Tijdschrift van het
Nederlands Radiogenootschap
DEEL 20 No. 4 JULI 1955
S ta n d a r d iz a tio n o f e le c tr o n ic c o m p o n e n ts
by H. W . Ghijsen *)
S U M M A R Y
In the first section a survey is given of the reasons w hich lead to s ta n dard izatio n in general an d of the a d v a n ta g e s to be obtained. In the f u r
th e r sections, the field of electronic com ponents is review ed. A n im p o rtan t in ternational activity is developed b y the relev an t T echnical C om m ittee of the In tern atio n al E lectrotechnical Comm ission ( I E C ) , w hich stan d ard izes along the line: definitions -—- m easuring an d test m ethods ■— simplification.
Special attention is paid to the classification on the basis of en v iro n m ental conditions of use. A short su m m ary of the progress up to n o w is given.
W hy stayidardization
I t m ay be w ell to consider the general aspects of s ta n d a r d ization before startin g on the subject of this article, especially for the benefit of those re a d e rs w ho are n o t in reg u lar contact w ith this field of technology. The m ost w id e sp re ad idea ab o u t the object of stan d ard izatio n is, th a t it is the a r t of m aking a re stricte d choice betw een a g re a t num ber of types. This, h o w ever, is only p a r t of the tru th . To p u t it in r a th e r general w o rd s : the object of stan d ard izatio n is to obtain economic a s sets for all p arties concerned by making a sensible choice from the multiplicity of possibilities. N o te the differences betw een these definitions. F irst of all, the "economic assets ', w h a te v e r th e y m ay be, and not the choice are the object, the la tte r only being the means. This m ust be k e p t w ell in mind.
Secondly, the first definition speaks of "ty p es" w here the second mentions "possibilities". This has the ap p earan ce of a pitfall an d will be explained w ith a simple example.
Transm ission shafts are functionally characterized by the
*) C en tr. N o rm . N . V. P h ilips’ G loeilam penfabrieken, E indhoven, H olland.
228 H. W . Ghijsen
m echanical p o w e r th e y can transm it. So having decided on the m aterial, and leaving aside such details as surface condition and roundness, the obvious thing to do is to lay dow n a series of diam eters and associated tolerances.
The same procedure might be followed for one or tw o oth er sh aft m aterials, b u t it a p p e a re d to be possible to cover the g re a te r p a r t of the need by one m aterial, viz. w ro u g h t steel for automobile and general engineering purposes. This forms an example of a stan d ard izatio n of types, commonly called sim pli- fication. The “ economic a s s e ts ” also become clear for this case ; the m an u factu rer will have a bigger tu rn o v er in less types an d the users will have less types in store, resulting in sm aller to ta l stock and sm aller investm ent. T hese savings can be p a rtly expressed in money.
This, how ever, is not the end of the story. The user is generally in terested to know w h e th e r w h a t he buys conforms to specification, in this case the sta n d a rd . Checking of dim en
sions does n o t raise undue difficulties, b u t checking of m aterial is certain ly more complicated. F o r th a t purpose, the dimensional sta n d a rd of the sh aft should be b acked by a m aterial sta n d a rd .
B u t here w e have only one “ possibility” of checking the d e livery. F o r example, the u ser m ay n o t be in terested in the composition of the m aterial he gets b u t in the stren g th only.
T herefore he m ay decide to c a rry out a mechanical test, such as a torque test, in stead of checking the m aterial composition.
I t will be clear th a t “m echanical te s t” in itself re p resen ts m any possibilities.
From this example it will be un d ersto o d th a t the ta s k of stan d ard izatio n is not only to m ake a choice of ty p e s , b u t also to lay dow n w h a t p ro p erties of an article will be te ste d ac
cording to exactly w h a t te s t m ethod. T here is a fu rth e r v e ry im p o rta n t task, viz. to define a c cu ra te ly w h a t is understood by the different term s applied to a certain article. H ere again stan d ard izatio n is a “ choice from all possibilities” . Going into d etail here runs the risk th a t you will la y aside this article because it's too dull, too theoretical. B u t w e just have to ta k e this risk because the im portance of good s ta n d a rd definitions cannot easily be over-em phasized and, furtherm ore, it will show one of the corner-stones of stand ard izatio n .
The opinion is often expressed th a t practice provides a suf
ficient grasp of the meaning of m ost if not all term s. E v e ry b o d y know s w h a t is m eant by resistance, inductance or w a tta g e .
Standardization of electronic components 229
This is correct, because these term s are derived from physics an d have been defined there. B u t as soon as you ap p ly these term s to an article, difficulties m ay arise. F o r instance, w h a t is the ra te d w a tta g e of a re sisto r? N o w it m ust be ad m itted th at, w hen the complete sta n d a rd for resisto rs is available, it m ay follow from the te st m ethods w h a t is to be u nderstood by ra te d w a tta g e. H o w e v e r, not ev ery b o d y is able to find out so read ily from the te st m ethod w h a t the definition is.
To find a definition is not difficult in this case. I t m ay run something like: the w a tta g e th a t the re sisto r is able to dissi
p ate continuously a t a given am bient tem p eratu re. C onsequently, a published figure for r a te d w a tta g e has no meaning unless the re lev a n t am bient te m p e ra tu re is also given and unless there is some und erstan d in g ab o u t how long “ continuous” is, th a t is a b o u t the life duration.
The above m ay be called a physical definition. The question arises w h e th e r it is ad eq u ate from the point of view of s ta n d a r d ization. L e t us suppose w e have to com pare tw o resisto rs of know n ra te d w attag es. F irs t w e have to find in the m anufacturer's d a ta sh eet the te m p e ra tu re corresponding to th a t w a tta g e . If these te m p e ra tu re s are not the same, w e have to find the p e r missible w a tta g e a t one tem p eratu re from the so-called d eratin g curves, i.e. the curves showing the p ercentage of r a te d w a tta g e as a function of the am bient tem p eratu re.
The a d v an tag e of such a physical definition is obviously th a t it ind uces m anufacturers to publish the w a tta g e rating to g eth er w ith the a p p ro p riate tem p eratu re, and so elem inates m isunder
standing. H o w ev er, the comparison of resistors of diffe
re n t m ake m ay still be a tric k y business and confusion is liable to arise. I t w ould be a g re a t a d v an tag e if it w ere possible to obtain agreem ent on one te m p e ra tu re to be used by all m anu
factu rers and users, say 40 °C. The definition w ould then in
clude all facto rs necessary for direct com parison and such a definition can be called a sta n d a rd definition.
I t will be noticed th a t the agreem ent on life d u ratio n has been left out. This is a v e ry difficult subject, to which sta n d a rd iz a tion cannot give a simple solution. O f course it is possible to ad d the w o rd s “for five y e a r s ” to “ continuous” in the definition, b u t this w ould not have much technical meaning. F irstly suf
ficient d a ta over long time perform ance are only obtained a t high cost and much trouble, and secondly, it w ould imply a long-term guaran tee. The only p ractical possibility is to p r e
230 H. W . Ghijsen
scribe in the sta n d a rd an endurance te s t of ad eq u ate length, say 1000 hours, u n d er severe operating conditions, and to spe
cify the maximum permissible d eterio ratio n a fte r th a t test.
So even s ta n d a rd definitions m ay not be strictly complete b u t for p ractical ev ery -d ay life th ey are satisfacto ry an d in d isp en
sable.
This shows an im p o rtan t aspect of the m atter. Definitions including the m easuring and te s t m ethods enable us to define the p ro p erties of an article completely. If done in the p ro p e r w a y , different m akes of articles are directly com parable and th ere can be no m isunderstanding b etw een m an u factu rer and p u rch aser or any o th er confusion. To p u t it in a n o th e r w ay :
“ all p arties speak the same la n g u a g e 1. H e re w e have an o th er aspect of the economic assets m entioned in the beginning, p e r
haps the m ost im p o rta n t one. To derive full ad v a n tag e from a sta n d a rd in this respect, a first condition is th a t it should be generally accepted and used. Again, stan d ard izatio n is n o t the aim b u t only the m eans to come to a b e tte r understanding.
A sta n d a rd m ay be com pletely sound technically, b u t w hen for some reason or o th e r it is n o t acceptable to some in terested p a rty , it is liable to increase confusion instead of decreasing it. The b est s ta n d a rd is the one on w hich general agreem ent can be reached, even w hen it is either incomplete or even not fully in accordance w ith the sta te of technical developm ent.
I t will be noted th a t in the above th ree distinct phases have been introduced. This has, how ever, been done in the w rong order. W h e n w o rk on a n ew subject is sta rte d , it is common practice to begin w ith definitions. As soon as sufficient idea has been form ed b y the committee as to the concepts th e y are talking about, so th a t during the discussions th ere can be no m isunderstanding, it m ay be good to leave the definitions aside for a while and to focus a tten tio n on the te s t and m easurem ent m ethods. These can be re g a rd e d as a supplem ent to the defi
nitions and in some cases it is easier to finalize the definitions a fte r the m easuring an d te s t m ethods have been discussed.
Finally, the simplification can be ta k e n in hand, comprising the laying d ow n of levels of minimum quality, of re stric te d series for nominal values and of dimensions. This la s t activity m ust be ca rrie d out w ith g re a t caution in o rd e r n o t to lay dow n rules w hich might interfere w ith future developm ent.
A p a r t from this, th ere are often one or tw o subjects such as colour codes an d general series of p re fe rre d values, w hich
Standardization of electronic components 231
are not quite covered by these three activities, b u t which are very suitable for stand ard izatio n .
O f course the relative im portance of the th ree activities varies from case to case. In general, the more com plicated a subject is, the more im p o rtan t the first tw o activities are. E lectronic
com ponents are [complicated.
To sum up the “w h y ” of sta n d a rd iz a tio n : the aim is to prom ote b e tte r understanding, especially b etw een m an u factu rer and user, to obtain ch eap er p ro d u cts of good q uality an d sm aller stocks and similar ad v an tag es by creating a well- defined “technical language” and by restricting the num ber of types. A very im p o rta n t fa c to r is the use th a t is made of a stan d ard .
Who standardizes
This m ay seem a r a th e r silly question, b u t it is of special im portance w ith re g a rd to the application of sta n d ard s. B est know n is the national and in tern atio n al stan d ard izatio n carried out by national stan d ard izatio n authorities such as A F N O R (France), B S I (U n ite d Kingdom), H C N N an d N E C ( N e th e r lands), and by in tern atio n al bodies such as I S O (In te rn a tio n a l
o
rganisation for S tan d ard izatio n ) an d I E C (In te rn a tio n a l E lectrotechnical Commission). This might be called “ general s ta n d ard izatio n ” , because a ll in terested p a rties are re p re se n te d in these bodies (m anufacturers, users, education, governm ent etc.). T here is, how ever, much activity in groups of in terested p arties also. T herefore, sta n d a rd s can be roughly divided into general sta n d a rd s, users' sta n d a rd s and m anufacturers' s ta n dards.
To s t a r t a t the b o tto m : a m an u factu rer m ay w a n t to s ta n dardize the r a w m aterials he uses. In principle he can do so w ith o u t consulting anybody, b u t if he is wise he will ask the advice of his suppliers. H e is m aking users' sta n d a rd s of a r a th e r limited application, b u t the s ta n d a rd s will norm ally have an obligatory c h a ra c te r in his facto ry and for his suppliers.
o
n the oth er h an d he m ay w a n t to stan d ard ize his products.N o w he m akes m an u factu rers' sta n d a rd s, which m ay have an obligatory c h a ra c te r for his facto ry b u t generally he will not have much chance to enforce these on his purchasers.
The m anu factu rers of similar articles in one country m ay come to g e th e r to follow the same procedure. In th a t case
national users' or m anufacturers' specifications arise, which as such do not have an obligatory ch aracter.
H ow ever, w hen a users' s ta n d a rd is accepted by a sufficient num ber of m anufacturers, it autom atically becomes an obliga
to ry s ta n d a rd for the supplier. N a tio n a l m an u factu rers' s ta n d a rd s are only recom m endations, b u t the value of these should n o t be under-estim ated, because in m any cases th ey form the first step to general standardization.
Finally, the same procedure can be rep eated on an in ter
national basis.
U sers' specifications are not only m ade by industry, b u t also by o th er groups of users, such as governm ent d ep artm en ts, adm inistrations, m ilitary authorities (service specifications) and b ro ad castin g companies. Since the "groups" of such users w ithin one co u n try are relatively small, agreem ent w ithin the group is as a rule easily obtained an d therefore these s ta n d a rd s autom atically become obligatory for the suppliers, w h e rea s in tern atio n ally d ra f
ted specifications of such groups are recom m endations to the n atio n al groups.
All these s ta n d a rd s can be tu rn ed into general sta n d a rd s by subm itting them to the a p p ro p riate general standardizing body (national or international) and discussing them w ith all in terested parties. G e n e ra l sta n d a rd s as such are recom m en
dations b u t can become obligatory w hen accepted by factories, companies etc. or w hen re ferre d to in a contract.
N o w let us re tu rn for a minute to the application of s ta n dards. Since the usefulness of a s ta n d a rd is equivalent to its application, the logical consequence for a facto ry m aking s ta n d a rd s for its own use is to m ake them obligatory to the g re a t
est possible extent. W i t h resp ect to national and international s ta n d a rd s it can be said th a t the ideal situation is w hen all m anufacturers and users ab an d o n th eir ow n sta n d a rd s in fa vour of the national or intern atio n al ones.
A simple w a y of prom oting the application of s ta n d a rd s w ould be to enforce them by law. In a dem ocratic society this w ould ob
viously be the w rong w ay, since only the in terested p arties are able to judge w h e th e r the s ta n d a rd is acceptable to them. A n exception should be made for sta n d a rd s involving the safety of persons or p ro p erty , b u t even then enforcem ent by law should be avoided w here possible because of the slow procedure of governm ent m achinery. S ta n d a rd s have to be k e p t in pace w ith technical developm ent and th erefo re revision is necessary from time to time.
232 H. W . Ghijsen
233
H ow electronic components are standardized
U p to W o r ld W a r II there had not been much activity in the stan d ard izatio n of electronic components. As a result, simi
la r com ponents of different m anufacturers w ere not electrically and mechanically interchangeable. This c rea te d enormous diffi
culties for the A rm ed Forces, w ho therefore sta rte d to d ra w up users’ specifications.
A fte r the w ar, w hen the m anufacture of m ilitary electronic equipm ent w as sharply reduced and w a s replaced by civil p ro duction, an a tte m p t w a s m ade to use the service specifications for industrial purposes.
This w a s not a complete success. In the first place, the m ilitary qualities laid dow n in the service specifications w ere not alw ay s suitable for industrial use, and, in the second place, w h a t in
d u stry needed w ere intern atio n al specifications. S tran g e as it m ay seem, the service sta n d a rd s existing a t the end of W^orld W^ar II w ere strictly national.
F o r this reason in some countries specifications w ere d ra w n up covering industrial and en tertain m en t components. In the U n ited Kingdom, these specifications are m ade by the R adio In d u stry Council (R .I.C .) which is an organisation of both electronic equipm ent m anufacturers and component m anufacturers.
C onsequently, these specifications are of a r a th e r general ch aracter. In the U .S.A ., com ponent specifications are m ade b y the R adio, Electronic and T V M a n u fa c tu re rs ’ Association (R .E . T .M .A .). The R E T M A specifications are m an u factu rers’ specifi
cations, which is possible because the component m anufacturers are m em bers of R E T M A .
Soon a fte r these activities w ere started , the subject w as also ta k e n up on a general international basis, viz. by the I n te r national E lectrotechnical Commission ( I .E C .) .
The constitution of I E C , which last y e a r held its 50th anni
v e rsa ry meeting, is as follows. The m em bers of I E C are w h a t are called the N a tio n a l Com m ittees, i.e. the national general electrotechnical sta n d a rd organisations in the m em ber countries, or special national committees form ed for th a t p u r
pose. A t the moment th ere are a b o u t thirty-five members.
A t the top of I E C is the Council, consisting of president, vice-presidents (all presidents of the N a tio n a l Committees), tre a s u re r and general secretary. The p resid en t is elected every three years. A t the moment it is D r P. D u n sh eath (U n ited Kingdom).
Standardization of electronic components
234 H. W . Ghijsen
The next body is the Com m ittee of Action, consisting of the president, 9 vice-presidents, tre a s u re r and general secretary.
They! fa ke the final decisions on m ost technical questions*
The technical w o rk p ro p e r is done by the Technical C om m it
tees and their Sub-C om m ittees. T here are now over 40 Technical Com m ittees, more th a n 10 of w hich have been installed since the la s t w a r. B ach Technical C om m ittee has a chairm an an d the se c re ta ria t is en tru ste d to a N a tio n a l Com m ittee, norm ally n o t the one to which the chairm an belongs. The S e c re ta ria t p re p a re s the docum ents to be discussed in the meetings, m akes the minutes etc. D istrib u tio n of docum ents and all o th er adm i
nistrative functions are ta k e n care of by the G en eral S ecretariat, the C e n tra l Office, residing in G eneva.
B efore W o r l d W a r II, m ost of the Technical C om m ittees either d e a lt w ith general technical subjects, such as nom encla
tu re an d symbols, or w ith p o w e r equipment, such as a lte rn a tors, insulators and sw itch gear. O n ly one C om m ittee w a s ac
tive in the telecommunication field, viz. T C 12 " R a d io C om m u
nications” which discussed a t th a t time safety requirem ents for b ro a d c a st receivers. The S e c re ta ria t w a s — an d still is — in the hands of the N e th e rla n d s.
W h e n the need w a s felt to do more w o rk in the electronic field, it w a s only logical to extend the scope of T C 12. In 1948 it w a s decided accordingly and four Sub-com m ittees of T C 12 w ere installed to handle the new subjects, viz.
1. m easurem ent m ethods for receivers, 2. safety,
3. com ponents and
4. electronic tubes and valves.
A t a la te r date, Sub-com m ittees for high-frequency cables, for piezoelectric crystals and for tra n sm itte rs w ere ad d ed to the list, the la tte r only quite recently.
o
n the o th er hand, the subsequent re-organisation caused a reduction of the num ber of Sub-committees of T C 12. Soon it a p p e are d th a t the Sub-com m ittee for electronic tubes an d valves could n o t re stric t itself to radio communication, since tubes find general application in the electronic field. T herefore it w as decided some y e a rs ago to convert the relative Sub-com m ittee into an independent Technical Com m ittee (No. 39). The same procedure w a s follow ed la st y e a r w ith respect to the S u b committee for components, which is now T C 40, absorbing the Sub-com m ittees for high-frequency cables an d for crystals.W hat is standardized in I.E .C . fo r electronic components
W h e n a com ponent is bought it is expected to give a certain p e r form ance for a certain time. I t is hoped th a t the perform ance will be as good as possible and the life indefinite, b u t if a t the same time the com ponent should be obtainable a t reaso n ab le cost, it is clear th a t a compromise has to be m ade som ew here.
N o w the concept of perform ance has m any aspects. In the first place the component is expected to have a certain p ro p e rty of a given value, let us sa y a resistance of 100 ohms which will sta y co n stan t over a sufficiently long period. Since nothing is absolute, it m ust be accepted th a t the resistance is not exactly 100 ohms b u t lies b etw een tw o tolerance limits, an d th a t it varies some
w h a t w ith the circum stances and w ith time. The la tte r m ay be called instability.
In the second place the com ponent will be subjected to a certain load, either continuous or cyclic or in term itten t e.g. 1 w a t t in the case of the above resistor. I t is expected th a t the instability is still acceptable.
In the th ird place the com ponents are subjected to m echanical and climatic stresses, by mounting them into an a p p a ra tu s, tra n sp o rtin g the a p p a ra tu s , heating during operation of the a p p a ra tu s and moist air during no-load periods. Still it is ex
pected th a t the com ponent will not b re a k dow n and th a t the instability will sta y w ithin reaso n ab le limits.
All these factors to g eth er govern the perform ance of the component, and its life du ratio n is obviously the time for which the instability during use rem ains w ithin pre - determ ined limits. The problem is to look for some means to estim ate the in stab ility over a certain period of use.
The first step on this w a y — ascertaining the p ro p erties of the com ponent — does not raise undue difficulties. T here are scores of m easuring bridges, dynam om eters, resisto r sta n d a rd s etc. which will give exact inform ation a b o u t the com ponent as it has been received. The only thing to be done is to specify the m easuring m ethods in such a w a y th a t the same re su lt is a lw a y s an d ev ery w h ere obtained w hen the same com ponent is m easured. Such details include am bient tem pe
ra tu re , load during m easurem ent, inaccuracy of the method, position of the com ponent etc. w ith o u t necessarily prescribing the in stru m en t to be used. F o r instance, the resistance of a carbon fixed re sisto r m ay be m easu red b y a W h e a ts to n e bridge
Standardization of electronic components 235
236 H. W . Ghijsen
or b y m easuring voltage and current, as long as the am bient tem p eratu re is b etw een 15 and 25 °C, the voltage a t the resis
to r term inals does not exceed say 25 V and the to ta l inaccu
racy of the m ethod used (including the influence of connecting w ires) does not exceed 0.5 °/0.
The next step is to verify th a t the com ponents will survive the handling previous to and during mounting. O n e w a y wo uld be to p u t a sample of com ponents into a setm akers fa c to ry and to see w h e th e r the resulting a p p a ra tu s are all right, b u t though highly p ractical this w ould not be a reproducible te st m ethod. F o r this purpose the com ponent specifications contain a num ber of m echanical tests, checking the stren g th of t e r minals, of mounting accessories and, w here necessary, the suit
ability of being soldered into the wiring. The bumping and vibration tests also belong p a r tly to this group of tests.
A fte r these tests, the com ponents m ust not have been dam aged an d the main p ro p erties should have changed to a negligible ex ten t only.
The th ird step is to verify w h e th e r the com ponent is suitable for the climatic conditions liable to be m et during norm al use.
A gain one m ethod w ould be to m ount the com ponents into an a p p a ra tu s and to a w a it com plaints over a num ber of y ears. I t m ust be sta te d th a t in m any cases this is the only w a y to obtain exact figures on the b eh av io u r of com ponents during norm al use, b o th for the th ird and the fourth step. H o w ev er, experience from both m anufacturers and users has m ade it possible to devise tests of reasonable d u ratio n w hich allow to make reliable predictions ab o u t the behaviour un d er p ractical circumstances.
The climatic te sts comprise : d ry heat, d ry cold, humid atm o s
phere, change of te m p e ra tu re and low air pressure, and the special t e s t s : mould grow th, corrosive atm osphere and dust.
Because these short-lasting tests are to be a m easure of w h a t the com ponent is going to do in its life, th e y are more an ex
aggeration th a n a simulation of practical circum stances.
A n o th er thing th a t is of im portance here is the sequence of tests. T here w ould be some logic in carry in g out every te s t on a se p a ra te lot of com ponents w hen each te s t rep resen ts the w hole life of the com ponents in a certain respect. O n the oth er hand, com ponents are subjected to all conditions of use simul
taneously. F o r th a t reaso n the com ponent specifications generally prescribe the following te st cycle to be carried out on the same components, the to ta l cycle rep resen tin g norm al life:
Standardization of electronic components 237
initial m easurem ents (step one) stren g th of term inations
soldering
change of tem p eratu re vibration and bumping d ry h e a t
humid atm osphere (first accelerated cycle) d ry cold
low air p ressure
humid atm osphere (fu rth er accelerated cycles)
A fte r this cycle, the change of the p ro p erties w ith resp ect to the initial m easurem ents should be w ithin certain limits. In m any cases essential inform ation is obtained by m easurem ents of the insulation resistance. The to ta l cycle tak es ab o u t 14 days.
A p a r t from this a “long-term humidity te s t” is carried out w hich m ay tak e up to 84 days.
In the fo u rth an d la st step it is tried to obtain inform ation a b o u t the resistance to electrical and mechanical loads. In some cases this can be fairly exactly established. F o r instance, w hen it is kno w n th a t a sw itch is o p e rate d tw ice a d ay on the average, then it is certain th a t 10 y e a rs life d u ratio n is rep resen ted by 7500 switching operations. H o w e v e r, in m ost cases the solution does not p resen t itself so clearly.
O fte n the life te st is based upon ra te d load u nder extrem e conditions of use (e.g. maximum am bient tem p eratu re). I t is obvious th a t such tests m ust be carried out for a long time to yield accurate results, b u t for some groups of com ponents it is know n from experience th a t the change of th eir p ro perties a p proaches a limit in say the first 1000 hours, and th a t failure
on the average occurs a fte r a multiple of th a t period. A test of 1000 hours then show s w h e th e r the instability is accep t
able and, if the num ber of failures during the te s t is very small, th a t the average life is sufficient.
E specially in the case of capacitors it is tried to accelerate the life te st by increasing the load, c.q. the voltage. I t m ust be stated, how ever, th a t except for D .C . p a p e r capacitors, very little is know n a b o u t the correlation b etw een norm al life an d te st life.
The problem of life d u ratio n is extrem ely complex and of u t
m ost im portance since it is closely re la te d to reliability. W h a t is k now n about it a t p re se n t is not y e t sufficient for sta n d a rd iz a tion and therefore this m a tte r is not any fu rth e r discussed
here. In I E C specifications ab o u t com ponents very little direct reference to it can be found.
W hat I.E .C . has published fo r components
A lthough not explicitly m entioned in the foregoing section it is easy to u n d erstan d th a t the w o rk is carried out along the line: definitions — te s t m e t h o d s —sim plification. F o r electronic components, definitions, m easuring m ethods and te st m ethods go to g eth er fairly well, b u t it is a r a th e r large step to simpli
fication. T aking into account the difficulty of obtaining in te r
national agreem ent and the relatively short time the Com m ittee has been active, it will be u nderstood th a t only one specification covering p a r t of this field has been published. In in tern atio n al
standardization, time is m easured in units of five y ears.
A p a r t from this, it w a s possible to reach agreem ent on some special subjects which w ere laid dow n in the following I.E .C . Publications :
N o. 62: C o lo u r code fo r fixed resistors
N o. 63: Series of p re ferre d values an d th eir associated tolerances for resistors and capacitors.
The first publication is a compromise b etw een the m ost im portant colour codes for resisto rs th a t a lre a d y existed. I t is expected th a t it will be in general use in the n e a r future.
The seco nd one gives th ree geom etrical series, based upon the same principles as the series of p re fe rre d values or R e n a rd series (R-series). The E-series h o w ev er have 6, 12 or 24 term s p er decade instead of 5, 10 or 20. Though generally reg retted , it proved to be necessary to stan d ard ize these exceptional series for electronic components because they correspond very closely to the values used in practice. The reaso n for this is the fact th a t the term ratio of the E24-series is 1.10, so th a t in the case of a 5 °/0 tolerance th ere are no gaps b e tw e en the tolerance ran g es of tw o subsequent r a te d values. This featu re m ay sometimes be of im portance in mass production.
The m ost im p o rtan t publication, w hich has been m entioned a t the beginning of this section, is N o. 68: Basic C lim atic an d M e chanical robustness T esting procedure for com ponents (B C M T ).
W h e n w o rk w as s ta rte d on com ponent specifications it w a s found th a t the electrical m easuring and te s t m ethods h a d to differ from one group of com ponents to the other, b u t th a t this is not the case for the climatic an d m ost of the m echanical tests, because the com ponents are used side by side in the same
238 H. W . Ghijsen
Standardization of electronic components 239
equipm ent and consequently are subjected to the same condi
tions of use. ^Therefore, all the tests indicated under “ step th re e ” of the foregoing section and p a r t of the tests of “ step tw o ” could be laid dow n in a general specification.
This, how ever, does not m ean th a t all com ponents are sub
jected to exactly the same tests. T h a t w ould not be logical, since th ere are internally hot a p p a ra tu s and relatively cool a p p a ra tu s, arctic climates and tropical climates, equipm ent for en tertain m en t and equipm ent for strictly professional purposes, etc. F o r th a t reason, each te s t of the B C M T is specified in tw o or more grades of severity, the m ost severe grade being num b e red 4 and the less severe grades 5, 6 etc.
The com ponent specifications indicate to w h a t grade of severity a certain type of com ponent shall be subjected.
I t w ould be possible to p u t all these figures indicating the te st severities in a row , startin g w ith the first te s t and ending w ith the last, alw ay s in the same order. Then a num ber of some ten digits is obtained telling how the component has been tested and also w h a t climatic an d m echanical conditions it is supposed to w ith sta n d in use. Such num bers are not very practical and th ey give rise to some 10000 possible types or ,,groups ’, as
I.E .C . calls them.
W h e n the first com ponent specification w a s d ra fte d — which covered p a p e r capacitors for D .C . — an a tte m p t w as m ade to solve the problem by m aking a limited choice of four groups.
A p a r t from the fact th a t this over-simplified the m atter, it w a s found w hen tackling the next specification th a t the limited choice there did n o t a t all correspond to the first one. So in fact, w hen thinking of all components there w ere still 10000 p o s
sible groups which did not m atch the idea of s id e - b y - s id e operation in the same equipment.
I t w as then proposed to ap p ro ach the m a tte r from the equip
m ent side and the follow ing very simple equipm ent classification w as g iv e n :
1. E quipm ent for high-altitude a irc ra ft
E quipm ent for norm al-altitude a irc ra ft an d h eavy-duty ground use
E quipm ent for industrial use
D om estic equipm ent (entertainm ent) 2. E quipm ent for tropical climates
E quipm ent for tem p erate climates S ealed equipm ent
240 H. W . Ghijsen
3. M axim um internal tem p eratu re of the equipm ent 55, 70, 85 or 100° C
This made it possible to reduce the above num ber to a three- digit num ber, leaving some 50 possible groups, some of which can be deleted because th e y are unrealistic.
The three-digit num ber is composed as follows. The first digit corresponds to the severity of the dry-cold te s t and a t the same time defines the severities of the following associated tests : bumping, vibration, ra p id change of te m p e ra tu re and low air pressure. All this corresponds to item 1 of the above classification.
The second digit corresponds to the sev erity of the d r y h e a t te s t and thus covers item 3 of the classification. T here are no associated tests.
The th ird digit indicates the grade of severity of the humidity te s t and is asso ciated w ith the mould g ro w th and sa lt mist test. I t corresponds to item 2 of the classification. A ccording to I.E .C . Publication 68, severity grade 4 of the humidity te s t is a te s t w ith a d u ratio n of 84 days which is especially suit
able for herm etically sealed components. S everity grade 5 has a d u ratio n of 28 d a y s ; such com ponents are suitable for m ost tro p ical applications.
C om ponents for use in en tertain m en t appliances in tem p erate climates need only m eet the tests of severity grade 6, lasting 7 days.
Th ere is still a fo u rth type of com ponent, i.e. for use in sealed equipment. A ccording to I.E .C . these are te ste d as com ponents for domestic use, b u t such com ponents need n o t have any m oisture protection on the condition th at, a fte r drying, th eir p ro p erties are closely equal to the original properties. I t is therefore possible th a t a fu rth e r severity grade will be a d d e d in the n e a r future.
I t has a lre a d y been pointed out th a t the highest severity grade of every te s t is indicated by 4, the n ex t lo w e r one by 5 etc. Thus the highest sturdiness group is designated by 444, indicating a com ponent suitable for te m p e ra tu re s from —55° C to +100° C and u n d er conditions of the g re a te s t humidity.
A complete survey of the group num ber system is given in the table below . The a d v an tag e of this system is th a t it can be un d ersto o d by com ponent users w ith o u t a d etailed k n o w ledge of com ponent testing. B ro ad ly speaking the first tw o digits define the extrem e te m p e ra tu re s b e tw e en w hich the
Standardization of electronic components 241
com ponent may be operated, taking into account the a p p ro p ria te derating. A p a r t from this, the first digit indicates the intended use (first p a r t of the classification). The third digit defines the humidity protection.
I t rem ains to be seen w h e th e r the system is sufficiently flexible to be suitable for the m ajority of components, especial
ly w ith re g a rd to the association of the m echanical tests to the minimum tem p eratu re rating. I t is fairly certain a lre a d y th a t the group num ber system can be used for n early all non
variable com ponents and it is expected th a t the group num ber system will play an im p o rtan t p a r t in the com ponent trad e, because it is re la te d to both component testing and the extrem e conditions of use the com ponent is expected to sta n d up to.
I t should of course alw ay s be rem em bered th a t it can be a g reat help to the equipm ent designer, b u t never a substitute for experience.
first digit second
digit third digit
test d ry
cold b u m p
ing vibr.
rapid temp.
c h a n ge
lowair press.
d ry h eat
mid-hu- ity
accel.
h u midity
mould
growth salt mist
4 -55 °C X X X 85
m b a r 100°C 84
d ay s 6
day s X X
5 -40 °C X X X 300
m b a r 85 °C 28
d a y s 2
d ay s - -
6 -25 °C X X - - 70 °C 7
d ay s - - -
7 -10 °C X - - - 55 °C
Finally a sum m ary of the subjects w hich are now u nder consideration is given.
The specification for D .C . capacitors has been accepted for publication.
The following specifications have been circulated under the six-months' rule, meaning th a t th ey are in a final s t a g e :
C eram ic dielectric capacitors (D .C . types) for te m p e ra tu re com pensation and general tuning purposes (so-called type I) E lectrolytic capacitors
The following subjects are in the discussion stage or will sh o rtly be subm itted for discussion:
242 H. W . Ghijsen
Fixed carbon resistors M ica dielectric capacitors
Shaft dimensions and fixing dimensions of variable resistors etc.
Q uartz oscillator crystals H .F . C ables
Plugs and sockets (both audio and radio frequency types) Radio interference suppression capacitors
Juli 1955 - Deel 20 - No 4 243
C h a r a c te r iz a tio n o f th e n o is e o f tu b e s a n d tr a n s is to r s b y fo u r m e a su r a b le q u a n titie s
by H. Groendijk and K. S. Knol *)
Lecture delivered before the Nederlands Radiogenootschap on November 5th, 1954
S U M M A R Y
T he noise of a neutralized triode is first calculated b y investigating the m echanism o f noise production an d then b y reg ard in g the triode as a lin ear four-term inal n e tw o rk . It ap p ears, from the first method, th a t the physical quantities connected w ith the generation of noise m a y not readily be determ ined. 1 he second m ethod show’s, how ever, th a t it is still p o s
sible to characterize the noise by four m easu rab le quantities and, if once these are know n, to calculate the noise factor. 4 his holds for a n y linear four-term inal n etw o rk , triodes, pentodes, transistors, etc. In general, no simple relationship exists b e tw e e n these quantities and the physical p r o p erties giving rise to noise. In the case of a triode the four noise q u a n ti
ties depend on frequency in a simple w a y .
1. Resistance noise.
Though the term noise is originally derived from acoustics it is now generally used for random fluctuations of cu rren t or vol
tage as occurring in resistances, tubes, tran sisto rs, etc. L et us first consider a resistance R (fig. la). I f w e do not connect a voltage source in series or in parallel w ith this resistance, a voltage still ap p ears to exist b e tw een the term inals. This voltage is continuously fluc
tuating in am plitude and polarity. F o r a finite time
*) Philips R esearch L ab o rato ries, N .V . P h ilip s’ G loeilam penfabrieken, E in d h o v en -N etherlands.
a
b c
F ig .l.
A resistance R an d its equivalent noise circuits.
244 H. Groendijk and K. S. Knol
interval this noise voltage m ay be expressed in term s of a F o u r i e r integral [1], W e next consider only the contribution to this integral of those frequencies w hich lie b etw een the fre quencies ƒ an d f + A f . The physical in te rp reta tio n of this p ro cedure is, th a t we p u t b etw een the noise source and the m easuring device a filter having th a t p a ssb an d w ith a re cta n g u la r ch a ra c teristic.
U p to very high frequencies the noise energy p re se n t in the frequency in terv al A f a p p ears to be p ro p o rtio n al to A f an d independent of the frequency f itself. The noise energy p ro duced by the resistance R m ay be calculated by regarding it as a voltage source (fig. lb ) w ith an in tern al resistance R and a noise e.m.f. e the m ean square value of which is
~? = 4 k T R A ƒ (1)
From T h e v e n i n ' s theorem it follow s th a t this voltage source is equivalent to a noise c u rren t source (fig. lc) having an in
te rn a l conductance g = l / R and a short-circuit c u rren t i w ith a m ean square value of
7 = 4 k T g A ƒ (2)
2. Tube noise.
W e shall now deal w ith the noise of tubes. L e t us first consider the physical aspect an d discuss the causes of noise in
side the tube. This is the w a y one began to stu d y tube noise w hen for the first time it becam e im portant. In section 5 w e shall consider this problem from a more form al point of view by startin g w ith the noise of linear four-term inal n etw o rk s. F o r signal an d noise voltages and currents in a r a th e r n a rro w fre quency b an d and w ith small am plitudes, amplifying tubes are indeed linear four-term inal n e tw o rk s and it will be show n th a t w ith this tre a tm e n t of tube noise certain difficulties appearing in the physical tre a tm e n t do n o t occur.
3. P hysical yioise sources.
To simplify our tre a tm e n t of the physical noise sources in an amplifier tube w e confine ourselves to a neutralized triode in a grounded-cathode circuit. By „n eu tralized ” is m eant th a t the capacitance b etw een grid and anode is tuned. As w e are only in terested in a lte rn a tin g cu rren ts and voltages w e omit the direct
Noise of tubes and transistors 245
73<D
• P—»
Üd
d>
dcr1
<u
Ud>
û .SPJS
«; • - - °
*o c ^M 9
C O "Ö «
^ .d .2 o J « ^ O - d e M Jd>4-> -M
• pH • p^
? ^
a "Ö r"d
<U .d (U
O b-l '*“*
C w>
^ « m__ d
• *s <u ^ -d 2
(M 3 .h O c
. u u d w
W)*b £ -d d u,
:r u § ^ o e ^g -d o
c ? u * ~ § - 73
CT3 O • ~ _ _ ,
. £ £ j § £ ^ d o £
^ M r > îü
W ci3 d
<ü -+-> • pH wn-i °
d c cr
u « o
U
<u .d
o <u
db£)
rHen
-dO
• pN
e 4) 73
g « V)O
» pN
_ _ CO
rdJ c ‘d o; S
^ JJ e C
^ d .t: d d
fi .2
d d nj d w
<u cr t, j <u
^ H H C ^
(Uü
Ud
73o
<U
bûcd -t-*
O>
<u
• 73pN
Od
<4H
O
- dd)
• pN
73 UU - a
O
o
nd«u
uu t o<-H 73C
tTS(h
bD
cd _o u 'd d) C/
1
I 9 f
I
- 7v!7\H-
r
9i
Z
a
b
c
d
e
246 H. Groendijk and K. S. Knol
c u rren t and voltage sources. F u rth e r it is easily proved by theory th a t the anode load does not affect the noise perform ance of a tube. F o r th a t reaso n the output has been short-circuited (fig. 2a). A cross the input of the tube a signal source (e.g. an antenna) an d a tunable circuit are connected having a to ta l adm ittance Y.
3.1. Low-frequency noise.
W e consider first the noise beh av io u r of the tube a t low frequencies. If w e p u t such direct voltages b etw een the elec
tro d es th a t a direct c u rren t is flowing in the output lead, su p er
imposed upon this direct cu rren t a fluctuating cu rren t iz is also p re se n t (fig. 2b). This noise cu rren t iz is caused by the electrons being therm ally em itted from the cathode. I t is called the shot noise current. A t low frequencies the tube noise can be ascribed to this effect only. The triode noise m ay therefo re be d e
scribed by giving the m ean square value of t1. H o w ev er, it is common practice to describe it in a n o th e r w ay. ix can be th ought of as being g en erated by a noise voltage i f S in the in
p u t lead, if vS is the m utual conductance. This voltage source m ay be characterized b y the value of the resistance giving the same noise e.m.f. This resistance is called the equivalent noise resistance R eq. If w e consider again only the contribution to the noise of the frequencies b etw een ƒ an d ƒ + A ƒ, w e have, acco rd ing to (1),
Yj/S2 = 4 k T R eq A ƒ (3)
3.2. Triode at higher frequencies.
The noise behaviour of a tube a t low frequencies is thus r a th e r simple, it being described by one quantity, the equivalent noise resistance. A t higher frequencies, how ever, the phenom ena become more intricate. T hen w e have to ta k e into account th a t% the tra n s it time of the electrons in the tube is no longer negli
gible as com pared w ith one period of the frequency considered.
3.21. T ransit time conductance.
To investigate the effect of the electron tra n sit time w e first p u t b e tw e e n grid an d cathode an a ltern atin g voltage w hich is
oise of tubes and transistors 247
so large th a t the noise curren ts do not play an appreciable p a rt. A n altern atin g electron cu rren t will then s ta r t from the cathode, and will induce a c u rren t in the input lead flowing from cathode to grid. A fte r the electrons have passed the grid! a cu rren t from grid to anode is induced (fig. 3). A t low freq u en
cies both cu rren ts in the input lead cancel. A t high frequencies, how ever, w e have to ta k e into account th a t these curren ts no longer have exactly opposite p h a
ses. A c u rren t is therefore produced in the grid-cath
ode circuit and it m ay be show n th a t the direction of this cu rren t is such, th a t the signal source has to supply energy. In o th er w ords, a t these frequencies there is an e x tra conductance over the input of the tube, the so-called tra n sit time conductance gx. W^e insert this conduct
ance in our equivalent circuit (fig. 2c). The effect of the in
duced grid curren ts on the noise is accounted for se p a rate ly in 3.23 for the reasons discussed there. W e th erefo re have to con
nect g% in such a w a y in the equivalent circuit th a t it acts only on the signal and not on the noise, i.e. it has to be placed before the fictitious noise voltage source tj/S.
In duced cu rren ts flowing in a triode w h e n an electron is going from cathode to
anode.
3.22. Feedback conductance.
T here is y e t a n o th e r conductance appearing a t high freq u en cies. I t is caused by the self-inductance of the cathode lead.
The output c u rren t flows through this lead, and as the la tte r is common to o u tp u t and input circuits an additional voltage is created in the input circuit. This feedback can be re p rese n te d b y a real adm ittance g z-. This holds for the signal as well as for the noise. F o r th a t reason g i has to be placed in the equi
valent circuit behind the noise source (fig. 2c).
3.23. H igh frequency noise.
Like the signal cu rren ts the shot noise cu rren t flowing b e
248 H. Groendijk and K. S. Knol
tw een cathode and anode successively gives curren ts in the ou ter leads from cathode to grid an d from grid to anode. If w e call these cu rren ts \iQ and i1 respectively, the cu rren t flowing into the grid is a noise c u rren t ig — ix — i0 . The w a y in which this noise c u rren t is produced, how ever, differs from th a t in which the induced signal current, tre a te d in 3.21, is generated.
In the la tte r case an a ltern atin g voltage is p u t b etw e en grid and cathode, influencing the electrons w hen th e y pass from cathode to anode. But the noise voltage source tJ S is not actu ally p re se n t; it is only a fictitious source used to r e p r e sent the shot noise c u rren t in the o u tp u t lead. Therefore, the induced noise c u rren t ig cannot be rep resen ted in the same w a y as the induced signal c u rren t b y means of the tra n s it time conduc
tance, so we have to introduce a se p a ra te noise cu rren t source ig (fig- 2d).
3.3. A eria l noise.
To complete our equivalent circuit we m ust still introduce the noise produced b y the re al p a r t of Y. Y is the sum of the ad m ittan ces of signal source an d input circuit (a coil and a capacitor). To simplify the form ulae w e neglect the conductance of the input circuit, as it is small com pared w ith th a t of the signal source. In m ost cases this is allow ed. W e w rite Y = g + j b t w h e re g is called the source conductance and b the circuit susceptance. F o r the first tube of a receiver, g is the tr a n s form ed an ten n a conductance. O u r considerations, how ever, hold for the second and following tubes as well.
The source conductance g produces resistance noise which m ay be rep resen ted b y a noise c u rre n t source is (fig. 2e). A c
cording to form ula (2) w e have i] = 4 k T g A f . 4. The noise factor.
In the equivalent circuit of fig. 2e the noise p ro p erties of the tube are re p rese n te d b y tw o noise sources in the cathode-grid circuit. The tube itself should then be assum ed to be noiseless.
H o w e v e r, we are generally not in terested in the noise itself, b u t in the ratio of noise p o w e r P n and signal p o w e r P s . A tube increases this value since the P n and the P s th a t are avail
able a t the input are amplified b y the same fa cto r an d the tube produces excess noise. So if w e com pare the ratio of the
oise of tubes and transistors 249
noise p o w e r P n and the signal p o w e r P s th a t are available a t the o u tp u t w ith the same ratio a t the input, the first one is largest. The quotient of these tw o noise-to-signal ratios is called the noise fa cto r ^ [ 2 ] . W^e have
P no P so
P ni Psi
P no
G • 1 mP ■
w here G = Pso • Psi is the p o w e r gain. P ni is pro p o rtio n al to the mean square of the noise c u rren t arising from the signal so u rc e :
PnojG , the o u tp u t noise p o w e r divided b y the p o w e r amplifi
cation factor, is the noise p o w e r w e should have to a p p ly a t the input of a noiseless tube in o rd er to get the noise p o w e r P no a t the output. So P no\G is p ro p o rtio n al to the mean square of the noise cu rren t produced by all equivalent noise sources a t the input together. This noise cu rren t is is H- ig H----Y g z),i
since the voltage source ~ in series w ith the adm ittance Y -I- gx i
is equivalent to a cu rren t source — ( Y + g t) in p arallel w ith Y + gx . W e have thus »S
B y substituting (5) and (6) in (4) w e obtain the form ula for the noise facto r
+ f r + ^ +<^7)}
ts•2
W e could now evaluate F if w e k new : (1) The shot noise zl/S* or R rg,
(2) The induced grid noise ig ,
(3) The tra n sit time conductance gx and (4) The correlation betw een ix and ig .
B u t if w e look for the tube d a ta in a handbook we only find R eq if even th at. The usual w a y to proceed is, then, to calculate gx from the dimensions of the tube. From the value of gx thus found, Yg m ay be calculated using a th eo retical for-
250 H. Groendijk and K. S. Knol
mula derived b y B a k k e r [3]. The correlation is n o t tak en into account a t all. The form ulae used are b a se d upon simpli
fying assum ptions, a.o. ideal triode. The tra n s it time conduct
ance can neither be m easured accu rately as w e are only able to m easure the to ta l input conductance gx + g i .
W h e n using tubes in circuits w e are n o t in the first place in terested in the physical causes of the noise in these tubes, b u t w e should like to have a t our disposal such d a ta th a t we could calculate the noise fa cto r of a tube for any input circuit.
This result has n o t been o b tain ed w ith the above con sid era
tions. Though w e have derived a form ula for the noise fa cto r w e cannot evaluate it since w e do not know several quantities occurring in this form ula and, above all, w e do n o t know for certain th a t o th er effects n o t ta k e n into account in the theory, m ay be neglected. A n investigation of the physics of noise p r o duction is, of course, very im p o rtan t if w e w a n t to m ake low- noise tubes. B ut w e do n o t concern ourselves w ith this problem here.
5. Noise o j four-term inal networks.
W e th erefo re shall s ta r t w ith a n o th e r m ethod. F o r small signal and noise am plitudes and in a n a rro w frequency band
Fig. 4.
A linear four-term inal n e tw o rk containing voltage and c u r re n t sources (a) is equivalent to a n e tw o rk w ith o u t internal sources {/?), b u t w ith a voltage source E and a c u rre n t source J across the input. T he second n e tw o rk is derived
from the first one b y omitting all the in tern al sources.
A f aro u n d a frequency f , tubes, tra n sisto rs, amplifiers etc.
m ay be re g a rd e d as linear four-term inal n e tw o rk s (fig. 4a). In the th eo ry of linear four-term inal n e tw o rk s w ith in tern al c u rren t and voltage sources [4] it is kno w n th a t the effect of these sources can be re g a rd e d as being caused by a voltage and a c u rren t source across the input (fig. 4b). So an active linear
oise of tubes and transistors 251
four-term inal n e tw o rk w ith internal sources is equivalent to the same n e tw o rk w ith o u t sources, b u t w ith these tw o sources E an d J connected across the input.
This theorem , w hich resem bles the theorem of T h é v e n i n for tw o-term inal n etw o rk s, w as used recently by B e c k i n g to rep resen t the noise behaviour of tra n sisto rs. H e show ed th a t w ith the aid of this theorem , a form ula for the noise facto r can be derived, containing only m easurable quantities.
W e shall not give a proof of this theorem here, b u t we shall only show its validity for the idealized triode th a t has been tre a te d in section 3.
5.1. The idealized triode as a jour-term inal network.
If w e com pare the equivalent circuit derived by the physical tre a tm e n t (fig. 2e) w ith the circuit in term s of E and J (fig. db), they appear, a t first sight, n early the sam e; how ever, there is one im p o rtan t difference, viz. in fig. db both noise sources are connected across the input of the n etw o rk , while in fig. 2e the noise voltage source is connected b etw een tw o p a rts of the input conductance both belonging to the tube. N o w this m ay be changed as follows. W e have seen a lre a d y th a t the noise voltage source t J S in series w ith ( V + g z ) is equivalent to a noise cu rren t source (iI/ S ) ( Y + gz) in p arallel w ith ( Y + g r).
So the to ta l im pressed noise c u rre n t is ts + tg + ( y + gz) This, how ever, m ay be w ritte n as
from which w e can infer th a t the equivalent circuit m ay also be d ra w n as in fig. 2f. If w e call the noise e.m.f. i J S = E and the im pressed noise cu rren t zg + (ixIS) gz = J y w e get fig. 2g w here both noise sources are connected across the input term inals of the tube ju st as in the form al noise circuit of B e c k i n g .
5.2. The noise fa cto r in term s o f the equivalent sources.
The noise fa cto r m ay be expressed in term s of E and J.
I t has been show n a lre a d y th a t F is the m ean square of the
252 H. Groendijk and K. S. Knol
to ta l equivalent im pressed noise cu rren t a t the grid divided b y Zj. H ence
77_ (it + J + Y . E f
•2
In this form ula occurs Y . E = g E + jb E . H e re jb E — b . j E is a noise cu rren t obtained b y first changing the phase of each F o u r i e r com ponent of the noise in the frequency in terv al A f considered by a q u a r te r of a period ol its ow n frequency and then multiplying it b y the susceptance b. Then we get for the mean square of Y . E the expression (g 2 + b2) E 2 as (j E )2 = E (the phase shift does not affect the mean square value of E ) and E . j E — O (the m ean of each sine-function multiplied b y the corresponding cosine-function vanishes).
is is not c o rrelated w ith eith er J or E . H o w e v e r, J an d E m ay be p a rtly correlated. So the mixed term s are zero except for 2 J . y E — 2 g J . E + 2 b J . j E . T hen w e get for F
F l i r + ( g + b * ) & + 2g . j ^ + 2 b . j rj E
•2
F o r C we have the expression (2) : tl = 4 k T g A f.
F orm ula (9) holds for a n y linear four-term inal netw ork. It contains four quantities characterizing the noise, viz. J ,
2 J E and 2 J . j E . These quantities m ay be determ ined by m easuring E for four different values of g + jb . This may be accomplished b y connecting a s a tu ra te d diode across the input.
The shot noise produced b y such a diode if a d irect cu rren t is flowing through it, is exactly know n, and it m ay be com pared w ith the to ta l equivalent im pressed noise cu rren t of the tube by m easuring the noise output p o w e r w ith this diode tu rn ed on an d off. If once the four noise quantities have been m easured in this w ay, F m ay be ev alu ated for any circuit or an ten n a adm ittance Y.
5.3. R esults o f the fo rm a l treatment.
By this form al tre a tm e n t of noise we have obtained tw o r e s u l t s :
(1) Though form ula (7) for F, derived b y the physical tre a t-
