modifying a Sanyo JCX-2600K

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
The Sanyo JCX-2600K is a stereo receiver made around 1979. The power amp section is a basic Lin amp, miller compensated. It makes 85wpc from a single pair of bipolar outputs running on a pair of +/-55V rails. In stock form it is rated for 0.01% distortion. Spice supports this figure; the distortion floor measures just below -80db in simulation.

I have one. I'm planning to recap the power supply and preamp, and to modify the power amp to make less distortion (-110db floor is the goal), less noise, and better reliability.

The output stage blew in the past. I'd like short circuit protection at minimum. Real SOA limiting is a stretch goal, this amp generally won't be asked to drive more than one watt per channel anyway.

The preamp on this unit is a no-feedback all-discrete design that simulates with a distortion floor 95db below signal or better, mostly low harmonics. That's good enough for now. If the power amp comes out nice, maybe the preamp will go in for mods in the future.

Attached is the original power amp schematic. Notice the weird VAS. Transistors Q704 and Q705 form a sort of two-transistor current source, their bases held at roughly constant and equal voltage, and so R707 saturates at a basically constant voltage. When Q704 conducts more, Q705 must conduct less and vice-versa. Ideally, collector currents are balanced at the LTP, and also at the Q704-Q705 pair. This places an upper bound on the current the VAS can sink, 10 or 12mA or so. That makes the protection circuitry's job easier, it doesn't have to source a lot of current to cancel the VAS when limiting the pull-down current.

One caution about this VAS is that if pot SVR701 is set too low, current through Q705 can be far too high. You can make it dissipate over 1 watt if you want, for as long as the transistor lasts.

I'll talk about possible mods in a future post.

Here's the starting-point schematic. It should be correct, except all transistors are replaced with modern ones for which spice models exist (from Bob Cordell's website or from the manufacturers.)
 

Attachments

  • sanyo_amp.png
    sanyo_amp.png
    76 KB · Views: 256
Here's one option, a relatively mild mod.

This design adds emitter degeneration to the LTP and makes Q704/Q705 into Darlingtons. It adds a protection resistor at the collector of Q705 to prevent it from conducting too much current. Compensation is changed to TMC.

This makes 2nd harmonic distortion at a -90db floor. Higher order harmonics are at -100db or below. That's more distortion that I had in mind.

Otherwise this design has its advantages.

It doesn't require many changes to the PCB. There are already jumpers that can be replaced with the LTP emitter resistors. And it's quite easy to replace a single transistor with a Darlington, if you buy the Darlington in a single package. That's like cheating.

Like the stock VAS, this VAS won't sink more than about 18mA. That helps the protection circuitry, which must cancel only 18mA when clipping the pull-down output current. That means the stock protection circuitry would be good enough.
 

Attachments

  • sanyo_mod3.png
    sanyo_mod3.png
    84.2 KB · Views: 231
This is a more aggressive option requiring PCB surgery. It can do a -115db distortion floor in the audio band at half power into 4 ohms. It has single-slope SOA protection for the outputs.

This has LTP emitter degeneration, a current mirror to equalize the LTP tail currents, a beta-enhanced VAS, and a fast-switch-off capacitor between the driver emitters. So far this is Doug Self's blameless recipe.

It has TMC compensation and a current limiter in the VAS. The limiter is borrowed from ostripper's wolverine. (Hence component names Ros1, Cos1, etc.) This is a nice limiter-- it's easy to find values that don't oscillate, it makes a nice flat clip, it still allows the VAS to pull its output voltage all the way down to about a diode-drop above the negative rail.

Again, the VAS current limiter will make the output stage protection circuit's job easier.

Unfortunately, the current limiter requires emitter-degeneration at the VAS transistor. That reduces VAS gain and overall linearity some, compared to a design with no VAS emitter resistor. I made the resistor as small as I thought would be safe.

The stock SOA protection is modified so that the pull-up and pull-down sides are not symmetrical. They shouldn't be! On the pull-up side, the circuit must cancel 4mA from the Q706 current source. On the pull-down side, it must cancel about 25-30mA that the VAS can sink. So the pull-down side has to tip-in a little faster to get the same clipping current level. And Q708 probably should be rated for at least 100mA, while Q707 can be tiny.
 

Attachments

  • sanyo_mod.png
    sanyo_mod.png
    91.3 KB · Views: 207
Build is complete on both channels. Sounds great! Lots of details and no etch, fuzz, or grain to my ears.

Here's the final schematic. There are a couple differences between the schematic and the real amp. The real drivers are 2SA1507 / 2SC3902 but there's no spice model for those. Otherwise all transistors are correct. "Lead1" through "Lead4" are wires leading to a little daughterboard, I modeled those as small inductors just in case it would make a difference. Ideally these should just be wires.

I was able to fit TO-3P devices in TO-3 sockets, using Aavid Thermalloy 4180G aluminum oxide thermal interface pads to mount them. I needed longer mounting bolts, the AlO2 pad plus TO-3P device is thicker than the old TO-3 and so the original bolts wouldn't quite grab the socket. I had to trim 2 or 3mm off the end of the base and collector pins to fit them in these particular sockets. It was worth it to get brand-new 30Mhz devices.

All caps in the signal path are film or C0G ceramics, except for C707 (the 220uF 'lytic in the NFB path) and Cef2 which is the Self EF2 switch-off speed-up cap. That one is a 1u MLCC which is not C0G. I don't think this cap ever has a lot of AC across it, and what AC it has probably doesn't resemble the signal too much, so my guess is that it doesn't really have to be the world's most linear cap. Not sure.

EDIT: one other minor mod I did was to swap the output inductor with its parallel resistor. That moves the inductor further from the traces that carry supply rail current to the output transistors. Sanyo to their credit used twisted pair for the supply rails, from the power supply to the amp PCB and again from the PCB out to the heatsink which should minimize induced magnetic fields. On the PCB the rails are parallel traces which in the original design are segregated pretty well from everything except the output inductor.
 

Attachments

  • sanyo_final.png
    sanyo_final.png
    79.7 KB · Views: 183
Last edited:
Neat project. :cool:

Note: The Q704/705 VAS in the original circuit "kinda-sorta" is a differential amplifier, with the idea possibly being to take advantage of the better distortion characteristics. I simmed it some time ago, and from what I remember it made the input pretty susceptible to common-mode distortion. It may be best suited for circuits doing pre+power at once (Av ~= 45 dB).

As far as the preamp section goes, I don't know what made you think "no feedback" - Q551/552 form a 2-transistor opamp circuit with plenty of open-loop gain, and closed-loop gain being set by R556/R555. The tone amp (Q553) also uses feedback, it's an inverting Baxandall type tone control if I see that correctly - Tone Defeat entirely bypasses it. (Which is a good thing as I suspect it may be the Achilles heel.) Finally, Q581 is an emitter follower with a decent amount of idle current and degeneration - with the decently high supplies it should have good linearity as-is, though you should be able to get somewhat better results still by adapting R591/592 to shift its output level up a few volts (5...10 V max to keep Ccb low and linear, and mind C591/592 polarity!), with R593 increased accordingly.
The mysterious Q582 merely serves a a switching transistor, engaging C595 to form a lowpass - the AM IF filter in this unit is not particularly steep (as in many run-of-the-mill AM sections) and needs all the help it can get to keep the adjacent channels at bay. Same goes for the high-frequency distortion components of typical AM detectors.

The main fault of the preamp is not distortion (did you sim with a worst-case source impedance of 25k?), but rather noise. R555 = 1k8, 100k volume pot... should be OK but hardly record-breaking. I would look at tweaking gain distribution.
R511 may be a good place to patch in a 6 dB or so preamp on a small extra board (needs good linearity at full line level input, while noise is not much of an issue, but supplies of +/-20.5 V have to be tolerated - this might be a case for the OPA2604 for once, or the less power-hungry LT1124, though a 2-transistor discrete affair may also do; in any case contemplate going with an inverting opamp circuit). You'd then drop gain in the existing preamp by about as much. If that doesn't reduce noise by almost as much, you may want to get gain back up by 3 dB while taking that much away from the power amp.

So you also tackled a KR-9400 on AK, huh? That buffer on the problematic PUSH SW A board actually isn't that different from the one found in this Sanyo, so I was wondering what the problem was.
Well, the Kenwood has lower supplies (+/- 14 vs. +/- 20.5 V), lower current in the EF (~2.5 mA vs. >4 mA), higher levels at the EF in practice, and most of all it has to drive a 10k pot vs. 47k worth of power amp. I would expect about 3 + 3 + x + 14 dB = 20 dB + x more distortion from these factors. You'd have to replace the 5k6 emitter resistors by current sources (which typically drops distortion by an order of magnitude) to save that one.
What I like better in the Kenwood circuit is that noise is better taken care of (showing the RF affinity of this company which still makes ham radio gear today) - it takes base current from ground rather than supply midpoint, which is roughly the same but could be a lot noisier. (And then there is the obvious advantage of a 2-stage volume control.)

Obviously there is nothing (in principle) keeping you from tweaking the Q581 buffer in the Sanyo with both a CCS for R593 and bias via ~120k from ground, especially when shooting for even higher levels in the preamp.

Do you have anything to actually measure distortion? I'd say you can get down to about -100 dB with good-quality consumer PC audio gear (under 200 bucks for internal cards, sometimes under 100 even). You'll need to fabricate 4 and 8 ohm dummy loads and attenuators in any case, and you may need the odd ground loop eliminator / "hum destroyer" (audio isolation transformer) or 1k resistor in the ground, but overall it's quite affordable fun when compared to a full-blown audio analyzer. OK, that one would have galvanically isolated ins and outs and would measure THD deeper down without an external notch filter - at one more zero in the price tag or more.
 
Last edited:
sgrossklass, thank you for the encouragement and feedback. I was wondering what the link between the AM tuner and preamp output was doing.

With no input signal on the Sanyo, there's an audible hiss that peaks at 6db below max volume and diminishes quickly on either side of that point. At sub-Spinal-Tap levels like -40db the volume pot impedance is closer to 1Kohm and the noise level is below the noise floor of the room.

Any preamp redesign should not make the noise level any worse.

Back in spice I added a 25K input impedance and 50K output load to the preamp. They don't make linearity worse. They make the noise sim a lot worse than zero input impedance as expected.

Loading Q552 and Q581 with stiff current sources improves linearity and doesn't increase noise.

Q551 has a noise/linearity trade-off. If you quadruple its tail current, you increase its gain and improve the linearity of the Q551/Q552 stage. This raises the noise level at preamp output from 220nV/hz^1/2 to 360nV/hz^1/2. I don't know why.

For fun I converted the input stage to a long-tailed pair with emitter degeneration and balanced tail currents, see the attached pic. This is about 10% less noisy. Distortion is very good, peaking at 130db below signal. There's no way all that stuff fits on the PCB. You would have to fit a crowded daughterboard somewhere. Or buy the opamp that this is starting to resemble.
 

Attachments

  • sanyo_preamp.png
    sanyo_preamp.png
    78.7 KB · Views: 154
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.