Band separation system

Nov. 12, 1935.

E. I. GREEN

2,020,409

BAND SEPARATION SYSTEM

<img file="US2020409A_D0001.tif" />

<img file="US2020409A_D0002.tif" />

7500 ~3000

<img file="US2020409A_D0003.tif" />

ATTORNEY

Nov. 12, 1935. e. i. green 2,020,409

BAND SEPARATION SYSTEM Filed Aug. 15, 1333 2 Sheets-Sheet 2

<img file="US2020409A_D0004.tif" />

<img file="US2020409A_D0005.tif" />

ATTORNEY

Patented Nov. 12, 1935

2,020,409

UNITED STATES PATENT OFFICE

2,020,409 BAND SEPARATION SYSTEM

Estill I. Green, East Orange, N. J., assignor to American Telephone and Telegraph Company, a corporation of New York

Application August 15, 1933, Serial No. 685,273

Claims. (Cl. 178—44)

This invention relates to· a method for dividing a band of frequencies and more especially to a method for dividing a frequency band by the use of phase discrimination without the use of electrical filter selection.

The usual method of selecting certain parts of a band of frequencies is to· use electrical filters. If, for example, it is desired to divide a band of frequencies into two parts, a combination of high and low-pass filters may be used. In the present invention a method is described for dividing a band of frequencies which does not essentially involve the use of electrical filters. This method of dividing a band of frequencies is based upon the general principle of separately modulating differently phased components of the same carrier frequency with the band to be divided. The different phase relations between the frequencies resulting from the two modulations are used as a basis for eliminating the band of frequencies either above or below the carrier frequency. The invention, therefore, provides a device to accomplish functions similar to those of wave filters and suited to· a wide variety of applications.

The features of the invention will now be better understood from the detailed exposition which follows, when read in connection with the accompanying diagrams, Figures 1 to· 5. Figure 1 shows an arrangement which might be used to divide a band of frequencies by phase discrimination. Fig. 2 illustrates in greater detail a possible embodiment of the arrangement shown in Fig. 1. Fig. 3 shows a phase divider for the carrier- frequency to be used with third order modulation. Fig. 4 illustrates a third order modulating circuit. Fig. 5 indicates the relative amplitudes of the voltages over a band of frequencies which can be obtained by means of the phase shifting networks shown in Fig. 2, or in Fig. 7. Figs. 6 and 7 illustrate band phase shifting networks.

The essential principles of the invention will now be demonstrated in connection with Fig. 1. Consider a band of frequencies from some source of signals such as SS. This band may be represented as sin Qit, sin ___sin q_kt, sin q_mt---sin q_nt (1)

Since the amplitudes of the frequencies composing the band do not affect the operation of the method, the amplitude indices have been omitted in expression (1).

The frequency intervals between successive components of the band may be of any width and may be equal or unequal. The band (1) may be located anywhere in the frequency spectrum.

However, in order to· avoid difficulties in obtaining the desired lower side band in the output 5 of the modulators, it may be convenient to have the band (1) located so· that

<td></td><td></td><td></td>

<td> is less than</td><td> \ 2π J</td><td> 10</td>

<td> is less than twice</td><td> qi . ο» 2tt ^d 2ir</td><td></td>

<td></td><td> qi q» 2τγ’ 2π</td><td> 15</td>

being the frequency at which division is desired and &lt;71 „„. On

2ir ^d being the lower and upper frequencies of the band. This avoids any overlapping of the input frequencies of the modulator with those of the lower sideband output. It is not essential, however, that this condition be obtained.

Assume that it is desired to separate the band (1) into two parts, the point of division lying between two of its component frequencies and %*

2?r 2tt

There is generated in the carrier frequency oscillator CO the frequency g-35

2tt such that

Qt&lt;g_a&lt;g_m(2)

Two components of this frequency differing in phase by 90 electrical degrees are obtained in ⁴⁰any desired manner by means of the phase divider PD. These components of the carrier frequency may be designated sin qst and ccs qst. Each of these components is then modulated with the frequency band (1) extending from sin qit ⁴⁵to sin qnt, in the modulators MDi and MD2, which may be of any suitable second order type. Only the lower sideband of the modulation process is retained in each case. The upper side- βθ band may be suppressed by filters such as LPi ⁰and LP2 of Fig. 1, or tuned circuits, or by phasing out in a single sideband modulator.

Assume, for the present, that second order modulators are used. As a result of the modula- 55

2,020,409 tion of sin q_st, the following band of frequencies is obtained in the output of modulator MDi:

cos (q₃—qi)t cos (q,—q₂)i cos (q_s — q_k)t (3) cos (q„ — q_s)t cos (q„— qi)t while through modulation of cos qst, there is obtained from modulator MD2:

—sin (Qs—qi)t —sin (q_s—qz'it —sin (q_s—qk)t (4) +sin (q_m—qs')t

-)-sin (q_n—qs) t it will be noted that the sign of the terms in the second series changes at the point where the division is to be accomplished. Next, as indicated in Fig. 1, the phases of all components from, modulator MD2, series (4), are shifted 90 degrees by the phase shifter PS without distortion of amplitudes relative to those of the other series. Thus there would be obtained a new series as follows:

—cos (qs—qilt ³⁰ —COS (Qs—Q2> t —cos (qs—qk~)t +cos (q_m—q_s)t ³⁵ +cos (q_n—q_s)t

Amplification is introduced by the amplifiers AMi and AM2 which have identical gain-frequency characteristics and introduce no phase _ά0 distortion. These amplifiers supply the primaries of transformers TRi and TR2. Two of the output windings, one from each coil, are connected series aiding and the other two are connected series opposing. Thus the output into line Li ₄₅ may be obtained by adding together the groups of frequencies represented by Equations (3) and (5) which will give the band of frequencies from

Qm Qs , Qn Qs ^to

The output into line L2 may be obtained by subtracting these groups since the windings of transformers TRi and TR2 oppose each other. This will give the band of frequencies from

Qs — Qk Qs — Qi ^t0

It will be apparent that these results will be obtained regardless of the phases or magnitudes 60 of the frequency components of the original band extending from in the treatment just given the use of second order modulators was assumed. That the same result may be obtained using modulators of the third order type will now be demonstrated. Consider again the band of frequencies represented by Equation (1). Referring to Fig. 1 this band of frequencies is introduced to the modulators

MDi and MD2 in the same phase, these modulators now being of any desired type of third order device. As before it will be assumed that the band (1) is to be divided into two parts, the point of division lying between the component frequencies

Τπ ^and 2Ϊ'

The carrier frequency oscillator CO generates a ^sfrequency

2tt such that , . , , 10 &lt;3tt&lt;2q'₈&lt;q_m (6)

Two components of this frequency differing in phase by 45 electrical degrees are obtained from the phase divider PD. These components may be designated sin q'st and sin 15 (α'.ί+ξ)·

Any odd multiple of 45 degrees [i. e., (2η-1)ϊ;

where n is any integer] may be used for the phase difference without essentially altering the treatment given. Each of these components is 25 then modulated with the frequency band (1) in modulators MDi and MD2, respectively. Again only the lower sideband is retained, other undesirable frequencies being suppressed with the low pass filters LPi and LP2, or by some other 30 suitable means.

As the result of third order modulation of the carrier frequency component sin q'st with the band (1) in MDi, the following frequencies will be produced: 35 +sin (2q’s—qi)t -j-sin (2q'_s—Q2)i --------------- (7) +sin (2q’_s—qk) t —sin (q_m—2q'_s) t40 —sin (q_n—2q’_s)t, while modulation of sin ί l saA will produce the following frequencies:

+cos (2q’s—qi)t

-i-cos (2q'_s—qz)t50 (8) +cos (2q'_s—qk)t

-j-cos (qm—2q'_s)t +cos (q_n—2q'_s)t55

It will be noted now that the sign of the terms in Equation (7) changes at the point where the division is to be accomplished. The procedure from this point on is the same as that considered under second order modulation. The phases of ⁶⁰one of these sets of frequencies are shifted by 90 electrical degrees and the two groups are then added or subtracted to eliminate the undesired frequencies. The frequency components resulting from the addition of the frequencies repre- ^{65 * * * * 70 * * * * 75}sented by Equation (7) to those of Equation (8) shifted 90 degrees in phase will now extend from g_m —2q', g_n—2g'_s

2ττ ° 2?r ’ 70 while subtraction will give frequencies from

2q'_s—qi 2q'_s-q_k

2π 2π

It is, therefore, plain that either second order or 75

2,020,409 third order modulation will give comparable results and either may be used.

In the disclosure so far the special case in which the relative phase shifts between the carrier components and between the sidebands are 90 electrical degrees (second order modulation) has been discussed. This case is the simplest and most useful one. However, it will be of interest to show briefly the effects of phase shifts other than 90 electrical degrees.

For purposes of simplicity consider only two frequencies in the band of frequencies to be divided,

Qk and gm 2ir’ one of which is lower than and the other higher than the frequency at which it is desired to divide the band,

2tt'

Let these frequencies have random phase angles, denoted by a and β, respectively. Then the expression for these frequencies may be written sin (q;_ct+ «) + sin (q_mi+/3) (0) where

Q,'_;&lt;q_s&lt;Qm· (¹°)

Now modulate with these frequencies two component currents of the carrier frequency,

21/ differing in phase by any angle Θ, which may be represented as sin q_sf, and sin (qh+0) (11)

Assuming second order modulators, the lower sideband frequencies resulting from modulation of the first carrier component with (9) will be represented as follows:

cos [(gs—“l + ^C0S[(Qm^—gs)i+0] (12) and that from modulation of the second component cos [(q_s — Qk)t—«+0] + cos [(&lt;7m q_s)i+^ 0]· (13)

If now we shift the phases of the components of (13) by some angle φ, we obtain cos [(q_s — Qk)t — ffi + 0 — ¢1 + _v, , .

cos [(gm— g,)t+6—0—φ]· (14)

Adding the lower sideband (12) to the other shifted by an angle φ (14) and combining we get:

cos^(0—Φ) cosi[2(g_s— Qk)t— 2α+0—¢1 + ZA cosi(0+<i>) cos|[2(q„—q_s)f+2/3—0 —¢1(15) ύ4

The two· terms of the above expression represent the two frequencies obtained by modulating the carrier frequency with the original signal frequencies and these frequencies, expressed by the last part of each, are and 2ir ΛττΟί greatest Interest, however, are the amplitude coefficients of these two frequencies which are cos|(0—¢)(16) and cos|(0+</>)(17)

In order to eliminate one of the sideband frequencies, one of these amplitude coefficients must 10 ( be reduced to zero, while the other remains. By inspection it will be obvious that this can readily be accomplished by making the sum or difference of Θ and Φ equal 180 electrical degrees.

If (0+0) = 180° (18) ¹⁵ expressions (16) and (17) become cos^(0—¢) = 2 cos (0— 90°) = 2 sin 0(19) ²20 cos|(0 + ^.) = 2 cos 90° = 0(20) or if (0-</>)=±18O°(21) cosg(0—¢)=2 cos±90° = 0(22) cos^(0 + </>) = 2 cos (0±9O°)= ±2 sin 0 (23)

By satisfying Equation (18) or (21) it is obvi- ⁸⁰ ' ous that either sideband frequency can be deleted, leaving the other. The coefficient of the remaining term, 2 sin 0, is of course, a maximum when 0=90°, which is the special case which has been described previously. In this case ψ is ³⁵either -(-90° or —90°, depending on which frequency it is desired to eliminate.

The frequencies used in this treatment,

Qfe j Qm 2^^and2+ are general expressions,

9½

2ir 45 being any frequency less than the carrier

-£-· and %” any frequency greater than the carrier. Thus 50 the derivation may be applied to a whole band of frequencies, the relations true for

Qk ^2π 55 being true for all the frequencies below the carrier, and those for ^2,7 60 holding for all the frequencies greater than the carrier. Thus, if the angle of phase shift φ is exactly the same angle at every frequency elimination of one of these bands of frequencies may be accomplished. 65

The expression (15) also indicates that the initial phase angles, a and β in this case, of the frequencies in the band to be divided are immaterial to the operation of the system, since they do not enter into the coefficient expressions 70 (16) and (17).

The exposition of the general case has been made assuming the use of second order modulation. Third order modulation could be used, and the derivation would be similar. The angu- 75;

2,020,409 lar phase difference between the carrier components in this case is doubled in. the modulation products just as is the frequency of the carrier. The amplitude coefficients, therefore, become δ i cos 2 (20' — Ψ)(24) and cos | (20' + φ)(25) and the condition for annulling, one. of the frequency bands is (20±0)=±18O°(26)

It will now be of interest to consider further some of the details of the arrangement discussed in connection with Fig. 1 in which only- 90-degr.ee phase shifts are employed. Referring to Fig. 2, a possible embodiment of the invention is shown.

²⁰ A source of signals SS is shown and transformers TRi TRz are employed to introduce signals to both modulators MDi and MDz in the same phase. A source of carrier frequency CO is shown. The output of this oscillator is impressed across a re25 sistance Ri and condenser Ci in series. The voltage drop across the resistance Ri is displaced in phase by 90 electrical degrees relative to that across condenser Ci, the desired relation for second order modulation. The voltage drop across ³⁰ the resistance Ri is applied to the modulator MDi by means of resistance Ri and that across Ci is introduced to MDz by means of resistance Rs. The reactance of the condenser Ci at the carrier frequency is made equal to the resistance Ri;

³⁵ that is,

Λ-τΛ- (27) where a_s equals 2tt times the carrier frequency 40

2ir

The components of the carrier frequency applied to the two modulators are, therefore, equal in ⁴⁵ magnitude and differ in phase by 90 degrees.

The frequency of the carrier oscillator

2tt ⁵⁰ is that frequency at which it is desired to divide, the incoming band of signals \2tt 2tt/

The modulators are of the balanced type so that the carrier frequency is suppressed in the output. These modulators are not necessarily of the vacuum. tube type as shown but may be any type of second order modulator, such as copper oxide ⁶⁰ units. Only the lower sideband is desired and, therefore, the low pass filters LPi and LPz are inserted after the modulators MDi and MDz, respectively. These Alters do not need to have sharp cutoff characteristics, because of the separation ⁶⁵ between the two sidebands, and might be replaced by condensers shunted across the outputs of the modulators. It is, of course, possible to produce only the lower sideband by some method of phasing out the upper sideband such as, for instance, disclosed in my Patents Nos. 1,719,052 and 1,744,044.

In case third order modulation is to be used, a carrier oscillator and phase dividing network similar to those shown in Fig. 3 should be sub75 stituted for the oscillator and network (Ri,Ci) shown in Fig. 2. The phase difference between the carrier voltage components should be an odd multiple of 45 electrical degrees, and the carrier frequency should be half that frequency at which it is. desired to, divide the frequency band 5 \2tt ¹ 2π)

Considering the network shown in Fig. 3, the carrier frequency generated in the oscillator CO' is 10 &lt;7'3 .

2π which is half of the carrier frequency

2tt used for second order modulation. The output of this oscillator CO' is impressed across a network of capacity C'i and resistance R'i in series. ₂θ The. voltage across the condenser C'i lags that across the resistance R'i by 90 electrical degrees. In order to obtain two voltages differing by an odd multiple of 45 degrees the reactance of the capacity C'i is made equal to part of the resist- ₂₅ance, that is (28) where w'_s equals 2π times the carrier frequency uw

2tt

The voltage across this capacity is then added to that across the portion.of the resistance .414 R'i. Since these voltages are equal and in phase quad- ⁰⁵rature the resultant voltage Ei will bear a phase relation to the voltage across the remainder of the resistance which is an odd multiple of 45 degrees, in. this case 135 degrees ί^3π /1· λ

I-4-radians )·

The magnitude of this component will be proportional to τ/2Χ(.414) or .586, which is the same as that across the remainder of the resistance (.586 R'i). The result of modulating the two components of the carrier ₅₀frequency q'_a2 π with the incoming band of frequencies will be 55 equivalent to that obtained when second order modulators are used as described in connection with Fig. 2. The modulator in this case might consist of one or more units of thyrite (a mixture of finely divided particles of carbon and clay), so A circuit arrangement for such a modulator, showing thyrite units TUi and Τϋζ replacing vacuum tubes such as VTi and VTz of Fig. 2, is shown in Fig. 4.

Referring now to Fig. 2, the lower sideband 65 from modulator MDi is impressed on the amplifier AMi by means of transformer TRs and that from MDz on amplifier AMz by means of transformer TRc. Amplifiers AMi and AMz comprise vacuum tubes VTs and VTe associated with trans- 70 formers TRs and TRe and chokes CHi and CHz. These elements should be substantially identical and the amplifiers are shown with common battery supplies in order that their gains may be kept equal. 75

2,020,409 s

As previously explained, regardless of whether second or third order modulation is employed, the amplified output of one modulator should be shifted in phase 90 degrees with respect to that 5 of the other. A possible method of shifting the phases of the frequencies composing one sideband 90 electrical degrees with respect to the corresponding frequencies of the other sideband is shown in connection with phase shifters PSi, and 10 PS2. The output circuit of amplifier AMi comprises a series arrangement of inductance, capacity and resistance, L2, C2 and R2. The output circuit of amplifier AM2 consists of two series circuits of 1,3 and Ca in parallel and a series resist15 ance Rs. These elements have values

Ζ/3=2Λ2, C3=—and 7?3 = 7?2It will be readily seen that these two networks ²⁰ PSi and PS2 present equal impedances to the plate circuits of the vacuum tubes VTs and VTe. The resonant frequency of these networks is adjusted to some frequency within the frequency band occupied by the lower side-band ²⁵ of the modulation. The leads to the trans¹ former TR7 are connected across a part of the resistance R2 and hence the voltage Vr is not shifted in phase with respect to the current. These taps should be so located as to make the ³θ voltage Vr equal to the output voltage Vlc of the lower network. The voltage Vlc will be shifted 90 degrees out of phase, however, as will now be explained.

' Obviously, the voltage drop across the con³⁵ denser C2 of the upper network lags the voltage drop across R2 by 90 degrees at all frequencies, while the voltage across' La leads that across R2 by 90 degrees. Since the voltage across C2 is inversely proportional to frequency, and that across ⁴θ 1,2 directly proportional to frequency, by reversing the sign of one of these voltages and adding it to the other there can be obtained a voltage which for a considerable range on each side of the resonant frequency has substantially the ⁴⁵ same magnitude as the voltage drop Vr across a part of R2 and differs from it in phase by 90 degrees. This will be clear from the curve Vlc of Fig. 5 which shows the two reactive voltages com. bined with one voltage reversed. In the upper ^δθ network the phases of the reactive voltages are of no importance as the output is taken from the resistance. In the lower network the output is taken from the reactances however, and the reactive voltages are added in reversed relation.

⁵⁵ This addition of the voltages across the inductance and condenser is obtained by the use of the parallel circuit, as indicated in network PS2. The reversal of the one reactive voltage with respect _eo to the other is brought about by reversing the order of the inductances and capacities in the parallel branches, the taps being taken at the junction points of the reactances in the two branches. The voltage Vlc is thus obtained by 65 superposing the voltage drops across the inductances L3 upon the voltage drops across the condensers C3 with their signs reversed. As will be shown below by a concrete example, this voltage Vlc may be made substantially equal, over a 70 considerable range of frequencies on either side of the resonant frequency of the L2—C2 circuit, to the voltage Vr which is the drop across a part of the resistance R2 of the upper network PS).

In order to obtain a concrete example of the 75 operation of the phase shifters, PSi and PS2 in

Fig. 2, let it be assumed that the resonant frequency (“=vfe) 5 is 1000 cycles, that the value of R2 is 1000 ohms and that Law and each have a value of 100 ohms at 1000 cycles. Let us assume for the moment that the same voltage E is applied to both the upper and the lower resonant circuits. If reactive voltages are reversed with respect to each other the magnitude of the combined voltages over the range ¹⁵of frequencies from 500 to 2000 cycles will be as shown by the curve Vlc in Fig. 5. The voltage Vr in this case, is selected as .22 times the voltage across R2, because that is about the average value of the voltage Vlc over the fre- ²⁰quency range plotted.

It is evident, therefore, in connection with Fig. 2, that by the use of the two networks PSi and PS2 we can shift one of the applied bands of frequencies exactly 90 degrees in phase rela- ²⁵tive to the other while producing only a small magnitude distortion over a fairly wide frequency band. The lack of precise equality of magnitude results in a small amount of crosstalk or interference from the unwanted band ³⁰into the output of the wanted band. It should be noted in connection with the method used to shift the phases of one of the sidebands resulting from the modulation, that the networks PSi and PS2 in Fig. 2 accomplish the desired result ³⁵only for a limited frequency band. This frequency band is limited by the magnitude relations between the resistive arid the combined reactive voltage drops. Too large a difference in the magnitudes of these voltages would im- ⁴⁰pair the performance of the system. From Fig. 5 it will be seen that in the band from about 500 cycles to 2000 cycles the resistive and reactive voltages are substantially equal for the circuit values chosen above. By the use of more ⁴⁵complicated networks or other phase shifting devices it would be possible to obtain more precise results with respect to the shifting of the phase of one frequency band relative to the other, and to extend the range of frequencies over which ⁶⁰ satisfactory results may be obtained.

Returning now to Fig. 2, these voltages Vr and Vlc are introduced to the amplifiers AM3 and AM4 by means of transformers TR7 and TRs. These amplifiers should have identical gain- ⁵⁵frequency characteristics. After being amplified, the side-bands are combined at the outputs of transformers TRs and TR10 by the switch SW. Addition of the Side-bands will produce the sideband frequencies corresponding to that part of ⁸⁰the original band which lies above the carrier frequency, that is, 65 the band &lt;7. &lt;7»—&lt;7» ~2T ^t0

By reversing the switch SW, the difference of the ?₀two sidebands will give the sideband frequencies corresponding to that part of the original band below the carrier frequency, or &lt;?«—&lt;ik. &lt;z,—&lt;n —---to —-—

2tt 2ir

2,020,409

In the systems illustrated in Figs. 2, 3 and 4, the carrier components supplied to the two modulators are shifted relatively either 90 degrees or 45 degrees, and the relative phase shift intro5 duced by the networks PSi and PS2 is also 90 degrees. If phase shifts other than these are to be used in accordance with the principles outlined in connection with Equations (9) to (26) inclusive, some modification in the networks em10 ployed will be necessary.

Any desired relative phase shift between the carrier components supplied to the two modulators may be obtained by using a carrier supply circuit of the type shown in Fig. 3. The particu15 lar distribution of resistance between the two halves of the circuit shown in Fig. 3 produces a relative phase shift of 45 degrees. By merely changing the ratio of the two parts of the resistance in the two halves of the circuit, relative 20 phase shifts either greater or smaller than 45 degrees will result. For example, by making the resistance in the upper section smaller and smaller, with corresponding increase of that in the lower sections, the phase shift will be increased until 25 in the limiting case, with zero resistance above and full resistance R'i below, the phase shift will be 90 degrees. On the other hand, if the upper resistance is increased and the lower resistance is decreased the phase shift may be made to ap30 proach zero.

To obtain a relative phase shift other than 90 degrees in the selected output bands of the two modulators, the lower network PS2 may be replaced by the network shown in Fig. 6, keep35 ing the simple series network PSi in the upper branch the same as in Fig. 2. In Fig. 6 it will be observed that a portion (aR) of the total series resistance R is removed from the common branch of the network, thus leaving the re40 mainder (1—a)R in the common branch, the removed part being in effect divided between the two parallel branches by including a resistance element having a value 2aR in each parallel branch. The three impedance elements in each 45 branch are arranged so their order is reversed in one branch.

By comparing this network with the network PS2 of Fig. 2 in which the reactance elements introduce a phase shift of 90 degrees between 50 the output voltage Vlo and the voltage across the resistance Ra, it will be evident that the effect of the resistance elements in the two parallel branches in Fig. 6 is to produce a phase shift of less than 90 degrees. This will be clear when 55 we consider that in network PS2 the drop across the inductance 2L2 in one of the parallel branches is 90° out of phase with the drop through the series resistance R3, while in Fig. 6 the drop through the resistance 2aR in series with induct60 ance 2L in one branch will be out of phase with the drop through resistance (1—a)R by an amount determined by the vectorial sum of the drops through resistance 2aR and inductance 2L.

From what has already been stated with ref65 erence to the action of the networks PSi and PS2 of Fig. 2, in which two separate voltages applied to the networks may be relatively shifted in phase 90 degrees, it will be evident that. by an analogous arrangement an applied voltage 70 may be translated into two separate voltages in phase quadrature. Such an arrangement is shown in Fig. 7. Here the applied voltage Vo is applied to a network somewhat like network PS2 of Fig. 2, said network having a resistance R 75 in series with two parallel branches each having inductance L and capacity C in series, these reactance elements being reversed in the two branches. By taking two sets of taps off as shown, a voltage Vr may be taken from the lower set of taps which is in phase quadrature with re- 5 spect to a voltage Vlc taken off the upper set of taps. By properly setting the taps these two voltages may be made substantially equal in magnitude over a considerable range of frequencies.

Having obtained by any of the methods previ- 10 ously described, separate frequency bands representing the two desired portions of the original band of frequencies, modulated to the carrier frequency, it is possible to translate either band to any desired position in the frequency spectrum, 15 including, of course, that originally occupied by said band. A single sideband modulator of any of several well known types may be used for this purpose, such as, for instance, those disclosed in my patents referred to above. 20

By additional steps of division in the manner already described, the original frequency band may be divided into any desired number of separate bands of frequencies. In this way my invention may be made to perform the functions of 25 one or more band filters, as well as those of high and low pass filters.

As compared with selection by the use of filters, where the dead space or separations between adjacent selected bands becomes greater as we go up 30 in the frequency spectrum, the method of selection herein disclosed has the advantage that the separation between bands is the same at higher frequencies as at low frequencies. Hence even if the design is such that the cutoff is not as sharp 35 as a band filter at low frequencies, it will give sharper selection at high frequencies.

It is scarcely necessary to discuss the applications of my invention. It has been disclosed as a method and means for performing functions simi- 4.0 lar to those of high and low pass filters. With additional steps the functions of band selection and band elimination filters may also be performed. It is obvious that a device capable of performing the above functions will have manl- 45 fold uses. One possible application, will however, be mentioned by way of illustration.

A possible application of this method of dividing a band of frequencies is in the reception of a single sideband of a double sideband transmission. 50 The arrangement of apparatus shown in Fig. 2 may be easily adapted for radio reception. The source of signals SS will for this purpose comprise a radio antenna with or without a radio frequency amplifier and tuned circuits for discriminating in 55 favor of the desired radio transmission. Both sidebands will be transmitted to the modulators MDi and MD2. The carrier frequency oscillator CO will generate a frequency equal to the carrier frequency of the radio transmission, or it might co even be derived from the radio carrier. The operation of the apparatus has already been described and it will be seen that by either adding or subtracting the two demodulated waves, at the output, the resulting audible frequencies obtained 65 will correspond to either the upper or the lower sideband of the original transmission.

While this invention has been disclosed as embodied in certain particular forms, it is capable of embodiment in other and different forms without 70 departing from the spirit and scope of the appended claims.