Home | Audio Magazine | Stereo Review magazine | Good Sound | Troubleshooting Departments | Features | ADs | Equipment | Music/Recordings | History |
Dimensions: Main chassis, (21 cm 11.5 cm x 16.7 cm); power supply, 6 1/4 in. W x 2 3/4 in. H x 3 3/4 in. D (16cm x 7cm x 9.5cm). Weight: Main unit, 4 lbs. (1.8 kg); power supply, 2.9 lbs. (1.3 kg). Price: $1,500. Company Address: c/o Mackenzie Laboratories, 1163 Nicole Court, Glendora, Cal. 91740; 800/423-4147. Philips calls the IS 5021 a sound enhancer; I call it a digital toolbox. It's an analog-to-digital converter, a digital-to-analog converter, a sampling-rate converter, a noise shaper, a digital compressor/expander, a digital fader, a digital noise filter, and a jitter reducer. It can also widen or narrow the stereo soundstage, create stereo-like effects from a mono program, suppress the ticks and pops of a vinyl record, and serve as a precise digital tone control. For all that, the IS 5021 is rather com pact. One small box contains its analog and digital electronics; an even smaller power supply provides +6.2, +17, and -17 volts DC through an umbilical cord that plugs into the back of the main box. Also on the back of the main chassis are stereo analog inputs and outputs (one pair of each), two coaxial digital inputs, and two coaxial digital outputs. (Optical digital connections are not provided.) All analog and digital signal connections are via gold-plated RCA jacks. The IS 5021's digital inputs accept consumer-standard S/P DIF (Sony/Philips Digital Interface Format) signals at any sampling rate between 15 and 50 kHz; a rear-panel pushbutton selects digital out put at either 44.1 or 48 kHz. When operating with digital signals, the 5021 reclocks and "de-jitters" the data. Sampling-rate conversion is performed by a Philips TDA 1373 chip, which supports 20-bit data. The chip's output rate is established by a crystal oscillator, whose stability determines the output jitter. If desired, the IS 5021 can perform what Philips calls Quantization Noise Imaging (QNI), a form of in-band noise shaping in tended to move quantization noise outside the audible range. Philips says that QNI improves the sound quality of 16-bit audio signals (such as those from CD), especially low-level signals, but that it yields no benefits with 20-bit recordings. Analog signals are likewise said to gain nothing from QNI, perhaps because they're converted to digital signals by a 20-bit Analog Devices MOD 79 A/D chip before being fed to the TDA 1373 processor chip. Subcodes embedded in digital input signals pass through the 5021 without being processed in any way. Signal processing is performed by a Philips PCF 5020D DSP chip, after which the digital signal is converted back to analog by a Philips TDA 1547 Bitstream IC. Although the IS 5021 has many functions, it's simple to use. In addition to left and right rotary volume controls, the front panel carries four black pushbuttons, two-gray pushbuttons, and two red pushbuttons. Three of the black controls ("Fade," "QNI," and "Effect") toggle DSP functions on and off; the fourth cycles through the three inputs ("Digital 1," "Digital 2," and "Analog"). The gray "Select" controls activate and deactivate other DSP functions ("Scratch," "Noise Filter," "Stereo Enhancement," "Compress/Expand," "Bass," "Treble," and "Spatial"). With the red "Adjust" controls, you can regulate the degree of each effect. Amber LEDs in the upper portion of the IS 5021's front panel show which input and functions are active. Any function's parameters can be adjusted during the first 6 seconds after it's selected; its LED blinks to show this. After 6 seconds with no adjustment, the LED glows steadily. A green "Lock" LED turns on when the 5021 has locked onto a digital input signal; a red LED indicates "Overload" of the A/D converter. The "Effect On/Off' button lets you check the sound with and without processing. When an indicator LED is flashing, "Effect On/Off' toggles only the function that's being adjusted; otherwise, it toggles all selected functions at once. When any function is in adjustment mode, its setting is shown on the LED level display. The rest of the time, this display, which dominates the front panel, shows signal level, using two strips of 26 LEDs: Green LEDs are used for levels from-60 to-4 dB, amber LEDs from-3 to-1 dB, and red LEDs for 0 and +1 dB. Although you're unlikely to confuse volume and setup indications, an amber "Volume Display" LED illuminates when the two LED strips revert to the task of level monitoring. The IS 5021's functions can be adjusted over quite a wide range. Bass and treble can each be adjusted by ±10 dB, and the noise filter can be set to any of 12 cutoff frequencies. You can choose from 20 "Fade" rates and 10 "Stereo Enhancement" levels. "Spatial" can widen or narrow the sound image by three steps either way. Compression and expansion are each adjustable in 10 steps, and the scratch suppressor's sensitivity is adjustable in 25 steps. Most functions can be used simultaneously, but there are restrictions. For example, you can use "Scratch," "Noise Filter," and "Stereo Enhancement" in combination with each other but not with "Com press/Expand." The LED arrangement and panel markings help you remember what can and can't be used together. Measurements In testing the Philips IS 5021, I fed it analog and digital test signals but measured only its analog output. When the 5021 is fed a digital signal, its "Volume" knobs control the analog output level and the 52-LED level indicator monitors that level. The indication therefore varies with the volume set ting, and the "0 dB" LED does not necessarily indicate 0 dBFS. With a 1-kHz input at 0 dBFS, I adjusted the "Volume" knobs to obtain a 2-volt output, which I de fined as my reference; the "0 dB" LED lit up at this level. I used this setting for all tests that I made from the 5021's digital input. (With "Volume" set 1 dB higher, the 5021's output amplifier clipped.) Fig. 2-THD + N vs. frequency. Fig. 3-THD +N vs. level. Fig. 4-Noise spectra for analog input (A) and digital input (B). According to a block diagram in the owner's manual, analog input signals pass through the volume-control stage prior to A/D conversion. This is the logical arrangement, enabling you to adjust the level so that the A/D converter is exercised as fully as possible without overloading on peaks and to use the level indicators as a guide. Yet my tests suggest that it is possible for the 5021 to clip severely without giving you any visual warning. The problem is not A/D converter overload. At an input level of 1.8 volts rms (slightly less than the 2-volt full scale output of most consumer digital products), distortion reached 1% and turning the volume down did nothing to reduce it. This suggests that a stage prior to the "Volume" control can be driven into clip ping without turning the "Overload" LED on or having the level display go into the red. I consider this a serious design flaw, al though the 5021 is hardly the first digital product I've tested that acted this way. You can avoid this problem, however. If you set the "Volume" knobs to "6" (just above their midpoints), the level display will indicate "0 dB" with a 1.6-volt input, the output will be 2 volts, and midband distortion will be well under 0.01%. If you use "Volume" settings of "6" or higher, the display will warn you of dangerous operating levels; with settings below "6," all bets are off. For all tests I made through the IS 502l's analog inputs, I turned the "Volume" controls to "6" and set the 5021's converters for 48-kHz sampling. For tests through the digital input, the sampling rate was 44.1 kHz. (For measurements made with either input, the left channel's performance is shown here; for most measurements, the right channel's performance was the same or differed insignificantly.) Figure 1 shows the IS 5021's frequency response with analog and digital signals. Note the greatly magnified vertical scale (±0.2 dB per division) and the negligible difference between the two curves. Clearly, the frequency response of the 5021's A/D converter is nearly perfect; in this respect, the 5021 processes analog signals just as well as it does digital signals. Figure 2 shows total harmonic distortion plus noise (THD + N) versus frequency at 0 dBFS. The reason the analog curve appears more irregular above 6 kHz than the digital curve is that the digital measurements are taken at relatively few, widely spaced frequencies, whereas the analog data is taken at 100 points across the frequency range. Thus, the analog curve is more likely than the digital curve to reveal problems caused by intermodulation between the signal and the sampling rate. Although I've tested a few digital products with somewhat lower THD + N at high frequencies than the 5021, I find little to complain about in either curve.
Fig. 6-Fade-to-noise test with QNI (A) and without it (B). Fig. 7-Crosstalk vs. frequency.
Fig. 9-Compressor and expander characteristics. When I tested the IS 5021's THD + N versus level for a 1-kHz signal, there was a small difference between left- and right-channel performance; the curves in Fig. 3 are for the worse (left) channel. Below the -10 dBFS level, the curves are nearly flat except for relatively minor rises in THD + N at about -40 and -45 dBFS. This time, the analog curve is slightly smoother, as a result of its larger number of data points. The analog curve lies above the digital because of residual noise in the analog input electronics and, possibly, in the A/D converter. The measurements in Fig. 3 were taken without the IS 5021's Quantization Noise Imaging. I took similar curves with QNI, but they lay well above the ones shown. (They were at -78 to -79 dBFS, just off the top of the scale.) Figure 4, which presents noise spectrum analyses with and without QNI, indicates why. Using QNI with the analog in put (Fig. 4A) causes a huge peak in the noise (28.5 dB!) near the Nyquist frequency (one-half the sampling rate); with the digital in put (Fig. 4B), QNI also substantially increases the noise in the octave from 10 to 20 kHz. The whole idea of in-band noise shaping is to move noise from the midrange and lower treble, where the human ear is most sensitive, to the high treble, where hearing is less acute. This doesn't change the total noise energy; it just shifts more of it to a region where it will be harder to hear, sort of like squeezing the air in a balloon from one end to the other. The problem with Philips's QNI noise shaping is that it doesn't seem to reduce midrange noise. Within the limits of experimental error, there's no difference in Fig. 4 between the noise with and with out QNI from 200 Hz to 7 kHz just the huge peaks in the high-frequency area (where they are, admittedly, pretty inaudible). This anomaly affects every item in "Measured Data" that includes the effects of noise. With a digital input, the IS 5021's A-weighted noise was 11.3 dB higher with QNI than without, quantization noise and unweighted dynamic range were 13.7 dB worse, and there was 6.4 dB less A-weighted dynamic range. (Although Philips does not claim a benefit from using QNI with analog input signals, I ran the tests anyway; the results were similar.) So why use QNI? Well, it does reduce the D/A converter's linearity error, as you can see from Figs. 5 and 6. Figure 5 reveals that linearity error is substantially smaller with QNI than without it, and the results in "Measured Data" confirm this. Figure 6 illustrates fade-to-noise linearity error. Again, the linearity error from -75 to -100 dBFS is smaller with QNI (Fig. 6A) than without it (Fig. 6B). (In all fairness, I've tested other D/A converters whose linearity at least equaled that of the 5021 with QNI, but they did not have the noise-related problems QNI seems to induce.) Figure 7 shows crosstalk versus frequency. The results are excellent for both inputs. The treble control's curves (not shown) were pretty standard, though the bass control's curves shelved below about 80 Hz. The controls allowed 10 dB of boost and cut at 20 Hz and 20 kHz, in 1-dB steps. Figure 8 shows how the cutoff frequency of the IS 5021's noise filter changes with different control settings. (Curves were taken at every setting; for clarity, however, only every other curve is presented.) These measurements were made with an analog source and 48-kHz sampling; the cutoff frequencies would scale down proportionately if 44.1-kHz sampling were used. Figure 9 shows the steady-state transfer characteristics of the IS 5021's compressor and expander when set for maximum effect, as well as a transfer curve with no compression or expansion. With maximum compression, the top 60 dB of the input dynamic range (horizontal axis) is compressed to a 30-dB output range (vertical axis). At input levels below-60 dBFS, the compressor re turns to unity transfer--i.e., the output tracks the input, decibel for decibel. With maximum expansion, the topmost 44 dB of the input dynamic range is expanded to 88 dB. Below that point, the expander attempts to return to unity transfer; my measurement was probably affected by residual noise. Use and Listening Tests The Philips IS 5021's analog out puts have plenty of drive voltage and a low source impedance, and its analog input impedance is adequately high. Interfacing it to your system can still be a problem, how ever, because the input will clip with analog input levels above 1.8 volts. You could fashion an attenuator so this will not happen, but with a $1,500 processor, you shouldn't need to. Nor should you need to restrict the volume control to half its range. Although there's no problem with digital levels, there is a potential snag with the IS 5021's digital connectors. Recently, I have noticed that some Japanese manufacturers of CD players are omitting coaxial digital connections and pro viding only Toslink optical jacks. I can't say I approve (wired connections usually have wider band width than the Toslink connections found in many consumer digital audio components), but if that's the way the elephants are moving, the rest of the animals in the jungle would be wise to adapt. The 5021's inability to accept an optical signal and deliver an optical output makes it less universally useful than it could be. Fortunately, this was not a problem in my setup, because my CD player, a Sony CDP-XA7ES, has coaxial as well as optical digital connections. I drove the IS 5021's "Digital 1" input from the XA7ES and then connected the 5021's analog outputs to one input of my preamp and the CD player's analog outputs to another. After I matched levels, I could compare the sound quality straight from the player and through the processor by toggling the preamp's input selector. The switch on the Bryston BP-20 preamp is silent, and a friend operated it for me, so the test was reasonably blind. Later, we swapped tasks so I could get a second opinion. Turning QNI on and off causes a momentary break in the sound. As a result, we could not make such near-blind comparisons of the IS 5021's sound with and with out this noise shaper. We therefore used the CD player's D/A converter as a reference and used the preamp's silent switch to com pare it to the 5021's D/A converter with and without QNI. On the whole, both of us preferred listening to the IS 5021 with QNI. This was especially true on piano recordings that have a fair degree of ambient "tail," such as Antonin Kubalek's Czech Miniature Master pieces (Dorian DOR-90121). To a lesser ex tent, the same was true of Evgeny Kissin's recording of Schubert's "Wanderer Fantasy" (Deutsche Grammophon 435028), although there's less ambience in this recording than Dorian captured when recording Kubalek in the Troy Savings Bank Music Hall. Neither of us heard much difference from QNI on the EMI re lease of Lalo's "Symphonie Espagnole" (EMI Classics CDS 55292), with violinist Sarah Chang and Charles Dutoit leading the Royal Concertgebouw Orchestra. One disc whose sound I preferred without QNI was Canteloube's Songs of the Auvergne, featuring soprano Dawn Upshaw with the Lyon Opera Orchestra led by Kent Nagano (Erato 96559). I thought the voice was cleaner and less "hairy" without QNI, but my friend did not react similarly. From these experiments, I conclude (at least tentatively) that, with the music I used and with the possible exception of the soprano voice, I find low-level nonlinearity more objectionable than a dollop of noise in the near-ultrasonic region. But how did the IS 5021 compare with the sound of my CD player alone? Except on the "Symphonie Espagnole," which did not seem to reveal differences, both of us consistently preferred the Sony player's converter to the 5021's, with or without QNI. The CDP XA7ES is a tough player to beat, so I re-ran the tests, using an older Sansui CD-X711 player (which was one of the first to use a MASH 1-bit converter). With QNI, the 5021 gave the X711 a good run for its money on the piano recordings; with QNI off, I'd give the edge to the Sansui. The 5021's compressor should come in handy when you're making tapes for your car; if the compression ratio is not pushed to extremes, the sound is quite listenable even without complementary expansion (which the IS 5021 will also be happy to perform). The action of the Philips IS 5021's bass and treble controls was precise and repeat able, and I found the alterations in tonal character they induced more to my liking than is typically the case. I was even impressed with the "Spatial" feature, which seemed to add some spread to narrow stereo recordings without making a mockery of the soundstage. The IS 5021's "Stereo Enhancement" function creates stereo effects from mono recordings. I tried it with the only mono CD in my collection that had not already been "stereo-ized," Chopin's Waltzes, played by pianist Dinu Lipatti (EMI Classics CDH 69802). I usually don't like what pseudo-stereo circuits do to solo instruments, but I was surprised to find I rather liked the results with the IS 5021. So I dug up some old LPs. When I say "old," I mean old. Down in the bowels of my collection, I found one of the first LPs I ever bought, Rachmaninoffs Piano Concerto No. 2 in C Minor, Op. 18, with the Austrian Symphony Orchestra conducted by Kurt Woss and Felicitas Karrer (a boy, judging from the picture on the record jacket!) at the piano. (This was on the Remington label and was "Factory Guaranteed" for "Complete Audible Range Reproduction.") Although the IS 5021 couldn't make a silk purse out of this relic, it made a valiant effort. By the time I got through doctoring the sound, the 5021's panel was lit up like a Christmas tree. A little treble cut and a little bass boost helped fix the (presumably nonstandard) equalization used in making the disc. The "Scratch" filter was reasonably adept at removing the big ticks; however, if I turned it up enough to tackle the minor ones, it punched more holes in the signal than I found acceptable. I therefore tried some aggressive noise filtering, which helped remove the hiss as well as the minor scratches; not much, if any, of the music was lost by this. (I expect that the " Audible Range" was less "Complete" back then.) Finally, "Stereo Enhancement" did a very nice job of expanding this old chestnut into stereo. As the above example might suggest, I expect that audiophiles will be interested in the Philips IS 5021 more for its special features than as an independent D/A converter. And I must say it's quite adroit at doing what it does. I was particularly impressed by its subtlety--a major virtue in this type of component. MEASURED DATA Analog Line Input Characteristics: Impedance, 49 kilohms; sensitivity, 0.34 volt for 0-dBFS output; overload, 1.8 volts. Analog Line Output Impedance: 450 ohms. Frequency Response: Analog or digital input, 20 Hz to 20 kHz, +0.02,-0.16 dB. THD + N at 0 dBFS, 20 Hz to 20 kHz: Analog input, less than 0.0398%; digital input, less than 0.0394%. THD + N at 1 kHz: Analog input, less than -83.8 dBFS from 0 to -90 dBFS and less than -89 dBFS from -30 to -90 dBFS; digital input, less than-87.5 dBFS from 0 to-90 dBFS and less than -90.4 dBFS from -30 to -90 dBFS. Maximum Linearity Error: Analog in put, 0.8 dB to-90 dBFS and 1.11 dB to -100 dBFS; digital input without QNI, 2.58 dB to -90 dBFS (1.72 dB at-100 dBFS); digital input with QNI, 0.47 dB to -90 dBFS and 0.99 dB to -100 dBFS. A-Weighted S/N re 0 dBFS for Infinity-Zero Signal: Analog input, 95 dB with out QNI and 86 dB with QNI; digital input, 105.5 dB without QNI and 94.2 dB with QNI. Quantization Noise: Analog input, -89.8 dBFS without QNI and -75.3 dBFS with QNI; digital input, -92.7 dBFS without QNI and -79 dBFS with QNI. Dynamic Range: Analog input without QNI, 91.2 dB unweighted and 93.3 dB A-weighted; analog input with QNI, 75.7 dB unweighted and 86.4 dB A-weighted; digital input without QNI, 94.5 dB unweighted and 97.9 dB A-weighted; digital input with QNI, 80.8 dB unweighted and 91.5 dB A-weighted. Channel Separation: Analog input, greater than 87.8 dB from 100 Hz to 20 kHz; digital input, greater than 91.6 dB from 125 Hz to 16 kHz. [based on review from Audio magazine, Aug. 1996] Also see: YAMAHA DSP-A2070 Digital Sound-Field Processing Amplifier (Sept. 1993) Yamaha DSP-1 Digital Sound Field Processor (Equip. Profile, June 1987) ============ |
Prev. | Next Top of Page All Related Articles Home |