Re: {Collins} Receiver test data, Sherwood



> properly digitized audio does not introduce discernible noise or distortion, aside from quantization noise (controlled by number of bits). Of course, frequency response (set by sampling frequency and filter) must also be designed for the desired specification. 

This discussion is relevant whether you are discussing digital music audio or direct digitization of the entire HF spectrum in a direct conversion SDR device. Direct digital conversion SDR is absolutely the cheapest way to make a receiver, especially if you leave out the input bandpass filtering circuitry. This cost saving will drive the market towards DDC/SDR whether or not real performance is better or worse than the performance of the technology which it replaces.

Certainly an assumption is that the Nyqvist frequency is met, and for direct conversion purposes, oversampled many times to reduce quantization noise.

However, in the case of digital audio, even uncompressed 16 bit 44.2 KHz sampled audio presents listenability problems for many critical listeners.

1) Even though vinyl LP’s were analog compressed to about 70 dB dynamic range (compared with CD’s supposed 96 dB) they still sound more pleasant to many people. Part of this may indeed be linked to the fact that one need not stain to hear  the soft diminuendo passages or get blown away by the loudness of the fortissimo passages (or worse yet, be forced to ride the remote volume control to keep things in range of comfortable hearing). 
2) Even though CD dynamic range is claimed to be 96 dB (106 dB with dithering and much oversampling), the reality is that any note with less than 4 bits representing it is so drastically distorted that it is annoying to the listener. That makes CD dynamic range realistically 12 bits, or 72 dB, not much better than analog pressings. A little dithering helps with the quantization noise but this reduces true dynamic range still further.
3) The hard stop at 20 KHz (in order to fulfill the Nyqvist condition with 44.1 KHz sampling) totally destroys the beat notes that occur when, say, a 19 KHz tone is produced by an instrument along with, say, a 23 KHz note by another. You SHOULD hear a beat note of 4 KHz in a live listening setting but you wont’t from a CD. Keep in mind that a violin’s highest fundamental is ~3520 Hz, but the violin produces overtones of up to 13 or so harmonic frequencies (and the relative intensities of these harmonics vary with violin type, age, player skill and many other things), with energy well out to 35 KHz. You can’t hear that but you can hear beat notes between those harmonics and other harmonics from other instruments nearby in frequency.

To make matters worse, most recordings are close miked and digitized at that point. Thus any beat notes that could have been picked up by the analog microphones are hard-limited out by the 20 KHz brick wall filter (or slightly higher if the studio is using higher sampling rate digitization for mastering purposes).

These same principles apply, although differently, to direct conversion SDR receiver designs. The good ones today certainly sample at pretty high frequencies (I believe Flex radios sample at rates over 200 MHz). This helps a lot with quantization noise and offers some degree of improvement in signal dynamic range (as opposed to noise dynamic range). Still, in the realm of direct conversion, I still see potential issues (hard clipping) with an SDR that has no input bandpass filters to limit clipping brought about by signals that can easily reach 120 dBm. 

The early DSP IF radios suffered from frequency response limitations in order to make working circuits with early DSP processing hardware. Most notable was the Ten Tec Omni 6 with its DSP. While I found it to be a great CW radio, I found the SSB audio to be unlistenable, at least for more than about 10 minutes.

Compressed digital (MP3, iPod, so-called HD radio etc.) take all of the best efforts digital recording processes and throw them away. This is far mare than zero counting to save space. Recording engineers deliberately leave out portions of the music in order to make space, living under the assumption that most listeners won’t know the difference. This is the same assumption some designers of SDR receivers will ultimately make in order to offer seemingly near- perfect specifications while not performing as well as a 50 year old 75S-3 on the air. 

As others have said, for most people, those not faced with presence of a +120 dBm interfering signal, the simplified direct digital conversion designs will be perfectly adequate. That is why many people are perfectly happy with their new IC-7300’s. Even though they score far better in the DR3 test than a 75S-3 (and most other radios, even better than an Orion), they go into clipping from strong out-of-band signals at signal levels with which a 75S-3 is perfectly comfortable due to its pretty good RF tuned preselector. 

Getting manufacturers to produce radios for those who DO have demanding needs will be the challenge. Metaphorically, DVD-based 24 bit 96 KHz sampled digital audio, which has technically been viable for over 20 years, has never been sold as a product because, according to the powers that be, 16/44 digital audio is good enough for most people and in response to those for whom it is not good enough, they say that market size does not justify the cost of carrying another format. Instead, what we got for DVD audio was 5 to 7 channels of MP3 quality noise - more quantity, not more quality. 

I think your 75S-3 is still going to sound pretty good to you even after another 20 years of direct DSP development have passed… Edwin Armstrong’s invention still has a lot of life left in it.

Gary 


This archive was generated by a fusion of Pipermail (Mailman edition) and MHonArc.