What kind of transformer is this ? (EEVBlog #1119)

[BravoV]

Just in case you haven't watch (HERE) just jump to 11:30 where Bart started to explain it if you're lazy to watch all of them.

Its used in the Cleverscope : CS448 which is an isolated scope. Its the power transformers to power the isolated scope front end, and claimed they are manufactured by them self, and claimed to have low CMRR.

Curious as I've never seen one like this before, photo captured & attached below.


What kind of winding is that ? I'm puzzled by how are windings go around that whole thing ? 

As highlighted by arrows, is that a 3 taps windings at both primary & secondary ?

Also regarding the magnetic material, the blue toroid is obvious, but at the surrounding black container, is that a magnetic material too ?

What are the advantages of this kind of build, say compared to plain ordinary EI core ?

Appreciate any other insight or comments.

[BrianHG]

The secondary, output, is on the toroid core.
The primary is lifted above the core, in fact, it isn't even in contact with the core.  This is needed to give you isolation at the electrostatic level and series capacitance prom primary to secondary < 1 pf.  Call this the ultimate galvanic isolation transformer.  When it comes to placing a GND of a probe on a 500v H bridge output, switching up and down in 10ns or less, you cannot bridge that swinging GND to anywhere else in the scope's electronics, or, other input channels.  Even a capacitance of 1pf from primary to secondary with a 500v swing at the 10ns rise and fall times would feed a disastrous signal right through a conventional coil sitting on-top of coil transformer.  You got to keep those primary and secondary conductors as far away from each other as possible.

Used to do this for audio DAC, not power, but data where you wanted to super isolate the SPDIF signal.
For power, we would use split bobbin transformers to get as far as possible from the mains AC.  Obviously, NO Y caps from mains to secondary.

The surrounding black material is plastic.  Something very low in capacitance, wont build up static electric charge, and is safe for a few Kv.

That black transformer case goes below the PCB, since you can see a cutout on the PCB.  There are 4 solder points on the PCB, not 3.  The center 2 on each side is the main transformer connection.  It looks like the outer 2 connections on both sides goes to a support wire holding down the entire apparatus.  This short outer conductor seems to go through a single loop in the transformer line.  Maybe it is a feedback loop.  I'm only guessing here.

I believe I once had my similar transformers made at Inductors Inc.  I had a 15kv AC hipot.  It was a custom job, nothing you can buy off the shelf.

[BrianHG]

With 4 solder contacts on each side of the transformer, and a twisted pair, it is possible the the inner 2 solder spots go to the twisted braid on the beginning of 1 coil while the outer 2 are coming in from the underside of the plastic case to the top side through 2 holes, to make it to the top contacts.  This would be my second guess, that it is just the way the outer 2 contacts are braided from the bottom side to the top of the transformer, then send out to the outer 2 pins.

Here are some other closeups, different angles, zoomed in with sharpening and a gamma boost to try to get a better look...

[BrianHG]

Labled...
Same goes for the right side, except, it is being done through a center hole in the plastic by the center pin, then twisted together, then spread out to the 4 pins on the right side.

[BrianHG]

My guess is on the primary, 1 wire in the twisted pair is power, 1 feedback, or, both conductors are in parallel.
On the secondary output, the twisted pair is wired in series at the output for a +/- tracking balanced power supply.  All the analog needs the +/- 5v or even +/-12v supply.

The turn count is also identical from primary to secondary, 5 loops on each.

[bloguetronica]

Not sure about the clearance distance, but on a first look, it seems to be huge for 1KV. I do recognize some isolators from Texas (SO packages bridging between two boards on the lower right corner).

P.S.: 0.5ppm is considered a very accurate clock? I'm off the hook, then! (just because I've designed a 0.025ppm clock gen, and I wasn't sure about the precision on that)

[BrianHG]

Not sure about the clearance distance, but on a first look, it seems to be huge for 1KV. I do recognize some isolators from Texas (SO packages bridging between two boards on the lower right corner).

Anything less on the spacing and a 100 MHz signal at 1Kv will bridge right through from secondary to primary.  We are not talking about a 60Hz 1Kv, we are talking about over 100MHz at 1Kv.  You better believe all that empty air space between primary and secondary is crucial.

As for the isolated function generator, it only goes to around 25Mhz, at a low voltage, so the digital isolators are good enough on that side.  It is also a low impedance output, not a 1 MegaOhm input reading small milivolts with a sensitive ADC.  You can believe that to achieve the posted scope isolation specs with no crosstalk, those special transformers were the sole solution other than making battery powered front ends.  Those long optical fiber for the data was also mandatory.

[bloguetronica]

Not sure about the clearance distance, but on a first look, it seems to be huge for 1KV. I do recognize some isolators from Texas (SO packages bridging between two boards on the lower right corner).

Anything less on the spacing and a 100 MHz signal at 1Kv will bridge right through from secondary to primary.  We are not talking about a 60Hz 1Kv, we are talking about over 100MHz at 1Kv.  You better believe all that empty air space between primary and secondary is crucial.

As for the isolated function generator, it only goes to around 25Mhz, at a low voltage, so the digital isolators are good enough on that side.  It is also a low impedance output, not a 1 MegaOhm input reading small milivolts with a sensitive ADC.  You can believe that to achieve the posted scope isolation specs with no crosstalk, those special transformers were the sole solution other than making battery powered front ends.  Those long optical fiber for the data was also mandatory.

I must confess I haven't seen the whole video. Battery power front-ends? So, basically they avoided isolated DC-DC converters and other such solutions. That is new and I was not expecting that. That seems to be extreme!

Kind regards, Samuel Lourenço

[T3sl4co1l]

It's not.  it's a common mode choke.

The giveaway is the pairs connect to each side, not across the barrier as a transformer.

The custom(!) bobbin is to minimize capacitance; the core may be nanocrystalline, for reasons well covered in this thread,
https://www.eevblog.com/forum/blog/eevblog-1104-omicron-labs-bode-100-teardown/
but for common mode choke purposes instead.

This gives maximum isolation impedance, to the highest frequencies possible.

At least, that's just my interpretation of it.

The "102" block nearby is probably another CMC, for higher impedance at lower frequencies (it's probably an off-the-shelf like a Bourns SRF0905), and will have a fair bit of capacitance (and maybe transmission line characteristics -- peaks and dips at high frequencies) at modest frequencies (low MHz), where the fancy CMC takes over.  Alternately, it could be a transformer -- the component itself could still be a CMC, they're just using it as an isolation transformer proper -- in which case the converter would be a forward converter, and given the low noise, probably a balanced, resonant or quasi-resonant type.

Actually, that wouldn't make sense, the isolation capacitance would be utterly awful.  Even if it's a power line choke*, the capacitance might not be great.  It could very well be a custom part with very low capacitance; this design certainly deserves it.

*A data choke is usually twisted pair or bifilar, wound on a toroid or other shape.  A power choke is usually two separate windings, on opposite sides of the core, giving much higher leakage which is useful for diff mode filtering.

Tim

[mzzj]

Not sure about the clearance distance, but on a first look, it seems to be huge for 1KV. I do recognize some isolators from Texas (SO packages bridging between two boards on the lower right corner).

Anything less on the spacing and a 100 MHz signal at 1Kv will bridge right through from secondary to primary.  We are not talking about a 60Hz 1Kv, we are talking about over 100MHz at 1Kv.  You better believe all that empty air space between primary and secondary is crucial.

Voltage is usually derated with increasing frequency, also in this case:

Maximum Differential Input Voltage

±1 kV, derated above 1 MHz.

Derated at 20dB/decade

Maximum Common Mode Input Voltage

±1 kV, derated above 10 MHz.

[T3sl4co1l]

Ya know, on closer look, that probably is a CT:CT transformer, and that's why they used twisted pair.  I didn't see the multiple connections at first and first-glance-assumed it was all one wire.

But again, just a guess until someone desolders one   (Gosh, what a sin that would be!  Would be great to peek inside those shields too though...)

Tim

[ArthurDent]

It appears to be a CT primary to CT secondary transformer with very high voltage isolation. It is wound with twisted pair like a lot of ham baluns. Here is a file photo of what would be one half of the pictured balun transformer.

[Kleinstein]

The toroid part very much looks like a transformer, build for capacitance. The 2 twisted wires are likely just for get 2 windings and maybe a split coil for the output. Some of the pictures are quite clear. The only thing slightly odd is that on top it is just some air space with no extra insulation material.

The extra ferrite block than might be extra common mode choke, especially for the very high frequencies, where even a few pF could be a significant.

[BrianHG]

Just to repeat, after analyzing the photo and listening to the interview, this is a power DC-DC converter transformer.  The secondary, power output is the twisted pair, on the left side, whose twisted pair is wound directly on the magnetic core at 5 turns.

On the outer plastic structure, there is the power input twisted pair going around the core, 5 turns, but spaced off the surface of the core to give that super low pf transmittance from 1 side to the other for the high frequency high voltage steps you see in power supply design work.  If a single twisted pair had the primary and secondary as the isolation, that 10ns 500v pulse would migrate through immediately.

The primary twisted pair is most likely wired in parallel for lower voltage.  The secondary twisted pair is most likely wired in series for double voltage, or, a +/- supply output.

All the data is on the 4 fiber-optic cables.  There is no data on this transformer.

[coppercone2]

what kind of power transfer do you get through that kind of construction at that size?

[BrianHG]

(I cheated here, they are actually using around 2.5 watts of power).....

Like I said, here is the ADC they are using, it consumes 1 watt max: ISLA214P50

Add another 1.5 watts for the fiber transmitter and analog front end opamps and I get approximately 2.5 watts, maybe even 3 watts.

[mzzj]

what kind of power transfer do you get through that kind of construction at that size?

from couple of watts to maybe 20-30 watts. Not much copper there.
1" toroid core is good at least to 100 watts with less radical winding

[BravoV]

BriangHG, thanks a lot for the comments & illustrations. 

On the outer plastic structure, ......

.....  There is no data on this transformer.

Its in the video on Dave's interview, Bart Schroder (the designer) said they took almost 1 year to develop the power section to power the scope front end including this transformer if I'm not mistaken, and mentioned too that its a custom in house job.

Regarding the outer black plastic surrounding the blue toroid core, I guess that is also a custom made isn't it ?

With this construction style, does the parasitic capacitance between the two windings (pri and sec) is relatively low too ?

For sure, Bart said it has lowest common mode noise.

[BrianHG]

Here is an example of a mass produced version of a super galvanic isolation transformer, a lot larger than the one I once had made for 10mb data around a decade ago:

Note that Cleverscope's transformer case looks like a home made 3D printed, and that plastic case holds the windings at specific locations and angles on both the inner coil and the outer windings to even further isolate the primary and secondary.  Remember, they are not trying to separate 60hz sine wave AC, or DC, they need to measure the weak small drive of a gate drive with that probe GND tied to the switching bridge with a 10ns 500v (this is like a spike intermittent radio broadcast) step up and down not to affect the adjacent probe channels which may sit on a different frame ground.  If the windings were any closer, that signal would bridge to the chassis & power of the Cleverscope's power system, then to the adjacent input channels.  Their only superior choice would be to use a big 25 watt laser diode to beam power to a photovoltaic cell using another fiber link with that huge empty space, for each channel.

[Hypernova]

Their only superior choice would be to use a big 25 watt laser diode to beam power to a photovoltaic cell using another fiber link with that huge empty space, for each channel.

Do people actually do that? At that point you may as well go for batteries.

[Kleinstein]

Besides a transformer and photovoltaic, there would also be the option to use an acoustic transformer: In the extreme case a glass rod with transducers on both sides. Efficiency is expected to be better than the optical way.

The optical way is not that bad anymore: red / NIR lasers are relatively efficient (up to about 50%) and also PV conversion just above the edge is more efficient than with solar light.  So I don't think it would need 25 W optical, more like 6-10 W, maybe 25 W of electrical input. I would prefer an acoustic transformer.

[BrianHG]

Their only superior choice would be to use a big 25 watt laser diode to beam power to a photovoltaic cell using another fiber link with that huge empty space, for each channel.

Do people actually do that? At that point you may as well go for batteries.

Here is a 1 watt transmitter/receiver system.
https://www.fiberopticlink.com/product/power-over-fiber-system-pof/#tab-id-4

[Hypernova]

Their only superior choice would be to use a big 25 watt laser diode to beam power to a photovoltaic cell using another fiber link with that huge empty space, for each channel.

Do people actually do that? At that point you may as well go for batteries.

Here is a 1 watt transmitter/receiver system.
https://www.fiberopticlink.com/product/power-over-fiber-system-pof/#tab-id-4

That is some hardcore gear for just 1W.

[SeanB]

Given that the black former is undamaged by soldering right next to it, I guess it is a custom made injection moulded part, probably made from PTFE with a glass bead fill, to get the right high temperature ability and also not build up a dirt film with time. either injection moulded or powder compressed in a mould, and I will edge to compressed as that is a lot easier on the die. Got to be hand wound as well, there must be a whole room with a few people in there sitting all day with these toroids, the PTFE coated twisted pair wire and carefully winding them, then checking them on some sort of test jig to get the capacitance both low and balanced between the halves of the windings.

Looks like the 5 turns wound on the toroid is the secondary, 5 turns arranged so as to be as far apart from the outer primary as possible, and as far as possible crossing at right angles to the windings. Primary is 4.5 turns, probably driven with a sine wave drive to keep the noise low, and probably is regulated on the secondary side by some further switching regulators to give the ADC and buffer stage supplies, after some full wave rectifiers to give 2 symmetrical supplies.

[max_torque]

silly question, but, given you only care about CMRR when actually making a measurement, and we know the high speed sampling means the device cannot actually stream continuously, then perhaps a (super) capacitor could be charged during the non measurement time by a simple relay connected DC/DC ??  Given the low power requirement for the front end, you wouldn't need that much "on time" to charge it?


(ie in effect, during your measurement period your front end IS battery powered!)