US1832366A - Electrical communication system - Google Patents

Description

Nov. 17. 1931.

H. w. DUDLEY 1,832,366

ELECTRICAL COMMUNICATION [SYSTEM Filed July 8, 1930 3 Sheets-Sheet l SOURGEOFSPEECH l rats/um, on RECE/WNG R 5M5 my 4/ R AL m A, i

.s- FIG IA'ENERGVLEVEL 1- 0- Loss T 5 -/vmar LEVELL Fla. 2

HIGH LEVEL mrmrmmcs l n El l E g LOW LEVEL INJERFERENCE FREQUENCY FIG. 3

Ul-T

mmuawcr //v VENTOR H. W. DUDLEY A TTORNEY Nov. 17, 1931. H. w. DUDLEY ELECTRICAL COMMUNICATION SYSTEM Filed July 8, 19:50

3 Sheets-Sheet 2 FIG. 5

FREQUENCY Fla. 4

(CUT arr a 5 H33 x855 NOISE F RE OUENCY FREQUENCY W 0 I c k .2 w a a .u m G m l U H 4 U F H 0 F c 2 0 m m a r M w. w r y w 1 33 H53 33 mswq M B 4 W 2 a a N B r 7 u w m 6 m H w l H C 0 F R F 5 c F a r E r P N m w W S 3Q mxmq 3Q HMS-3 m U W m H V, n m w m 2 E .G m G U I P O F M F F m T M E E N r M M K s s 35 fixwq 3! H53 INVENTOR By H. W. DUDLEY A TTORNEV Nov. 17, 1931. H. w. DUDLEY 1,832,366

ELECTRICAL COMMUNICATION SYSTEM Filed July 8. 1930 s Sheets Sheet s A, .,s A, FIG-9 A, f

I, H I; I I I 1 l WEST EAST LINE .3 LINE L QH i i A, R A, A, s A,.

A, FIG. /0 A, A,

H H H '1 a f l A, R f1; 5

A, Pk FIG. k 4

1, 5/? H H H H 5/? 1 l l I I 3 I l SENDING SEND/N6 I REGE/V/NG RECEIVING EQUAL/ZER 11] ii] EQUAL/ZEP A, 1 I

INVENTOR H. W. DUDLEY BY/WW4- A TTORNE) latented Nov. 17, 1931 ain-m.

PATENT oFFIcE HOMER W. DUDLEY, OF EAST ORANGE, NEW JERSEY, ASSIGNOR T BELL TELEPHONE LABORATORIES, INCORPORATED, or YoiaK, n. Y., .A- CORPORATION on N W YORK uncritical. GGMMUNICATION sYs'rnM Application filed my 8; 1920. Serial n5. 466,460.

Thisinvention relates to Wave transmission systems, as for example, telephone systems.

Objects of the invention are to decrease deleterious effects of disturbing energy and to improve overall. response-frequency characteristics.

Often signals must be transini ted through a noise or interference zone, that is, through region in which noise or disturbing currents enter or are set up in thetranslnission system. llllethods have heretofore been. proposed to increase the signal-to-noise energy.

ratio obtaining at the receiver. One of these prior methods was to amplify the signal Wave before transmitting it into the interference region. May 31, 1927; Hitchcock ?atent 1,507,178, September 2,1924.) The noise and signal after emerging'from the interference region were attenuated sufficiently to reduce the signal energy level at the receiver to the desired value and to correspondingly reduce the noise energy level at the. receiver. By this method, if the signal amplification (and therefore the margin of energy infavor of the signal) be suiliciently great, the noise energy can be reduced to a negligibly small valuewithout effacing the signal. That is, the effect of the disturbance can be made negligible if suflicient energy is used at the sendiug end of the system and sufficient uniform or distortionless attenuation at the receiving end. However, limitation on the practice or such method is that restrictions are imposed on the sending energy. For example,

in many systems a sending end amplifier or other transmitting apparatus is operated at its maximum permi 'ble load (for the shape of the Wave it is transmitting), regardless of disturbing energy that may enter or be set up in the system between the sending end and the receiving end. In such cases, if greater amplification of each frequency component (Latour Patent 1,6303%:6,

troduces distortion or noise. This distortion limits the advantage to be gained from increasing the sending end power of each frequency component with the given apparates. in other Words it is of no avail to reduce the disturbing elfect of a given interference energy if in doing so the system becomes so overloaded that, for example, because of harmonic generation (and other distortion frequency or distortion component generation), the signal suffers a degradation that more than oil-sets the advantage of the reduction of the effects of the given interfering energy. 1

However, it: has also been pointed out heretofore (Hartley Patent 1,787,843, De cember 3,1929) that, even When the load limit of the sending apparatus has been reached (for the shape of the Wave it is transmitting), the deleterious effects of noise upon the received signal can in many or most still be reduced, by changing the Wave shape. The frequency distribution of signal energy of time-averaged signal power sent from transmitting station generally is not uniform. For instance, as indicated by the curve showing the energy. frequency distribution of average speech, page 79.0fH. Fletchers Speech and Hearing, D. Van Nostrand Company, New York 1929, the frequency spectrumof the energy of normal speech is not uniform; nor is it uniform after .modification by the microphone or other sending apparatus. To the contrary, the time-averaged value of the power in speech is greater for some frequencies than for others. That is, the frequencyspectrum of the signal contains frequencies of compara tively lOW energy level and frequencies of comparatively high energy level. (A high energy level occurring at any particular frequency region doesnot necessarily imply high sound intensities in the region, but may .result from either ada'rge number. ofspeech distribution of the noise energy is not uniform over the utilized frequency range, the maximum value of single frequency timeaveraged power or of single frequency energy occurs at a different frequency for the signal than for the noise. A. method heretofore proposed for so changing the wave shape of signals as to increase the amount that noise effects can be reduced, is to change the frequency distribution of the signal energy (as for example, speech energy) before the signal is sent through the sending end amplifier or other apparatus that otherwise might be too greatly overloaded.

According to this method heretofore proposed, this change in the energy-frequency relations in the speech before transmission may be effected by the use of a network having transinission-frequency characteristics such that the network discriminates against the transmission of components in the frequency regions of certain signal energy levels and in favor of the components in the frequency regions of certain other signal energy levels. (For example, it has been proposed that under certain conditions the frequency zones discriminated against be those in which the interference or noise energy to which the line is subjected is small and those frequency zones which the network favors be those in which the interference energy is relatively high.) A network at the receiver having a characteristic complementary to that of the transmitter network then restores the speech to its original energy distribution and at the same time attenuates the noise. It is found in practice, that an increase in intelligibility by the use of such complementary networks can be effected if the networks are given the proper characteristics. Instead of making the net work at the transmitter and receiver exactly complementar they may be so designed as to compensate for distortion in the system such as that due to unequal attenuation of the different signal frequencies. For example, if desired the overall system attenuation obtained by adding the individual attenuations of the transmitting and receiving terminal apparatus, and considering amplification as a negative attenuation, may be made the same for all frequencies in the entire frequency range which it is desired to transmit.

It has heretofore been proposed that the predistortion of the signal (that is, its distortion before it is transmitted to the interference or noise zone), be made such as to give it a distribution of energy with respect to frequency that is substantially the same as the distribution with respect to frequency of the noise to which the signal is subjected A particular proposal has been that when the noise-producing interference has a substantially uniform distribution or energy with respect to frequency, and before entering the interference zone it passes through vacuum tube amplifiers which have a limited load carrying capacity, the predistortion of the signal be made such as to obtain HPPIOKl' mately the same energy level for all of its frequencies, in order to bring the energy level of each of its frequencies up to the maximum uniforn'i energy level for the apparatus included in the .ystem. (It has also been proposed that, to override a noise having a narrow frequency range, the predistor tion of the signal be made such as to provide a particularly high signal energy gain for the interfering frequency range.)

As indicated above, then, it has already been proposed to overrioe interference by in creasing the sending energy of the signal and moreover, make the increase elective with respect to frequency so that overloading will be less than were the sending energy merely amplified uniformly over the unutilized frequency range; and as also indicated above, it has heretofore been particularly proposed to give the sending energy a frequency distribution like that of the noise energy, or equalize the sending energy to a flat energyfrequency characteristic when the limiting factor determining the sending energy is overloading of amplifiers and when the noise energy (or the average noise power) of each frequency in the utilized frequency range is substant ally the same so that the interfering energy has a flat energy-frequency character. However, those proposals assume a fixed circuit loss and a lined upper energy level at the different frequencies in the utilized frequency range. In accordance with the present invention, account is taken of the effect upon the overall transmission clr acteristics of a communication system wl .b may be produced by varying its characteristics of transmission loss versus frequency, upper limitation on signal energy versus frequency, and lower limitation on signal energy versus frequency, and by means of the ii'ivention the most favorable relation between these three characteristics is established. For brevity, these three characteristics will he referred to herein as T, U, and L, respectively. '1 is the loss between the portion of the s em in which U occurs and the portion in which L occurs.

In accordance with tl e invention it proposed to send the power into the exposed region with such a frequency characteristic that the interference (if any) up at the high level end of the system, (as for example, interference set up by overloat'ling) of the same order, measured in interfering effect, as is the interference picked up at the low level end of the system. In particular, for the case of speech it is proposed to take account of interfering eff ct and consequently, in view of variation with frequency in discause the energy level of any given frequency that would. result in the predetermined contribution to the interfering effect is not independent of the energy level at other frequencies. However, by assuming such independence, a characteristic can be determined readily which is a close approximation to the characteristic U For brevity, this ,unroxii' e characteristic will be ded U By then taking account of the olved in the asum )tion of the in.-

just mentioned, from U the stic U can be determined to the requisite degree of accuracy.

For any given signaling system, as for eX- ample, a speech transmission system, havmg any given characteristic of interferlng ener 7 versus frequency, there is a characterisic of lower limit of message energy versus frequency, such that, were the predistortion of the messae'e energy such as to obtain this latter characteristic (with the given T of the system), the loss of signal quality resulting "rom the disturbing): energy of one frequency of the utilized frequency range would he a predetermined amount, equal to the quality loss resulting from the disturbing energy of any other frequency of the utilized range. For brevity, this characteristic will be designated L and the characteristic of the actual lower ev'el of message energy versus frequency will be designated L L will ary with the quality desired, it being higher lor hotter quality.

In one specific aspect the invention is a mi system in economic equilibrium as ards U, L and T.

in another specific aspect the ii'ivention is a sigmdinst; system having means for rendering its characteristic U substantially the same as,or more nearly of the same form as its characteristic U regardlcls of what the form of U may be and regardless of the frequency distribution of the disturbin energy to which the signals are subjected and the resulting characteristic L F or example, the characteristic U of the message energy may be made to approach the characteristic U when U is not flat but is, for instance, of the form mmu oned above as resulting; from its deto 11 iticn by the limiting factor of ores"- talk into other circuits, or is, for instance, of a form. n entioned hereinafter as resulting from its determination by the limiting factor of modulation effects; and moreover, such approach to a characteristic U that is not flat may he caused either when the frequency distribution of disturbing energy to which the system is subjected and/or the resulting characteristic L is flat or is not flat, and furtheri may he so caused either when T is fiat or when it is not flat. By the way of further example, the form of the characteristic U of message energy may he made to approach that of the characteristic U when L is not flat and either when U is flat or when it is not.

In another specific aspect the invention is a signaling system in which'the value of U2 is chosen so that the definitely determined interferin effect received by the distant subscriber (or recording instrument) due to high energy level elf-cots balances the definitely determined interfering effect due to the low energy level effects in such a way as to make for a minimum total interfering effect. In man cases this will mean that equal interfering effect results from the high level and low level interference. In such cases U is a compromise between U and (L +'l).

In still another specific aspect the invention is a signaling system in which U is set for other than intelligibility reasons, for example a breakdown voltage limit, in which case there is a variety of choice of U for nuilti-frequency signals and U is chosen to lit the particula U that leads to minimum interference effect due to low level energy.

if U and L could he made the same as U and L resocctively, at the same time, then each small or elemental frequency band of energy in the utilized frequency range would contribute equally to the degradation of signal quality, the degradation resulting from both the limiting factor that primarily determines U and the limiting factor that primarily determines L This ideal case is then very simple. 'lowever, this usually is not practicable; for practical considerations limit the degree to which it is feasible to control U T and L Therefore, it is often desirable to make U depart somewhat from U in order that L may approach L more closely. For example, when the characteristic obtained by plotting (L +T) versus frequency lies above the characteristic U at some frequencies, or in other words when the characteristic obtained by plotting (U -T) versus frequency lies below the characteristic L, at some frequencies, then U is made a compromise between U and (L +T). In general, the compromise value of U is made substantially the average of U and (L 'tT) at each frequency, unless U and (L -t1) differ considerably in which case U is taken with due regard to the fact that a great increase in decibels can be made in a small quantity without making it appreciable compared to a large quantity.

A feature of the invention is the reduction of masking effects of interfering waves of given frequencies in the utilized frequency range upon signaling waves of neighboring frequencies in that range, by discriminating against transmission of the interfering waves and the signaling waves of the given frequencies relative to transmission of waves of the neighboring frequencies. This method does not depend upon alternating the ratio of signal energy to noise energy at any frequency, but reduces masking effects sufiieconomizing equalizers or networks having non-uniform frequency-response characteristics. has been found that for certain typical communication telephone circuits,

there is little d'fierence between the transmission loss frequency characteristic of the equalizer that would be required to convert the electrical output of the telephone subset to one uniform over the equalized frequency range and the transmission loss-frequency characteristic of the equalizer that would be required at the subset to render the air-toair transmission,loss-frequency characteristic of the system flat (assuming the two ends of the system alike as regards equalizer equipment). Therefore, in accordance with the invention, the average of the two characteristics may be taken as the. characteristic to be used for both the equalizer for the sending end energy and the equalizer for the receiving end energy, and theninstead of employing a-separate sending and receiving equalizer at each end of the circuit, only one equalizer need be employed at each end, the-one equalizer serving as both the sending end equalizer and the receiving end equalizer for that end of the system, when sending from that end and receiving at that end, respectively.

Other objects and features of the invention will be apparent from the following description and claims.

Fig. 1 of the drawings is a circuitdiagram of a one-way signal transmission system em- 'bodying a form of the invention;

2, 3, l, 5, 6, 7A, 2 13, 70, 8A, 8B and 8C show curves for facilitating explanation of the invention;

Fig. 9shows a four-wire system embodying a form of the invention;

Fig. 10 shows a twowa-y two-wire circuit en'ibodying a formo-f the invention; and

' 11 shows another form of two-way,

the invention. I v F 1 shows a system including a source 1, of signal energy (as for example, a telephone transmitter), some of which is to be transn'iitted to a signal receiving circuit or me ns 2 (as for example, telephone recei er) over :a path or transmission medium which includes in someportion of it at least one re 'ion of exposure (For example, a single region of exposure is shown.) Before this region Gf'EXPOSUlG are sending end equalizer and one-we 1 amplifiers A and after are receiving end equalizer R amplifiers 21,. Such a system may .11 its j-haracteristics U, L and T, referrezil abo as shown in F 2, for example. The lower limitation L is determined in view of the effects of the exposure conductors.

vfor. the standard of transmission required; thus, if noise is picked up the speech or other signal energy must be kept above a certain level as compared to the noise or intelligibility will be impaired too much. Between the sets of equalizers and amplifiers, is the transmission-loss T which is definite quan- "tity of known characteristic for any given typeof system. Somewhere in the system the upper limitation on energy level, U, is present, either in the sending end amplifier or in a portion of the system in the exposed region, presumably in the first part of it before the energy level has been attenuatedby a portion of the transmission loss T Q If this upper limitation U is set primarily by some limiting factor that degrades the signal (for example if it is primarilyset by modulation), then thelevel of the signal energy that primarily determines U (for example the energy of the modulation products) can be represented by some such curve as N. I

As indicated above, any of the characteristics U, Land T can be varied to improve the quality or figure of meritof the system, by varying properties of the system upon which depends the cost ofthe system. An example noted above is an improvement in T obtainable by decreasing the resisance of As noted above, a sli 'ht improvement in this respect usuallygives an incrementalimprovement at each frequency inthe quality or figure of merit ofthe system. Tcan be represented by a family of curves, each successive curve corresponding to increased expenditure and to increased improvement; similarly with U and L. "The definite relation of U, L and T that, as noted above, yields any given value of quality or merit of the system with a minimum value of cost, or in other words, with the system in economic equilibrium, obtains when the improvement of the characteristics or variables U, L and T has been carried as far in the case of each individual variable as is required to give a common rate of change of .cost with quality or merit.

.'Let (U T. L') costof system as afunction of -U, T and .L; and Q (U. T. L) =qua1ity or merit as a function of U, T and L. 7

Then economic equilibrium exists when iii obtain improvement, each method of Variation should be considered. hus suppose U can be varied in 1* ways, T in s and L in i giving 1' families 01" U curves, .9 of T and if of L then the above equations become %QE 65S QU 83B d li 6&5 6T dU dQ dU 6Q dU dQ; 6T 6Q 65; GT 633 6L 5T 6Q dL 6Q 6L 8Q The process discussed here may become complicated it an attempt is made to set up the equations of the general systems. However, graphically it is relatively simple to consider pairs of values U l; T T; L L, such that each pair corresponds to about the same increment of cost or quality and compute the other increment. It the ratio of the increment is approximately the same then the system is near the economic equi ibrium point whi e if one ratio is materially diil'erent from the others, a cheaper system of the same Q, is obtainable by properly adjusting the values of Tl, T and L.

The system such as that shown in Fig. 1 may be, for example. a submarine telephone cable system having its U L T, U ch aracteristics. referred to above, of the form indicated in Fig. 3. In determining the IL characteristic to u e for designing the sending end equalizer for a given system which has a given T characteristic) it should be borne in mind that, although the svstem usually is such that it is not practicable to have (U -T) equal to L1. ideally that relation should obtain: for then U could be made equal to U and consequently (Uy- Tl would equal T1 or, in other words L wou d equal L The eoualitv of L and L would mean that at each utilized freouency the actual low value of message level was just hio'h enouorh to keen within the desired limit the contribution of the low level noi e energy of that frequencv to the sisrnal degradation; and the equality of U and U would mean that at each utilized frequency the actual upper value of'message energy was inst low enough to keen within the desired limit the contribution that the greatne s of the maqnitude of the message en rgy of that freouency caused to he made to the signal de rradation. The deal condit on is that in which L equals L U euuals U and approi'imatelv eoual signal dec'radation resu ts from the limiting factor (for example overloadin z'l that nrimarilv determines L and the limiting factor (for eitampl inte terencel that primarilv determines L If at each instant the mpressed siq'ual electromotive or other force at the po nt of exposure were sinusoidal. then the U and L. cur es could readilv b determined. The limitation L is primar v set by interference pick-up and conseouentlv L should be at each frequency a certain number of decibels above the energy that would be masked by the interference, this number depending on the desired standard. If this number of decibels cannot be obtained at each frequency then it must be e'l'l'ectively ohtainec over the frequency range by having suliicient excess or impr venient at some frequencies to make up for the deficiency at other frequencies. iVhen U is primarily determined by overloading, U (which is the same as U Where only one frequency present at a time) would be determined by the amount of overloading the system perini. ible to the ear or other intelligence receiving lf); into account the variaoevice r tion that the rcceivin device given in quality of transmission for the same amount of overloading at di'l'li'erent frequencies and also the dill ercnce that the receiving device requires in inaslc ratio at dililerent frequencies for a njiven igure of merit.

For the case of signal energy such as that from speech or music, at any instant instead ot a single frequency, a set of harmonic frequencies is present except, of course, during the silent interi i. For any given system the value oi? il at any frequency is now dependent to some extent at least on its value at other frequencies. Thus, it it is important that a given frequency be received in good volume. '1' en this 1' t dt may be attained to a slightly r extent if the energies at other trecic To simare decreased somewhat. he prob m let us first assume that this 7, ect is negl h account oi the changes required for the actual conditions (Case H).

Case 1'T72.c eralue of U at (my frequency (itemized to be inrZep-enalcnt 0f the "nah. ailowesl for other i-er uc'ncz'es 2'. a. U U For speech, then, the desired sending; end equalir r characteristic is obtained as follows illneenerig' irequency characteristics of all sounds to be transmitted are plotted and. an envelope drawn over these. The equalizer niinimini'i loss capable of correcting" this envelope to the shape of U is then desia'ned. Next, sufidcient amplification added to bring the level up to U For transmisiou of signal energy other than that of speech a corresponding procedure would he employed so that deta led discussion for such cases is unnec ry. Several questions arise in this procedure under this Case I, which w l be discussed below:

It mpossihie to get energv-frequeiicy charai ics oi a l sounds by all people so thin r that is iairl v representative must All. i. k The up ruin ei energy versus l rency for four male am. two female "l'\"l lav (h'azula l a l a in the .l Review. l

i in a par icunr design which led l d The is a satisfac ory: jji'filifidlll't. on a nnnner ble {Case I) and then ta e voices of voices Would, of course, be very desirable Whether theaverage or the envelope be taken. As it is only thecurve shape (and not the ab solute values) that is of importance here, the average and envelope values lead todesigns that do not differ very much. If it ie assumed that difi'erent frequencies have practically the samedistribution of relative levels,

theneither average or limiting energies lead to the same results. This assumption while not precise, is satisfactory for the practical design of the sending end equalizer.

2. The proper standard to set is in general one on the basis of intelligibility but sometimes one on the basis of naturalness or a compromise between intelligibility and naturalness. This would bring; in the-Question of the frequency of occurrence of the different sounds and also perhaps the listeners familiarity with them. Accordingly, it would be very diflicult to follow this procedure precisely. However, for practicalpurposes approximately this is effectively done When the average curve of energy versus frequency is taken for a large number of speakers for c0nnected speech. because then unusual characteristics of individuals as Well as uncommon ei'ierg'y distributions due to unusual Words are averaged down.

3. After equalizing and amplifying. the desired characteristic U is obtained. his does not mean for a uniform characteristic that in speech energy corrected to this energyfrequencv characteristic. at any instant the distribution of speech power over the utilized frequency range is uniform. However, it signifies that the distribution will be more nearlv so than in the present commercial telephone systems. If the frequencv distribution of instantaneous speech power did not change from instant to instant, there would be nothing to differentiate the various tem- DGI'Hl energy elements, and consequently no intelligence could be transmitted except by the effect of starting and stopping energy flow). When the limiting; factor determininn U for a circuit is transfer of energy to other circuits. that is. production of cro sta k or noise in other c rcuits. in general t s characteristic is less flat than when primarily determined by overload, and slopes in many cases in such a Way as to indicate that less eneroy can be transmitted at the higher freouencies in the utilized rano'e tli. lower frequencies, for eoual interferin effects, i. e. for equal intelligibility contributions to the disturbed circuit at each utilized freuuency.

The relative contribution of vowels consonants to the standard-of quality establ" Vowels are ished is of some importance. eceived satisfactorily in'general on account of higher energy levels and ener 'v concentration at the lower frequencies. T he design can be made to help the consonants thee pense of the vowels, if it is desired to increase intelligibility at the expense of naturalness. The energy for consonant sounds at most frequencies below 3,000 cycles is considerably less than that for vowels. At higher frequencies the energy of the consonant sounds is relatively more important. Thus, some slight gain in intelli ibility can be introduced by an equalizer design based on energy distribution in consonant sounds. However, increase of intelligibility Wit 1 such a system is at the expense of naturalness of sound and therefore such a system is not desirable in most cases.

In Case I (as in the case of transmission of a singlefrecpiency at any instant), When the upper limitation is not such as to produce interference energy that degrades the signal (for example when the upper limitation is a voltage limitation because of dielectric properties of insulating mater'al), the sending end equalizer plus amplifier system has been determined by two things only: (a) The energy level. of the signal at each frequency upon reaching the beginning of.

the exposure; and 6) The characteristic U at this point. Thus, in this case, as Well as in the case of single frequency transm sion at any instant, the design of the sending end equalizer is determined completely by con ditions existing at the beginning of the exposure. In particular, it is noted that the transmission loss characteristic over the es.- posed region (thatis, the characteristic'T) does not affect it nor does the limiting factor (for example, interference) that primarily determines L f the receiving end eou izer is designed to give, in conjunction W n the sending end equalizer, a certain air-to-air characteristic, either flat or typical of similar circuits, then the receiving end equalizer characteristic is also determined independently of that limiting factor, so that sign of'the optimum equalizing arran for minimizing the effect of the. factor is independent of that limiting the sending end equalizer being also independent of the loss T over the region of exposure.

Case I! 1lc2f'2mZ speech t)(Z-HSPPRZSS'ZO?l-.TI1 the case of speech transmission the assump tion that for the characteristic U1 the energy level for each frequency is independent of the levels for the other frequencies very good first approximation, so that We can consid r the procedure as that of determining the two equalizer characte 'cr (i. e., the sending equalizer eh i and the receivingequalizer characteristic), as discussed under Case I .d then modifying; these slightly as ch ussed later to fit better the speech transmission circuit in the particular problem. From computations of the energy components of representative vowel:sou11ds,it appears for the vowel sounds that after correction to give a uniform (i. e., flat) characteristic, the di'fierences Between the total energy of the vowel sound (that 1s, its energy over the whole utilized frequency ran 'e and the n Xunum ener v of the voxvcl h 1 an sound at any single rrequency averages approximately .lecibels. From this figure it is seen that the er 2* of the assumption,

l \1 Y-\ i" *l n i l ve Q 7531 (vase a )OXC}, .nao lac equa 1M1. loss no at any frequency vas to be chosen though energy at other frequencies did not ailect the proper value of equalizer loss at that frequency appr inmates an average of about (3 decibels. This is int very large. When the transmission i055 in the lQfllOll of ex- L energy limitation L posure ant. the lone are of such nature 15 permit utilization of most. of this nount. the c rreetion would be more important than is the case in many systems Where only a small fraction is Oitainable.

Fig. 6 shows a tvpicl. which, in improvin ing the effect characteristic U, of a given i 1 mission (decreasan adthe speech energy is hat a the sending end of the exposure it wil. slone o; frequenc es after aversin; t region. This is. indicated by ie ecuosed the solid lines U, and (ll- 5) Elome gain then can be obtained bv tron mittinsz the higher frequencies at slio'htly higher level, at the expense of appreciably lower levels for the lower frequencies. The dotted curves labe ed modified U, ana modified (U indicate this. In this cas a; the high frequency end masking serious. @ome improvement is gained here although this 'n in part offset by increasing masking" at the low frequency end. However, as the masking; was at a much lower level at the low frequency end, some netaain is obtained. The gain in a circuit of this sort by making the sending energy level other than n," 'lently less than is evic the average decibels) given prerourh es-imated figures for a falling in this case mie'lit level. Deal: the sending 16$ elf-active gain, taking e of n asl'ing of norain from When the submarine c? l factor t L L, has a decided r row frequency ran. es, then be expected in these an rections, since at 5 more grain can e di e to these corirecue miss the energy Y i l 1 ,1)? l it is ueghgihle in quantity, thus permitting the realization of the 6 decibels gain. Approximat ly two decibels more may be realized in those cases in which the effect of the changed s 'lcctrum, at the point of the circuit vrhcre the. limiting factor that primarily determines the U enters, is to permit more energy for the same loss of intelligibility or other standard of performance. Similarly, in other cases tl e opposite effect may cause a loss of approximately two decibels. As an QTiLli'lljlG, if the dist u'bing etl'ect of the H ird harmonic produced by overloading is the limiting factor that prima 'ily determines the in a speech transmission system then the peel-t it 3006 cycles would have less harmful ellect than peak at a lower frequency, since the harmonics would be produced in a range in which there is little speech energy, While a peak at 350 cycles would have more harmful ei'iect the third harmonic .vould occur in a most valuable speech region. Another effect to be considered in designing the equalizers the shape of the masking; curve. In general, interference energy at an frequency masl lit only intellig bility from that frequency, but also that from others, particularly, from those adjacent. If the maskingcur e is higher in so 1e frequency range we may infer that inter iencc ener; in that range is l'iigher and is tl'ierefore ma ing the contributions of adjacent freouenc The receiving end equalizer It c.

signed to give additional loss over this frequ ncy range in order to obtain some not advantage, that is. to obtain more arhrai'ltage from the decreased mashing at the adjacent frequencies than lost by e sing the frequency distortion. By tl is method We do not change the s' nal to ratio at any frequency, but decrease the masking of frequencies that are adjacent to the high mag; nitude noise f equencies. Examples of this are shown in Figs. 7-A, Tl3 and 7-C for an ideal case of a frequency range of heavy interference pick-up and in Figures 8;l, 8B and 8-C for the common commercial case of interference increasing with f'equency.

In Fig. 7-A the curve 10 for received energy level of speech shows the speech en ergy flat as it approximates over a limited frequency range in a high quality system. Portion 12 of the curve 11 for received euergy level of interference shows 1 frequency range of high interference leve in the utilized speech frequency A gain telliglbility is obtained by sufficiently attenuating in the receiving: equalizer over this frequency range (or hy sha 'ilingr the energy curves in th e cud. ampl and equalizers to tiatt iYGtl i ference energy so that ence energy Will be as s l V in Jig. 7B, as then the interference enan be deergy, in: the frequency range: in. which it had the higher level shown at 12 in Fig. 7.-A, no. longer has" that high level, but is. sufliciently low to. not greatly mask the adja- 7--A there would be appreciable masking of the frequencies just abovev and. below the high level portion 12 of curve 11. The equalizer attenuation (or the receiving amplifier and equalizer correction) which flattens (or changes) curve .11 .to curve 13 also changes curve 10 to curve 14, which: has a depressed portion over the frequency range of the high level portion 12 of curve 11. Fig. 7-C shows a lesser correction than Fig. 7B. Curves 13 and 14 correspond to curves 13 and. 14. Curve 13' has an ele-' vated portion 12 over the frequencyrange of portion 12 of curve 11; and the receiving equalizer attenuation, being insuflicient to entirely remove the elevation 12 from the interference energy level curve, produces a speech energy level curve 14 that has a depression to 15 less deep than the depression 15 in the speech energy level: curve 14. Figs. 8A, 8B and 8C correspond, respectively, to Figs. 'Z-A, 7B and 7C, but in Fig. 8A curve 21 shows the interference energy increasing with frequency, typical of many commercial systems- A gain in intelligibility is obtained by sufficiently attenuating in the receivingequalizer R (or by shaping the energy curves in the receiving end amplifiers and equalizer) to flatten the received interference energy so that the received interference energy will be as shown at 23 in Fig. 8B, as then the high frequencies'in the utilized frequency range have sufficiently 'low interference energy to not greatly mask the lower frequencies in the utilized frequency range, whereas in- Fig- 8-A there would be appreciable masking of these lower frequencies. The receiving end correction which flattens or changes '6 curve 21 to curve 23 also changes curve 10 to curve 24.

Fig. 8--C shows a-lesser correction. thanv Fig. 8-B. Curves 23 and 24 corresponding respectively tocurves 23 and '24 as curve 13 and 14. The best results are obtained when the signal degradation caused by the added overall: signal distortion equals that caused by the interfering waves; that remain after the insertion of. theadditional attenuation,

and equals that arising from the upper lim-v the receiving equalizer R is such. thatv the: sending. end: acoustic correction. is. removed 46* the. rest. of. the systeim from: the sending cent utilized frequencies, whereas in Figand 14' correspond respectively to curves 1 3 end of the exposure to the listener is corrected; (2) To give the system the same airto-air. characteristic as. a corresponding systemwith no region ofexposure, the receiving equalizer R is complementary to the sending end equalizer and contains in addition equalization for transmission over the range of eX- posure and (3) A condition between these two conditions is sometimes desirable. particular, the compromise between flat speech level and flat masking level for the. observer as discussed previously isv often desirable.

The specific formindicated by Fig. 3 for the U L T and. U curves is merely illustrative.- The system. of Fig. 3 may be a system having various other forms of these curves- For example, instead of being a submarine telephone. system having its U1 and L curves non-uniform with'respect to frequency, as indicated in Fig 3,. it may be a submarine telegraph cable with characteristics such as those shown in Fig. 4 or a telegraph land line with characteristics such-as: thoseshown in Fig. 5.

In Fig. 4, L is fiat and T and U rise with frequency. L may herebe determined primarily by a uniform noise energy level indicated in Fig. 4. IVith such a noise levelv the characteristic U may be substantially flat since-thereceiving device may be an ordinary type of telegraph relay. U may here be de. termined primarily by cross-talk to adjacent cables. In Fig. 5 L falls with frequency rise. T riseswith frequency and U is substantially flat. L may here be determined primarily by a'noise energy level which deacreases with frequency as indicated in the figure, the receiving device being a telegraph relay. U may here be determined primarily by amplifier overloading.

Two purposes can be served by equalizin 1 an exposed system (1:): Improvement ofthe air-to-air response characteristic and (2.) Decrease ofthe effect of the interference en- 11o ergy pick-up in the exposed region. The enposed region in a portion ofv the circuit where the energy level of interference or; more strictly, a masking level is not sufficient ly'low compared tothe level of the communi- 115 cation energy at that portion. Thus, in a system having considerable attenuation it may be that the" interference energy pick-up is greater in other portions than it isin the portion where the interference energy is a 120 limiting factor. In other'words, it is not the absolute level of the interference energy, but its relative energy at different frequencies compared to the intelligence-transmitting energy'that must be considered. Circuit limi- 125 tations, particularly singing greatly affectsthe choice of networks and would often make. desirable an arrangement that otherwise is not; some cases the problem is one of fitting the corrective network in a circuit al.- 1 30 ready in existence without disturbing its stability. In other cases, a design of a new.

circuit is to be made with the correction network as a portion of it. In the latter set of cases, there is more latitude.

1 in Fig. 1, it is sometimes desirable both to use an amplifier before the equalizer so that the equalizer does not attenuate the energy to too low a level and to use an amplifier after the equalizer so that the amplifier will not need to carry too large a load. In general, at the receiving end, an amplifier will be used before the equalizer to build up the energy level before it is attenuated further. In many cases a single amplifier will be suflicient, particularly at the receiving end where the level required may be less than at the sending end. Figs. 1 and 9 to 11 shows circuits with a sending end or receiving end or combination type equalizers, S, R or SR respectively. The

'J- sending end equalizer S is to be designed so as to give energy equalization to correspond with U at the beginning of the exposure. The receiving equalizer R is to be designed for one of the three purposes: (1) To correct for the rest of the circuit so as to give a flat overall characteristic (R (2) To compensate for the loss characteristic of the sending end equalizer (R (3) Something intermediate between R and R (R The several circuits are discussed below: 7

1. Simple one-way circuit, F ig. J.In this case the circuit is preferably suitable if properly designed so there are no limitations on circuits developing as in practice this correj sponds to program transmission circuits.

These are already of high quality so that the possible improvement is not so great as in circuits used in commercial telephony, but on the other hand any improvement is, of

a: course, much more valuable relative to the number of circuits involved.

[1. 4-wire circuits, Fig. .9.Most long distance two-way circuits are four-wire in part. Much greater total amplification is then possible for a given intelligibility, and a region of exposure may occur in one or more likely both of the one-way branches. Fig. 9 indicates the insertion of the equalizers of Fig. 1 in each of the one-way branches of the 4- wire circuit extending between lines L and L For convenience the lines L and L may be regarded as a line west and a line east, respectively. Hybrid coil (bridge transformers) or other means for changing from oneway to two-Way transmission and vice versa are shown as blocks designated H. Three cases to be distinguished here according to the type of receiving end equalizers used are: (1) If R is used the net loss in either branch is unchanged and therefore the circuit stability is not affected; (2) If R is used there is a net gain at some frequencies in either branch so that singing may occur, unless the circuit is balanced better than was required previously. For commercial telephone systems, the added gain inserted is at low and high frequencies, there being a loss over some intermediate range. As repeater circuits are more likely to be unstable at the higher frequencies singing may occur if high net gain is obtained at these frequencies, unless a voice operated device is used for changing from one-way to two-way transmission and vice versa; A compromise between R and R that is, a receiving end equalizer of the may be used. For example, the added gain in the lower frequency range can be retained and some of that at the lower end of the high frequency range also, thereby improving the air-to-air characteristic and at the same time not changing the circuit so much that singing will result.

Ill. Two-way transmission circuit, F 29. 10.For two-way transmission over a twowire circuit the system of Fig. 10 should be used. If two-way amplifiers are used at each end of the exposure, equalization can be made at them. It may be that no added gain can be permitted at any frequency, in which case as in the four-wire circuit discussed, R would cause singing, R would not and a compromise R could be found that would give a part of the advantage of R without its singing limitation. If no repeater is used in the circuit then perhaps some extra gain can be had without singing, permitting a better R type of equalizer.

I V. Unircrsal equalizer, F ig. 11.In some cases it is desirable to use the same equalizer for transmission in both directions. The transmitting end equalizer and receiving end equalizer for the case of flat upper limitations on the energy and flat air-to-air characteristic differ very little for typical commercial telephone circuits and the average of the two characteristics may be taken as a universal characteristic for use whether sending or receiving. The circuit that should then be used is shown in Fig. 11 wherein the equalizers are designated SR.

What is claimed is:

1. A signal transmission system comprising two portions related in that one is subject to interfering waves whose interfering effect can be decreased by increasing the levels at which the signaling power is impressed on the system whereas the other produces interference power which increases with the increase of signal power levels, and means for impressing the signals on the system at such power levels that the interference effect produced by said interference power is substantially equal to the interference effect produced by said interfering waves and that the total interference contributed by each frequency band ofany given small width in the utilized frequency range is the same.

2. A system for transmitting signaling waves lying in a given frequency range through a region of interference, said system comprising means for so distorting the signaling waves before they reach the region of interference that, for the signaling waves in said frequency range, the average energies at the difierent frequencies are substantially different but contribute predetermined equal amounts to signal degradation resulting from the greatness of the magnitudes of those energies.

3. A two-way telephone transmission system for which the optimum values for the energy of the components of different frequencies in speech waves at a point in said system are substantially equal for the frequency range to be transmitted, said system comprising a telephone transmitting and receiving set having a higher transmitting efliciency for some frequencies of said range than for other frequencies of said range and having a higher receiving efliciency for some frequencies of said range than for other frequencies of said range, a common sending and receiving network fed from and feeding said set, said network having its transmission efficiency so varying with frequencies as to produce approximately said optimum relation and to effect such compensation for the variation of transmission efliciency of the other parts of the system with frequency as to render the overall transmission efliciency of the system the same for the frequencies to be transmitted.

In witness whereof, I hereunto subscribe

Images