[Halld-cal] FCAL CW bases

[Paul Smith]

Hi Fernando,

I have no objection to a board redesign which moves the diodes further  
apart.

I tested one of the 2225 AVX capacitors with small wires soldered to  
the ends and just sitting in the air with no board or coating.  It was  
soldered with water soluble flux which was then cleaned with distilled  
hot water under pressure and then ethyl alcohol.  There was never any  
flux on the faces or sides, only a little on the end caps where the  
wires were soldered.  So, it should be extremely clean.  It sparks  
somewhere between 1500 and 2000 volts.  Sparks are across the sides  
between corners.  If you look at the end cap metallization, it is not  
exactly smooth at the boundary where the metallization meets the  
ceramic.  After a spark, evidently a sharp point gets burned off since  
it will then withstand a slightly higher voltage before sparking  
again.  Humidity was ~55%.

The next thing I plan to try is epoxy-ing the diodes and humisealing  
the capacitors on one of the existing boards.  I think it is important  
to learn as much as possible about the problem areas before redoing  
the board layout.  If this works we can consider avoiding the use of  
epoxy.

I sent a pdf of the present board layout a few weeks ago.  I can send  
it again if you don't have it.  I can send the new layout when I have  
it done.

Paul



On Jun 21, 2010, at 1:54 AM, Fernando J. Barbosa wrote:

> Hi Paul,
>
> As we have discussed in the past, placing the diodes farther apart  
> (larger than the diode's pad-to-pad length) will prevent arcing on  
> this part of the board (the limitation will then be within each  
> component rather than component-to-component).
>
> There are a number of different types of epoxies and for different  
> applications, as you know. The problem with epoxy is that it does  
> not allow for re-working the circuitry, over time it de-laminates or  
> cracks (especially with high levels of radiation but this is not the  
> case here) and it is impossible to match the CTE of epoxy to that of  
> ceramic. Because epoxy is also slightly hygroscopic, arcing will  
> occur shortly after de-lamination. You are correct that thick coats  
> of epoxy are used on some components for protection from mechanical  
> stresses and environmental factors but its purpose is not to prevent  
> arcing, even though it may be rated for HV operation.
>
> The proper coating for HV should be acrylic-based (like the Humiseal  
> you have used) or Parelene. I prefer Humiseal because it cures  
> quickly, has very little out-gassing, is excellent in high humidity  
> environments and it allows for re-working of the circuitry.
>
> The problems I have encountered that lead to arcing have usually  
> been due to surface residues (cleanliness is the most important  
> issue), improper pad spacing (the pad-to-pad spacing[edges] under  
> the component should be about the same as the component length,  
> sharp edges, component spacing and defective components, in this  
> order. Of course, humidity is a big factor. IPC does not address HV  
> conditions properly, in my opinion, as it is concerned primarily  
> with assembly of PCBs for most commercial uses.
>
> We had some problems with HV in the FDC and the vast majority of  
> these were due to cleanliness. It turned out that the epoxy that  
> held the wires in place was contaminated from cleaning the solder  
> residues which also softened the epoxy; one component was placed too  
> close to another; residues on the PCB also caused quite a lot of  
> problems. These problems have been resolved by proper cleaning and  
> we have re-designed the boards more conservatively. We have also  
> applied a conformal coating of Humiseal to HV capacitors outside of  
> the gas volume because of humidity.
>
> The HV capacitors are 1825 from AVX rated at 4kV and operated at  
> over 2kV (1825JA221KAT1A). These are smaller than the ones you  
> mention and so I would look very carefully at the cleanliness of  
> your assembled PCB. These SMT components should operate at the  
> specified HV provided by the manufacturer for SMT assemblies - the  
> disclaimer is just an acknowledgment that operation of the component  
> may be compromised if humidity is high, for example.
>
> In my opinion, the CW bases should be robust by design as we  
> consider at least 10 years of operation and the cost of re-design  
> should not be that significant compared to production and  
> maintenance costs.
>
> I will be glad to check your layout if I can import your file.
>
> Best regards,
> Fernando
>
>
> Paul Smith wrote:
>> Hi Fernando,
>>
>> I'm not on the calorimetry email list.  Please feel free to forward  
>> to the list if appropriate.
>>
>> Comments in-line below:
>>
>> On Jun 18, 2010, at 11:28 AM, Matthew Shepherd wrote:
>>
>>>
>>> Hi Paul,
>>>
>>> I'm not sure you are on the calorimetry list -- see the note below  
>>> from Fernando.
>>>
>>> -Matt
>>>
>>> Begin forwarded message:
>>>
>>>> From: "Fernando J. Barbosa" <barbosa at jlab.org>
>>>> Date: June 18, 2010 11:24:28 AM EDT
>>>> To: Matthew Shepherd <mashephe at indiana.edu>
>>>> Cc: Hall-D Calorimetry <halld-cal at jlab.org>, "Fernando J.  
>>>> Barbosa" <barbosa at jlab.org>
>>>> Subject: FCAL CW bases
>>>>
>>>> Hi Matt,
>>>>
>>>> I have a few concerns about the plan for the CW bases you  
>>>> suggested at the GlueX meeting and regarding the HV sparks  
>>>> observed on the present design:
>>>>
>>>> 1. HV sparks across components - Due to close proximity of  
>>>> components, HV sparks or leakage are observed. The present  
>>>> proposed solution is to pot with epoxy. This seems to prevent HV  
>>>> sparks.
>>>>
>>>> My concern is that this is temporary solution but not a cure for  
>>>> the inherent problem. I really don't want to rely on potting with  
>>>> epoxy (long term issues) and a conformal coating should only be  
>>>> used as a preventive measure.
>>
>> This problem is between adjacent diodes in the CW multiplier  
>> chain.  If one follows the IPC "rule of thumb" of 500 volts for a  
>> 0.1 inch spacing on uncoated FR4, there shouldn't be a problem.   
>> However, even though the pads have round corners, the diode leads  
>> have square corners which aren't always completely covered by  
>> solder.  Even though conformal coating is supposedly good for >1500  
>> volts/mil, this seems to be the case only if the electric field is  
>> transverse to the coating.  As we reported a few weeks ago, sparks  
>> will form under the coating in a plane parallel to the board  
>> surface, presumably between the board and the coating.  Evidently  
>> the coating doesn't completely bond to the FR4.  Slots, etc are  
>> possible, but at extra cost.  This problem could be solved by  
>> moving the diodes further apart, resulting in a longer board.
>>
>> What are the long term issues with epoxy you refer to?  Radiation  
>> damage to the epoxy?  Isn't epoxy used all the time in chamber  
>> construction, etc?  Aren't through hole HV capacitors and resistors  
>> potted/molded in epoxy by the manufacturer?
>>
>>
>>>>
>>>> 2. HV sparks are observed across one HV capacitor - the capacitor  
>>>> is rated for operation at 2,500V but it cannot hold 1,500V. The  
>>>> proposed solution is to use a low viscosity epoxy to under fill  
>>>> the capacitor and prevent sparking.
>>>>
>>>> A capacitor that is sold and rated at 2,500V should withstand  
>>>> that same voltage or the manufacturer has a big problem. Most  
>>>> likely the PCB pads to which the cap is soldered to are too close  
>>>> or have sharp edges - this limits the maximum operating voltage.  
>>>> Alternatively, you could choose a larger capacitor with the pads  
>>>> spaced farther apart but this may not be necessary.
>>
>> This capacitor is an AVX # 2225WC103KAT1A  The data sheet states:   
>> "may require protective surface coating to prevent external  
>> arcing".  The length is 0.22", so the clearance is way less than  
>> the 500 volts/0.1 inch rule of thumb.  The arcing is observed NOT  
>> between the pads (which do have round corners) , but most  
>> definitely between the metallization of the capacitor terminals  
>> along the sides of the capacitor (between the ends).  Epoxy  
>> underneath the capacitor doesn't solve the problem; it also needs  
>> the sides and top covered with epoxy as well in order to withstand  
>> the rated 2500 volts.
>>
>> The previous version of the CW base used a Johanson 302H51W103KY4  
>> for this component; this was a 3530 case size.  It was rated for  
>> 3000 volts and would withstand 2500 volts if the humidity was 50%  
>> or lower.  Humidity levels higher than 60% would cause sparks at  
>> 2000 volts.  Unfortunately, this capacitor is no longer made, and  
>> in fact Johanson now has capacitors rated up 6000 volts in the 2225  
>> case size.  They also include the disclaimer "may require a surface  
>> coating to prevent external arcing".
>>
>> AVX does make a 3640 case size capacitor; unfortunately it is 4X as  
>> expensive as a 2225.  To my knowledge this is the only option for a  
>> physically larger surface mount HV cap.  Several manufacturers make  
>> 2500 volt rated capacitors in the 2225 case size.
>>
>>>>
>>>> I strongly suggest that the base be re-designed to space the  
>>>> components and pads farther apart to provide for reliable  
>>>> operation by design, without a conformal coating. Once that is  
>>>> verified, a conformal coating is highly recommended because of  
>>>> moisture and dust.
>>
>>
>> Assuming we can find a surface mount capacitor physically large  
>> enough to withstand it's rated voltage without any sort of coating,  
>> the next issue is with the 1 Gohm feedback resistors.  The previous  
>> version of the CW base used a Victoreen (now Ohmite) MC102821007F,  
>> in a 5025 case size.  Unfortunately, this is no longer a standard  
>> product and is only available as a special order.  The present  
>> version of the base uses an Ohmite HVF2512T1007FE.  This is rated  
>> at 2500 volts in a 2512 case size, so has slightly better clearance  
>> than the 2225 capacitor, but still well in excess of the 500 volt/ 
>> 0.1 inch rule.  I haven't actually seen sparks across this resistor  
>> since they form across the capacitor first, but have also coated it  
>> with epoxy both underneath and top and sides in order to get it to  
>> work.  Venkel does make a 1 Gohm resistor in a 5020 size, but it is  
>> 6X as expensive as the Ohmite 2512.
>>
>>
>> To summarize, the diode clearance issue could be solved by a larger  
>> layout.  The 2500 volt rated resistors and capacitors are difficult  
>> to get and expensive in larger case sizes and still won't work  
>> without a coating unless the humidity is very low.  Conformal  
>> coating doesn't seem to be adequate for insulation parallel to the  
>> board surface and sparking can occur underneath.  Epoxy does work,  
>> but if you think there are long term issues, the only alternative  
>> for producing a layout that will operate without any type of  
>> additional insulation would be to use non surface mount components  
>> for the parts that see the entire 2+ kV across them.  I suppose  
>> another alternative might be to use 2 components in series so that  
>> each only sees half the voltage.
>>
>> Because of the block to block spacing, it is not possible to make  
>> the PCB any wider.  It is possible to make it longer.  This works  
>> out fine for increasing the diode spacing, but doesn't help much  
>> with the other parts.  It might be necessary to go to more layers  
>> to use the longer board to advantage with larger, non surface mount  
>> parts or doubled parts.
>>
>> I would point out that CW bases made by Hamamatsu (for example) are  
>> completely potted.  Obviously, I will do whatever is required to  
>> pass the JLab review, but changing the layout enough to allow it to  
>> operate without additional insulating material is going to be  
>> expensive in terms of engineering time and parts cost.  Is this  
>> requirement going to apply to HV distribution boards on the  
>> chambers, for example?  I know the ATLAS TRT drift tubes had to use  
>> epoxy on their HV distribution/blocking capacitors.  The  
>> fundamental problem seems to be that HV surface mount components  
>> have extremely optimistic voltage ratings.  Presumably the  
>> capacitor dielectric can withstand the rated voltage, but they must  
>> test them in an oil bath or something other than air.
>>
>> I am planning to visit a new contract manufacturer next week (our  
>> previous CM in Indianapolis went out of business).  I will discuss  
>> the manufacturing issues involved in coating HV components with  
>> them, whether with epoxy or some other material.
>>
>> In any case, your comments/suggestions are extremely useful, and I  
>> will keep in touch,
>>
>> Paul
>>
> <barbosa.vcf>