Sangean HDT-1

The Sangean HDT-1 is the first inexpensive HD Radio home tuner. It receives all AM and FM HD Radio modes, including multicasts, as well as analog AM and FM, including C-QUAM AM stereo. This review is for firmware version 1.2F. See the end of the review for version 1.4F changes.

The HDT-1X is reviewed here.

The HDT-1 is 17" x 10" x 3", a size compatible with ordinary stereo components. It weighs 5¾ lbs. The FM antenna input uses a 75Ω F-connector, while the AM input accepts wires. The HDT-1 comes with an AM loop and an FM folded dipole with integral 300Ω:75Ω balun. RCA jacks provide audio output. The detachable two-wire line cord has a polarized plug.

All controls are pushbuttons. The display is a high contrast, yellow-on-blue LCD.

Under the Hood

Inside the mostly empty enclosure are processor, interface, and power supply boards. The tuner uses digital signal processing for all reception. The processor board, made by LG Innotek, uses the TI chipset. A shielded RF front-end module feeds the DRI8201 IF processor. This chip has a high-speed 12-bit A/D converter, I/Q mixer, and three programmable FIR filters. It passes the baseband signal to the TMS320DRI350, which implements the AM/FM detection and HD Radio decoding algorithms. The board includes a Samsung K4S643232H 8-megabyte dynamic RAM, PCM1803 24-bit A/D, PCM1782 24-bit dual D/A, and LTC3411EMS voltage converter. The underside of the board contains several additional chips, including an Atmel microprocessor for non-DSP functions.

The processor board contains two microprocessors. Sangean customizes the Atmel firmware, while iBiquity programs the TI DSP chip. An extended-info screen displays the two firmware versions. When the code for the two processors is not updated together, bugs may arise.

The processor board mounts to an interface board that interconnects everything. It contains 14 electrolytics and three ICs. One, disabled in standby, regulates 14 V down to 12 V. Two NJM4580 op-amps drive the audio output.

The power supply board provides one unregulated 14 V and two regulated 3.3 V outputs. The power switch controls one 3.3 and the other outputs are always on. The board holds two additional circuits. The AM RF amplifier includes a JFET, adjustable RF transformer, trimmer capacitor, varactor diode, and varactor-drive op-amp. The audio output network consists of 2.2kΩ in series with 10 µF for each channel, with .001 µF across the RCA jacks.

This is the simulated hardware frequency response. The blue curve is for an IEEE 185-1975 load (100kΩ || 1000 pF), while the red curve is for the load (10kΩ || 330 pF) specified in an iBiquity test document. The circuit simulation includes the output network as well as additional components on the interface board.

When I finally applied power with the top cover off, a bright blue blinking LED startled me! Its blink rate is constant and its purpose is inscrutable.

Features

The operating feature I find most useful is the HD Radio carrier-to-noise-ratio display. It is quite helpful when aiming an antenna. CNR accounts for inaudible interference to the digital sidebands from adjacent-channel signals. Its maximum does not necessarily coincide with maximum signal strength.

The bit-error-rate display behaves strangely and the numbers make no sense. Evidently this function requires the transmission of a special test signal.

The signal-strength indicator shows the RF level in 18 steps on a numerical display and bargraph. On FM it stays pegged at 18 until the signal drops to 34 dBf, 10 dB below where stereo channel blending begins. Modulation near 15 kHz drops the reading regardless of RF level. On AM the indicator is rather unstable on an unmodulated carrier and varies wildly with any modulation.

The spectral display has ten bands centered at roughly 55, 77, 122, 261, 522, 1045, 2090, 4180, 8360, and 16720 Hz. The dynamic range is 70 dB.

The HDT-1 can display the HD Radio transmission mode. This is useful for spotting extended-hybrid modes MP2-MP7 that can cause higher levels of HD Radio self-noise in analog tuners.

There is no way to force analog reception of HD Radio signals. This is a great drawback when the analog signal sounds better than the digital, as is often the case on AM.

Though dimmed in standby, the LCD is still very bright in a dark room. Power the HDT-1 from a switched AC outlet to kill the display (the clock resets with no power, but station memories persist overnight). An alternative is to remove R17 on the interface board. This 1.5kΩ resistor dims the LCD backlight in standby; without it, the backlight is off. R17 is just to the left of the backlight connector (red and black wires). You can unsolder R17 or crush it with dikes. In standby, ambient frontlight provides enough LCD visibility to set the clock. The backlight uses LEDs and should never require replacement.

Tuning by entering a frequency with the numerical pushbuttons is handy.

The tuner displays RDS information. The decoder requires a signal level of roughly 24 dBf. It incorrectly decodes Mexican and many U.S. callsigns.

The HDT-1 comes with a cute credit-card-size remote control. It is an attractive Cotswold blue with well marked, widely spaced, tactile pushbuttons.

Analog FM

The HDT-1 uses a comprehensive (some would say heavy-handed) strategy to combat noise.

To begin with, as the signal weakens the tuner automatically blends the stereo channels. Blending begins at 44 dBf, when S/N is 58 dB. At 42 dBf separation is 20 dB, and by 39 dBf it is just 10 dB. Too little channel separation remains at 50-dB quieting for a stereo sensitivity figure to be meaningful. But since the noise rises approximately linearly in this region, an estimate would be 8 dB weaker than where 58-dB quieting occurs, or 36 dBf.

In addition to the blend, the HDT-1 begins to roll off the highs below 29 dBf with a single-pole noise filter. The filter corner gradually decreases as the signal level drops, leveling out at 2.9 kHz at 17.6 dBf. The image above shows the audio output for 15 preemphasized tones from 1 to 15 kHz. The traces are for high and low RF signal levels. Because the high-end rolloff occurs for mono as well as stereo signals, it may reduce intelligibility when DXing.

The HDT-1 has a curious initial power-on state. When turned on the first time after applying power, stereo blending begins at 50 dBf and noise rolloff at 35 dBf. In addition, the tuner mutes signals below 15 dBf in eight steps, with the audio level down 20 dB at the final step. After cycling power, the blend and rolloff thresholds are 6 dB lower and no muting occurs. The muting seems innocuous. I could not get the tuner to mute on an unoccupied channel when connected to an indoor or outdoor antenna.

Co-channel and multipath interference outfox the noise mitigation system. Either can make a signal above the blend threshold unlistenable. Forcing mono could clean things up, but the HDT-1 provides no way to do this.

I wasn't able to directly measure monophonic quieting sensitivities because the audio passband isn't flat at low signal levels. But using its equivalent noise bandwidth, I calculate that the noise filter reduces noise by 6.3 dB. The signal level required for quieting 6.3 dB beyond spec should be that needed for a flat passband.

For the following measurements I used IEEE 185-1975, modified as noted here. I used the test equipment listed here.

Mono 30-dB quieting sensitivity     8.9 dBf  (estimate)
Mono 50-dB quieting sensitivity     15.5 dBf (estimate)
Stereo 50-dB quieting sensitivity   36 dBf   (estimate)
Mono THD, 1 kHz                     0.04%
Stereo THD, 1 kHz, L+R              0.05%
Stereo THD, 1 kHz, L or R or L-R    0.3%
Stereo separation, 1 kHz            50 dB
Mono S/N, 65 dBf                    76 dB
Stereo S/N, 65 dBf                  73 dB
Capture ratio                       1.4 dB
Capture ratio, stereo 30 dB         12.6 dB
Capture ratio, stereo 50 dB         32.6 dB
Adjacent-channel selectivity        63.5 dB (noise limited)
RF intermod                         84 dBf  (97.7 + 98.5 -> 96.9)
RF spur                             87 dBf  (92.2 -> 96.9)
Stereo pilot injection for lock     0.75%
Deemphasis error                    L +2.0/-0.8 dB, R +1.9/-0.9 dB
Audio level referred to 600 mV      +8 dB (1.5 V)
Latency                             118 ms

30-dB quieting sensitivity is somewhat like IEEE usable sensitivity, except that I used 36.3-dB quieting to compensate for the noise filter and I did not modulate the signal while measuring 3% residual noise plus distortion. Although no distortion was apparent, background noise did increase with modulation so the usable sensitivity figure will be higher.

To compensate for the noise filter, I quieted the output 56.3 dB to measure 50-dB mono quieting sensitivity. The resulting figure is a few dB better than that of most tuners I measure, as is the stereo sensitivity.

Stereo capture ratio (no IEEE definition) is how far below a 65-dBf unmodulated stereo signal a 100%-modulated stereo signal must be to obtain the specified quieting.

The spectacular adjacent-channel selectivity is very noticeable on the air. It is far better than that of any analog-IF tuner I've ever used, including tuners with two 110-kHz ceramic filters. The filter adapts independently to interference on each side of the tuned frequency. In addition, the digital-IF filter yields much cleaner audio for multipath-laden signals with adjacent-channel interference. See the HDT-1X review for details.

RF intermod is the 50-dB quieting level for a third-order intermodulation product. RF spur is the 50-dB quieting level for an untuned signal. Both measurements are made in a way that sidesteps the phase-noise sidebands of the tuner and signal generators. I made these measurements on an HDT-1X, but the figures should apply to the HDT-1. The RF intermod figure is similar to what I've measured for other synthesized tuners with good front-ends.

I had no trouble with RF overload at my location. But others with several FM transmitters nearby have reported severe overload with the HDT-1. The problem seems to be due to local-oscillator spurs or IF A/D overload, not ordinary front-end intermod. At signal levels above 80 dBf, the HDT-1 responds to untuned single signals. Zero to several spurs occur and their pattern completely changes when retuning 0.1 MHz. The HDT-1 RF spurious response is 9-33 dB worse than the figures I've measured for other tuners.

The audio output level is very high. I built a little pot-in-a-box variable attenuator so I could run A/B listening tests with other tuners. A fixed attenuator allows no compensation for the widely varying HD levels I found on the air.

This is the frequency response above 1 kHz for each channel. Although I didn't notice it when listening to the HDT-1 alone, the treble boost is audible when comparing the tuner to one with flat response. (This curve is for version 1.2F only. See the HDT-1X review for the 1.4F curve.)

When I overdeviated 50%, the audio initially was very distorted on peaks. Then slowly, in little steps, the sine wave straightened itself out until it looked normal. I believe this is the IF filter adapting to a wider signal spectrum.

This is the distortion spectrum for the factory alignment for 1-kHz, left-channel, stereo modulation deviated 75 kHz (9% pilot). After realignment, the second harmonic dropped away and the third popped up to -50 dB (0.3% THD).

Analog AM

The AM antenna is a 4" x 5" rotatable loop with six single-layer turns. Its inductance is 12.2 µH and its Q is 70 at 1000 kHz. At lower frequencies the loop exhibits two nulls in opposite directions. Higher in the band there is just one null, which indicates omnidirectional E-field pickup about equal to the bidirectional H-field response. For AM measurements I connected a signal generator terminated in 50Ω directly to the antenna input terminals.

At 1500 kHz, the signal required for noise down 30 dB from a 90%-modulated 1-kHz tone was -94 dBm. At -52 dBm the noise was down 70 dB.

1-kHz THD at 30% to 90% modulation was 0.3% or less at -50 dBm. THD reached 1% at -30 dBm and nearly 10% at -15 dBm (about 40 mV at the antenna terminals). Using the loop, Lyle Henry reports intermod near downtown Los Angeles where several AM transmitters are located. The RF amplifier is the main culprit. Its output waveform is visibly distorted at signal levels above -25 dBm. Even at -70 dBm, internally generated harmonics from a 500- or 750-kHz test signal are clearly audible at 1500 kHz.

The amplifier is a 2SK242-T4 source follower with 33kΩ from gate to ground. To rebias the amplifier to handle larger signals, connect 12 volts through 150kΩ to the gate. This is easily done by soldering to component leads on top of the board. One end of the 150kΩ goes to the right side of R308. The other end goes to the front lead of C318 or C319. This modification increases large-signal handling about 5 dB and reduces distortion slightly at lower signal levels. Sensitivity drops about 1 dB at the high end of the band and is unaffected elsewhere.

This image shows the audio spectrum to 10 kHz with the antenna terminals open. Low-end rolloff, sloping deemphasis, and an impressive brickwall IF filter are all evident. The IF filter here yields an audio high end of about 4.5 kHz. (This curve is for version 1.2F only.)

This shows the IF filter yielding an audio high end just beyond 8 kHz for complex modulation with many harmonics. You can watch the passband corner increase in little steps after you apply modulation.

For a wide IF filter passband, wideband modulation is necessary and adjacent-channel interference must be low. If you listen carefully you can hear the IF filter slowly and smoothly adapt, wider or narrower depending on conditions, after you tune in a noisy station. On strong signals the adaptive algorithm is rather conservative, often yielding a narrower passband than seems necessary.

This compares AM frequency response with standard NRSC-1-A AM deemphasis. The curves match within 3 dB from 250 Hz to 5 kHz, which is better than most tuners. However, the early bass roll-off is pronounced on music. On speech it isn't as noticeable. I took this data at 30% modulation. The curve varies somewhat with modulation level. (The curve is for version 1.2F only. See the HDT-1X review for the 1.4F curve.)

AM latency is 125 ms.

AM uses synchronous detection. During selective fading on skywave signals, there is none of the squawky distortion an envelope detector makes when the carrier drops so low that instantaneous downward modulation exceeds 100%. Along with the brickwall IF filter, the synchronous detector makes listening to skywave signals at night a pleasure. I noticed a few minor problems for very weak signals. As the tuner unmutes you can occasionally hear the carrier fall into lock. The detector may not lock if the adjacent channel has a very strong signal, and it may lose lock during prolonged deep fades.

The HDT-1 takes up to 30 seconds to switch to AM stereo for a weak or fading signal, with no indication on the LCD. On bench tests it was picky about the pilot level, refusing to decode unless it was within a 2-dB window.

I connected a spectrum analyzer to the output of the RF amplifier and coupled a tracking generator to the AM loop with a small inductive probe. This setup let me see the front-end resonance vary as I changed the tuned frequency. The factory alignment was close and needed only a small tweak (another HDT-1 was spot on). Resonance is much narrower at the low end of the band so accuracy is most important there. Tracking was off as much as 30 kHz at midband in both HDT-1s, with on-frequency response down about 3 dB. The varactor drive voltage is neither uniform nor monotonic, with backsteps at 840, 930, and 1340 kHz.

HD Radio

Using an RF step attenuator on an FM HD Radio signal with high CNR, the HDT-1 switched reliably to HD at 29 dBf. This is 7 dB below the stereo 50-dB quieting point.

At stronger signal levels I used a signal generator to interfere with one digital sideband. As I increased the interference level, CNR dropped 3 dB, plateaued there for 30-40 dB, and then gradually decreased. I conclude that when it can, the HDT-1 decodes both digital sidebands and combines them coherently; otherwise it uses just one sideband.

Depending on the station, I found HD Radio signals to be bandlimited at anywhere from 15 to 20 kHz on FM. On AM I never saw an HD Radio signal with audio beyond 15 kHz.

The HDT-1 provided my first opportunity to hear HD Radio. Having been involved years ago in the design of low-bit-rate digital voice systems with serious artifacts, I was prepared to dislike the sound. When I finally heard it, I was amazed how similar digital and analog sounded and that no encoding artifacts were obvious. I could hear tonal differences between analog and digital, but they varied greatly from station to station. Digital sounded consistently brighter than analog on some stations, while on others the reverse was true. One station had close tonal balance but very different stereo soundstages. I had to constantly revise my opinion of HD Radio sound quality as I tuned around. I finally realized that what I was hearing was most likely determined by the way the stations had set their transmit processors, both analog and digital, not by the digital encoding algorithm. But on the best stations, percussive sounds were consistently more impulsive and lifelike on HD, probably due to preemphasis-induced limiting in the analog transmit processor. On stations with heavy analog processing, impulsive bass notes were much cleaner on HD and sounded much louder. I found the complete lack of background noise surprisingly compelling. I didn't realize that I had become accustomed to low-level analog noise on several stations I listen to regularly.

AM HD Radio was a different story. I heard pronounced speech artifacts on every station. I finally came across one or two male speakers whose voices didn't sound funny, but these were exceptions. Background noise does go away on HD and the audio bandwidth does broaden, but most of the time the sound is just too unnatural for my ears. Music may sound better, but I never heard any that lasted long enough to evaluate.

Alignment

To align the FM front-end, pick a clear frequency near the 97.5-MHz geometric center of the band. Set the level of a stereo signal source (an iPod FM modulator will do) in the region where stereo channel blending rapidly commences. Adjust the four slugs in the front-end module for maximum separation, which corresponds to maximum signal in this region. The rearmost slug is the mixer transformer. It affects second-harmonic distortion and final stereo separation. Because minimum distortion, maximum separation, and maximum signal are nearly coincident, I simply adjust for maximum signal. You can tweak the rearmost slug to optimize distortion or separation if you like. Alignment improved the sensitivity of my HDT-1 2.5 dB at the low end of the band, 2 dB in the middle, and 0.5 dB at the high end. At band center another HDT-1 improved 3 dB, and a prototype HDT-1X improved 4.5 dB. Sensitivity at the band edges was within 0.5 dB of that at band center.

Caution! I know of two persons who pressed too hard or went too far while rotating one of these tiny slugs, popping it beyond the threads deep into the form where it spun uselessly. Use a tool that precisely fits the tiny slot in the core, apply the smallest downward force necessary to achieve rotation, and don't go too deep. I use a small metal screwdriver that I remove before checking the response.

The IF output signal is on the leftmost pin of the front-end module. I found the alignment method described above to be much more sensitive than tuning for maximum IF output. But you can use the latter technique if you don't have a stereo signal source.

To align the AM front-end, connect the signal generator to a small loop and loosely couple it to the HDT-1 loop. Separate the loops by a foot or more. Adjust transformer T301 at 600 kHz and trimmer TC301 at 1400 kHz for maximum IF output. Do not connect the signal generator directly to the AM antenna terminals. The inductive loop is an integral part of the front-end tuned circuit. You can use off-the-air signals for alignment as long as you rotate the loop to keep the IF level below saturation.

Version 1.4F

1. Audio output level is about 700 mV.
2. AM frequency response is different. See the HDT-1X curve.
3. FM deemphasis is different. See the HDT-1X curve.
4. BER is displayed in scientific notation but remains meaningless.
5. CNR is displayed in tenths of a dB.
6. Fast repeated keypresses are lost.
7. Frequency scanning is much slower.
8. AM varactor drive voltage no longer backsteps.
9. Pressing a single digit recalls a memory. The keystroke must be brief.
10. Pressing a digit for two seconds instead of five overwrites a memory.
11. Stereo channel blending begins at 50 dBf instead of 44 dBf.
12. Stepped muting is always active.

Mismatched Firmware

Problems

A summary of the problems I found with the HDT-1 is here.

More is here.

Updated May 11, 2008