Here's a bit of a puzzler.
This is the voltage regulator for the main +14V in a Kenwood KT-80 tuner:
Note compensation capacitor C87 (560p) below Q10.
This is the same in a 1-2 year newer KT-900. Almost identical topology and values, basically the same part types - 2SC945 or similar for the small transistors and 2SD330 for the big series transistor. (Current draw could be at least 50% higher though, and the 68 mA shown is decidedly not accurate, as are a number of things on this schematic. For one, R154 is actually a 5.6k, and voltage is 13.5 V and change.)
Note how C96 is now a .01 = 10n, as confirmed by the parts list.
This is the only model with this regulator topology that's taking such a drastic approach to compensation. No other model using it uses anything nearly as big (KT-615, 815, 80, 1000: 560p; KT-1100: 2200p or 0p). Feedback resistors and hence voltages tend to be of very similar values: 5k6/6k2, 4k7/5k1, 5k6/6k8, 4k7/3k9, 4k7/4k7.
Question of the day: Why would they have done this?
Most things are connected to this main rail via RC filtering usually using 100 ohms. The only thing with a lowish-impedance connection is the frontend, which sits after series silicon diode D5 alongside the IF strip and has LC filtering in the form of 1 µH and 100µF || 10n || 1n. (A bit of 10.7 MHz IF might be getting onto the rail via 100 ohm Wide bandwidth switch resistor R9, but I can't imagine it would be that much at this fairly early stage and preceded by a 10n to ground, plus I already swapped the poorly dimensioned R9 for a 470 ohm. The optimum value would be 560-750 ohms according to sim, but I only had 470R and 1k on hand.)
The PCB layout is quite different between KT-80 and KT-900, mind you.
I intend to port over my feedforward mod from the KT-80 (which incidentally works a treat for DC regulation)...
....and was wondering whether some attention to compensation may be warranted and how to check for potential abnormalities. (Have fancy multimeter and old 10 MHz analog scope.)
This is the voltage regulator for the main +14V in a Kenwood KT-80 tuner:
Note compensation capacitor C87 (560p) below Q10.
This is the same in a 1-2 year newer KT-900. Almost identical topology and values, basically the same part types - 2SC945 or similar for the small transistors and 2SD330 for the big series transistor. (Current draw could be at least 50% higher though, and the 68 mA shown is decidedly not accurate, as are a number of things on this schematic. For one, R154 is actually a 5.6k, and voltage is 13.5 V and change.)
Note how C96 is now a .01 = 10n, as confirmed by the parts list.
This is the only model with this regulator topology that's taking such a drastic approach to compensation. No other model using it uses anything nearly as big (KT-615, 815, 80, 1000: 560p; KT-1100: 2200p or 0p). Feedback resistors and hence voltages tend to be of very similar values: 5k6/6k2, 4k7/5k1, 5k6/6k8, 4k7/3k9, 4k7/4k7.
Question of the day: Why would they have done this?
Most things are connected to this main rail via RC filtering usually using 100 ohms. The only thing with a lowish-impedance connection is the frontend, which sits after series silicon diode D5 alongside the IF strip and has LC filtering in the form of 1 µH and 100µF || 10n || 1n. (A bit of 10.7 MHz IF might be getting onto the rail via 100 ohm Wide bandwidth switch resistor R9, but I can't imagine it would be that much at this fairly early stage and preceded by a 10n to ground, plus I already swapped the poorly dimensioned R9 for a 470 ohm. The optimum value would be 560-750 ohms according to sim, but I only had 470R and 1k on hand.)
The PCB layout is quite different between KT-80 and KT-900, mind you.
I intend to port over my feedforward mod from the KT-80 (which incidentally works a treat for DC regulation)...
....and was wondering whether some attention to compensation may be warranted and how to check for potential abnormalities. (Have fancy multimeter and old 10 MHz analog scope.)
Last edited:
Rather than querying Kenwoods parts choice on its own , the answer you seek is in recognizing what the circuit provides. You should spot its a Darlington pair, arranged as a capacitance multiplier. The capacitance values at C86 and C87 they deemed appropriate, to decrease ripple sufficiently from the R/C network presented at the base, for supplying subsequent circuit stages,
The later 10nf value is better still for ripple reduction. Kenwood carefully looking at the Darlington pairs hfe. Substituting their transistor choice with others as your third diagram shows, has it appears not factored how much ripple reduction was firstly achieved by Kenwood's choices.
The later 10nf value is better still for ripple reduction. Kenwood carefully looking at the Darlington pairs hfe. Substituting their transistor choice with others as your third diagram shows, has it appears not factored how much ripple reduction was firstly achieved by Kenwood's choices.
The real oddity to me is L16 in series with the reference Zener diode in your second image.
I wouldn't worry to much over the whys and wherefores of what is really a classic text book implementation. It may that the design suffered from demodulation of high field strength RF signals (used in locations near powerful transmitters) and all this is just a belt and braces approach to force absolute stability under all possible conditions.
(Service manuals of this era and type often have lots of different variants shown as well, often with differences applicable to certain geographical regions)
I wouldn't worry to much over the whys and wherefores of what is really a classic text book implementation. It may that the design suffered from demodulation of high field strength RF signals (used in locations near powerful transmitters) and all this is just a belt and braces approach to force absolute stability under all possible conditions.
(Service manuals of this era and type often have lots of different variants shown as well, often with differences applicable to certain geographical regions)
Might be a simple as saving money on parts by standardizing on 10n, eliminating the 560pF.
The inductor might be to compensate for falling transistor gain at HF, or it could just be to reduce ground noise.
The inductor might be to compensate for falling transistor gain at HF, or it could just be to reduce ground noise.
This capacitors 560p & 10nF define load regulation transient response. So decision is based on what is connected to output, switching / high current load demands low capacitance, same time for low current capacitance may be increased to minimize noise.
I tested this topology driving cap value up to 510 uF (!!!) and was able to get < 0.2 nV/sqrt(Hz). 10 nF likely to show ~2-3 nV noise.
I tested this topology driving cap value up to 510 uF (!!!) and was able to get < 0.2 nV/sqrt(Hz). 10 nF likely to show ~2-3 nV noise.
I also think it's for improved high-frequency noise filtering (/ avoiding ground pollution), in addition to the RC. I can't make out the markings terribly well, looks like red-red-silver(/gold)-black or orange-orange-silver(/gold)-black, so perhaps 0.22/0.33 µH or 2.2/3.3 µH.
Wouldn't be the first suspected instance of "let's bodge it till it works" in this model then. There are various indications of it being rushed and on a budget. (I have unfortunately been unable to find a schematic for the similar Japan-only Trio KT-990 that succeeded the KT-900 there in 1982, that would have been interesting.)
The optimist in me wants to believe that they may have spotted how much the voltage was changing under variable load (this model has a bunch of blinkenlights, and output can vary by 80ish mV) and tried to smooth out the transitions. In which case implementing the feedforward mod would make the bodge redundant.
120 kHz loop gain sounds like quite a lot, until you look at the RC time constant involved (5k6 * 10n = 56 µs) and realize that output impedance would start rising and going inductive around 3 kHz. Noise-wise it probably isn't a bad thing, even if the circuit is very quiet to begin with. (Which it also has to be, since B+ effectively acts as audio ground at the stereo decoder output. Granted, the SNR spec is 88 dB(A) at 75 kHz deviation / 0.75 V, that's roughly 30 µV(A) worth of output noise.)
I am also concerned because this may affect how much if any rail bypassing my planned amplifier circuit for the AM department would need. The most recent iteration in sim looks like this, R7/C4 values being mighty tentative:
The stock circuit is using a passive RC highpass (0.47µF * 15k || 0.5(110k||47k||Rin_amp)), and using a headphone amp with 10k input impedance (instead of the 47k that the accompanying KA-900 would have provided) is kinda messing with the low end, plus I can't imagine a KA1197 was designed to drive an effective 3.8k.
(I know my PSRR on the bottom end is a bit meh, but it should easily be good enough still, as in hum 80+ dB below signal on AM. Also, I am using the ON Semi LM358 model, which is way better than the crude one that was available before.)
Anyway, it's wild how quickly the regulator circuitry evolved at the time. The 1977 KT-7500 still had to make do with basic zener-followers. By 1983, the KT-1010 / BASIC T2 and KT-770 synthesizers sported precision type regulators with 4558-class opamps. OK, those disappeared again afterwards and were replaced by more traditional-looking designs in subsequent years, perhaps noise performance was not satisfactory or it was just too expensive or both. When opamps did return (e.g. KT-880D, KT-5020), they were M5223Ls, which it seems are plain LM358s. Otherwise the preference seems to have shifted to adjustable voltage regulator ICs (M5231TL, M5230L).