X-ray images of precision resistors

[branadic]

Hi,

this thread is related to X-ray images of precision resistors.
All images are free to use by anyone License
First some G.R. 8G16D, a complete LTZ1000 set.
Construction: Evanohm supposed to be spot welded to Alloy180, Alloy 180 connected to copper leads ...

-branadic-

[branadic]

Next, some images of Burster resistors I got from Alexander B.

Type 1148 +-0.10% 50 Ohm
Type 1152 +-0.01% 247,038 Ohm
Type 1168 +-0.1% 1 Ohm
Type BP 2A 8/78 +-0.05% 3 kOhm

Construction: Manganin or Isaohm brazed? to copper leads

-branadic-

[branadic]

Here are images of UP805, also a complete LTZ1000 set.

Construction: Evanohm rod butt welded to copper, Evanohm wire CD welded to Evanohm rod

-branadic-

[ataradov]

Is there a specific reason they separate them into two sections?

[SeanB]

I would guess so they can wind them in 2 opposing directions, to minimise the inductance of the unit.

[Twenty4Pack]

I've wondered how these were constructed - absolutely fascinating - thank you very much for posting the images.

[ataradov]

I would guess so they can wind them in 2 opposing directions, to minimise the inductance of the unit.

Yes, this makes sense.

[branadic]

I would guess so they can wind them in 2 opposing directions, to minimise the inductance of the unit.

...One other 'myth' I'll pop while I'm at it, the misbelief that reverse winding the pi has a significant effect on overall inductance, it doesn't, measurements indicate that while there is a small reduction in overall inductance, it is not really that significant when it comes to AC signals.  The coupling between pi is quite loose and therefore has little cancelling effect because they are next to each other in series so the flux lines do not cancel significantly, not to mention that the Q of resistors is exceedingly low...

-branadic-

[ataradov]

Even if it has very little effect, why not do it? It is not like it will increase the price significantly.

[babysitter]

By a magic numeric procedure you split the number of turns that you would need otherwise in half, for faster processing and easier counting ! 

[glarsson]

The image 8G16D-12kR is an odd one. There are clearly not an equal number of turns on the two "coils".

[branadic]

By a magic numeric procedure you split the number of turns that you would need otherwise in half, for faster processing and easier counting ! 

My best guess is that this is done for symmetry reasons. They split the length of the wire into two equal halfs, both half of the bobbin are wound simultaniously starting from the middle of the wire.

-branadic-

[branadic]

The image 8G16D-12kR is an odd one. There are clearly not an equal number of turns on the two "coils".

As already mentioned in the T.C. thread this is because the 12k were fabricated by stripping some 18k resistors.

-branadic-

[Edwin G. Pettis]

To Branadic......thank you for quoting me but you forgot to give credit where credit is due.

The reverse pi winding myth has been around a very long time, mainly a marketing thing.  Actually, the real reason for multiple pi is for voltage ratings, the enamel coating on wire is quite thin by necessity, therefore using more than one pi reduces the voltage applied across each pi and reducing the likely hood of punch through of a pin hole in the enamel.

Bifilar winding is not commonly used in PWW resistors, it is not friendly to production, it is used in some special instances.

The unbalanced resistor from G.R. was much more likely a 24K unit, as one of the pi is practically empty leaving 12K, if it had been 18K, about half of the wire would have been left in the near empty pi as they would only have needed to remove 6K from it.


I'm with holding further comment on the Bursters at this time.

[741]

Hi

Nice, clear images 

Would it be possible to do this for chip ceramic capacitors? - I recall some feature on this forum about cracks in them. It would be interesting to see the plate layer stacking, but tricky to get a good picture, as they are such small devices.

PS - What is your equipment for taking these X-ray photos - Old CRT and real photographic film maybe?

[Magnificent Bastard]

Is there a specific reason they separate them into two sections?

There are 3 reasons that I can think of:

1) Multiple 'pi' sections reduces inter-winding capacitance.

2) Multiple 'pi' sections increases the ability of the resistor wire insulation to withstand higher voltages.

3) A slight reduction in inductance (assuming here that the wire is wound in opposite directions alternately between sections).

A bifilar winding would reduce the parasitic inductance, but would it would also maximize the inter-winding capacitance and would reduce the maximum withstand voltage of the resistor.

Within reason, the more 'pi' sections, the better; but the number of sections should always be evenly dividable by 2.

[Edwin G. Pettis]

The Burster resistors are solder terminated.  While I noticed Burster puts a top temperature of 85°C operating (at zero watts), they are really only rated to 60°C, wattage plus ambient, in short they're meant to be operated at room temperature, especially if running closer to rated power.  What happens to Manganin above 60°C?  The TCR increases rapidly and sustained temperatures above 60°C can cause permanent changes in TCR.

[branadic]

Would it be possible to do this for chip ceramic capacitors? - I recall some feature on this forum about cracks in them. It would be interesting to see the plate layer stacking, but tricky to get a good picture, as they are such small devices.

In principle yes, as the voxel size is <10µm. But I currently won't make images of chip capacitors, as I think there is enough material shown on the web.

What is your equipment for taking these X-ray photos

It's a HMXST 160, X-TEK Systems Ltd., Tring, Herts, UK.

-branadic-

[Conrad Hoffman]

I like manganin because I think it's easier for a DIY person to get a decent result with it. Not as good as the 800 alloys, but there are many pitfalls there. There's lots of NBS/NIST history on the use of manganin. Or, if you don't like resistors, apparently it makes a dandy explosive blast sensor. On the reverse winding thing, I've made measurements and agree the benefit on inductance is minimal at best. There are way better geometries for low inductance, but probably not suited to minimal size and sensible production methods. Something (IMO) really interesting is you can buy inexpensive "non-inductive" power resistors for audio use that have two paralleled concentric ribbon windings. They seem fine on the surface, but interact in such a way as to dramatically increase THD at high frequencies. Not sure, but I think one of the windings acts as a metallic core for the other and saturates or goes non-linear somehow. Love the X-rays. I had heard a story that somebody (the military?) X-rayed some Julie resistors and rejected them for having contamination.

[branadic]

Obviously kWeld is a good solution to butt weld and spot weld, if some additional effort is put into it:

https://www.keenlab.de/index.php/portfolio-item/kweld/



-branadic-

[MisterDiodes]

That weld looks like it's only barely done...I know it's just a first attempt, but have you tried pulling that wire off?  Does the wire or the weld break first?  If the wire pops off then that was not a join at all.

Looking at the weld from the top is only a very small part of the story - You really have no idea unto you slice that joint open and look for the actual weld penetration and metal grain structure at the join zone...  It all gets a -lot- trickier with resistance wire.

[branadic]

This weld isn't made by me, as you might already expected. And you can be sure that if this would have been made by me, that I would have already performed a pull test to measure the force nessecary to lift the wire off as well as an analysis under the microscope of the weld interface itself. And I wouldn't have used a steel rod nor a copper wire.

-branadic-

[babysitter]

Clearly a case of Hahn*********ism ?

[branadic]

Clearly a case of Hahn*********ism ?

No, scientific acting

[Edwin G. Pettis]

I knew the photo was from the keenlab.de website since I went there out of curiousity, as I've said in the past, you cannot CD weld two dissimilar metals together and get it right.  I am talking large differences in tensile strength for one thing.  That copper wire is merely stuck (mashed) on there because of the heat and weld pressure, it will come off without too much effort.  Looking at the interface between the copper and iron will show the lack of a true weld.  That photo is a rather poor choice for showing their welding capabilities.  Their CD welders are probably just fine for welding similar metals like any other CD welder.

This is nothing new, all of this trial and error happened decades ago when dental braces came into being and they needed a welder to do the work.  CD welding was tried on Evanohm >70 years ago with I don't know how many different combinations of metals trying to match it to standard copper leads, at one time, even stainless steel was used plated with copper...ugh.  Sure it welded but made terrible resistors.  All of the 'successful' (to one degree or another) interfaces between Evanohm and copper leads has been in production for decades, you're just repeating history if you're trying to interface Evanohm and copper leads again, nothing new to find here.

[Magnificent Bastard]

All you need to attach the AWG#16 Evanohm "plug" to the AWG#20 copper wire is a DIY version of one of these:

http://www.percussionwelder.com/

I contacted the engineer that designed the machine.  He said that it is DC, not AC or RF.  So, reverse engineering what they are doing on their website should not be too difficult; here is a video introduction to the technology (and there also is a video playlist on this machine):



There is a mechanical device that positions the 2 wires together, with a lever that first moves them apart, then back together but with a small solenoid that "slams" the wires together at just the right moment.  The videos do not explain the process exactly right; I think they are hiding the details on purpose (they want to sell you a welder, not tell you how to build one).  If you are going to weld Evanohm to copper, a plasma-arc is absolutely essential-- nothing else gets hot enough to properly join the materials.  If this were a continuous process, then it would be called "TIG" [Tungsten Inert Gas] welding, but it is a pulse of a measured amount of energy; not a continuous arc, and there is no tungsten electrode (the 2 wires are the work pieces and also act as the electrodes); also there is a mechanical percussion that TIG does not have.  That said, much of the information on TIG welding dissimilar materials applies here also.

Imagine that you have a power supply that puts put a high voltage at very low current (a large resistor in series).  When the 2 wires are touching, this current flows between the wires.  When you pull the two wires apart a short distance, an electric arc forms; but because the current is very low (say, 10mA to 100mA) it really doesn't heat either of the wires very much; it is more of a "pilot arc" to keep the gas in between the wires (either air or an inert gas like Argon) at a low impedance.  Now, lets say the wires are about 2mm apart at this time, with this pilot arc happening.  If we have a large capacitor charged to a high voltage, and then suddenly (through a solid-state switch like an SCR) connect that capacitor to the two wires, there will be a powerful arc formed, and because of the very high current, the arc temperature can exceed 8000K.  Probably, the negative arc voltage should be connected to the Evanohm wire, with the positive lead connected to the copper wire.  This plasma-arc melts the ends of both wires, and at the right moment, the wires are moved back together mechanically, and simultaneously slammed together with the solenoid.  This entire process happens in 0.5ms to 2ms (depending on the requirements).  The switch between the capacitor and the arc is a high voltage high current SCR (which also prevents back-feeding the pilot current into the capacitor until the switch is turned on).  The energy in the capacitor is 0.5*C*V^2, so you control the energy both by controlling what voltage you charge the capacitor to, and the value of the capacitor(s) for the discharge.  The voltage for the pilot arc is simply a high (isolated) voltage (say 100V) going through a high-wattage resistor.  The SCR should be rated to 400V at very high amperage.  The capacitors should be multiple polypropylene film type, on the order of 100uF (but I haven't tried this or done the math; that may be too much or too little capacitance).  You can make a switched arrangement that will parallel 1-10 capacitors to get more range out of the setup.  If you are always welding #20 copper wire to #16 Evanohm wire, then the capacitor value and charge voltage can be fixed.  The capacitor charge current needs to be shut off before firing the SCR so that the SCR will eventually turn off.  Once the SCR turns off, the charger can then be reconnected to the capacitor (using some kind of relay).  There needs to be some kind of SCR driver that can provide a fast pulse to the gate of the SCR, and a solenoid driver too.  All of this could be controlled by an Arduino or something similar.  Oh; and because copper wire oxidizes rapidly at high temperatures, there needs to be an Argon tank, regulator, and solenoid valve that controls the shielding gas at the weld point.  The gas is only on just before the weld, and then is shut off just after the weld, so one canister of gas can be used to make many thousands of lead sets.  The video does not show using a shielding gas; if you don't use the Argon shielding gas, then the copper will oxidize, and it will still have a mechanically good weld, but it will have a lot of copper oxide in the weld which will create a high thermal EMF connection (which would be very bad).

The price for one of these welders is > US$10000 (and that's just for the butt-welder), so it must be a DIY project.

[MisterDiodes]

...None of which really applies to maintaining a precision low TC joint in resistance wire.  You don't want to "slam" anything together or create noise-inducing ridged heat affected zones around the weld.  The alloy distribution in the resistance wire must be remain absolutely undisturbed if you're building a precision resistor - otherwise the apparent total TC and noise from one weld can be many times higher than the rest of the resistor.

A much more delicate & controlled weld process is involved if you want to maintain low TC.

[Edwin G. Pettis]

To reply #25,

Yes, a percussion welder is what has been used for decades to make the lead assemblies, light bulbs, vacuum tubes and a host of other things that use dissimilar metals.  In concept, the percussion welder is relatively 'simple' but in actual use, things get rapidly complicated.  Technically, if you bought one of these setups, you could eventually make rough lead assemblies, quite expensive but it would work.  There are mechanical parameters which must be observed precisely if you intend on making lead assemblies for precision wire wound resistors.  Sloppy results isn't going to work very well in practice, ask me how I know.

The main reason most resistor manufacturers buy their lead assemblies from a company that makes them by the millions for lots of various customers, is the cost of the precision welding and handling equipment to make these assemblies in larger quantities, that is over $20,000 minimum and you need operators who can set them up properly to produce lead assemblies to spec. and that doesn't always happen, even well experienced manufacturers can have problems with consistency.  Of all the resistor houses I am familiar with, none of them make their lead assemblies in house.

There is a heck of a lot of details that goes into making a PWW resistor that requires knowledge and experience to execute properly, when you're trying to make resistors with TCR <<10 PPM/°C, the details and execution are exceedingly important.  I'm not trying to discourage anybody from experimenting but if you're going at it with the idea of making your own LTZ resistors, forget it, at the very least it will be way more expensive and the results will be disappointing with inconsistency bugging you all the way.  I'm talking from a position of history and knowledge of the resistor industry, you are simply repeating what has been tried and done before.

Experiment with alloys and materials that are friendly to experimenters and have fun seeing what you can produce, you just might be able to produce a bit less than 10 PPM/°C with Manganin or Cupron without trying to acquire much more expensive welding equipment and frustration.  You can buy a lot of PWW resistors for the cost of a welder even if it is DIY (don't forget your time should be worth something).

[Magnificent Bastard]

@Edwin:

Well, since nobody in the resistor industry is willing to teach us how to properly build a PWW resistor, then we have to figure this all out on our own.  It might take some time, but there are many of us working on this, and most of the engineers on this forum are intelligent; so if we simply put our minds together, we can develop a process that is inexpensive and friendly to hobbyists to build a few PWW resistors ourselves.  I don't think anyone on this forum (other than yourself) wants to be in the resistor business; but it is very satisfying (and more rapid) to do things ourselves.  Also, we can try things that the main stream resistor houses would not want to do, because of lack of volume production.

Yes, we know it will be difficult, and really scarrrrrry to build our own resistor leads, and to make our own bobbins, and to wind our own resistors.  We will get it wrong; (a lot!); but that's OK as long as we don't make the same mistake twice.  If we never try, then it is certain that we will never know how to do it.  By sharing our experience and knowledge in this forum, the learning process can accelerate exponentially.  We will get there; you will see!  Thank you for sharing the knowledge that you have (even if you do appear to hold some information back for business reasons).

[branadic]

Managed to make some further x-ray images of resistors.

-branadic-


Some additional comments: So what is all that Ultraohm Plus myth about?

In The last half-century: Wirewound resistors Part one and The last half-century: Wirewound resistors Part two we can read about the history of resistors and the position Edwin G. Pettis is claiming for Ultraohm Plus resistors.

In the Ultraohm Plus "datasheet" we can read the following:

Proprietary design and manufacturing procedures allow us greater freedom in build and greater reliability in the end product. Our materials are chosen for use, purposely not matched, as other manufacturers claim. They compare apples and oranges to show that they're matched - we know they can't, so we purposely mismatch. Our design is welded, else it can't get through the process - theirs is a mechanical joint which will pass all testing in the plant, then fail in the field. If our part isn't welded, it can't leave the manufacturing area. The design is self-inspecting. An open crossed-wire weld system, the joint can be inspected visually and electrically, and if it's not made, the part is scrapped. It's strictly a go-no go process. Not so with the competition - they make a mechanical joint first, then pretend to weld, and let the mechanical component take over. That is the reason for MIL-STD-202, power conditioning, and thermal shock testing. It's really shocking the number of non-welded partss, just mechanically fastened that pass these tests. If they show large shifts in resistance value (∆R's), then the manufacturer's answer is to elevate the bake temperature, bake them for a longer time, etc. to stabilize the wire. In fact, the wire is already 1,000 times more stable than the resistor unless it's been wildly mistreated. The error is in the connection of the resistance wire to the lead.
This fact has been proven by ultraohm plus. Our "stabilization bake" requires 8 hours @ 150 degrees C. It does not remove stresses from the wire. Instead, we try to stress the wire above what the customer will ever use it. The upper operating limit of use is 145 degrees C. Once we've stressed the wire over the limit, the user can't touch it and his operating stability is guaranteed. For those customers operating over the full MIL range, we'll cycle a part 5 times through the thermal shock cycle, and once prestressed, again it won't fail. The design has been tested and refined over a period of 15 year, and is in use in numerous units in the field. It has been tested by major users of resistors in jet engine fuel controls, and found to be superior.
Power conditioning (burn-in) is redundant testing in our parts - a waste of time. They're either good or they're not. Final inspection of plain old DCR will tell you immediately - no further testimg required.

Please feel free to judge the construction of the different resistors yourself, how the wire is attached to the leads and how strain relief is realized by the x-ray images given above. E.g. 8G16D shown above and the construction explained in their datasheet uses strain relief and what looks like a proper weld.
Use the search function of the forum to read about reliability of Ultraohm Plus resistors and experience of forum members with them too. Once the resistors left the factory the manufacturer can't guaranty for anything and it is known that the welded joins tend to break when bending the leads due to missing strain relief.

In Group buy: LTZ1000 precision resistor set we can find the following quote:

Some may point to the LTC5400 chip for very low tracking TCR, to some degree that is quite accurate but the type of resistors inside have wide tolerances from chip to chip, significant noise and 1/f noise as well, PWW resistors have neither.  So there is a tradeoff for getting low TCR with higher noise and you also have more trimming to deal with because of tolerance and limited values.

Here is the full text for completeness:

I can appreciate the problems with shipping to various countries especially in the EU, for customers outside of the USA, the average 1st Class shipping charge is about $14,90, domestically it is about $4.15 for 1st Class.

If your group does a ‘bulk’ purchase, I will extend the discounts to everyone based on the total of each line item ordered, the minimum quantity for discount is 10 and that is 15%.  It would help if everyone contacts me directly through my e-mail (pettiseng@q.com) rather than the blog’s system.

I will leave it up to you and the group to determine which [e.g.] two weeks they want to order in; of course I’ll need to know the dates.

1.    Everyone will need to identify to me that they are part of the group purchase.
2.    I will determine the discounts based on the final tally of values ordered.
3.    I will notify everyone of the final pricing and individual shipping charges shortly after the [e.g.] 2-week period ends.
4.    Most customers use PayPal, my account is: pettiseng@q.com
5.   Production time will depend on the final quantities; I will apprise customers of the expected time.

Current base pricing for type 802 resistors, 1-9 is:

120R0   $7.19
1K      $6.65
12K      $7.72 [~45°C heater set point, only for NON A version at roomtemp]
12.5K   $7.85 [~52°C heater set point]
13K      $8.03 [~60°C heater set point]
70K      $9.07

±0.1%, ±3 PPM/°C maximum, 0.250” D x 0.375” L

I do not recommend using anything less than 12.5K for the LTZ circuit, you will not achieve any significant decrease in aging characteristics and it is better to have more than a few degrees overhead for temperature control.  It is the internal temperature of the LTZ that you are regulating, not the external environmental temperature; the LTZ is a power in – power out device.

Technically, it would have been best if the temperature ratio pair had been put next to each other on the PCB so they could be better thermally coupled but since that isn’t possible, thermal air drafts will need to be blocked to minimize differences.  NOTE: do not encase the LTZ to the point of no thermal exchange with the air as this will cause the LTZ to exceed the internal regulation loop and it will lose thermal regulation.  This is a known characteristic; you need to block air drafts around the leads, both top and bottom but not the body.  Do not use any hydroscopic material inside the LTZ box, you don’t need anything that absorbs humidity and holds it.  Of course you can also include a package of silicate to help with that.  I am not saying that any particular components are humidity sensitive; it just isn’t a good idea to have humidity building up.

A voltage booster circuit is much more complicated than the LTZ board; many things beyond the resistors contribute to the output voltage.  Thermals of course, the op-amp, the op-amp amplifies thermals and TCRs, thermal coupling between the resistors, and noise to name a few things that will affect the output.  While I have made resistors for this purpose that has exceptionally low tracking TCR, it took exceptional effort to achieve in the short term.  Tracking TCRs down to 0.1 PPM/°C were achieved but this was for an industrial customer where the effort and cost were justified, the customer actually did the final processing because he had the equipment to do it.  He also went to some extreme to minimize thermals.

For the average user and some reasonable processing, you could expect something around ≤0.5 PPM/°C or so, if the design of the booster circuit is done correctly and carefully you could achieve very good thermal equilibrium thereby minimizing further the TCR of the output.  One possible way is to use a small oven to stabilize the booster circuits so that temperature becomes a minor problem.  The voltage booster circuit could easily end up costing more than the LTZ circuits.  Look at the Fluke 732A/B to see the lengths they went to to stabilize the outputs, quite complex.  The LTZ board does not need this sort of effort because it is basically internally heat regulated and is not aided by external ovenizing.

Some may point to the LTC5400 chip for very low tracking TCR, to some degree that is quite accurate but the type of resistors inside have wide tolerances from chip to chip, significant noise and 1/f noise as well, PWW resistors have neither.  So there is a tradeoff for getting low TCR with higher noise and you also have more trimming to deal with because of tolerance and limited values.

You are correct in that there will be a power imbalance between the ratio resistors in the booster circuit (just like the heater circuit in the LTZ), just having resistors with the same TCR will not get rid of that problem, that is where some of the thermal problems come from, they require very good thermal coupling between them to start with, such as copper tape wrapped around them.  If you decide to trim the output, there will be additional resistors to contend with including a trimmer, this should be a Bourn 3290 for best stability (yes you have to shop around, many places are asking a high price for it but you can find them at lower pricing if you look hard enough, the Vishay trimmers are inferior no matter what the specs claim).  If you can tolerate a more fixed output then you can eliminate the hassle of trimming, the drawback is you will need to calculate the booster ratio resistors values to a closer tolerance (no you really don’t have to go to ±10PPM).  Your resistor values will depend on your circuit design and what you want at the output so I can’t really advise on specific values, just how to implement them.

You do not need to have exactly 10.0000000Vs at the output to calibrate with, anything fairly close will do as long as the value is known, whether it is a manual or automatic adjustment calibration.  The stability is really more important which brings me to another topic, long term drift.  If you haven’t read the post I put in metrology section for definitions, it is a good place to start.  All freshly made resistors have drift which tends to decrease over time naturally, depending on the user’s patience, one can either wait some months for the drift to settle down or purchase the so-called PMO or enhanced processed resistor which takes some of the initial drift out of the resistors.  On average, again depending on circumstances, resistors will settle down to fairly low drift after about a year (it might be more or less) to low single digit drift without PMO, depending on the enhanced process, it could remove the lion’s share of the initial drift to possibly single digit drift per year.  Note that your LTZ is going to take at least that long or longer to really settle down to its lower drift rate per year.  Precision takes time and in many cases it cannot be rushed as rushing can have its own long term consequences.

For the two resistors [for 10V boost] you mentioned, I would suggest ±0.01% tolerance:

4.8K      $10.04
12K      $10.13

Ultrohm Plus
125 Vista Grande Dr.
Grand Junction, CO 81507-1427
970-242-4929
pettiseng@q.com

As proven by Nikolai Beev in "Measurement of Excess Noise in Thin Film and Metal Foil Resistor Networks" but also measurement results provided by him here with the following image being share with the forum:



some resistor networks including LT5400 exhibit extremly low 1/f noise <-60 dB or even <-70 dB, while at the same time they provide extremely good t.c. tracking. There is the myth that precision wirewound don't exhibit 1/f noise. While this can be true in principle for some special treated resistors, this isn't true in general. There are many contributors to 1/f or excess noise, such as impurities in the wire, cracks/strain in the wire from winding, noise from joining the resistive wire to the leads, ...
Most of the myth that wirewound resistors exhibt no excess noise is coming from a time when measurement capabilities were limited and excess noise was measured as defined by MIL-STD-202-308. With modern test equipment, capable of measuring very low signal levels over a long period of time one is able to spot even very small noise indices. Stay tuned for results that are coming soon.