MASTER TMNG AND TOF MODULE- G. Mazaher Stanford Lnear Accelerator Center, Stanford Unversty, Stanford, CA 9409 USA SLAC-PUB-66 November 99 (/E) Abstract n conjuncton wth the development of a Beam Sze Montor (BSM) for the Fnal Focus Test Beam (FFTB) at SLAC, we have bult a general purpose tmng devce wth capabltes useful for many dfferent applcatons. The Tme Master conssts of a fast clock, a large memory loaded va a PC, and a tme verner (analog) wth &bt resoluton. The Tme Master generates an arbtrary pattern of pulses on 6 dfferent channels (up to 26), wth a resoluton of l/2* tmes the clock perod. The clock content s stored n another memory to measure the tme of up to 6 channels, wth a resoluton of /2s tmes the clock perod (frequency s set at 0 Mhz), usng a tme-to-ampltude verner. The data stored n the memory s accessed va a PC. The depth of the memory for pattern generaton s bt.s (2767), equal to the depth of the tme measurng part. The devce s self-calbratng, smply by prescrbng a pattern on the output channels, and readng t nto the tme measurng secton. The total clock length s 24 bts, equvalent to 4 ms of tme at 0 Mhz frequency. Therefore, the resoluton s of the order of 2 bt.s (.e., 24 bts of clock plus 8 bts of verner). NTRODUCTON One of the requrements for the Fnal Focus Test Beam (FFTB) s the measurement of the beam sze at the Fnal Focus. A method for dong ths to submcron accurac) requres very accurate tmng (fracton of nanoseconds) of long (hundreds of mcroseconds) tme ntervals. n order to measure ths tme nterval, we chose a method whch has very good lnearty and a large dynamc range;.e., the use of a hgh-frequency oscllat.or wth a counter (Ref. ). The resoluton obtaned by ths method s proportonal to the oscllat.or frequency (20 ns for 0 MHz). To obtan the hgh resoluton requred, we further subdvde the clock perod wth an analog tme t.o amplt.ude converter, read by an 8-bt flash ADC. The combnaton of a clock counter plus verner has the lnear dynamc range of the countng method and the hgh accuracy of the analog method for a short range nterval (less than 00 ns) (Ref. 2). PRNCPLE OF OPERATON The prncple of operaton and a smplfed block dagram of the Master Tmng and TOF module are presented n Fg.. As shown n the block dagram, t conssts of two separate sectons, the Tme Master and the Tme of Flght (TOF). The crcuts descrbed n ths secton have been desgned usng predomnantly ECL 0 KH components to acheve good tmng resoluton and mnmze propagat.on delay. * Work supported by Department of Energy contract DE-AC0-76SF00. Tme Master Ths crcut generates an arbtrary pattern of pulses on 6 dfferent channels (up to 26 t conssts of a 24-bt synchronous counter (bnary) wt! a free runnng clock. The output of ths counter s converted to Gray Code to mnmze the transton errors to &l/2 least count. The resultng Gray Code s followed by a 24-bt word comparator. Pattern generaton data (n Gray Code) are wrtten n consecutve locatons of a 24-bt x 2 K ( bt) pattern memory (PAT MEM) va a PC. The content of ths memory s then converted from TTL to ECL through a 24-bt TTL-ECL latch and s fed to the word comparator. The ntalzaton and start of the pattern generaton s by a start command from the keyboard, or by an external pulse. When a word s recognzed, a pulse s generated and sent to a 9bt memory address regster (MEM ADD) and trggers an 8-bt programmable delay unt (Tme hlaster Verner). The MEM ADD s ncremented each tme a word s compared. The delay data for each pulse s loaded to an 8 x 2 K memory (VER MEM) va a PC. The full scale settng of the tme verner (programmable delay, AD 900) s equal to the clock perod. Output of the programmable delay unt s sent to a l-to-6 decoder. whch s addressed by an 8 x 2 K channel dentfcaton memor! (CHD MEhl). Output pulses are avalable ether n Kl\l or TTL levels. The mnmum pulse separaton between two successve pulses s 40 ns, lmt.ed by the dead tme of the crcut. Tme of Flgllt (TOF) As s shown n the Fg. block dagram, TOF accepts 6 dfferent tmng pulses (up to 26), and assgns to each tmng pulse a channel dentfcaton number and tme nformaton. The tme nformaton s expressed n a 24-bt Gray Code, thus lmt.ng the transton errors to & /2 least count whle latchng data from a runnng clock. The ratonale of usng a Gray Code and detals of ts generaton are gven n Ref.. The tme data and channel dentfcaton data are stored n a 24 x 2 K TOF tme memor> (TOF TME MEM) and TOF channel dentfcaton memory (TOF CHD MEM) respectvely. We use a tme verner crcut to measure the fracton of clock perod between a tmng pulse and the followng clock pulse plus one. Ths crcut conssts of a tme-to-ampltude converter and an 8-bt hgh-speed Flash ADC (AD 902, 0 ns converson tme, l/zbt LSB lnearty). The dgtzed data are stored n the TOF Verner memory (TOF VER MEhl). stored n the memory are accessed va a PC. The data Prototype Board A wre wrapped prototype of the Master Tmng and TOF module was constructed, and conssts of two boards: a Fast Bus sze board to accommodate 6 C s, and a small sze board for TOF Verner. n order to ensure optmum performance, basc hgh-frequency desgn rules were used around ECL OKH components, and especally for the TOF Verner crcut that uses an AD902 Flash ADC. See Fg. 2. and Presented at the EEE Nuclear Scence Symposum Medcal magng Conference, Santa Fe, NM, November -9, 99.
r-------- TOF TOF 2 TME MASTER ; -_. --.-----..-, Control > Control Logc Logc TTL to ECL, 8 - ChO : 0-9 70Al / - Ch Fgure. Prncple of operaton and smplfed block dagram of the Master Tmng and TOF module. 2
(4 ClockCount ~20 ns lo-9 706AQ Fgure 2. Pcture of Master Tmng and TOF module showng wre-wrapped board, Verner board, and nterconnectons. Test and Performance The crcut has been tested usng the Tme Master part and an external pulse generator. Fg. (a) shows the pulse separaton dstrbuton between two channels n clock count unts. As expected only one bn s flled. Fg. (b) shows the verner dstrbuton for the same sample; notce that the sngle channel sgma of the verner data (~ = ps) s the combnaton of fluctuatons from the Tme Master and from the TOF. The values of Sgma gven are the usual; u = d-the least count s 90 ps for all the data presented. n Fg. (c) the verner dfference between two channels s hstogrammed. n ths case we have ~72 = u &, because we add two sgmas n quadrature. Fgure 4(a) shows the pulse separaton dstrbuton between two channels, usng an external pulse generator. We have manly two bns flled, because the two pulses are randomly dstrbuted wth respect to the nternal clock. Fgure 4 b) shows the verner dfference dstrbuton between two c 6 annels for the same data sample of Fg. 4(a). There are two dfferent peaks, correspondng to the dfferent pulse separatons. Fgure (a) shows the verner dstrbuton for one channel, usng the external pulse generator: dll (v)ldv, where dn(v) s the number of the event, wth a verner value wthn the nterval [v, v + dv]. Ths dstrbuton should be flat because the pulses are randomly dstrbut,ed wth respect to the nternal clock: dn(t)/dt = C(C an arbtrary constant,) () where dn(t) s the number of events wthn the nterval [t,t + dt]. ndeed, assumng that the verner value s lnear wth tme: v = K t. Wth K a relatve constant by dfferentaton, we obtan: dv = C dt. Substtutng ths n () gves: dn(v)/dv = C/K. The devaton from a horzontal lne s a measurement of the non-lnearty of the TOF system whle convertng 2 r 9 (b) 22 2 Pulse Separaton (Clock Counts) Count-90 ps 0 ps Verner Dstrbuton (Counts) 26-9 0 lo-9 70y2 Verner Dft erence Dstrbuton (Counts) Fgure. Usng Tme Master crcut, dstrbutons of: (a) Pulse separaton between two channel unts of clock perod; (b) Verner for one channel; (c) Verner dfference between two channels. the tme to a dgtal number. f we now ntegrate the verner dstrbuton wth respect to the varable verner v, usng (l), we obtan the relaton between the tme and the verner value v accordng to the followng formulas: (v) = dn(v)/dv + dv = C + dt(v) = C *t(v). llll ml Fgure (b) shows (v). Ths functon s used t.o remove the non-lnearty of the system, snce we know that for each
r Clock Count =20 ns (a) 20 g 20 Pulse Separatons (Clock Counts) 2 z Count =90 ps (b) 24 -z = 0 7 9 2 0-9 VERNER 70A6 Fgure 6. Tme versus Verner usng Tme Master and a varable delay. Count = o=20 ps 90-68 76 70w Verner Dfference (Counts) 09, Fgure 4. Usng an external pulser, dstrbuton of: (a) Pulse separaton n unts of clock perod between two channels; (b) Verner dfference between two channels. 80 (a) 682 Events 2s C 6 68 99 z Not Lnearzed Tme Dfferences (Counts), () Count = 90 ps L (a) & a=9 ps Verner Dstrbuton (Counts) 2-68 94 90-9 70, YS Lnearzed Tme Dfferences (Counts) Fgure 7. Usng an external pulser, dstrbuton of: (a) Tme dfference (not lnearzed) between two channels; (b) Tme dfference (lnearzed). 26 70E.N Verner (Counts) 2 0-W Fgure. Usng an external pulser: (a) Verner dstrbuton of one channel; (b) Tme versus Verner functon obtaned by ntegratng Fg. a. verner value v, the correspondng true tme s (v) = (v)/(tmzz), n clock unts; wth the condtons that (vmn) = 0 and t(vmaz) =. The same non-lnearty has been measured by usng the Tme Master plus a varable delay. The results are shown n Fg. 6. The two methods gve t.he same results, wthn the errors. The non-lnearty can be removed, usng the calbraton procedure descrbed above.
Fgure 7(a shows the hstogram of the tme dfference (not lnearzed between two channels. The resoluton s c = 20 ps. Fgure 7b shows, for the same data sample, the dstrbuton of the lnearzed tme dfference. The resoluton n ths case s mproved greatly wth respect to the prevous case, u = 90 ps. We are measurng two dfferent pulses so we can estmate the absolute resoluton by measurng one pulse as u = 90 ps/f = 6 ps. The measured resoluton s constant wthn flo% for pulses separatng from 00 ns up to 0 ms., SUMMARY We have developed a stand-alone module nterfaced wth a PC, wth the capablty to ether generate or accept G dfferent tmng sgnals (up to 26), wth an accuracy of less than 00 ps n measurng tme n the range of 00 ns to 0 ms, at a clock frequency of 0 MHz. ACKNOWLEDGEMENTS The author thanks Dr. F. Vlla for helpful dscussons and techncal assstance, K. Bouldn for assstance n the constructon of the unt, and Dr. S. Wllams for wrtng software to test the module. Specal thanks to Dr. P. Hamond for software support and data analyss. REFERENCES. D. Porat and S. Wojcck, Fast synchronoous Gray counter, Nucl. nstrum. Methods 69 (980) 24. 2. uha Kostamovaara and Rsto Myllyla, Tme-tc+dgtal converter wth an analog nterpolaton crcut, Amercan nsttute of Physcs, Rev. Sc. nstrum. 7 (ll), November 986.