Adventures with FM Radio

FM radio was my first introduction to High Fidelity Audio, and I still love it.  Local radio stations play local live performances and other music that can't be obtained elsewhere.  Professional DJ's curate streams of music from endless sources that rarely grow tiring (on the mostly noncommercial stations I listen to).  You may not love FM still or yet, but still I think I've made some interesting discoveries investigating FM radio that may apply in other circumstances.

My chief way of listening to FM radio now is the Kitchen Tuner, because it is connected to the highest antenna I have now, the MD whip antenna at the peak of the roof, with grounding installed by licensed electrician.

But more about that later.

This really starts with playing FM in the living room.  20 years ago I found an amazingly good indoor antenna location on the front wall of the house, near the ceiling and the door, for KPAC, the local noncommercial classical music station I listen to most.  Curiously, this one location was far better than anything else in the house.  As much as I preferred to listen to FM in the bedroom, having a tuner in the bedroom was never very useful, that seems like the worst location for an FM antenna, which is all the more curious because it's actually closer to where my favorite stations are being broadcast from.

Since all my audio sources must feed a digital processing network, analog sources like FM must be converted to digital.  For a long time, I've had many ways to do this:

1.  The Tact 2.0 RCS digital preamplifier I have has the analog input board, which converts analog sources up to 1.6 volts full scale to digital, with approximately 16 bits of resolution according to John Atkinson.  I have generally avoided these analog inputs except for testing purposes.

2.  I've played and streamed music to the living room system through Sonos since about 2006.  I currently have a Sonos Connect node as one of my main digital sources to the Tact.  Sonos ZP80, Connect, and similar nodes have line-inputs, which you can set to "uncompressed" and then play them on any other Sonos playback jack or device.  This was one of the #1 reasons I got Sonos in the first place.  So I could listen to the living room tuner, which had by far the best antenna inside the house (because of location, and for no reason I've ever been able to figure) in the bedroom, and later other rooms, including the living room itself.  The analog input on the Sonos seems to be better than the one on the Tact, and can take well over 2V inputs.

3.  Through the Lavry AD10 I purchased in around 2008 or so, when I started getting serious about the Living Room System.  The Lavry is a $1488 professional digital Analog to Digital converter, with 24 bits nominal resolution, and dynamic range of 117dB dynamic range (120dB weighted) and distortion below 0.0009%) designed by the analog to digital converter designer famous for having previously been lead designer on the legendary Pacific Microsonics Models One and Two.  It wasn't his ultimate effort at the time (or since), that has always been the Gold series, and this is only the Black series.  The Gold series had about 10dB more dynamic range, or now even better.  The Gold at the time I bought my black was legendary, based on real self-calibrating resistor networks, though far more complex than mine and possibly requiring more service.  My AD10 has been absolutely reliable and I've always expected it to stay that way.

However, this approach to listening to FM had a downside back for the 10 years or so I was using my Krell FPB 300 mostly.  It was first with FM radio sources that I experienced a mysterious "shutdown."  This may have been just coincidence but I figured FM radio had curious DC shifts that drove the Krell into unsustainable high current class A modes when not much was actually being delivered to my Acoustat speakers.  But I don't really have any explanation for it, I just basically rarely used the Lavry AD10 to listen to FM radio.  I considered it dangerous.  So it hadn't really been in mind much recently, even though I now use amplifiers that are perfectly reliable and never shut down, especially the Hafler 9300, designed by Dr James Strickland--who also designed the Acoustats, and also the Aragon 8008BB.  As it turned out, the problem with the Krell shutting down was a problem with the Krell itself, that I tried to have repaired, but I now believe stems from a combination of over-aggressive biasing and the lack of availability of the original transistors.  If I had been lucky, I would have gotten Transistor Service in the second channel at the time they did the first.  Now it's impossible to get that service, so it has one good channel and one problematic channel.  My long term dream is to reprogram the computer for lower and safer bias levels.  But even though I'd eliminated the shutdown problem by moving on to other amplifiers, I was still avoiding listening to FM radio through the Lavry.

Now, this works very differently generally than people surmise.  People think that because you may start with what they perceive as a "low resolution" source like FM radio, it therefore doesn't matter so much how you convert to digital, because digital is so much better.

But this is to fail to appreciate that the noise and junk that is included in the output of an FM tuner (which might be similar to a phono system in these regards) is best reproduced--or attenuated--perfectly.  If the crud isn't reproduced perfectly, the newly mangled crud is even more objectionable.  The upshot of this is that it may be just as important--if not more--to encode crappy sources in high resolution as better sources.  This is a general principle that goes beyond digital encoding to all forms of signal transmission.  Crappy sources must be reproduced better to sound decent.  It's also possible to attenuate the crud, but that must be done nicely rather than crudely.  This is the opportunity where a well engineered noise filter and buffer might be useful.  Classic FM tuners, even my legendary Pioneer F-26, was a rather mediocre output stage (which I've sometimes wondered already does some of the filtering that makes it sound so good compared with most other tuners).

4.  Recently, to time-shift programs on FM radio, I purchased a Marantz PMD 580 digital recorder.  This was the final step in a long series of time-shifting arrangements.  Until 2020, I was recording FM radio programs on my cassette recorder, an auto-reversing Nakamichi 550 (once again, the more reliable second down model from the Dragon--notorious for needing costly repairs).  So the way this worked is, programs were played on my living room tuner, whose output went into the Sonos line input, through the Sonos network and out through the ZP80 in the bedroom #2 position, which I've dubbed "Turntable and Tape", and direct into the Nakamichi.  I used Chrome Type 2 tape and Dolby B.  Some of the recordings I made by this ridiculously complex setup sounded remarkably wonderful, in fact I felt that almost all of them sounded quite good.  Somehow the Cassette Tape recording filters the FM in such a way that it sounds even better, or so I think.

But it still irked me that I could not make recordings directly to digital, which is virtually lossless.  I originally planned to make such recordings on the Masterlink digital recorder in the bedroom.  But it refused to record from the Sonos digital output, claiming "Copyright Violation."  Sonos apparently sets that flag for all analog input sources.

5.  At least I figured I could record the analog itself directly on a new Marantz PMD 580 I bought this year for the living room.  I mainly intended this as a replacement for the Nakamichi 550 in FM time shifing.  I figured it HAD to be better.  I was bypassing a whole serious of Sonos conversions, first to-digital and then from-digital.  And then recording on cassette tape, an obviously fidelity limited system (but pretty damned good sounding when done well, I must add, and especially for FM radio sources).

I could the play back directly to the living room system as well, and in nominal 24 bit 48 kHz quality.  Not the 96kHz I prefer, but I think 48kHz is nearly as good, and the 24 bits is the more important part anyway.

I could also play live radio directly through the Marantz, with the recorder in pause, letting it do the digital conversion.  This ought to be better than the Sonos conversion by being 24 bit and 48 kHz.  As well as playing back time shifted music.

It sounded very clear with no added noise and reasonably good.  But it did have a sort of exaggerated sibilence that digital is mostly unjustly now infamous for.  It sounded electronic, hifi, etc.  I was lusting for my old Cassette Tape playback, even coming back from the Cassette Player once again through Sonos to the Living Room.

So then, it occurred to me to try the Lavry again.  I've put Emotive XSP-1 Gen 2 unit number two, the nearly new one, into the Living Room, recently.  I knew I had Input 2 available, and it was fairly easy to hook it up to the tuner Variable Output (I use the Fixed Output for Sonos) with gain turned all the way up.

The sound was so much better, it was an immediate revelation.  I will never go back to the unbalanced analog input on the Marantz again.  Using the Lavry through the Emotiva is just worlds better.  No "digital" sound at all, just more like it's not there.  No argument, really, we can see it should be from the specs involved. The Marantz has 20dB or so less dynamic range (S/N is rated at 91dB) and higher distortion (rated at 0.01%).  I think those are more important than the additional fact the Lavry is digitizing to 96kHz rather than 48kHz, but that's one more detail.

But, but, I was inspired to go further.  There was still a bit more boom, pop, and "transistory" sound to be that short of perfection.  I was inspired to try something very different.

I was inspired to try one of my rarely used Musical Fidelity X10-D V3 tube buffers.  I was inspired by how good the cassette recording sounded, as compared with the Marantz.  Perhaps the output of my Pioneer F-26 FM tuner needs a bit of conditioning before digital conversion.  It had really been designed to feed into analog preamps, not digital converters.

My current feeling is that this is exactly correct.  The X10-D in the above system, prior to the Emotiva, makes it just that much more pleasant.  I feel good enough to keep listening longer, and even to other stations I'd hardly bother listening to.

It seems in particular the X10-D V3 removes some low frequency out-of-phase garbage and very low frequency instability and popping from the tuner.  Excess pop is greatly reduced, so I can crank the level up more.  Also the highs sound more relaxed, with a reduction of apparent sibilence.

These are the kinds of things you might or might not expect from the V3 specs, 10 Hz-100kHz.  Note that John Atkinson measured actually a slightly rising response to 10Hz and extension considerably beyond 100kHz for the first version.  However, we know that somewhere before DC, the X10-D by necessity (as the RC coupled tube amplifier that it is) adds 3 6dB/octave poles on the downside, and something like that on the upside.  That filters out very low frequencies close to DC, the flapping around that FM tuners tended to have in the old days when they were good.  The X10-D is by necessity a filtering device.  However, the filtering is way outside the 20-20kHz nominal range.  Is that important?  Yes, I think it is.  I wish more digital processors had controls that allowed you to set cutoffs at very low or high frequencies, as that is often more useful than prosaic filters at audible frequencies.  The crud is often best filtered out with out-of-band filters, and that helps reduce later in-band effects.  In band filters typically make matters worse more than better.  Somehow I can do the in-band filtering in my head better, just not the out-of-band filtering.   It seems your head can remove the in-band noise better than a filter, but less so for the out-of-band noise.  Or perhaps that's just my imagination.

Atkinson also measured in the original X10-D a loss of stereo separation at very low frequencies.  He blamed the power supply.  Many changes happened between that model and the one I have.  But if some of that loss of stereo separation also exists in my unit, it might also help the FM low frequency rumbling and popping.  I had an insight "maybe this thing was actually designed for FM radio, achieving effects similar to a output stage on a premium tube tuner that sounded good, but then they tried to sell it for the more popular digital systems."  I had never found it to be useful on digital system (though, perhaps, I never tried much either).

Reducing the load on FM tuner outputs to a negligible 300K is also desirable.  This prevents the tuner from wasting any output current, avoiding any kind of rippling in the power supply.  As well as reducing distortion from the output circuitry, and improving bandwidth, which are automatic with lower impedance.  I've always thought the reduction of sense-of-strain (distortion and delayed affects through power supply) to be more important than the increase of bandwidth when input impedances are lowered on less-than-high-current outputs.  But I now figure this to be imporatant factor on most analog connections.  Input impedances 100K ohm and higher are desirable.  This may not be a factor with $10K 2000's vintage Accuphase and the like super tuners, which may have high current output stages of comparable quality to the reset of the tuner. 

I'm not sure how well I'd do proving this with blind experiments.  I have wired up a switch to switch back and forth, and my tendency is to stick with the X10-D V3 circuit.  I have to simultaneously remember to correct for the 0.9dB of added gain that the X10-D provides, and assuming my V3 has a similar amount of gain, by flipping the gain on my Emotiva by 1.0dB (I wished, like Levinson, the Emotiva had 0.1dB level adjustments for audio reviewing purposes, and could memorize them for particular inputs...nobody but Levinson does such things, instead Emotive spent big money on home theater tricks I never use...I'd like polarity mono and stereo reverse controls too, but unwilling to spend the big buck on Levinson, which doesn't even spec as good as Emotiva until the top stratospherically priced Levinson models either...but I continue to lust for a ML 32.)

Endless tube buffers available on ebay from Chifi companies might serve a similar noise-filter/buffer as the X10-D V3.  But I don't think many if any of them are engineered to the same high fidelity standards, with vanishingly low distortion, and essentially flat response in the "audible" range.  People are more interested in gimmicks like being feedback-free.  Feedback-free tube circuits inject familiar tubey colorations.  I'm not interested in those anymore.  The X-10 V3 is basically neutral and is a very high fidelity device.  The noise of the X-10 V3 is even lower than that of the Marantz PMD 580, plus it is nice analog noise, not correlated and awful digital noise as is generated by the Marantz but suppressed to total inaudibility by good converters like the Lavry AD10 (which are invisible to me).  I think there were issues in 90's era digital encoders, where they thought they were being smarter with one-bit and the like systems which measured better.  But they weren't being smarter, it was a case of too-little-knowledge, and later sigma delta systems up the the present are far better.

So, along with all this transformation of FM in the living room, I'm headed towards a similar transformation in the Kitchen.  I moved out the Sansui D99X, a rarely recognized sleeper that actually represents Sansui's last and greatest digital tuner.  (It was highly recommended in an Audio Magazine review.)  It uses Sansui's proprietary Walsh decoder--which they invented.  Just before the company went bankrupt.  The tuner looks cheap, identical to a $49 model (but heavier).  But it's special, as it turns out.  Very very quiet, noise just completely disappears nearly always.  Sansui clearly uses a few tricks to achieve this sound, as they also did in the preceding X-1, as well as superlative FM engineering.

I had just moved that into the kitchen to finally test it.  I was planning to donate it to Goodwill.  It had been in my collection for 15 years or so, but never used.  It looked so boring, compared with my other super tuners.  And nobody ever talks about it, I just happen to know the history from reading old magazines, including the article written by Sansui Engineers about their new Walsh Decoder in the Journal of the Audio Engineering Society.

But once it was clear how good the Sansui was, I wasn't willing to move it out again.  I did recall that the Kenwood L-1000t it was sitting atop was an even more ultimate supertuner, being the last super tuner made by Kenwood, whose design was praised by David Rich in the Audio Critic.  Problem is, my L-1000t has a horrible drift problem, so bad I wonder if James Bongiorno wasn't getting confused when he said the Sony 730ES has a horrible drift problem (I've never noticed any in mine).  After about 6 hours on time, you must re-tune every station by 0.025 Mhz (which is possible on this tuner) to keep stations from sounding distorted.  But when the tuning is fixed like this, the Kenwood sound so fabulous you could confuse it with CD, perfectly transparent.  By comparison, the Sansui, as well as most other tuners, including super tuners sound dark.  Kenwood tuners generally sounded too bright, but the L-1000t gets it exactly right, for once, in their final super tuner.

I hadn't listened to the Yamaha TX-1000U in ages, but I remembered liking it.  It's essentially Yamaha's last super tuner, though they made a virtually identical TX-2000U in a fancier box.  It was never appreciated as much as Yamaha's legendary analog tuner, the CT-7000, but other than using the digital front end (which simply cannot be as good as an analog front end) it's more advanced in other ways, using a proprietary analog multiplier for FM stereo decoding, the ultimate best way of doing it, but only copied by a few other companies.

The Yamaha sounds damned good.  Both more lively and more earthy than the Sansui.  We've moved out of the shade and into the light, but still noise does not intrude.

I hope to figure out how to fix the drift in the Kenwood.  I cleared my bench in the first time in a year (it was plied up with junk after my Acoustat repairs and a few others last year) and deposited the massive L-1000t on top of it.

As long as I'm using the Yamaha, or the Sansui, or many other tuners in the kitchen I also need an X10-D V3 for the kitchen, where I'd like to send one copy of the tuner audio signal to Sonos to enjoy elsewhere in the house much as I do for the living room tuner.  This works out terribly using Y adapters, I've determined, because most tuners sound much worse into a lower impedance.  This is another perfect spot for an X10-D.  Whichever output sounds better for listening in the kitchen will go to the Yamaha or future home theater receivers, and the other output sent to the sonos, so their loads don't add up.  The Yamaha has 47k impedance and Sonos about 38k, but combined that's getting close to 20k, which is undesireable for most analog tuners.

I had two X-10D V3's but only one power supply.  The second was an eBay mistake from years ago.  I thought I could just buy a 24VAC supply, which I did, but then found you needed a center tapped 24VAC.  I just bought an upgraded ChiFi 24V supply designed for Musical Fidelity products.  It will be interesting to see if this actually makes the X10-D sound better, or perhaps some of the "goodness" on FM is due to the relatively limited 500 MA power supply imposing limits of its own and thereby reducing the "popping."

The specs for the X-10D V3 show a lower high frequency specification (60kHz) than the original X-10D (100khz).  For "noise filter/buffer" purposes, the lower cutoff may actually be better.

It was Mitch Cotter who designed and sold a "Noise Filter Buffer" audio device (the M A Cotter NFB-2).  I was a technician at a high end store when it was introduced--it was considered a very big thing.  It was primarily intended for use in a phono playback system.  M A Cotter also produced turntable bases, preamps, and moving coil cartridge transformers.  We used all of those in our store.  However, the NFB-2 was not marketed especially as a phono correction device, but as a device needed by all good audio systems to remove noxious out-of-band signals.  Mitch Cotter was a technical consultant for Marantz for the development of the Model 10 tuner.  

The M A Cotter devices were very costly when they were originally sold in the 1970's, and still today.  Sadly there seems to be absolutely no technical information about the devices available, such as cutoff frequencies, slopes, or other details.  It looks like a blue brick, and I wouldn't be surprised if the circuit on the inside was potted.  To operate one of these, you not only had to buy the device itself, but also a large power supply unit capable of powering 5 devices. Today the power supplies are the rarest and most expensive items of all.   Cotter must have had some good ideas, but they are all pretty much lost to most people now, except the very rare users of his devices.  But since he had previously worked on the 10B tuner, noted for good sound, it's possible he started thinking about noise filtering and buffering with FM tuners.

The original X-10D had two tube sockets and you could try different tubes.  The X-10D V3 has a "pencil" 6122 tube soldered in.  These tubes were designed for missile navigation, are very rugged and rated to last 100,000 hours.  That's almost worth the bother of soldering them in.   The V3 had very slightly lower distortion along with the lower 60kHz cutoff.  It's easy to imagine the smaller tube has slightly higher inter-element capacitances and therefore the lower cutoff.  Musical Fidelity may have bought the last set made by Phillips, but there are other versions.

I was inspired to buy a third minty looking X-10D V3 so I can have a spare unit for experimentation, repair, and possible modification.  I currently imagine have one in living room and one in kitchen for the tuners in those locations.  They work for either sound improvement, "splitting", or both.  Even if a tuner provides two outputs (as some do, "Fixed" and "Variable" usually) it is generally the case that one of those outputs is better.  With an X-10D V3 I can split the good output two ways to go to the local system and also the nearest Sonos Connect input to be available in other rooms, and all without worrying about excess loading on the tuner  And the use of the "buffered" and "unbuffered" outputs on the V3 will also be decided in which I consider most important, but likely with the better sounding output going to the local system since the other is just going to Sonos anyway.