Multipath Simulation for SSB FM Stereo

[cgould]

From listening to the tuner after playing those tests, I can easily identify what it does that reminds me of mp3. On fairly dense (modern) music recordings, that hard mono switching kicks in dozens of times for brief moments over, say, a 20 second period, momentarily collapsing the stereo image. Mp3 (at around 128 kbps) does something similar (along with other things unique to MP3). How often the Sony tuner does this varies with the music content, but dense (mid / high frequency stuff) really seems to upset it.

If the blending was smooth vs. switching at extremely coarse steps, I probably wouldn't be so distracted by it. It would be a bit annoying, but not so distracting. (To me, anyway.)

This happens at strong RF signal levels. I was using an FM exciter near the tuner, so there were no reception issues. I would guess that weaker signals would be stranger.

As Frank noticed too, low levels revert to mono as well. On my radio, it does not blend, just abruptly switching to half stereo or mono.

-C

Cornelius, I'll put my Sony back on the bench and see what it does with low-level audio. In the meantime, could you send me a brief recording of the MP3-like sounds?

Brian

 

[FFoti1]

From listening to the tuner after playing those tests, I can easily identify what it does that reminds me of mp3. On fairly dense (modern) music recordings, that hard mono switching kicks in dozens of times for brief moments over, say, a 20 second period, momentarily collapsing the stereo image. Mp3 (at around 128 kbps) does something similar (along with other things unique to MP3). How often the Sony tuner does this varies with the music content, but dense (mid / high frequency stuff) really seems to upset it.

If the blending was smooth vs. switching at extremely coarse steps, I probably wouldn't be so distracted by it. It would be a bit annoying, but not so distracting. (To me, anyway.)

This happens at strong RF signal levels. I was using an FM exciter near the tuner, so there were no reception issues. I would guess that weaker signals would be stranger.

As Frank noticed too, low levels revert to mono as well. On my radio, it does not blend, just abruptly switching to half stereo or mono.

-C

Corny,

In the above scenario you described, this is happening in DSB mode?

-Frank Foti

 

[cgould]

Frank,

This is standard DSB FM Stereo mode...

-C

From listening to the tuner after playing those tests, I can easily identify what it does that reminds me of mp3. On fairly dense (modern) music recordings, that hard mono switching kicks in dozens of times for brief moments over, say, a 20 second period, momentarily collapsing the stereo image. Mp3 (at around 128 kbps) does something similar (along with other things unique to MP3). How often the Sony tuner does this varies with the music content, but dense (mid / high frequency stuff) really seems to upset it.

If the blending was smooth vs. switching at extremely coarse steps, I probably wouldn't be so distracted by it. It would be a bit annoying, but not so distracting. (To me, anyway.)

This happens at strong RF signal levels. I was using an FM exciter near the tuner, so there were no reception issues. I would guess that weaker signals would be stranger.

As Frank noticed too, low levels revert to mono as well. On my radio, it does not blend, just abruptly switching to half stereo or mono.

-C

Corny,

In the above scenario you described, this is happening in DSB mode?

-Frank Foti

 

[satech]

From listening to the tuner after playing those tests, I can easily identify what it does that reminds me of mp3. On fairly dense (modern) music recordings, that hard mono switching kicks in dozens of times for brief moments over, say, a 20 second period, momentarily collapsing the stereo image. Mp3 (at around 128 kbps) does something similar (along with other things unique to MP3).

You're thinking of "Intensity Stereo" MP3 encoding, typically used for bitrates of 96 kHz and less. Intensity Stereo sums all the audio content above 4 kHz to mono, and then gives it directional steering, which the decoder can use to reconstruct a crude approximation of the stereo effect. This works OK for voice and music which has all the highs mixed dead-center, but for wide separation music, you'll definitely notice the loss of high-end stereo separation, and often hear the treble chattering and wavering around the stereo image.

 

[cgould]

Yes. You are right.

Keep in mind that the Sony Tuner isn't anywhere as bad as that codec...it just does its thing just enough to catch my attention, and reminds me of that effect.

-C

 

[k6sti]

Cornelius, I gather that you're not going to send me a sample of the MP3-like sound. So let's try it this way. I have recorded what I think is a sample of dense sound:

http://ham-radio.com/k6sti/samp1.wav

This is a 20-second sample from a local station recorded at 32 kbps (to reduce the file size). Tell me if you hear the MP3-like artifacts. If not, I'll try another sample.

Peak/RMS for this sample is about 11.5 dB. Is that dense?

Brian

 

[FFoti1]

Cornelius, I gather that you're not going to send me a sample of the MP3-like sound. So let's try it this way. I have recorded what I think is a sample of dense sound:

Brian,

I can't speak for Corny, as he's out of the office today. But here's a situation where if you had the same test setup, audio processor, modulator/exciter, and antenna, you would be able to create the same results very easily.

-Frank Foti

 

[cgould]

I'll check it out when I am back home.

I am out dealing with other things since late this morning.

Been posting via the I-phone. Once I can be back to devote some
Time to figure a way to highlight the effect, I will post. When it is material
I am infinitely familiar with, I can hear it right away. Other material I might notice, or
Maybe not.

I need to find a method to remove such issues of familarity for anyone else who would like to hear this too. That will take time...especially since I am out of the office, away from my test lab.

I'll see what I can scheme up, and if I can do something that I know highlights the issue, and can share, I'll post!

-C

Cornelius, I gather that you're not going to send me a sample of the MP3-like sound. So let's try it this way. I have recorded what I think is a sample of dense sound:

http://ham-radio.com/k6sti/samp1.wav

This is a 20-second sample from a local station recorded at 32 kbps (to reduce the file size). Tell me if you hear the MP3-like artifacts. If not, I'll try another sample.

Peak/RMS for this sample is about 11.5 dB. Is that dense?

Brian

 

[k6sti]

Frank, if you'll loan me a processor with SSB, I'll be happy to thoroughly test a variety of high-performance FM tuners in addition to the Sony and report the results. A couple of the tuners have VSB postdetection filters and one has a quadrature multipath monitor, both of which assume a DSB signal. For an exciter I can use the SCA input of my ST-1000A, or I can use an HP 8640B signal generator, which has very wideband FM and extremely low RF phase noise.

I have a feeling that the SSB system has yet to be rigorously tested by a tuner enthusiast with decent lab equipment. Here's your chance.

Here's what's in my lab:

http://ham-radio.com/k6sti/lab.txt

Brian

 

[FFoti1]

Frank, if you'll loan me a processor with SSB, I'll be happy to thoroughly test a variety of high-performance FM tuners in addition to the Sony and report the results. A couple of the tuners have VSB postdetection filters and one has a quadrature multipath monitor, both of which assume a DSB signal. For an exciter I can use the SCA input of my ST-1000A, or I can use an HP 8640B signal generator, which has very wideband FM and extremely low RF phase noise.

I have a feeling that the SSB system has yet to be rigorously tested by a tuner enthusiast with decent lab equipment. Here's your chance.

Here's what's in my lab:

http://ham-radio.com/k6sti/lab.txt

Brian

Brian,

Once we're past the NAB Convention in mid April, we should be able to do this. Right now, our stuff is either heading to customers or getting ready to head to Vegas for the trade show. I think you'd be a helpful asset as we further investigate this topic.

Thanks for your request.

-Frank Foti

 

[k6sti]

OK, Cornelius. I think I know what you're hearing.

I put 1 kHz into the left channel of my ST-1000A and reduced the audio level. The Sony delivered a very well-separated stereo signal down to some very low level where the tone finally moved over to the center. The level was so low I didn't bother to measure it. But when I used either an oscillator rapidly sweeping between 1 and 5 kHz or a broadband noise source, the tuner blended to mono at much higher audio levels. My ST-1000A does not have the wideband filter option and its separation is really good only at 1 kHz. But it's not too bad higher in frequency. For 100% modulation I measured 52 dB of separation at 1 kHz and 37 dB at 5 kHz. I'm sure the drop at 5 kHz is due to the ST-1000A, not the Sony. When sweeping a tone from 1-5 kHz, I measured 40 dB separation. Now, as I decreased the audio level, I measured these separations: 28 dB at -10 dB, 11 dB at -20 dB, and 3 dB at -30 dB. The separation reduced smoothly with no abrupt changes. I also tried using a wideband noise signal and the separation dropped similarly. But with the noise I did hear a few rather abrupt center sounds on noise peaks if I adjusted the input level just so. For both the swept tone and the noise, the higher frequencies blended toward the center first.

Now comes the interesting part. I then connected the output of another tuner tuned to a local FM station and used it as a nonpreemphasized audio source. When I connected both channels to the ST-1000A, it sounded normal (except for the lack of preemphasis). But when I disconnected one channel, I heard a low-level sort of sputtering or tinkling sound in the center. At times I'd describe it as "phasey." When I reconnected the missing channel, I could not detect a hint of it. I don't know whether it was masked by L+R audio now at the center or whether the tuner only makes these sounds for higher-frequency, lower-level, single-channel input. I'm wondering whether your tests were all done with audio on only one channel. If so, I have no doubt that this is what you heard.

I'm going to make a left-input-only recording so you can confirm that this is the sound you're talking about.

Brian

 

[k6sti]

I've replaced samp1.wav with the new sample. This one is at 22.050 ksps. I just reviewed it with my desktop speakers and I can't hear the funnies unless I turn my balance control all the way over to the quiet channel. Headphones really do help. I restored the missing channel a couple times during the brief recording so you could hear what it sounds like in normal stereo. Be careful not to get blasted if you've got the quiet channel turned way up to hear the funnies.

http://ham-radio.com/k6sti/samp1.wav

Brian

 

[cgould]

Here is what I hear.

Just did a quick experiment, and got it to reveal the funnies. In this case, wear headphones to hear the dead channel. Used some aggreessive processing on this Adele track, and viola! The radio revealed some of what I hear. With two channels, the noises get louder...loud enough for me to hear it all the time, and the stereo to mono shifts (while subtle in single channel) become more pronounced in stereo content.

There were times when I pulled the channel, and the noise was briefly WAAY up there (at the levels I'm used to hearing them), then the smarts in the NR algorithm dynamically suppressed it all to the levels you hear here.

Clip here: http://www.cgould.com/sound-files/SonyLRDemod.wav has both channels with the right channel pulled.

This clip: http://www.cgould.com/sound-files/SonyRightChannel.wav is the isolated right channel. I did not play with the audio gain at all. Figured anyone interested could do this.

Again, these artifacts are greatly suppressed in single channel mode by the NR algorithm, but in full stereo, they are more pronounced.

These clips are Double Sideband Mode, BTW...

-C

 

[cgould]

Brian...

YUP!!! That's it!!!!!

On that classical piece, it is really well behaved in stereo. On contemporary music in stereo, the artifacts can get pretty loud / annoying.

-C

I've replaced samp1.wav with the new sample. This one is at 22.050 ksps. I just reviewed it with my desktop speakers and I can't hear the funnies unless I turn my balance control all the way over to the quiet channel. Headphones really do help. I restored the missing channel a couple times during the brief recording so you could hear what it sounds like in normal stereo. Be careful not to get blasted if you've got the quiet channel turned way up to hear the funnies.

http://ham-radio.com/k6sti/samp1.wav

Brian

 

[k6sti]

I can hear something way down in the quiet channel of your sound sample, Cornelius. It sounds pretty much like what's in the other channel but it is somewhat distorted. From time to time I can hear some whistle-like sounds come and go. The level is much lower than what I produced here and the sounds are not nearly as odd.

I just realized that I had observed and documented low-level funny sounds for single-channel stereo some time ago. They are mentioned in my XDR-F1HD writeup. One day a station I was listening to lost one stereo channel. I turned up the volume on the quiet channel to see what the stereo separation was like. I heard some very weak funny sounds that I described as reminiscent of aliasing, which I think is a pretty good description.

I also just realized that my ST-1000A is not really suitable for these kinds of tests. As I mentioned, it does not have the wideband stereo-separation option. This is a complex, flat, phase-linear postgeneration filter that suppresses the L-R harmonics at 114 kHz, 190 kHz, etc. The filter in my unit works fine for a 1-kHz test tone, but it is not flat enough for higher frequencies. I just measured a difference of 4 dB between the tones generated at 23 kHz and 53 kHz when I feed it a 15-kHz signal. This imbalance means that it is generating a quadrature signal. The signal would be much stronger at 15 kHz than at 1 kHz, where 37 kHz and 39 kHz are within a couple tenths of a dB. I don't know exactly how the noise reduction algorithm would respond, but I'm sure it sees the varying quadrature power. That may be why it is blending some low-level higher-frequency sounds. So the upshot is that I don't think my single-channel tests with the XDR-F1HD are representative of the way the tuner would respond to a high-quality DSB signal with a low quadrature component.

What did you think of my revised sound sample? Is that more or less what you're hearing?

Brian

 

[k6sti]

Our messages are crossing.

OK, next question: Do you hear the funny sounds on a normal two-channel signal? (I'm talking DSB for the whole discussion.)

Brian

 

[k6sti]

Rereading your message, I believe the answer to my last question is yes. If that's the case, please record a sample with both stereo channels active so I can hear what it sounds like.

Do you happen to know the level of any residual quadrature signal in your stereo generator for normal DSB stereo? I don't imagine this is something you would normally be concerned with since ordinary receivers don't respond to it. But perhaps you've checked.

Brian

P.S.--I forgot to mention that I pulled out my tuner that has a quadrature multipath monitor. This is just a 38-kHz oscillator phased 90 degrees to the one in the stereo decoder. It detects what's in the quadrature channel. I can switch its audio to the tuner outputs. The idea is that you use the quadrature signal to rotate your antenna for minimum multipath, which is very audible in the quadrature channel even when masked by program audio in the in-phase channel.

I wanted to see what the tuner thought of the reconstituted broadcast audio from my stereo generator. It didn't sound funny, but the quadrature audio level wasn't all that low. When I put an antenna on the tuner and checked several local stations, the residual quadrature audio level was much lower. This confirmed what I thought about the flatness of my stereo generator's output filter. I then remembered that I had aligned the multipath monitor (phase shift and L+R rejection) on off-the-air signals, not my stereo generator, because I was aware of the filter flatness issue. It's been several years. This stuff is all coming back to me piece by piece . . .

 

[cgould]

Our messages are crossing.

Yes! They are!

OK, next question: Do you hear the funny sounds on a normal two-channel signal? (I'm talking DSB for the whole discussion.)

Brian

I'm talking DSB for the whole discussion too.

Yes, I hear them in normal two channel mode...only it's much worse. If you don't know what to listen for, you just might miss it. Once you "key in on it" you can peg it every time.

As I said, (and I think you observed this too) there are moments you can catch the NR algorithm "with its pants down", so to speak, and the artifacts will be way louder for a moment when pulling a channel, then they dynamically settle down to the levels heard in our clips. It is that "pants down" moment that I hear all the time in normal L/R mode on a fair amount of (contemporary rock) music content. Other times, the tuner sounds OK.

-C

 

[cgould]

It's the DSP stereo generator in the Omnia.11 Nothing funny going on in DSB. It's practically a mathmatically perfect implementation of a DSB SC stereo generator.

I first noticed this when I got the tuner working for a radio station with a different brand processor....so it ain't the stereo generator...though I totally see what you're getting at here.

I'll have to carefully go through my library to find something that anyone could pick out the crud in normal LR mode. The test I did was something I was able to do quickly with what I had on hand @ home. Yes, I have an 11 at home. LOL. Just not the rest of the lab to go with it like I do @ the office.



-C

Do you happen to know the level of any residual quadrature signal in your stereo generator for normal DSB stereo? I don't imagine this is something you would normally be concerned with since ordinary receivers don't respond to it. But perhaps you've checked.

Brian

 

[k6sti]

OK. Please let me know when you come across a good two-channel example with funny sounds. As I mentioned, I've never heard anything strange from the XDR-F1HD on ordinary signals. If it exists, I want to report it in my writeup.

See the P.S. I added to a previous post. I'm not going to do that any more because added remarks can get lost in the crossfire!

Brian