SYSTEM DESIGN NOTE

SDN0007.03 - Radiation Shield Design

 

Prepared by	Date	Approved by	Date	Rev.	Rev Date
J Elias	6/17/99	N. Gaughan	6/17/99	B	5/12/00

1. Introduction

This document discusses the design of the GNIRS radiation shields. In particular, it discusses provisions for “floating” (passive) radiation shields and the need for “active” (cooled) shields. The purpose of the radiation shields is to limit and control heat radiated into the cooled mass of the instrument; exclusion of photons from the instrument interior is accomplished by the optical bench’s light seals, though the radiation shields will provide some attenuation.

Revision A includes a more complete discussion of MLI as an alternative to rigid floating shields.

2. Floating Shields

“Floating” or passive shields are intended to provide isolation of the cold mass and thereby reduce the amount of heat that must be removed by the cryocoolers. Radiative heat that must be actively removed needs to be kept below 100 W in order to carry out the last stages of cool-down with reasonable efficiency (see SDN007.02).

2.1 General Considerations

The formula for heat transfer between two parallel plates whose spacing is small compared to overall dimensions is given by:

q = e_1-2s (T₁⁴ – T₂⁴)

where

e_1-2 = 1/(1/e₁ + 1/e₂ - 1)

and e_n and T_n are the emissivity and temperature of surface n. The constant s is the Stefan Boltzmann constant. Values for sT⁴ are given in Table 1, below, for some temperatures of interest.

Table 1 – Black-Body Radiation from Surface (W/m²)

 

Temperature (K)	Radiated Energy (W/m2)
300	459.
293	418.
273	315.
253	232.
200	91.
150	28.
100	5.7
80	2.3
60	0.7

If the area of the dewar shell is roughly 10 m², and there is no intervening shielding, for an emissivity of 0.1 for the dewar shell and of the cold structure the heat transfer for 300K ambient (60K internal) is roughly

Q = 10 x 0.053 x (459 – 0.7) W = 243 W

If a “floating” shield is placed between the dewar shell and the inner structure, one can apply the formula above to calculate the temperature of the shield and the heat flow, with the results:

T_s = 2^-1/4T_d

q = e_1-2s T₁⁴/2

Adding additional shields gives the results, for n shields,

q = e_1-2s T₁⁴/(n+1)

Ts,i = {(n+1-i)/(n+1)}^1/4T_d

where the radiation from the structure is assumed to be negligible.

Some representative values are given below, assuming emissivity of 0.1 and ambient temperature of 300K. At least 2 shields are required to keep heat flow under 100 W, as discussed above.

Table 2 – Shield Temperatures and Radiative Heat Flow

Number of Shields	Heat Flow (W/m2)	Temperature of Shield N	Temperature of Shield N-1
2	8.1	228	271
10	2.21	165	198
20	1.16	140	166

2.2 Conduction Effects

Floating shields will either be thin, fairly rigid shells with some sort of spacer to maintain separation, or multi-layer insulation, which usually is built up with some sort of separating layer.

The separating material will in either case provide some conduction losses; for the MLI it may have enough area to increase the effective emissivity as well.

For the case of rigid shields, one can assume spacers of area ~1 cm square and 1 cm long between the shields, about 1/m². The conductivity of G10 around 250K is about 0.8 W/m/K, so heat loss through a spacer would be 0.8 x 0.01 x 0.01 / 0.01 = 0.008 W/K. Between shields 1 and 2 in the two-shield case, this gives a heat flow of about 0.34 W per spacer, which is modest compared to the radiation.

The effects of the spacer material for MLI are more difficult to calculate. First of all, if the effective emissivity of the MLI is increased, the radiative heat flow calculated above (10 and 20 shield cases) will increase roughly in proportion. Second, because the temperature differential between shields decreases only as the fourth root of heat flow, the conduction losses tend to become relatively more important as the number of layers increases. If the thermal conductivity between layers of MLI is similar to that for rigid shields, the conduction between shields 19 and 20 would be about 0.21 W/m², which is starting to become significant compared to the radiative heat flow.

The assumption that conductivity between layers of MLI is similar to that from 1 cm spacers is not a good one, but it is hard to decide whether it is better or worse. However, MLI is an established technology, so it is fair to conclude that it will work at a reasonable level: it should be possible to get heat flow through an MLI-based radiation shield under 3 W/m², but probably not under 1 W/m². This implies total radiation loads on the cooled structure with MLI in the range 10-30W.

Published data (e.g., IR Handbook) on MLI typically describe performance in terms of conductivity (which is not surprising). Values in the range of 40-60 mW/m/K are quoted for blankets with 20-40 layers/cm. Thus a 30-layer blanket 1 cm thick would have a conductivity of perhaps 5 mW/m²/K for a resulting heat load of about 12 W. This is not quite double what one would expect from 30 layers with minimal conduction. 10 layers distributed in a 1 cm blanket would have a conductivity around 7 mW/m²/K for a 17 W heat load – actually somewhat better than the 10-layer case in Table 2.

By way of comparison, a 2-layer rigid shield would provide about 81 W load.

2.3 Variability

The numbers in Table 1 show that the range of radiation heat input that GNIRS will experience (warm day in the lab vs. cold day at the telescope) is about a factor of 2. If the heat flow through the shields is largely radiative, this same factor of 2 will be present in the heat input into the cold structure. If there is a significant conductive component, the variation will be less.

For the simple 2-shield arrangement, the variation is almost 40 W. As shown in SDN007.01, if this change in heat is removed by the cryocoolers, the head temperature

varies by almost 3K, and the cold structure must vary by at least this must if it is not controlled. If MLI is added, variations in heat input are several times smaller. (Note, by the way, that the RGD 5/100 coolers have a cooling curve that is less “stiff”, so the sensitivity to heat input is almost 2x greater if four of these heads are used.)

2.4. MLI vs. Rigid Shields

Rigid shields offer one significant advantage, which is that they do not trap air (or water) and hence will allow the dewar to be pumped out faster than if MLI is used. Although there is no formal requirement (from Gemini) on pump-down time, it is clear that there is little advantage to shaving a few hours off cool-down if the time to pump out the system increases by the same amount. The cool-down time studies (SDN007.02) gave a time to get from 80 to 60K of about 14 hours with radiation input equivalent to 2 floating shields. This time would be reduced to perhaps 6 hours if an efficient MLI shield were used – but unless the increase in pump-down time is under 8 hours there is no net gain to cycle time.

Rigid shields also can be modeled more easily, which means that one need not put a safety factor into the design (number of shields).

Finally, rigid shields can be handled directly (albeit with some care).

The primary disadvantages of the rigid shields are the space and weight they occupy. A shield of area 10 m² and thickness 1 mm made out of aluminum will have a weight of 27 kg, ignoring all spacers and supports. Hence provision of two shields will add 55 kg or so to the instrument weight. This produces radiative loading of ~80 W; to reduce this to 40 W one would need 5 shields for a combined weight of 135 kg.

Also, if the spacing is roughly 1 cm. (much closer would complicate installation) the use of multiple shields restricts the available volume.

MLI does trap air, leading to slower pump-down. This indicates that if it is used, some care in design and layout of the blankets is needed to minimize this problem. In addition, MLI should not be used in excess to keep the trapped volume as low as possible.

In addition, the difficulties in modeling imply a need to provide some margin for error in performance in the design.

MLI also requires mechanical support, and is presumably handled (after installation) by handling the support. The instrument design must make it possible to remove the radiation shields for access without having to replace the MLI.

MLI clearly weighs less. Measurements of a commercial 5-layer blanket give a weight of

roughly 250 g/m² (measured by Ken Don, 19 Nov. 98), or 2.5 kg for a 5-layer blanket covering the dewar surface. The total weight of the MLI would be <10 kg unless one wanted 20 layers or more. Some support would be needed, probably comparable in weight to a single rigid shield, unless the MLI is attached to the dewar shell or to the inner, active shield.

The baseline chosen is to use 2 floating shields, as the performance of these can be specified more easily – i.e. we know it will meet requirements. The final decision should be deferred until design of the dewar shell is undertaken and a detailed weight budget is available. Space, weight, and handling considerations will drive the final decision.

3. Active Shield

Thermal gradients and their variability were discussed in SDN007.01. The analysis showed that the radiation from the passive shields would tend to produce both unacceptable gradients and gradient variations if it was coupled directly to the bench structure, with the collimator probably the most vulnerable element.

Implementation of an “active shield” surrounding the cold structure appeared to be an obvious solution to this problem.

3.1 Active Shield Properties

The active shield would be coupled directly to the cryocooler heads, and would be thermally isolated from the bench structure (though mechanical coupling is of course OK). It would also provide a means of cold-stationing wiring, support struts and fill tubes for the pre-cool system (if provided).

Thermal gradients in the active shield are only a concern if they lead to “hot spots” that produce significant radiation onto the inner cold structure.

For a simple model of the shield as a cylinder with uniform heat input along its length that is cooled through its center (circumference), the temperature difference at the end is:

DT = QL/2Ak

where

Q is the heat input in W (half of total into cylinder)

L is the half length of the cylinder (nominally 1 m)

A is the cross-section

k is the thermal conductivity

This is half the value that one would have if all the heat were injected at the ends. If 6061 Al is used, an appropriate value for k is around 100 W/m/K, but there is no reason not to use a higher conductivity alloy (e.g., 6063) where values around 250 W/m/K can be obtained. If the cylinder has a diameter of 1.1 m and a thickness of t mm, and Q = 40 W, we get, for 6063 Al:

DT = 22.5/t K

If the thickness of the shield is 1.5 mm (roughly 1/16 inch), the gradient would be ~15 K. The weight of such a shield is roughly 41 kg.

The shield needs to be connected to the cryocoolers somehow. It is important to recognize that the higher-conductivity aluminum alloy has similar conductivity per unit weight to that of copper, so that one should simply examine heat flow through the shield to head, if it’s attached directly.

If one models the shield in the vicinity of the attachement as a flat plate, and assumes all the heat arises externally, the temperature gradient is:

If there is a single connection to a pair of heads dissipating 40 W, and shield parameters are as above, the formula becomes

DT = 17 ln(x/x₀) K

The gradient between 1 m radius (where the assumptions are not really valid) and 20 cm is about 27 degrees; between 20 cm and 5 cm it is another 24 degrees. This latter gradient could be reduced simply by thickening the shield over 40 cm diameter about the cold head connection. These considerations suggest that the overall gradients within the shield can be held to around 30 K. If the shield is made thicker by an additional 6 mm at 2 locations 40 cm, diameter, the added weight is about 4 kg.

One still needs to connect the shields to the cryocoolers, and there are a couple of options. One would be to bolt the shield to the cryocoolers directly – vibration of the shield itself is not an issue, so long as it does not couple to the rest of the instrument. If this is viewed as too high a risk, one could use copper straps. These can be quite short. For a length of 10 cm, and other properties as in SDN007.02, the resistance per strap is 3.3 K/W. If we take nominal head temperature as 35 K and assume a shield gradient of 30 K with a maximum allowable temperature of 100 K (implies radiation into cold structure <5 W for shield emissivity of 0.1), the allowable temperature gradient across the straps is 35 K. If the heat flow is 80 W, this implies a need for 8 straps. This is not a lot of copper (1/2 kg or less) so it would make sense to be conservative here 16 straps would keep the shield temperature (in principle) everywhere under 85 K.

Total weight of the active shield, including the provisions for connection to the cold heads, is then about 46 kg. This does not include any weight in mechanical supports (stand-offs).

4. Mechanical Issues

4.1 Assembly

The radiation shields surround the cold bench structure. Thus to work on the instrument it will be necessary to remove the shields – or at least have some form of access through them. Furthermore, all wiring and support trusses must pass through the shields.

Two approaches are possible for mechanical assembly (at least).

In the first, the shields are attached to the dewar shell. Logically, they would separate into three parts, corresponding to the two end caps plus the middle bulkhead structure. In this approach, it would make sense to have all wiring cold stationed on the middle shields, which would remain in place in the bulkhead. Connectors would allow further disassembly if required, but the objective would be to avoid having to remove the middle shields for mechanism access.

The passive shields on the end caps would not need to be mechanically coupled to the middle shield in any way (one would like to avoid enormous gaps, of course). The active shields would have to be connected to the cryocoolers; one could do this either by connecting to the middle active shield or by direct connection to the cryocoolers.

In the latter case in particular, one might consider having the “middle” shields much less in area than the bulkhead, with the “end” shields protruding past the end caps into the bulkhead. Connections would be made or unmade through access ports in the bulkheads.

The second approach would be to attach all the shields to the inner structure, in which case they would be removed separately from the dewar end caps. Access would obviously be easier, and could be configured from the point of view of instrument access rather than dewar shell removal. On the other hand, the shields would be less protected and would require separate handling.

4.2 Weight

If the shield design consists of 2 floating shields plus an active shield, total weight will be approximately 100 kg, plus mechanical supports.

A minimum weight design would be one in which there are no rigid floating shields and an MLI blanket is supported by the active shield. The total weight for this would be ~55 kg, and some light-weighting of the active shield might be possible since heat input into it would be 2-3 times less. This might reduce total weight to as little as 40 kg, although the shields would have greater fragility and the design and fabrication effort would be greater.

5. Vacuum Issues

If the shields actually fully enclose the instrument structure, the only way for air to be pumped out from inside will be through the hole in the shields for the entrance window. A more direct path should be provided, in the form of a hole in the shields whose diameter is somewhat greater than the vacuum valve, which is located at the pump port. This will provide direct access to the interior volume, but will not produce significant additional radiation loading. If we assume 30 cm2 area, and unit emissivity since this is a hole in the dewar shell, the radiation at 300 K will be 1.4 W (a radiation baffle could be added to reduce this by a factor of 10 or so).
 

 
   
 

If you have any questions or suggestions regarding this website, please contact Melissa Bowersock.

NOAO Intranet Services

NOAO Copyright  Statement

National Optical Astronomy Observatories, 950 North Cherry Avenue, P.O. Box 26732, Tucson, Arizona 85726,
Phone: (520) 318-8000, Fax: (520) 318-8360