ANALOG DEVICES AN-329 APPLICATION NOTE ONE TECHNOLOGY WAY e P.O. BOX 9106 e NORWOOD, MASSACHUSETTS 02062-9106 617/329-4700 Dynamic Performance of CMOS DACs in Modem Applications by Mike Curtin and Matt Smith INTRODUCTION inthe new high-speed modems manufactured to meet the V.32 and V.33 standards, it is of prime importance to be able to produce a high-quality carrier signal. The D/A con- verter used to produce this needs excellent dynamic char- acteristics; harmonic distortion must be typically less than 70dB. This application note evaluates three Analog Devices CMOS DACs when used in this application. It ex- plains how to get the best performance from each DAC and looks at the requirements for deglitchers. The note is intended to provide the information necessary for modem designers to evaluate these DACs and assess. their suitability for particular systems. TEST CIRCUIT AND CONDITIONS In both the V.32 and V.33 standards, the carrier signal fre- quency is 1800Hz. The D/A converter digitally constructs a composite signal with update rates of 9.6kHz in V.32 sys- HP300 COMPUTER GPRIO Rw BUFFER CONTROL 1 ' PROGRAM- D/A MABLE RAM CONVERTER}o+ DE i SPECTRUM COUNTER UNDER IGLITCHER, | ANALYZER $= / TEST ' Vv \ I t I J tems and 14.4kHz in V.33 systems. One of the fall back rates for both standards is 7.2kHz. Figure 1 shows the test circuit used to evaluate the distortion performance of the DACs. The system generates an 1800Hz sine wave using the three DAC update rates already mentioned (14.4kHz, 9.6kHz and 7.2kHz). To do this the HP300 Series computer generates the digital values for the sine wave based upon the output frequency and the update frequency. These are then loaded into RAM. When the timing control logic block is activated, it sequences through the RAM at the predetermined update rate and loads the digital words into the DAC to produce the 1800Hz sine wave. This is then fed into the spectrum analyzer via the de- glitcher which may or may not be used depending on the glitch performance of the DAC under test. The spectrum analyzer shows the spectral content of the output sine wave, and output distortion can be calculated from this. Le TIMING AND CONTROL LOCK ee1 CONTROLLED CRYSTAL CLOCK Figure 1. Harmonic Distortion Test Circuit DIGITAL-TO-ANALOG CONVERTERS 8-207DEGLITCHER When the code changes in a current-steering CMOS DAC, there is a capacitive coupling of charge across the switches in the DAC. This causes an injection of charge nto the lout line which in turn causes a voltage spike or glitch to appear at the output of the current to voltage amplifier. When the DAC is being used in sine wave con- struction, the presence of these output glitches results in increased harmonic distortion. This harmonic distortion increases as the glitches increase. 2.2k0"' 4 Vin 2.2k0' D1 YQ $1 o - Vout 4 + AD ADG201A S| 711 KN EQUIVALENT Int NOTE METAL FILM RESISTORS 1% TOLERANCE. FROM TIMING AND CONTROL LOGIC Figure 2. Deglitcher Circuit DAC UPDATE SIGNAL r20ps ! not SWITCH HOLD DRIVE , SIGNAL SAMPLE | jo 15ps Figure 3. Deglitcher Timing In order to overcome the problems caused by glitches, it is possible to use a deglitcher. Such a circuit is shown in Figure 2 with its timing in Figure 3. The digital value loaded to the DAC changes when the DAC update signal goes low. The switch in the deglitcher is turned on 20ys after this, causing Vin to be sampled. This delay allows plenty of time for the DAC output to settle. The deglitcher samples for 15s before going back into hold and waiting for the next DAC update. At the slowest sampling rate of 7.2kHz this means that the signal must be held for up to 125s. This presents no problems for the circuit. The con- trol signal for the deglitcher can be derived from the DAC update signal with a dual monostable. Alternatively, if a microcomputer with on-chip timer-counter is used to load the DAC, then the deglitcher control signal can be set up as an output from the controller. The deglitcher of Figure 2 is used in some of the following circuits with excellent results. It is up to the system de- signer to decide if he can achieve acceptable performance without the deglitcher. The results given in this applica- tion note will help in making this decision. 2 Ano NICITA!L TA ARALOAL COAMVERTERGS AD7537/AD7547 In Current-Steering Mode The AD7537/AD7547 are dual 12-bit DACs, packaged in narrow 0.3, 24-pin DIPs or 28-terminal LCCCs and PLCCs. Power consumption is low (100mW typical). The only dif- ference between the two devices is in their loading struc- tures. The AD7547 has a 12-bit paraile! structure while the AD7537 has (8+ 4) loading. Figure 4 shows the AD7537/ AD7547 set up in the standard current mode for + 5V out- put. This circuit is used with the system of Figure 1 to pro- duce an output sine wave with 10V pk-pk amplitude and 1800Hz frequency. +15V 20K42" NOTES METAL FILM RESISTORS 1% TOLERANCE. 2CONTROL CIRCUITRY OMITTED FOR CLARITY. Figure 4. AD7537/AD7547 in Bipolar Current-Steering Mode The distortion figures for this circuit at the three update rates (14.4kHz, 9.6kHz, 7.2kHz) vary from 69dB to 73dB. This is shown in Table |. To improve on this you can use the deglitcher circuit of Figure 2. This cleans up the output waveform and gives improved distortion per- formance which is shown in Table II. The improvement gained by using the deglitcher is about 10dB which shows that the glitches in the DAC have a considerable effect on the ac performance. APPENDIX A contains a selection of spectral responses obtained with different circuits. All of those shown are for the 14.4kHz sampling frequency, making comparisons between the circuits more meaning- ful. Figures Al and A2 show the spectral responses for Figure 4 with and without the deglitcher. When the de- glitcher is used, the level of the 2nd harmonic drops from 70dB to 90dB. Update Rate Distortion (dB) | 2nd Harmonic (dB) 7.2kHz -73 -73 9.6kHz -69 -71 14.4kHz 69 -71 Table |. Distortion Performance of Circuit in Figure 4 Update Rate Distortion (dB) | 2nd Harmonic (dB) 7.2kHz 85 85 9.6kHz -78 -86 14.4kHz -76 90 Table Il. Distortion when Using Deglitcher of Figure 2AD7537/AD7547 in Voltage-Switching Mode Another way of getting improved distortion performance from the AD7537/AD7547 is to use it in the voltage- switching mode. Figure 5 shows the circuit diagram. The DAC connections have now been reversed with the refer- ence voltage applied to the Iouta terminal, AGNDA grounded and the Vrera terminal as the output. Rega is not used and is tied to louta to prevent stray pickup. The glitches at the output of this circuit are much smalier than in Figure 2. The point where the glitches appear (lout) is now connected to a low impedance point (AD580 output) which can absorb the current glitches without producing voltage spikes. The resulting major-carry glitches from - Figure 5 are typically 20nV-secs compared to 200nV-secs for Figure 4. When operating in the voltage-switching mode, the DAC linearity degrades as the reference volit- age increases. For this reason, the reference input in Fig- ure Sis limited to + 2.5V, giving an output signal of + 2.5V. +15V ? 20kn"' 20k" +E //Prea Voo 5 ho Eour louta ner: + Your, ADS80 AD7837KN7 AD711_s800Hz JH AGNO AD7547KN KN -e NOTES METAL FILM RESISTORS U 1% TOLERANCE. 2CONTROL CIRCUITRY OMITTED FOR CLARITY Figure 5. AD7537/AD7547 in the Voltage Switching Mode Table Ill shows the distortion figures for the voltage- switching circuit. The results are the same both with and without the deglitcher and Figures A3 and A4 are the spectral responses when sampling at 14.4kHz. The de- glitcher makes no difference to the performance of this circuit, verifying the absence of glitches in the outputs. However, if you compare the distortion results directly with those for the standard current-steering circuit plus deglitcher, you can see there is a slight degradation. Though the glitches are much smaiier, the linearity per- formance in the voltage mode is degraded. (See the AD7537/AD7547 data sheets for typical performance.) This degraded linearity will increase distortion in the out- put signal. The distortion figures for the voitage-mode cir- cuit are still better than -70dB making it suitable for many modem applications. AD7245/AD7248 The AD7245 and AD7248 are voltage-switching, 12-bit DACPORTs. Each contains a reference, 12-bit DAC and output amplifier. They differ only in their loading struc- ture; the AD7245 is a 12-bit parallel load device while the AD7248 has a byte loading structure. When connected.as in Figure 6, the AD7245/AD7248 can be programmed for a sine wave output as previously discussed. The output signal range is + 5V. 1020" oe 20nF +18V G G rer | out noFrs Voo Veae - vi OO: ner pac + Vour 1800Hz MET a FILM RESISTOR 2 METAL FILM RESI AD7245/48JN" 1*. TOLERANCE. -o- 2CONTROL CIRCUITRY DIGITAL [OGND AGND Vas OMITTED FOR CLARITY. WPUTS 15V Figure 6. AD7245/AD7248 Connected for Bipolar +5V Output Table IV shows the distortion results obtained from this circuit, and Figure AS5 is the spectral response. The distor- tion is caused by slew rate distortion in the AD7245/ AD7248 output amplifier. If the deglitching, circuit of Fig- ure 2 is cascaded with Figure 6 and used as a sample/hold amplifier, there is a marked improvement in performance. The output is now sampled at a time when the siew rate effects are over and distortion is determined by the SHA output amplifier (AD711). The distortion results with this setup are shown in Table V, and Figure A6 shows the spectral response for the 14.4kHz sampled signal. Update Rate Distortion (dB) | 2nd Harmonic (dB) 7.2kHz -76 -76 9.6kHz 57 -71 14.4kHz -61 -69 Table IV. Distortion of AD7245/AD7248 in Figure 6 Update Rate Distortion (dB) | 2nd Harmonic (dB) 7.2kHz -82 -82 9.6kHz 74 -83 14.4kHz ~75 ~ 82 Update Rate | Distortion (dB) | 2nd Harmonic (dB) 7.2kHz -75 -75 9.6kHz -74 -82 14.4kHz -72 -79 Table Ill. Distortion Performance of Figure 5 with and without Deglitcher Table V. AD7245/AD7248 Distortion Performance Using Circuit of Figure 2 DACPORT is a trademark of Analog Devices, Inc. DIGITAL-TO-ANALOG CONVERTERS 8&~209AD7538 in Current-Steering Mode The AD7538 is a 14-bit CMOS DAC and is of interest to the designer who needs somewhat better than 12-bit system performance. The device is packaged in the same narrow 24-pin package as the previous DACs. 20K?" +18v 20k 9 > Ww = 3 4 L apr 112 e00Hs AD712 Veo Rog 1 KN 1)Vnee four? - Oks? aa NOTE: AD7536JN? V2 METAL FILM RESISTORS Ves (26 a0712 12 TOU DIGITAL KN OMITTED FOR CLARITY. Figure 7. AD7538 in Bipolar Current-Steering Mode Figure 7 shows the AD7538 in the standard bipolar current-steering configuration. Table VI gives the dynam- ic performance of Figure 7 when used on its own without a deglitcher. The performance is a reflection of the larger D/A glitch impulse which this part exhibits over the previ- ous 12-bit devices. When the deglitcher of Figure 2 is used there is a major improvement in the performance as can be seen from Table Vil. THD is now down to a level of 86dB for the 14.4kHz sampled signal. Figure A7 and A8 ~ are the spectral responses for Figure 7 with and without the deglitcher. +15V 20ka" 20ki2" aA Vour 22.5V +E V2 1800Hz AD712 KN JH NOTES METAL FILM RESISTORS E 1%o TOLERANC 2CONTROL CIRCUITRY OMITTED FOR CLARITY. Figure 8. AD7538 in the Voltage Switching Mode Update Rate | Distortion (dB) | 2nd Harmonic (dB) | 7.2kHz 56 -56 9.6kHz -57 65 14.4kHz 54 ~57 Table Vi. Distortion Performance of Figure 7 Update Rate | Distortion (dB) | 2nd Harmonic (dB) 7.2kHz -88 88 9.6kHz -84 -89 14.4kHz -86 -89 Table Vil. Distortion Performance Using Deglitcher AD7538 in Voitage-Switching Mode When the AD7538 is operated in the voltage mode, the ref- erence driving lou; must have extremely good dynamic characteristics (i.e., response to a changing load). The nominal impedance which it is driving is 6kQ and this changes with the DAC code. As well as this, the glitches which the reference must absorb at Iour as the code changes are large. The AD580 is not capable of delivering this performance and must be buffered by the AD712 shown in Figure 8. The distortion results are in Table Vill. The spectral responses for Figure 8 are shown in Appen- dix A. Figure AQ is the response without deglitching and Figure A10is the response with deglitching. 2 91f AIPITAL TA AMALAR COAAN/EDTEDRS Update Rate Distortion (dB) | 2nd Harmonic (dB) 7.2kHz - 75. -75 9.6kHz -75 -79 14.4kHz -72 -78 Table Vill. Distortion Performance of Figure 8 DISCUSSION OF RESULTS The method used to calculate the distortion figures for the various circuits was to feed the digitally constructed sine wave directly into the spectrum analyzer without any fil- tering. Then, the distortion was calculated as the rms sum of the distortion components. This may seem to be a somewhat cumbersome approach. However, if a distor- tion meter was used, there would be very severe output filtering requirements. For example, when sampling at 7.2kHz, even if we wanted to measure only up to the 2nd harmonic (3.6kHz), a filtes with cutoff at 3.6kHz and suffi- cient attenuation at 7.2kHz to eliminate the clock fre- quency would be necessary. The best way of practically evaluating the circuit distortion is to use the spectrum analyzer. For each of the update rates, components up to half this rate were summed to caiculate the circuit distor- tion. In this way, all harmonics of interest are included. In practical modem systems, the output signal from the D/A converter will be followed by a filter section. This low- pass filter will nominally have a cutoff frequency of 3.5kHz for V.32 and V.33 systems. it will remove clock compo- nents and other unwanted noise from the carrier signal. The output filter will also attenuate all but one of the har- monics at the output. The one that does not get attenu- ated is the 2nd harmonic (3.6kHz) and so it is of special in- terest to the modem designer. For this reason the 2nd har- monic in each of the circuits tested has been listed sepa- rately in the tables. In analyzing the performance of the various circuits, the best results were obtained from the AD7537/AD7547 and AD7538 when both were operated in the current-steering mode and the output was deglitched. The 2nd harmonic level was equal for both of these (89d8), but in terms of THD the 4D7538 was superior (- 86dB versus 76dB for the 14.4kHz sampied circuit). In practical terms this means that if the system designer is using a 14.4kHz clock-rate, his fitter requirements will be less if he uses the AD7538.Attenuation of the higher harmonics isnt as critical as with the AD7537/AD7547 and so a lower order filter is pos- sible. If the designer wants to eliminate the deglitcher, then the best circuits to use are the AD7537/AD7547 or the AD7538 in the voitage-switching mode. Both of these give good results with a THD of 72dB and 2nd harmonic level of 78dB inthe 14.4kHz sampled circuit. The AD7245/AD7548 achieves good performance when the circuit of Figure 2 is used with it as a sample/hoid. It also has the advantage of an on-board reference. THD for this circuit is 75dB and the 2nd harmonic level is 82dB. CONCLUSIONS Each of the CMOS DACs discussed in this application note ig capable of delivering better than 70dB distortion per- formance when synthesizing an 1800Hz signal with the stated update frequencies. Some, but not all, of the DAC configurations require deglitching to achieve this. There are varying levels of performance among the devices which deliver better than 70dB THD, ranging from _-72dB for the AD7537/AD7547 in the voltage mode with- out deglitching to 86dB for the AD7538 in the current- steering mode with a deglitcher. it should be remembered that the results contained in this application note are typical and were obtained from a range of devices taken randomly. Several production/fab- rication lots were sampled. However, the results are meant to show typical performance only and do not guarantee that this will be met in ali cases. APPENDIX A Spectral Responses of Figures 4, 5, 6, 7 and 8. Four =1800Hz. Update Rate = 14.4kHz. Responses are shown both with and without the deglitcher circuit of Figure 2. ite ae) D) pC aad Zz a Z an | | = a | Zz Zz | | = Figure Al. Spectral Response of Figure 4 (AD7537/ AD7547 in Current-Steering Mode) without Deglitcher REF 3-0 din MARKER 1 801.7 Hz pte 7a eT Cal eed ro Ld STOP 12 000.0 Hz LP 4 RBW 10 Hz VOW 30 Hz 3 aK) 1 ae Figure A3. Spectral Response of Figure 5 (AD7537/ AD7547 in Voltage-Switching Mode) without Deglitcher REF -5-8 dBa 10 dB/DI a Figure A2. Spectral Response of Figure 4 with Deglitcher | |_| | = | | | | aa eo he STOP 12 000.0 Hz Ds ee) ae Figure A4. Spectral Response of Figure 5 with Deglitcher DIGITAL-TO-ANALOG CONVERTERS 8-211STOP 12 000-0 Hz ST 238 SEC Figure A5. Spectral Response of Figure 6 (AD7245/ AD7248 in Bipolar Output Configuration) without Deglitcher ita ee ee MARKER 1 001-7 Hz pete ra a a RANGE 25-0 dBn 21.9 dBa 1 H =e ms a a Taryn tae Tle RBM 10 Hz ST 238 SEC Figure A7. Spectral Response of Figure 7 (AD7538 in Current-Steering Mode) without Deglitcher REF 20.0 dBa MARKER 1 802-7 Hz 10 4B/01 RANGE 20.0 48a 17.7 d = ie ES Coes U ann5 areqe ges STOP 12 000.0 Hz ST 238 SEC ca Figure A9. Spectral Response of Figure 8 (AD7538 in Voltage-Switching Mode) without Deglitcher 2.919 PUICITAL.TO_ANAIOR COANERTERS TU ne - irene coe CI ST 230 SEC Figure A6. Spectral Response of Figure 6 with Deglitcher aa ees aL! pe Pa a START 106 +0 Hz STOP 12 000.0 Hz cold ty ou VBW 30 Hz ST 238 SEC- md Figure A8. Spectral Response of Figure 7 with Deglitcher REF 20-0 dBa MARKER 1 601.7 Hz 20.0 dBa 17.0 48 - _ | | re TART 100.0 Hz CLE abe! als ra Figure A10. Spectral Response of Figure 8 with Deglitcher