REPORT 974

THE SUPERSONIC AXIAL-FLOW COMPRESSOR:

Bv AsTHrs KAXTEOTTITZ

SUMMARY were possible to design efficient axial-flow compressors with

An investigation has been made to explore the possibilities of supersonic velocities into the blade rows. The tip speeds
axial-flow compressors operating with supersonic velocities info required to achieve these supersonic relative velocities and
the blade rows. Preliminary calculations showed that very the accompanying high pressure ratios are not in excess of
high pressure ratios across a stage, together with somewhat those currently used in impulse-turbine practice (about 1,600
increased mass flows, were apparently possible with compres- ft/sec). A study is therefore being made at the Langley
sors which decelerated air through the speed of sound in their Aeronautical Laboratory to explore the possibilities of super-
blading. The first phase of this investigation, which has been sonic axial-flow compressors. Introductory general informa-
reported in NACA ACS L5D20, was the development of efficient tion on the supersonic aerodynamics involved in the present
supersonic diffusers to decelerate air through the speed of sound. study can be found in reference 1.
The present report is largely a general discussion of same of the Experience with the high wave losses that usually occur
essential aerodynamics of single-stage supersonic axial-flaw in supersonic flows might lead to the supposition that the
compressors. In the supersonic flow about isolated bodies, efficiency of such a compressor would inevitably be low. It
large energy losses usually occur due to wave systems which must be noted, however, that these high wave losses are
extend far from the bodies. Supersonic flow entering a cascade usually associated with extended wave patterns. Busemann
is considered and, in this case, the possibility of entirely elimi- thus showed and Ferri (reference 2) verified that it was
nating this extended wave system is demonstrated; thus, no possible to design a biplane in which (in the absence of lift)
reason for supersonic compressors to be necessarily inefficient the waves originating at one airfoil were canceled at the
is apparent. The conditions that occur as the flow through the opposite airfoil. In this case, Busemann was able to elimi-
compressor is being started are discussed and a hypothesis as nate the extended wave pattern and, theoretically, all of the
to the type of transonic flow which will be encountered is wave losses associated with the thickness of isolated super-
proposed. sonic airfoils. This idea can be extended to the case of an
As an approach to the study of supersonic compressors, three infinite cascade and it can be shown that, if the entering
possible velocity diagrams are discussed briefly. Because of Mach number is sufficiently greater than one, the cascade
the encouraging results of this study, an experimental single- can also produce lift (or turning) without the production of
stage supersonic compressor has been constructed and tested in an extended wave pattern, the strong waves being entirely
Freon-12. In this compressor, air decelerates through the confined to the region between the blades. Thus, there is
speed of sound in the rotor blading and enters the Stators at no a priori reason to suppose that supersonic compressors
subsonic speeds. A pressure ratio of about 1.8 at an efficiency will necessarily be inefficient, since the usual source of high
of about 80 percent has been obtained. losses in supersonic flows can be avoided.
Preliminary consideration of the velocity diagrams suitable
INTRODUCTION for use in a supersonic compressor showed that, in order that
advantage be taken of the high pressure-rise possibilities, it
The occurrence of supersonic yelocities in the flow about would be nec'essary to decelerate the air through the speed
conventional axial-flow compressor blades results in large of sound in the blading. The first phase of this study (re-
energy losses. For this reason, axial-flow compressors are ported in reference 3) was the development of diffusers to
generally designed to operate with relative air velocities into decelerate air efficiently through the speed of sound. This
the blade rows low enough to avoid supersonic velocities. study indicated that a normal shock in a diverging passage
This restriction has limited the pressure ratios obtainable is necessary for stable flow and methods for minimizing the
with axial-flow compressors to about 1.2.5- across a stage intensity of the shock were developed. Maximum passage
(rotor and stator) and has meant that many stages are re- contraction ratio which could be used was found (for the
quired to produce the pressure ratios needed for gas turbines, conditions which would probably prevail in a supersonic
turbojet units, and superchargers. It is apparent that axial-flow compressor). Diffusers were designed which
great increases in the pressure ratios per stage, with accom- recovered over 90 percent of the kinetic energy in a supersonic
panying savings in weight and size, could be obtained if it stream up to a Mach number of 1.85. It therefore appears
I This report was originally issued as NACA A.CR L6D02, April 1946 and until recently has been subject to security regulations;
47S
474 REPORT 974—NATIONAL ADVISORY COMMITTEE FOR AERONAUTICS

that supersonic diffusers can be made with efficiencies com-

parable with those obtained with good subsonic diffusers.
Because of the encouraging results of the supersonic-diffuser Shock waves
investigation, a single-stage supersonic axial-flow compressor
was designed and built at the Langley Aeronautical Labora-
tory of the NACA. .
The first question to be settled in the study of supersonic
compressors is whether the axial velocity should be subsonic
or supersonic. A supersonic axial velocity would mean that,
as the compressor is started, it would be necessary to form a
normal shock in the annulus ahead of the compressor and to
move this shock downstream and finally into the rotor.
The possibility exists that, in this case, discontinuities in the
performance of the compressor would occur when the shock
entered the rotor. At the present time, it is not clear how
any advantages can be gained by making the axial velocity '••••Cascade
•-.Cascade
of flat plates
supersonic (except perhaps in the case of supersonic aircraft of flat plates
in which it might be desirable to avoid the losses inherent in
'•••Mach wave
decelerating through the speed of sound in the inlet). The originating at B
decision was therefore made to study the case of subsonic Entrance region...
axial velocities first.
SUPERSONIC CASCADES
ENTRANCE REGION • Entering air

When the axial component of the velocity entering a cas-

-.Entering air
cade is below the local speed of sound, waves started by a por- (a) (b)
tion of each blade can move ahead of the rotor. It will be (a) Axial velocity too low. (b) Axial velocity too high.
shown that, in a,supersonic compressor, the axial velocity is
FIGURE 1.—Schematic illustration showing manner in which volume flow adjusts to caseado
controlled by these upstream-moving waves. The part of geometry for a case with a straight cntranco region AB. The only steady flow is that
the blades from which upstream-moving waves can originate with the entering velocity parallel to AB.
will be called the "entrance region." Consider Mach waves Undisturbed flow
starting from the rearward side of the blades (that is, rear- direction; M=l.52-
.Shock wave
ward as the rotor advances). Schematic illustrations of Streamline.. „Expansion wave
waves originating in the entrance regions of supersonic cas-
cades are given in figures 1 and 2. Waves originating up-
stream from the point B in figure 1 can move ahead of the
rotor; thus, the region AB of the blading will constitute the
entrance region.
The simplest case to be considered is that in which AB is a
straight line. In this case, no waves move ahead of the rotor
except those given off by the leading edge of each blade
FIGURE 2.—A graphical construction of the extended wave pattern ahead of a supersonic
(if no strong compression waves are started immediately be- cascade with a curved entrance region AB. The construction starts by the choice of tlio
hind B). The steady flowfor this case can be shown to have point B, which determines the conditions far from the cascade. In this construction, the
surface AB is a 10" arc of radius equal to 1.38 AA' and the surface at B makes a 30° angle
a velocity vector (in rotor coordinates) parallel to AB. with the stagger Jine AA'. The energy contained in the pressure waves that move ahead
Consider first the case where the axial velocity is lower of the blades Is enough to heat the incoming air about 1° F (for inlet stagnation tempera-
ture 520° F absolute).
than would be required to satisfy this condition (fig. 1 (a)).
In this case, an expansion wave corresponding to flow around Thus, the axial velocity and hence the volume flow at a given
a corner will be given off from .the leading edge A. A similar rotational speed into the compressor is controlled entirely by
wave would be given off from the leading edge of each blade. the slope of the line AB in the supersonic operating condi-
Bach of these waves will accelerate the flow in the axial tion. Also, in the steady operating condition, no waves will
direction until the relative velocity is parallel to AB. Simi- move ahead of the rotor; the extended wave pattern dis-
larly, in the case where the axial velocity is too high (fig. 1 (b)), appears and all the waves are confined to the passages be-
an infinite series of oblique shock waves will issue from tween the blades. These conclusions do not apply to
the leading edge of the blades and these waves will reduce the cascades, of a solidity so low that the point B would fall
axial velocity until the relative velocity is parallel to AB. behind the trailing edge of the blades.
 THE SUPERSONIC AXIAL-FLOW" COMPRESSOR 475

It should be pointed out that waves originating from the A first approximation to the energy carried by these waves
forward side of the blades (that is, forward as the rotor ad- can be found, once the velocity distribution at the entrance
vances) do not move ahead of the rotor. The shape of the (or exit) of the cascade is known, if the disturbances are
forward side will therefore not affect the entrance velocity assumed to be small and standard relations from the theory
so long as the velocity entering the blades is supersonic. of sound are applied. The sound energy transmitted across
(See, however, the subsequent section entitled "Starting an element of surface dA in a time dt is ApAVax dA dt (refer-
Conditions.") If, for example, leading edges of the blades ence 4), where Ap is the fluctuation from average pressure and
are wedges, the entering velocity will be parallel to the rear- AVax is the fluctuation in the velocity normal to the surface.
ward side and an oblique shock will be given off from the This expression, which is evaluated in the appendix, is for
forward side. small disturbances generated by a surface moving with the
In practice, the use of a curved entrance region may per- mean motion of the fluid; however, if the fluid has a mean
haps be desirable. In this case, waves will be given off, in velocity T« through the radiating surface as in the case under
general, from the whole region AB and will move ahead of consideration, the length of the column of sound radiated in
the rotor as is illustrated in figure 2. These waves must a unit time would be reduced by a factor 1—Max-
represent equal expansions and compressions if the condition la. order to evaluate the effect on efficiency, these losses
is to represent a steady state.1 When the compressions and can be expressed as a temperature rise AT by setting the
expansions do not balance each other, the average axial energy dissipation per unit mass flow to equal cpAT, where
velocity will be changed" by the net resultant waves (as in c, is the heat capacity at constant pressure. The gas enter-
the case considered previously where AB was a straight line) ing the rotor will be heated through the temperature rise AT
until the waves balance each other. If these waves are to by the waves in the entrance region. The stagnation tem-
cancel, they must represent equal and opposite deflections of perature at the end of the compressor will be raised by AT
the air stream to a first approximation. The air entering times the ratio of the final stagnation temperature to the
the blades at A must therefore be parallel to the surface at B. temperature before the rotor. This increase in the final stag-
The air entering the rotor at other points between the blades, nation temperature can then be compared with the tem-
however, and thus the average entering air (that is, along perature rise across the compressor to evaluate the effect on
the line AA') will not, in general, be parallel to the air at A. the efficiency. It is shown in the appendix that
A graphical method of calculating the wave pattern and the
average axial velocity for this case is given in the appendix. AT= JtV* . (1-2UW
CpMas COS 6
The wave pattern found for a particular case is shown sche-
matically in figure 2. where Max is the axial Mach number, 8 is an average angle
between the waves leaving the cascade and the stagger line
LOSSES DUE TO WAVES MOVING AWAY FROM THE CASCADE
(defined in fig. 3), and A TV is the mean-square fluctuation
In cases where the entrance region is curved, compression in axial velocity at the entrance to the cascade. The wave
and expansion waves move ahead of the supersonic cascade. losses are then roughly proportional to the square of the
If these waves are not reflected by obstacles (such as another departure of the entrance region from a straight line. Eval-
cascade immediately upstream), they will be propagated up- uation of these energy losses for the wave system shown in
stream until the compressions and expansions neutralize figure 2 may be of interest. By graphical integration, it is
each other. The energy carried by these waves will be dis- found that AVax2 is 9,100 (ft/sec)2 and, with the shock angle
sipated (largely by shocks) and this dissipation represents a (23°) used as 8, there is obtained AT=1° F for air. These
loss in efficiency. Wave losses will also occur when detached
"bow" waves are formed ahead of a supersonic cascade (pro-
.-Entering velocity
duced by a finite leading-edge thickness or by too large a
wedge angle). These bow waves eventually will also be
neutralized by expansion waves and this wave pattern will
similarly involve an energy loss. Again, in supersonic cas- .-Approximate direction
.Mach angle / of pressure waves
cades where the air decelerates through the speed of sound,
a normal shock will always be present. There are several
circumstances in which this normal shock will not be con-
tained within the passages formed by the blading. The
following calculations show that, in this case, large losses
appear, as a result of the strong wave pattern propagated
away from the rotor. Wave losses of the same kind can
occur when air leaves a cascade at supersonic speeds.
i It should, be noted that a small part of an expansion wave striking a shock is reflected. In FifiCEE 3.—Diagram illustrating the approximate wave-loss calculation. The component
these considerations, these reflections are neglected; thns an error is introduced in cases where of the entering velocity normal to the wave direction is approximately the local speed
strong shocks are present. of sound. All velocity changes are normal to the waves.
476 EEPOBT 974—NATIONAL ADVISORY COMMITTEE FOR AERONAUTICS

losses will not be very large so long as all the waves are •Undisturbed flow •Expansion
direction waves
attached to the blades. For cases in which detached bow
waves are formed or in which a normal shock is not contained
within the blade passages, the root-mean-square variation in
axial velocity can increase to several hundred feet per second
and the resultant wave losses are prohibitive.
The losses considered in.this calculation are only those
upstream of the line at which the mean-square variation in
axial velocity is known. For example, if bow waves are
present, the losses due to the portions of these waves behind
the leading edge of the cascade are not considered.
In the event that it is necessary to calculate losses due to a FinuRE i.—Hypothetical wave pattern for cascades in the transonic region. As the enter-
system of strong waves more accurately than this first approx- ing Mach number is increased, the detached waves more closer to the blade leading edges;
thus, the energy radiated from the cascades is reduced. A. mechanism of this kind permits
imation, the mean-square variation in axial velocity can be the air to pass through the cascade even when the undisturbed velocity Is closo to sonic.
calculated some distance ahead of the cascade where the The velocity crossing the line AA' Is partly supersonic and partly subsonic; thus, the
average mass flow per unit area can Increase to allow for the thickness of the blades.
strength of the waves will be somewhat reduced. If this
loss is then added to the losses occurring downstream from A bow-wave system would have two effects on the perform-
this line (which can be found directly from the known shock, ance of the compressor: first, the wave losses considered
strengths), a more accurate estimate of the energy losses can previously would occur and, hence, the efficiency would be
be obtained. reduced; and second, the axial velocity and the mass flow
STARTING CONDITIONS
would be lower than would be expected for a supersonic
It is important in planning any type of supersonic flow, operating condition.
and especially flows.involving deceleration, to examine the As the rotational speed of the compressor—and conse-
mechanism by which the flow is to be set up. For example, quently the relative velocity entering the blades—is increased,
for the supersonic diffusers investigated in reference 3, the a weaker wave system (bow waves closer to the blade leading
maximum amount of contraction of the supersonic passage edges) is required to permit the passage of the air through the
that could be used was found to be determined by the starting blades. Thus, as the rotational speed was increased, it
conditions. In this case, a normal.shock formed ahead of would be expected that a favorable effect on the efficiency
the diffuser and the critical condition was with the shock at would be obtained and that the relative velocity at entrance
the entrance. It was necessary to ensure that, in this condi- to the blading would become more closely parallel to the line
tion, the mass flow passing through the shock could also pass AB. (See fig. 1.) If this starting mechanism is correct, it
through the throat of the diffuser. In the case of a super- would also be expected that the starting of a supersonic
sonic compressor, the starting mechanism is much more com- compressor would not occur discontinuously but that gradual
plicated and is not at present fully understood. Speculation changes would occur as the bow waves approached the blade
on this subject, however, may be of use in providing guidance leading edges. Kesults on the experimental supersonic
for further experimental work. compressor described subsequently tend to support these
The axial velocity entering the compressor will be assumed three conclusions based on the bow-wave hypothesis.
to be always subsonic. As the rotational speed of a com- In order for the bow waves to become attached, the air
pressor is increased, the relative velocity into the blading crossing the line A'C (fig. 4) must be able to pass through the
will reach and exceed the local speed of sound. It is clearly narrowest section of the blading. When the bow waves are
not possible to have a uniform stream at the local speed of nearly attached, the conditions just ahead of the downstream
sound entering a cascade of. blades,of finite thickness since leg of a bow wave will be close to the operating supersonic
in this condition the mass flow per unit area is already a' conditions. The mass flow in this condition can be approxi-
maximum. Consider, however, a system of detached bow mately calculated from this fact. Attachment can take
waves formed ahead of each blade and compensated by ex- place only if this mass flow, after experiencing the losses in
pansions starting from the blade surfaces, such as those shown the downstream leg of the bow wave, can pass through the
schematically in figure 4. Now, along the stagger line AA' narrowest section of the blading. The losses in the down-
the average mass flow per unit area is everywhere less than stream leg of the bow wave will be of the order of magnitude
maximum and it should be possible thus to allow for the of normal shock losses at the entrance Mach number.
space taken up by the blades. Such a bow-wave system has Perhaps, therefore, a good first guess as to the ratio of A'C
been observed to form in the flow about a stationary sub- to the minimum passage area that could be used in super-
sonic cascade even before the speed of sound is reached ahead sonic compressor blading would be the contraction ratio
of the cascade. The flow about an isolated body appears to found previously for supersonic diffusers (fig. 3 of reference 3).
be fundamentally different from the flow into a cascade in This hypothesis is at present hardly more than a speculation
that the bow waves will not move indefinitely far ahead of and much more work will have to be done before the starting
the cascade when the entering velocity is close to sonic. mechanism is fully understood.
 THE SUPERSONIC AXIAL-FLOW COMPRESSOR 477
One other necessary condition is that the wedge angle be compressors will be to high power machines, where compact-
made small enough to allow attachment of the bow wave ness and light weight are important. For these applications,
(fig. 2). Note that the air must turn through an angle equal it will be desirable to make the mass flow large. Maximum
to the: total wedge angle in the case of a straight entrance mass flow is obtained when the velocity at the entrance to
region. the first rotor is axial and at the local speed of sound. The
Present indications are that compressors operating with maximum mass flow is thus obtained without the use of
detached bow waves will not be very efficient. In order that entrance vanes. Other things being equal, however, it is
the bow waves become attached to blades with a practicable possible to obtain an increased pressure rise by the use of
thickness near their leading edges, the relative Mach number entrance vanes which turn the air against the direction
must be considerably above 1; thus, compressors operating of rotation. Entrance vanes can certainly be of use in
with relative Mach numbers near unity will probably not be equalizing the energy input at the root and tip of the blades.
of much practical interest. The design of entrance vanes does not now seem to present
any new problems and, therefore, these vanes have not been
THE DESIGN OF SUPERSONIC COMPRESSORS considered in the designs presented herein.
In order to stimulate thinking about supersonic com- For the cases discussed, only the tip flow will be considered
pressors, a few representative single-stage designs are dis- and it is not yet known how far the blades can be continued
cussed briefly. toward the axis (that is, how large the annular air passage
Preliminary remarks.—In general, the energy input per can be made) before serious structural or aerodynamic
unit mass of an axial-flow compressor stage can be shown difficulties intervene. Thus, it is not yet clear how large the
from simple momentum considerations to be the product mass flow of supersonic compressors of a given diameter can
of the blade peripheral speed VB and the change in tangential be made. However, it seems quite certain at present that
velocity A.VT. All this energy appears as heat producing the mass flow can be made larger than that of any existing
subsonic axial-flow compressors. In the designs considered,
a stagnation temperature rise AT— where cB is the heat the axial velocity has been made subsonic. In the discussion
capacity at constant pressure in mechanical units (6,006 that follows, the passage-area changes can be assumed to be
ft-lb/slug/°F for room-temperature air). provided, where necessary, by varying the annulus areas
If the efficiency of the compressor is 100 percent, the pres- from the front to the rear of the blades.
sure ratio per stage is Design 1; supersonic relative velocity entering and leaving
rotor and subsonic stator.—The first design considered
P.R.
/T0+ATy»rR (design 1) avoids the necessity of decelerating air through
the speed of sound in any part of the compressor. The
pressure ratio that can be realized in this type of compressor
is seriously limited. In order to keep the mass flow through
the compressor high, the entering axial Mach number must be
where T0 is the initial stagnation temperature and R is the fairly high—say, of the order of 0.8. The static pressure
gas constant for air. If the adiabatic efficiency is TJ, ahead of the rotor will therefore be 0.66 times the initial
e lR total pressure for axial entry. If the initial-air stagnation
P.R =(l + v VBAVT\ » temperature is 520° F absolute and the tip blade speed is
c„T ) 0
1,600 feet per second, the Mach number entering the rotor
It was pointed out in the section entitled "Supersonic Cas- blades will be about 1.72. If the maximum pressure ratio is
cades" that, in order to permit the use of blades with sufficient to be obtained, it will be clearly desirable to obtain the
thickness to be structurally practical, the Mach numbers used highest possible static-pressure ratio in the rotor since the
must be considerably above 1. For this reason and because static-pressure ratio in the stator is definitely limited. In
the pressure ratios obtainable increase quite rapidly with the order to increase the static pressure in a supersonic passage
blade speed available, the blade speed in these typical designs without decelerating through the speed of sound, the passage
is assumed to be 1,600 feet per second at the tips. Since tip must be contracted. The maximum contraction ratio found
speeds of this order of magnitude have been used in practical from figure 3 of reference 3 that permits the air following a
turbines, which have been run at high temperatures, it seems normal shock at a Mach number of 1.72 to pass through the
quite likely that the construction of an axial-flow compressor passage is 1.15. In the supersonic operating condition, the
to run at this speed will be practical mechanically, at least Mach number after a contraction of this magnitude would be
from the centrifugal-force considerations. 1.51. If this value, then, is the Mach number leaving the
The inherent advantages of supersonic compressors are: rotor, the static-pressure ratio across the rotor would be
first, high pressure ratio per stage, and second, high mass flow. about 1.37. (See reference 1.)
The chief apparent disadvantages • are the mechanical A second limitation for this design is due to the subsonic
difficulties usually associated with high-speed machines. It cascade behind the rotor. Subsonic cascades have been
seems, therefore, that the first applications of supersonic designed for high efficiency up to a Mach number of 0.8.
478 REPORT 974—NATIONAL ADVISORY COMMITTEE FOB AERONAUTICS

At this Mach number, the total pressure entering the stator The results of reference 3 clearly indicate that the decelera-
is 1/0.66 times the static pressure.^ entering the stator. In tion through the speed of sound in the rotor passages must
the absence of any losses, the total-pressure ratio across the be accomplished by a normal shock. The efficiency of this
single-stage compressor would be 0.66X1.37XQ-gg=1.37. compressor then depends upon how much the total-head
and separation losses accompanying this normal shock can
A velocity diagram for this case is given in figure 5. This be reduced. These losses can be reduced if the Mach num-
calculation indicates that, the maximum pressure ratio which ber immediately ahead of the normal shock can be lowered.
can be obtained from supersonic compressors which do not The Mach number ahead of the normal shock can be lowered
anywhere decelerate the air through the speed of sound is by starting compression waves as soon as possible behind
limited to the same,order of magnitude as that obtainable the entrance region. The tests showed that starting com-
with subsonic axial-flow compressors; and that the chief pression waves by the inclusion of a concave region on the
advantage of such a compressor would be in its high mass- rearward side of the blades—that is, rearward as the com-
flow possibilities. pressor rotates—immediately behind the entrance region
Design 2; deceleration through the speed of sound in considerably improved the efficiency of the test compressor.
rotor and subsonic stator.—The design of blading to de- (See fig. 7.)
celerate air through the speed of sound has not been given 'Velocity entering rotor
sufficient study to make clear what can or cannot be done. ! (stationary coordinates),
I 341.9 fps;A£*0.8;
An indication of the type of rotor flow that will be desirable 1 T-461' F abs;
for maximum pressure rise can be obtained from considera- p=0.6S6 aim
tion of a compressor in which the air enters the stator with
the maximum tangential velocity for which efficient subsonic Velocity entering rotor
stators can be designed. For a stator Mach number, of 0.8 (rotating coordinates),
1808 fps; M-I.7I9;
and a stagger angle of 60°, the pressure ratio across the T'46f F abss
p'0.656 atrn
compressor can be calculated.
Assume a blade tip speed of 1,600 feet per second and an
inlet stagnation temperature of 520° F absolute. From the
stator design conditions, the ratio of the tangential velocity
entering the stator (VT) to the local speed of sound can be
found. This ratio leads to a known value W/V^ä where
Ti is the stagnation temperature behind the compressor.
It was shown previously that
\v Rotational
\ speed,
T2=T0- VBAVF
\ 1600 fps
s
Velocity entering stator "'Velocity leaving rotor
(stationary coordinates), frotating coordinates)
883 fps; M'O.8; I66ZS fps;M-l.506;
where, in this case, AVT can be replaced by VT. This T-505" F abs) T'SOS' F abs;
equation can be solved for the conditions given and yields p* 0.89S aim p-0.89S aim
Ti=754 ° F absolute. The pressure ratio for 100 percent effi- FIGURE 5.—Velocity diagram for supersonic compressor 1. In tbls design, air cntcrsjimd
leaves the rotor at supersonic speed and the Mach number entering the stator Is limited
ciency is 3.67 or for 80 percent efficiency 2.93. The pressure to 0.8. Starting considerations limit greatly the amount of deceleration which can bo
rise determined by these limiting conditions is unaffected by used in the rotor and the 0.8 stator Mach number condition severely limits tho amount
of turning in the rotor. Thus, the energy input and the pressure rise (even for the
the axial velocity entering the rotor. This axial velocity isentropic flow assumed here) are small.
will therefore be chosen as 0.8 of the local speed of sound, /Velocity entering rotor
whichwill yield a high mass flow. The states of the air at (rotating coordinates),
1808 fps, M'l. 719;
various points now can be readily calculated, if losses are Velocity entering rotor T'46/"F abs-,
neglected, and are given on the velocity diagram (fig. 6). (stationary coordinates), P"0.656 atm
842 fps-, M'O.8-,
This calculation indicates that a compressor of this type T* 461''F obst
.Velocity leaving rotor
p - 0.656 atm^
would have considerable advantages over subsonic axial- (rotating coordinates},
882 fps, M'0.686,
flow compressors. The most significant problem presented T'668°F abs;
is in the design of rotor blading to decelerate air from a p--8.404 atm
Mach number of 1.72 to a Mach number of 0.68 and simul-
taneously to tum the flow through an angle of 7.36°. The
preliminary work on supersonic diffusers reported in ref-
erence 3 was undertaken to study this problem. It was
found possible to design supersonic diffusers with efficiency Velocity entering stator
comparable with that of good subsonic diffusers. Machines (stationary coordinates), ^Rotational speed,
1014 fps, M'0.8, 1600 fps
decelerating air through the speed of sound were therefore T'668'F abs:
p"2.404 atm
considered to have considerable promise and construction
FIOOTE C.—Velocity diagram for supersonic compressor design 2. In this design, tlwjulr
of an experimental supersonic compressor of this type was enters the rotor at supersonic speed and the Mach number entering the stator is limited
decided upon. This compressor and preliminary test results to 0.8. The pressure ratio obtainable by a machine of this type is restricted to about 3
by this limitation. This velocity diagram assumes isentropic flow throughout and [or
are described briefly later in the present report. this case the pressure ratio would be about 3.67.
 THE SUPERSONIC A3IAIr-ELOW COMPBESSOK 479
The results of reference 3 showed that very slow diver- similarly would be expected to be controlled by the back
gence of the subsonic passage was necessary to obtain high pressure so long as the shocks were located in portions of
efficiency in the supersonic diffusers tested. A machine of the blading where the passage area was diverging down-
this type would consequently be expected to require passages stream. The losses in the normal shocks would thus be
of considerable length—compared to the distance between reduced as the back pressure on the compressor was in-
blades—in order that the passage divergence might be slow creased. The maximum efficiency would be expected to
enough to prevent serious separation losses. occur just before the normal shock was forced out ahead of
In the tests of reference 3, it was found that the position the compressor blading. "When the normal shock is forced
of the normal shock in the diverging portion of the diffuser out ahead of the blading, a strong extended wave system is
was stable and was controlled by the back pressure at the set up ahead of the compressor and large losses result. This
subsonic end of the diffuser. In the supersonic compressor condition will most likely produce a result similar to stalling
(design 2), the position of the normal shocks in the blading in subsonic compressors.
Design 3; deceleration through the speed of sound in the
stator.—The third design to be considered (design 3) provides
deceleration through the speed of sound in the stator. This
type has been given very little study and not much can be
said about it at present. (No velocity diagram is presented
for this case.) If deceleration through the speed of sound
is permitted in the stator, there are no limitations on the
usable velocity diagrams that are apparent at present. The
pressure ratio obtainable will, as far as can now be foreseen,
depend only on the tip speed available. Of course, in this
design—as in design 2—it will be necessary to design passages
that decelerate or turn the air flow without serious separa-
tion losses. A difficulty that this design introduces is that
the solidity of the blading used will have to be considerably
higher than the solidities ordinarily employed in axial-flow
compressors. For example, if a rotor blade design that
turns the air back to the axial direction relative to the rotor
can be used, the tangential velocity imparted to the air by
the rotor would then be equal to the tip speed. If the initial
stagnation temperature and blade speed are again 520° F
absolute and 1,600 feet per second, the final temperature
coming from such a compressor would be 946° F abso-
lute. Assume an efficiency of 80 percent; the pressure
ratio would then be 5.85. Such a compressor will probably
present new starting problems and these, as well as efficiency
considerations, will determine the velocity diagram used.
This type of compressor will need a great deal of study and
Thickness added to
reduce Mach number it is hoped that more attention can be given to this design
of normal shock in the future.
PRELIMINARY RESULTS ON" THE EXPERIMENTAL
SUPERSONIC COMPRESSOR

An experimental supersonic compressor, which is designed

to decelerate the air through the speed of sound in the rotor
and to operate with subsonic stators, has been constructed
and preliminary tests have been run. The compressor is
designed to operate at a tip speed of 1,600 feet per second in
air, with an entering stagnation temperature of 520° F
absolute. The tests were actually run at comparable Mach
numbers in Freon-12. (As far as is known, Freon-12 can be
used to simulate the flow of air, at least qualitatively.)
This gas has a velocity of sound about half that of air and
many of the structural problems could therefore be avoided
En tering- air in these preliminary tests.
direction The compressor was originally intended to have approxi-
FIOUEE 7.—Cross section at pitch, of blades of experimental supersonic compressor. In mately the velocity diagram of design 2 discussed previously.
order to prevent flow separation, the annulus in which thft blades run was made
to contract downstream. The first tests, however, disclosed that the passages formed
936646—51- -32
480 EEPOBT 974—NATIONAL ADVISORY COMMITTEE FOB AERONAUTICS

by the rotor blading diverged too rapidly in the subsonic Laboratory to construct a compressor with thicker blading
region and serious separation losses resulted. It was neces- and with a higher solidity in an attempt to realize the
sary to reduce the rate of this divergence and the annulus performance predicted in the discussion of design 2.
was therefore made to converge downstream. This change (3) The volume flow entering the rotor was close to the
reduced the passage area at the exit of the rotor blading theoretical value and was virtually independent of the back
and the Mach number leaving the rotor blading was there- pressure on the compressor in the supersonic operating con-
fore about 0.97 instead of 0.68, as intended in the original dition in agreement with the theory of the entrance region
design. This modification decreased materially the pressure of supersonic cascades discussed previously.
ratio obtainable from this compressor. (4) The efficiency of this compressor could be increased
In order to minimize the starting difficulties, the blades, by starting compression waves on the rearward sides of tlio
which are shown schematically in figure 7, were made very blades (that is, rearward as the compressor rotates) just
thin (thickness, about 2 percent chord). Since the tests behind the entrance region. Compression waves wero
have been run, it seems that these difficulties were over- started by adding thickness to the rearward side of the blades,
emphasized and that thicker blading can be used. In as shown in figure 7. Improvements in efficiency of several
these blades the entrance region was nearly straight and the percent were made in this way.
necessary thickness (0.015 in. near the leading edge) was
CONCLUSIONS
provided by a 10° wedge. The blade length (2 in.) was
made one-quarter of the tip radius so that the theoretical Supersonic axial-flow compressors are capable of much
volume flow through this machine would be about as large higher pressure rises per stage and probably somewhat
as any previously obtained in axial-flow compressors of higher mass flows than are possible with subsonic compres-
similar diameter. As remarked previously, it is not yet sors. The extended • wave system, which is usually re-
known how great the blade length can be made compared sponsible for the large losses in the supersonic flow about
to the tip radius, but no great difficulties appear to have isolated bodies, can be eliminated in the case of a supersonic
been introduced by the blade length selected. Entrance cascade and thus there is no apparent reason to indicate that
vanes ahead of the rotor were used to help equalize the supersonic cascades would necessarily be inefficient.
energy input supplied by the root and tip of the blades. A hypothesis as to the type of transonic flow that will occur
Tests were made by surveying total head, static pressure, as the compressor is brought up to operating speed is pro-
and flow direction in front of and behind the rotor. The posed. It appears from this hypothesis that the starting of
test results do not include any losses that might occur in supersonic compressors designed for subsonic axial velocity
the stator behind the rotor. The results obtained thus far will not involve any discontinuities.
indicate that: The volume flow entering a supersonic compressor (in its
(1) No discontinuity occurs in the performance of this operating condition) is independent of the back pressure.
compressor in passing through the transonic region with the A single-stage experimental supersonic compressor that
blading used. All indications are that the transition from a was tested in Freon-12 showed a pressure ratio of about 1.8
subsonic type of flow to a supersonic type of flow occurs at an efficiency of about 80 percent. No discontinuities in
smoothly. The data in the transonic region provide some performance while the compressor was being accelerated
support for the bow-wave starting-mechanism hypothesis could be observed. A theoretical method of calculating the
discussed previously. operating volume flow was presented and agreed well with
(2) Pressure ratios of the order of 1.8 with efficiencies of experiment.
about 80 percent are obtainable with this compressor in its
present form. It should be. remarked that considerable
warping occurred during the fabrication of the thin blading
of this compressor and that this warping resulted in large LANGLEY MEMORIAL AERONAUTICAL LABORATORY,
inaccuracies of construction. Plans are consequently being NATIONAL ADVISORY COMMITTEE FOR AERONAUTICS,
made in the Fluid and Gas Dynamics Section of the Langley LANGLEY FIELD, VA., March 8, 1946.
 APPENDIX
WAVES CREATED BY A CURVED ENTRANCE REGION

Graphical construction of the extended wave system.—All small disturbances generated by a surface moving with the
waves involved in the flow ahead of the cascade originate on mean motion of the fluid; however, if the fluid has a mean
the rearward surface of the blades and are therefore of the velocity Va through the radiating surface as in the case under
same family. All the oblique shocks present will be assumed consideration, the length of the column of sound radiated
to be of sufficiently low intensity to permit the flow to be in a unit time would be reduced by a factor l—MttX. If
considered isentropic to a first approximation. These facts the disturbance is produced by pressure waves—such as
allow the flow to be constructed graphically as follows: those shown in figure 2—all velocity increments are normal
First, assume a point B (fig. 2) on the rearward side of the to the waves, there being no velocity change in the plane of
blade to be the end of the entrance region. The choice of the waves. Since the component of velocity normal to
this point B will be shown to determine the rotational speed the wave front is approximately the local velocity of sound
(or more accurately, the ratio of the rotational speed to the a, the pressure jump Ap across a wave front will be
speed of sound). Draw the line BA'. This Mach wave is
the last in the entrance region. The Mach angle—and hence, Ap=p aA Ta(l — MK)
the local Mach number—can now be determined by measur-
where p is the ambient density and AVa is the change in
ing the angle between BA' and the blade surface at B. If a
the velocity normal to the wave front in crossing the wave.
real flow is to be represented, this angle must be less than 90°. From figure 3, it can be seen that ATa^ in this case is given
The Mach lines originating at points between B and A can
by AVax=AVa cos 8; therefore,
now be drawn. (The velocity along AB is assumed to be
everywhere supersonic.) It has been shown that, for AF«
Ap=pa -
steady flow, the air entering the rotor at A must be parallel * cos 8
to the surface at B. The wave (in the case illustrated, an The energy dissipated per unit mass entering the compressor
oblique shock) originating at A can then be drawn. If this is then
process is repeated for successive blades, a complete flow
pattern can be found. The compression and expansion p dA
AptfloMaxJ cos 8
waves interact and eventually cancel each other some dis-
tance ahead of the rotor. From the construction given in where p„ and a0 are density and sonic velocity, respectively,
figure 2, the Mach number and direction of the air far from in the undisturbed stream and A is the annulus area.
the cascade can be found in rotating coordinates. The Since the analysis is restricted to small disturbances, it will
direction of the air far from the cascade in stationary co- be permissible to assume constant values of density, sound
ordinates will be known from the known setting of stator velocities, and 6; these quantities thus can be taken out of
blades ahead of the rotor. (In the absence of any stator the integral. Equating this power to a heat input cpAT
blades, it would be assumed that the undisturbed velocity gives
was axial.) It is now possible to complete the velocity
triangle and thus to find the rotational velocity and the A7E
AT--
undisturbed axial velocity and volume flow. Max cos 0e„
losses due to an extended wave system.—The losses where
involved in the waves that move ahead of or behind the rotor AK !
can be calculated once the wave pattern is known. A first
approximation to the energy carried by these waves can be
=iJAVafdA
REFERENCES
found by assuming them to be small disturbances and
applying well-known relations from the theory of sound 1. Taylor, G. I., and Maccoli, J. W.: The Mechanics of Compressible
(reference 4). The sound power transmitted across a surface Fluids. Vol. Ill of Aerodynamic Theory, div. H, W. F. Durand,
.A (chosen perpendicular to the axis) is given by ed., Julius Springer (Berlin), 1935, pp. 209-250.
2. Ferri, Antonio: Esperienze su di un biplano iperacustico tipo
Busemann. Atti di Guidonia, no. 37-38, 1940.
fApAV^dA 3. Elantrowitz, Arthur, and Donaldson, Coleman duP.: Prelimi-
nary Investigation of Supersonic Diffusere. NACA AGR
L5D20,1945.
where Ap is the fluctuation in pressure and Al^is the fluctua- 4. Rayleigh, (Lord): The Theory of Sound. First American ed.,
tion in velocity normal to the surface. This expression is for vol. 11, sec. 245, Dover Publications (New York), 1945, pp. 15-18.
481