i.e. one, where the ends are connected.
A transmission line is usually modelled as a length of distributed inductances connected to ground by distributed capacitances. The inductances in this case are magnetically uncoupled, though, which is true for e.g. a long and thin solenoid or a straight wire, but that doesn't hold for a toroid with a core. Actually, if you e.g. split a single layer winding on a toroidal core into to 2 halves, you'll have a nice 1:1 transformer. The lowest frequency mode will be one that drives the 2 windings in opposite direction, which doesn't work for an ideal transformer. AFAIK this mode can only be excited if there is some leakage inductance. So the resonance frequency will depend to a great extent on how tight the windings are around the toroid. It looks like most of the action takes place between the winding and the core and not inside the core.

I don't have much experience with CTs. Is what I say correct?

Title: Re: CT performance investigations (Pearson and DIY)
Post by: T3sl4co1l on February 04, 2020, 06:26:24 PM
Yes, indeed the leakage inductance of a toroid, at high frequency, is due to the asymmetry of the fields.  And that's why you get seemingly weird behavior, like nulling the resonance when the source current is in a loop perpendicular to the burden resistor connection.

Similar fields manifest if you use a CMC differentially.  Which you normally do, in an EMI filter.  That is -- consider the common mode choke, using a toroid core, with two windings on opposite sides.  They are phased so the common mode is in phase (high impedance) while the differential phase is opposed.  Leakage inductance is fairly high in this configuration, i.e. the differential mode doesn't completely cancel out.  If you plot the fields from these, it looks very much like two bent solenoids, beside each other, opposing.  All the leakage field is external (since after all if it were in the core it would properly oppose!).

Maybe that's not too useful without a diagram... alas, my apologies.

Title: Re: CT performance investigations (Pearson and DIY)
Post by: Uspring on February 05, 2020, 10:45:02 AM
Thank you for your explanation. No need for apologies, I managed to imagine the fields of two opposing bent solenoids before my brain turned to something looking similarly.
SimplePortal 2.3.6 © 2008-2014, SimplePortal