Home [en] [de] /

FM Tuner Tweaks and Repairs

Intro

After nigh on 2 decades, I recently rediscovered my passion for tweaking FM tuners (tuner tuning, if you will). My old tuner page in German, UKWellensalat, was last touched in 2007 and I don't really dare disturbing the dead, so I initially started writing about this in English-language forums and then added entries to my music and audio "blogalike", but it all got a bit much, hence this page.

New Mods for the Kenwood KT-80

Motivation

You may not be aware that yours truly has a history as a bit of an FM tuner enthusiast from back in the 2000s. (My hobby trajectory basically went shortwave → FM → music.) From those days, I kept a Kenwood KT-80 in the bedroom – while not a world-beater, it's a neat little analog tuner of above average build quality that is kinda technically interesting (due to sporting a pulse count discriminator), performs decently and sounds good (well, after bridging or replacing the 10µ/16V output coupling caps, which have never seen any bias voltage in decades and would be degraded and leaky), all the while being of manageable complexity. I'd call it "Pulse Count Detector, Essential Edition". These were sold from about 1979 to 1982 and then as the black KT-80B in 1983/84, which by the standards of the time was an eternity, so they are fairly common on the used market.

Mods for a better set of matched filters (of limited interest nowadays, as the parts have sadly long gone EOL) and moving the muting onto another button have previously been documented.

I wanted to address the following issues in my unit:

  1. The sole incandescent lightbulb, an 8 V / 50 mA affair lighting up the power button, had long since been burned out when I got my unit – I strongly suspect it was run out of spec from the factory now. Meanwhile, the series resistor in the power supply was frying away close to 2 W while supplying about 6.6 mA to a single green dial pointer LED, with pronounced mains (50 Hz) flicker at that. (I now think that a leaky zener diode was making things even worse.) When the whole tuner only draws 10-12 W from the mains, that's quite a lot.
  2. Out of curiosity I was poking around with a multimeter and noticed that regulated voltage was quite variable depending on load (i.e. number of LEDs lit). A variation of about 92 mV was observed despite the implementation of load-side voltage sensing; besides, the voltage varied between about 15 V idle and 14.9 V under load, a fair bit above the nominal 14 V noted in the schematic.
  3. The AFC always struck me as a bit more "sticky" than I like.

So I fired up LTspice, made a decently accurate model of the power supply and lighting and regulator circuitry and developed some mods. Here is the result:

[LTspice schematic of lighting and regulator circuitry with mods]
KT-80 lighting and regulator circuitry with mods.

The mods in detail:

KT-80 Voltage Regulator Mod

[LTspice schematic of regulator circuitry with mods]
KT-80 regulator circuitry with mods.

The mediocre voltage stability turned out to be largely caused by mediocre suppression of input voltage fluctuations near DC. The culprit is zener reference RC filtering via resistor R117 (see above schematic). The current through this part approximately equals that through R115 + R116, so the voltage drop over it is approximately

ΔVR117 = R117/(R115 + R116) × (Vunreg - Vreg - 2Vbe)

which as you can clearly see is decidedly not constant when unregulated input voltage changes! While the variation is only in the double-digit mV, times a gain of 1 + R119/R120 (about 2.2) it's enough to account for the bulk of load variation seen. The zener itself proved rock stable in comparison.

In the spirit of KISS, I added some feedforward resistance to pull up the feedback node voltage along with input voltage. RC filtering had to be used to avoid worsening AC PSRR performance. In sum, a total of 3 inexpensive passives plus some wire will do the job, which can be inconspicuously installed under the PCB.

Since ground is readily available at the wire bridge next to R119/120, Rff1 and Cff can be installed right there. Then install Rff2 at D16/17 junction and run wire between them, avoiding the mounting tab at the side. Fix wire with a dab of hot glue or somesuch.

With the values of 68k/10k/68k that I settled on, load variation went from -92 mV to +1 or 2 mV – going with 75k for Rff1 would arguably have been preferable, but eh, good enough. (I hadn't bought any of these, just 56k, 62k and 68k.) It's probably a bit overcompensated for pure input voltage changes, but definitely way better than stock. Perhaps not concidentally, frequency stability is about an order of magnitude better than for either my KT-1100 or KT-900 (but unfortunately I didn't check it pre-mod).

Voltage came down to about 14.23 V with a bit of negative tempco, much closer to what they had originally intended (I am pretty sure someone messed up the calculation when designing the regulator).

The same voltage regulator architecture also finds use in:

although those with many blinkenlights, notably the KT-900, would be most obviously affected. (Now that guy is a whole different story in and out of itself… *foreshadowing*)

KT-80 No-Flicker LED Lighting Mod

I had already tried to simulate and improve upon the lighting circuitry back in 2007, but didn't have any good ideas at the time how to tackle the high AC voltages. Well, now I did – what about a capacitive dropper?

[LTspice schematic of lighting circuitry with mods]
KT-80 lighting circuitry with mods.

I settled on a slightly oversized capacitor and zener regulator, and in hindsight am glad that I did because voltage regulation turned out to be excellent, about 7.95 V with only 10 mV of load variation once 5 mA worth of warm white LED was added for power button illumination. That's actually better than the main voltage regulator stock… precision lighting. ;) OK, there is about 15 mV of initial warmup drift, but still.

The following had to be done:

Note, it's a clear LED and in hindsight I could have tried diffusing it, although the result is not bad as-is. Also, 5 mA is plenty bright (actually too bright when it's staring you right in the face) and half of that would arguably still be plenty sufficient.

Since finagling the power supply board in and out of there is a bit tricky, I would very much consider replacing the main filter cap while you're in there. I didn't and later noticed signs of incontinence. Post-mod, rectified voltage rose by 2 V, reaching a tad over 24 under light load, which is a bit hard on a 45-year-old 25V part. The original is a 16x25 mm size (d x h), though you could install something a bit bigger still in both diameter and height… I found some 2200µ/35V caps in 16x32 locally, and after installing one of those I was kind of glad that I didn't go with a 3300µ/50V in 18.5x35.5, as things turned out to be fairly tight as-is. Definitely whip out the insulating tape for the top when going with one of those. The original glue is a bit of a bear to get off, scrape off what you can. I didn't actually find any major leakage, only the glue on the cap struck me as softer than I'd have expected. Electrically the part still appears to be in perfect working order.

Mains power consumption with all mods has been reduced to 5.8…8.5 W. You could arguably drop this by another third or so with a different transformer of lower output (you really don't need 22-24 V unregulated for a 14.2 V supply), but since this one doesn't appreciably hum I'm not in a particular hurry to replace it. Plus, a soft, slow transformer protects my dropper capacitors and the small stock rectifier diodes (only a 0.6 A type). (The secondary measures about 10.3 ohms across, the primary about 450.)

Should you merely want to replace the original lightbulb (you can apparently get kits), I would strongly recommend the following additional mods:

This, inspired by the KT-900, should keep both the bulb and zener running in spec. You would still have the flickery dial pointer LED, of course. That could still be taken care of e.g. by running it from regulated +14 V with higher series resistance, but things would no longer be as minimally invasive then.

AFC Mod

The main AFC time constant is determined by (R74 + R75) × C72. So I tacked on an additional 47µ/25V below the board at C72, with similarly long legs, pointing away from the frontend just in case. That's literally it.

Listening to the LO signal with AFC turned on now reveals a fair bit less low-frequency content. There is also a pronounced difference in hum levels between this and the KT-900 now, which was already somewhat there before (I think the '900 has a fair few hours on the clock) but has seemingly widened with the new, bigger filter cap installed. Update: A new filter cap for the KT-900 has brought hum levels roughly back in line again, although for the newer model it's still more of a hum whereas for the KT-80 it's more of a buzz; maybe the KT-900's 100µ/25V could stand being changed as well.

Tips and Tricks

The stereo decoder VCO alignment procedure outlined in the service docs is quite convoluted. If you happen to have a fancy multimeter with a frequency counter function, odds are you'll be able to follow the procedure outlined in the KT-900 service manual, i.e. measure from the VR1-R28 junction to ground. If you do not have an RF generator, turn on REC CAL for a much more steady oscillator frequency. Incidentally, according to the HA12016 datasheet, separation should be maximum for a free-running frequency of about 75.1 to 75.6 kHz, but note that the VCO also drifts down when warming up.

Of course, on a tuner like this that'll enable stereo pretty much as soon as possible, you can easily do VCO alignment by ear and not be very far off at all (aim for the middle of the range of working stereo, then quickest locking on a weak signal after momentarily enabling REC CAL).

↑ Back to topics

Entry last modified: 2025-06-07 – Entry created: 2025-05-24

The Kenwood KT-900 AM/FM Tuner Project

Overview

The Kenwood KT-900 is a neat slimline analog tuner with digital frequency counter made in 1981. It is kind of a cross between the older KT-80 or KT-615 and its bigger brother KT-1000 and thus also features a pulse count discriminator.

Now that I have a physical unit to go with the service docs (it even came with the ever elusive AM loop and a similarly elusive user manual), I want to bring it back up to spec and turn it into the awesome band cruiser it could be.

[KT-900 in muted lighting]

Here are some pros and cons I noticed:

Service Doc Inaccuracies and Hardware Bugs

While the KT-80's schematic only had one major issue (the incorrect output connections at the stereo decoder IC), the KT-900's is downright blooper city, indicating a bit of a rush job. I have found the following inaccuracies, plus even a straight up hardware bug:

Work Log

My unit seems to have a fair few hours on it. The plan is to make it an awesome band cruiser in my local, fairly high-RF environment.

Work already carried out

TO DO

Tips and Tricks

If you happen to have a fancy multimeter with a frequency counter function, odds are you'll be able to easily realign the stereo decoder VCO, as your multimeter with its high-impedance input will be VTVM/SSVM and counter rolled into one. (The service instructions assume a lab counter with a 50 ohm input, which would completely throw off the frequency if you were to attach it to the circuit directly, hence the voltmeter as a buffer.) With this assumption, follow the procedure outlined in the service manual, i.e. measure from the VR1-R28 junction to ground.

If you do not have an RF generator, turn on REC CAL for a much more steady free-running oscillator frequency. For some reason, this function does not defeat the VCO, which in turn is completely unaffected by the ca. 430 Hz cal tone.

Incidentally, according to the HA12016 datasheet, separation should be maximum for a free-running frequency of about 75.1 to 75.6 kHz, but note that the VCO also drifts down when warming up, so given how warm the KT-900 gets, aligning to 75.8-76.0 kHz may be just about right.

Of course the VCO alignment is one of those that can generally be dialed in pretty well by ear already, even with no special tools at all. Even on a tuner that won't enable stereo below a certain signal level you can still determine where the range of working stereo ends (preferably on a station as weak as the unit will permit) and aim for the middle of that range, and odds are you won't be very far off. You can also momentarily enable REC CAL and observe how quickly the stereo indicator lights back up afterwards – obviously the closer the oscillator is to 76 kHz to begin with, the quicker the PLL will lock.

AM RF input circuit tracking alignment points may be worth playing with; I peaked it up at 1458 instead of the usual 1400 and have been quite happy with evenness of sensitivity towards the top end.

Mods!

WIDE IF switching diode bias

You know I couldn't leave R9 alone (originally a 100 ohm fusible btw, heaven knows why). While the 470 ohm resistor I had on hand isn't quite the 560-750 simulation reckons it should be, 3.7/6.0 mA is a fair bit closer to the 3.2ish Kenwood usually seems to have been going for than the original 4.8/8.4 mA. It'll never be quite right for both diodes anyway, and 1k would have landed me sub-3 mA.

S-meter

Looking at potential reasons for the optimistic S-meter on FM, I found the LED driver input amp to have more gain than it should. According to its datasheet, amplified voltage needs to reach 3.5 V to light the 7th segment, but that already happens at 0.23 V, indicating a gain of about 15.2 when according to resistor values it should be 9.333… instead, which normally would require 0.375 V. (In a way that is quite lucky, as I haven't seen more than 0.36 V yet, which is 2.81 V out of IC2, so the 7th segment would actually never get lit if the circuit worked entirely as intended!) Amplified voltage can reach 5.5 V. I suspect 150k resistor R135 has drifted quite thoroughly (must be a good 260k if so); at least 16.2k looks close enough for an in-circuit measurement of 18k R138.

Now I actually like S-meter action on AM just fine, it certainly doesn't need to be any less sensitive, plus our suspected culprit R135 is a bit buried next to L19 and wouldn't be the easiest to get ahold of (component legs were bent over quite thoroughly during production, so things won't come out without a bit of a tug). So I decided to change FM S-meter series resistor R31 (68k) instead, which forms a voltage divider with R136 (10k). A calculation reckons the ideal value to be about 109.5k, I went with a 100k I had on hand. A lot of my previous 7s are still 7s so range still doesn't set any records, but at least the previous 7 threshold should now be down to about 5 instead.

I just read a 1980 midrange tuner review, and insufficient S-meter range seems to have been shockingly common at the time, with the KT-80 of all tuners (max at ca. 300 µV, consistent with DD0OT's 51 dBµV) still being among the very best, and I don't exactly consider that a world-beater. Standards must have improved substantially in the following years, much like they did for tuner performance in general.

Stereo threshold mod

The previous mod can backfire on you as stereo operation is tied to the 4th S-meter segment being lit and the stereo threshold is higher than necessary to these ears to begin with.

If you want stereo at (almost) all times, you can unsolder and pull out the end of wire bridge J31 facing R142/IC6 pin 12, and run a short wire or spare component leg from the other end of J31 over to the R141/IC6 pin 9 junction instead, which should drop the threshold to 1 S-meter bar (plenty low enough for anyone, I would reckon). Even just unsoldering J31 already seems to work as well (while a connection to J29 weirdly doesn't).

↑ Back to topics

Entry last modified: 2025-06-07 – Entry created: 2025-05-30

Kenwood KT-1100 No AM Repair

Here's another from my stash of tuners from the olden days. I had done a bit of tweaking on it back in the 2000s (mostly S-meter and center-tuning discriminator alignment), but it had been living in my parents' basement for almost a decade before I brought it back recently. Most of the musty smell thankfully went away after a good wipedown with some vinegar, and contact cleaner plus an old toothbrush made quick work of the oxidation on the varicap wiper contacts.

Borrowing the KT-900's AM loop to try out the fairly fancy AM section, I noticed that it did not actually work any more. After a brief blip of activity, the frequency counter settled on 550 kHz (which I assume is some sort of default, since 550 kHz = 1 MHz - 450 kHz IF). I also noticed that the FM LO remained on.

In order to get down to this problem, I finally had to understand how AM power works in this model. You see, the highlighting in the schematic only leads you back to a 10k resistor on the base of a switching transistor that provides power to the FM frontend when AM B+ is off. So that led me to suspect that the AM section didn't have any power, but where was it supposed to be coming from then? Well, there is an innocuous-looking branch going to the cathode of diode D15, from which it turns out there's a connection all the way back to the FM/AM mode switch. Bingo!

So then I proceeded to locate D15 inside the actual unit, which seemed a convenient place to measure and bisect if nothing else. On AM, the multimeter indicated 13.5 V on the anode and 1.7 V on the cathode. So the regulated supply clearly made it to the diode, but there was nothing going through it as 11.8 V is far from your normal silicon diode forward voltage drop. I quickly verified with the thermal camera that it wasn't getting hot either (which may have indicated a short in the AM section), but nope, nothing warm at all. Should it really be this easy?

I only had some massive hulking 1N5406s on hand, so a trip to the local components shop was in order, from which I returned with several types of diodes. The old D15 is a type that can sustain 0.3 A and 30 V of reverse voltage (1N4148 for comparison: 0.2 A / 75 V), so playing it safe I settled on a 1N4936 quick recovery silicon rectifier (1 A / 400 V), the intermediate option between a 1N4148 and 1N4007.

Now Kenwood always bent over their component legs super well, so D15 put up a bit of a fight, but nothing that a desoldering pump and needlenose pliers couldn't overcome. The new diode soldered in without issue, and a quick test revealed 12.8 V after D15 and a working AM section! Phew. (The 1.7 V seem to be a result of current from the aforementioned 10k resistor to switching transistor base affair.)

I then proceeded to realign the AM section as per service docs (note, they leave you to figure out which trimmers are TC3/5/7 on the varicap, but it isn't too hard to figure out as touching the LO part will detune it), which proved quite necessary as alignment actually turned out to be a fair bit off even on the bottom end.

Unfortunately, receiver performance turns out to be a bit of a mixed bag. It is as hot as the presence of a tuned RF stage would imply, however I have a hard time getting rid of 9 kHz hets entirely even when switching to NARROW mode, which activates additional audio filtering. The base IF filtering – a 4-element SFP450F 12 kHz filter plus IFT – is best described as "barn door". Good for audio fidelity on strong local stations in 10 kHz territory, I assume (and it does sound good), but not very useful to me with a band that only comes alive in the evening. I actually prefer the more basic KT-900 with its much narrower filter. Anyone got an SFP450H or SFP450G to spare?

↑ Back to topics

Entry last modified: 2025-06-06 – Entry created: 2025-06-06

© Stephan Großklaß 2025. Commercial duplication of this content, including eBay descriptions and similar, with prior permission only.

Contact me