DMM7510 Coax transformer

[Echo88]

Do we have someone here on the forum who owns a DMM7510 and would be willing to make better pictures of the included coax transformer?
Maybe it can be lifted out of the DMM without destruction and also some measurements done on it, for example determining the switching frequency, resulting secondary noise, cable type and ferrites used.

References:
https://youtu.be/uvgJ2zAxgAY?t=992
https://data.epo.org/publication-server/document?iDocId=4725394&iFormat=0

[Kleinstein]

The transformer really looks like the drawings in the patent, especially fig 7

Given that they have a low number of turns (looks like 2 turns for 6 V) on the secondary (the enamel wire), chances are the coax is more normal with a single center conductor and not a shielded twisted pair.  So the transformer would be like  1+1 turns on the primary and 2+2 wires on the secondary (possibly even als just 1+1 with 2 wires in parallel).

Chances are the high voltage are made with a charge pump - that is at least how I have it with my DCDC converter. 2x6 V is reasonable for a charge pump to get something like +-6 V +-11 V and +-22 V.

To ease the soldering I would no be surprised of coax is even a fancy PTFE isolated one.  The cores are quite large in cross section (the corres add up) and thus likely not a high frequency needed.  This also depends on the core material - especially of they use high saturation nonoperm and not regular ferrite. There are quite a few electrolytic caps - suggesting more like a low frequency (e.g. 20 kHz not to be hearable) for the charge pump.

The coax outer conductor would be the shield (output side) winding. There is no primary shield, but with the coax the symmetry would be pretty good.

[Echo88]

My guess so far:
Triaxial cable made from PTFE/similar to make it unproblematic to solder the shielding to the pcb like you said.
Looks like ferrites, i would expect smooth plastic cases around the cores if they would have used Nanoperm/Vitroperm.
Drive frequency 20-100kHz.

After reading the patent it seems to me they claim theres no displacement current since all windings/shields would have the same voltage across them. Does that sound correct?

Also they mention that the shield braid covering factor plays a big role, so id assume that semirigid triaxial cable would be best in this case.

Why they claimed that this transformer variant has lower leakage inductance i dont know.

Time to build a prototype 

Edit: Question: Does semirigid triaxial cable even exist or is there only the standard coax variant?

[Kleinstein]

There should be essentially no displacement current, as the coax / triax cable does not really see the magnetic core. It behaves just like a straight cabel, so just the normal coupling capacitance between the 2 shields. The induced voltage is the same in all layers and thus the same voltage difference along the cable.
 
Triax would have a slight advantage, but for the first test just 2 x coax could be good enough, especially if the winding is symmetric.

I don't think the shield coverage should be so critical. With 2 x coax or triax there are 2 shields and the parasitic coupling capacitance through the shield would be small. In addition it would be rather constant over the length and thus not much net effect.

A 1 st rest could use a relatively low voltage or higher frequency if one does not have a large core.
I would start with something like rg179 or RG316  which are reasonably avaiable PTFE isolated cables. Just make sure not to get the version with steel core, as the resistance does matter.

I am not so sure if the leakage inductance is so low. It depends on the wires used. It would be low with a kind of shielded twisted pair.
The relevant leakage inductance is between the halves of a split primay. Leakage inductance to the secondary may even have a positive effect.

The required frequency / possible voltage*time can be calculated from the core dimentions.

[niner_007]

Is it really a transformer, it might just be an inductor

[Echo88]

Nope, its a transformer as according to the linked patent.
Short test with a triaxial cable and ferrites that i had on hand confirms it works as expected (no shields connected so far), see attached picture.
More tests regarding common mode noise need to be made after i layout a suitable test pcb for the used LT1533.

Since the LT1533 is a push pull driver i needed to make a coil tap, which isnt really doable with a triaxial cable since i cant access the inner shielding easily. Could of course use another IC to omit push pull topology, but i like the adjustable slew rate very much.

[jbb]

It’s a completely different approach, and I haven’t used one (yet :-) ) but TI have released an interesting new chip, the https://www.ti.com/product/UCC25800-Q1]UCC25800-Q1[/url].

It’s a little driver IC for LLC resonant auxiliary supplies. It should do soft switching (reduces dV/dt) and actually wants leakage inductance.  Wuerth makes a transformer (part 750319177) where the primary and secondary are spaced wide apart and stray capacitance is <1pF. The combination of limited dV/dt and very low capacitance might be helpful.

[openloop]

It should do soft switching (reduces dV/dt) and actually wants leakage inductance.

If I remember correctly, leakage inductance causes severe voltage spikes at the output rectifier due to reverse recovery.

[Kleinstein]

The LLC converter is a different concept than the hard push pull driver like LT1533 or SN6505. One could still use the same type of transformer, just with an additional inductor on the output side. The UCC25800 is so far only preproduction and limited availabilty.


The transformer in the DMM7510 has 2 coaxial windings and this may be to get the center tap for push pull driving.

With the relatively high effort for the windings it would make sense to use the same coil for both directions. As the drive voltage would be relatively low anyway, I would prefer a half bridge driver. The limted slew rate is nice, but there are alternatives to this (e.g. drive from a power OP).

[Echo88]

Voltage spikes produced by Diode reverse recovery time is eliminated by a snubber across the transformer secondary in the end.
The SN6505 does spread spectrum and slew rate limiting, but it seems the slew rates are quite fast compared to the 1.3µs slew rate for lowest noise with the LT1533?
The suggested UCC25800-Q1 seems really nice, i cant really remember what the reason was that bothered me when i looked at it last time in this thread: https://www.eevblog.com/forum/projects/favorite-high-isolation-capacitance-dc-dc-modulestransformers/msg3628930/#msg3628930

https://www.ti.com/product/UCC25800-Q1
https://www.ti.com/lit/ds/symlink/ucc25800-q1.pdf?ts=1636631618327&ref_url=https%253A%252F%252Fwww.google.com%252F
https://www.ti.com/lit/ug/sluuce8a/sluuce8a.pdf?ts=1636700531220&ref_url=https%253A%252F%252Fwww.startpage.com%252F
https://www.ti.com/lit/ds/symlink/sn6505b.pdf?ts=1636699043396&ref_url=https%253A%252F%252Fwww.google.com%252F
https://www.analog.com/media/en/technical-documentation/data-sheets/1533f.pdf
https://www.analog.com/media/en/technical-documentation/application-notes/an70.pdf Page 8 lists the noise vs. slewrate

[Kleinstein]

The spread spectrum of the SN6505 is very anoying, as the modulation frequency is quite low (some 50Hz). The slew rate looks quite fast, but so far may circuit works. It may also work with a split bobin. Leakage inductance only has to be low between the 2 parts of the split primary. The other problem is the currently poor availablity.

The UCC28500 is currently only available from Ti directly. The need for the correct amout of series or leakage inductance can be a bit unconvenient for a small quantety. The choice of the transformer is not so easy. Here the simple unregulated forward converters are easier.

[Echo88]

Forgive me, i know this thread is two years old.
Just hoped someone in the meantime got a DMM7510 which might lend itself to a teardown for detailed pictures of the coax transformer, layout and driverchip, but of course i understand the refrain from opening a calibrated and working 7.5 Digit DMM.
So far i havent come across a damaged DMM7510 or other info sources.
Can trade against teardown pictures of a nice Digistant 6706 or Thorlabs TED4015...
After that i could do a suitable layout for it, like im trying right now for the pickering transformer: https://www.eevblog.com/forum/projects/lt1533-lt3045-low-noise-dcdc-pcb-suggestions-needed/msg4743629/#msg4743629

[Kleinstein]

Dave has a video with a few pictures of the transformer.  It looks like the transformer is rather close to the patents.  The clock for the converter may go a bit unusual way, with the clock coming from the main ADC side. This makes some sense to have the DCDC in sync with the ADC.

[Echo88]

He did also make a few good photos of it from the outside, sadly no detailed ones as the transformer is encased/is a module?.
https://www.flickr.com/photos/eevblog/albums/72157651830909112
I have to look up my measurement results from the last attempted build two years ago or so.
The new pcb should at last make it easier to test such stuff.
Cant hurt to have detailed info on how Keithley did it though, the used ferrites would also be interesting.

[Echo88]

I did some more delving into the transformer patent and tried to get as much details from the photos/video of daves teardown.
Here are my thoughts:

-they appear to use the transformerstyle like in Figure 9 of the patent, with the detail [022]
"In another embodiment, shown in Figs. 8 and 9, an isolation transformer is shown using two triaxial cables wound in parallel around a magnetic core. One triaxial cable is placed on the top of a circuit board and the other triaxial cable is placed on the bottom of the circuit board. For ease, the triaxial cables are referred to as the top cable and the bottom cable."
while Figure 8 uses all 4 external wires for the secondary and in this case the external wires are apparently used for two bipolar secondaries

-they appear to have two independent bipolar secondaries, one for marked +-6VAC with two diodes and one capacitor and the other unmarked but surrounded by lots of diodes and capacitors (6 Diodes and 7 caps visible on each pcb side). The diodes sit on one side of the board and the caps beneath them on the other side, the pcb is very symmetrical.
I assume its a voltage cascade stage due to the assumption that the "zero displacement current" is only working when all conductors/shields on the transformer have equal amount of turns, limiting the achievable secondary voltage based on the primary voltage...but maybe im completely wrong here.

-id assume they use the LT1683 PushPull-driver as its having the necessary drive power and uses external mosfets like mentioned in the patent, also there are some low esr caps visible in the vicinity of the assumed driver.

-ferrite type is unknown and complete core length is ~6cm, outer diameter of each core is ~3cm, inne diameter is 1cm, i would assume nanocrstalline high Al cores would be best here to achieve low frequency usage, but that would be expensive

References:
https://youtu.be/uvgJ2zAxgAY?t=992 here the diodes/caps can be seen and the supply-footage starts
https://youtu.be/uvgJ2zAxgAY?t=2938 Pushpull driver caps visible
https://youtu.be/uvgJ2zAxgAY?t=2840 supply-footage continues
https://youtu.be/uvgJ2zAxgAY?t=3047 backside visible with +-6VAC marking
https://data.epo.org/publication-server/document?iDocId=4725394&iFormat=0

[Kleinstein]

The choice of the core depends on the frequency. The size of the core with multiple rings and thus quite some "iron"-area helps a little.  With the rather limited number of turns that can be used with the coax I would expect a somewhat higher frequency. The nonocrystaline material is somewhat limited in the frequency (quite a bit below 100 kHz, maybe 30 kHz), though it can use a high field.
For the core material I would expect high µ ferrites, like those used in CM chokes. So nothing very special.

The equal turns for primary and seconday should only apply to the shield. So with a cable with more conductors one could have a turns ratio like 1:2 or maybe 1:4.
With a relatively high frequency one could use a charge pump circuit on the output side to get higher voltages without too much effort. The size of capacitors may give a hint on the frequency for the charge pump.
For my DVM board I have a 2x5.5 V transformer output and than get +-21 V from a charge pump and +5.2 V directly.


AFAIK the DMM7510 uses some RF coax cable with PTFE isolation to ease soldering. For a few more turns one may consider a shielded LAN cable to get something like 4 pairs with a shield.

[Echo88]

Jep, high µ ferrites and quite a bit higher frequency compared to the pickering variant should do the trick here, even with a LT1533
They mention two and three core double shielded cables in Figure 2 and 4, i have never seen such cables before.

Im thinking about isolating regular semi rigid coax on the outside and then putting it in a soft copper pipe to produce semi rigid triax cables, as this would provide 100% screening compared to regular triax cable. Have no idea if that´ll work mechanically though.

Attached is an adapted pcb-version. I hope i got all pads for the windings in the correct order.
As this transformerversion isnt shielded by regular grounded shields theres no possibility to bypass CM-currents with a CM-capacitor in parallel to the transformer as fas as i can see.
The ferrites are attached via zipties, therefore the rectangular holes next to the ferrites.
So far no shield above the transformer.

[Echo88]

https://youtu.be/9JinSfCKuNQ?t=1005
While casually watching Marco Reps video about low leakage DCDCs again ive seen another coaxtransformer-implementation that i missed/forgot about the first time i watched it.
Branadic found out its the mainboard from the Keithley 2450 
https://xdevs.com/doc/Keithley/2450/img/20180414_180031.jpg
https://xdevs.com/doc/Keithley/2450/img/20180414_180128.jpg
https://xdevs.com/doc/Keithley/2450/img/20180414_180148.jpg
https://xdevs.com/doc/Keithley/2450/img/20180414_180227.jpg pcb-backside with FFC blocking the view
https://i.ebayimg.com/images/g/zjMAAOSwoNVh3vID/s-l1600.jpg a K2450-board offered by ebay-seller "elektrorus2023" (no affiliation) were the backside if visible
https://i.ebayimg.com/images/g/4yUAAOSwoJxh3vEc/s-l1600.jpg another offered K2450-board with better visibility and butchered coaxtransformer
Seems to be a Variation of Figure 7 in the patent: Triax with middle tap used as primary and a coax used as secondary.
Topology and driver chip unknown so far.

Another special ferrite transformer variation can be found in the Keithley 6220/6221
https://xdevs.com/doc/Keithley/6221/photo/kei6221_nake.jpg
In this case a UC3825 + mosfets as transformer driver.

And another one in the Keithley 2636B https://xdevs.com/doc/Keithley/2636B/20140426_212414.jpg

[eplpwr]

Regarding the eBay pictures of a K2450 board, I think the seller is wrong and it's a K2460/2461 board; the 2450 has triax rear connectors but the K246x has the orange connector pictured. That would imply that the special transformer is in the 100W+ range, so a bit beefier than the K2450.

[MegaVolt]

When repairing the k2612 power supply, I took photos and drew diagrams. Maybe this will be useful. (Scheme is at the beginning of the topic)

https://www.eevblog.com/forum/metrology/keythley-2612-teardown-and-repair/msg3231840/#msg3231840

[branadic]

Echo88, do you already have first results that you would like to share?

-branadic-

[Echo88]

  branadic wants me to spill the beans

Some time after starting the CM-measurements on the Potcore-pcb-design i noticed same flaws in the measurement, therefore i`ll correct those and do the measurements again. For this whole project i'll open another thread then.
Corrections for the next measurements:
-take scope-screenshots with different bandwidths for every configuration, as the perceived noise differs between scope and SA (like done in https://x.artofelectronics.net/wp-content/uploads/2019/11/9xp14_low-noise_isolated-pwr.pdf
-measure every configuration with SA, scope (in 20MHz and Full BW) and HP 3403C RMS meter
-using a Hammond case to shield the measurement completely while using BNC-feedthroughs (no simple BNC-jacks with attached coax-cable wy pigtails which worsens BNC-shielding) as measuring the pcb on the table introduces external noise in the more sensitive ranges

DM-noise isnt measured right now, since i would first need to repair my LNA for that. I will do it later though.
LT3439 is now used instead of the LT1533 since it can handle the 12Vin (roughly, no exact requirement) 12Vout @ 200mA spec, without overheating when the slewrate is increased to 1µs or more.
I kept the LT1533 in the potcore and Pickering design though since its still usable when drawing less output power.

I attached the measurement results so far (missing scope/RMS meter measurements), schematic for the potcore design and the gerber files for the three test pcbs which use different transformers: P26/P30 potcore transformer, Pickering/Teflon transformer and Triaxial/Coaxial transformer (like in the DMM7510).
The potcore-design works and is tested, while i havent yet received the Pickering and Coaxtransformer-pcbs and therefore cant guarantee anything here.
Maybe someone wants to join the project and replicate my measurements or better yet: one up me and produce even lower CM-noise results by using better techniques 

[branadic]

Maybe someone wants to join the project and replicate my measurements or better yet: one up me and produce even lower CM-noise results by using better techniques 

As you know I have replicated the DC-DC converter with Pickering transformer and LT1533. I did some very first measurements on it before, powering it from a linear wall adapter and using my DSA (Advantest R9211E), but that measurements were different to yours:

Used measurement equipment:
Siglent SVA1032X used as Spectrum Analyzer, connected directly via 0,5m BNC-cable to the 50R CM-noise measurement shunt J3
Siglent SDS2354X Plus Oscilloscope used to determine the Slewrate of the LT3439, which is measured with a ground spring directly between a transformer primary pin to primary-ground
Battery Yuasa 12V 7Ah

So if we agree on how to measure things (simplified measurement instruction scheme), I'm happy to contribute.

-branadic-