Current Source replacement of 2SK34 in SX-780 PSU

[rcs16]

I do not buy that NTE is garbage, I have proof, I used TO-220 drivers in my Pioneer SX-950 20 years ago and they are still fine.
but NTE parts should be used as a last resort since there are usually many cheaper/better options.
for the jfet match Idss is one of the most important spec's in order to match a jfet. Pinchoff voltage is another. Since this application is basically DC abs max voltages and power dissipation are things to match. If you look at many Sansui high end gear, they used a simple bjt/diode/resistors as a constant current source for this linear regulator supply startup application..
Echowars led/bjt CCS is another option.

 

[xconnectiv]

is this cucuit work with SX-1980 power supply ?

 

[markthefixer]

is this cucuit work with SX-1980 power supply ?

NO

The sx-1980 has a MUCH higher voltage that calls for that specific EchoWars circuit, not a CRD or JFet.
depending upon the transistor ratings, it can work and be safe at some insanely high voltages.
The original (and individually selected) JFet from the factory was always in jeopardy, running on the ragged edge.

 

[merlynski]

I was referred to this thread from another thread. There were difficulties about the CRDs and their application in an SX-780. I modified the prior schematic, and post it here in hopes it will help any future viewer.

 

[ItsGarand]

I happened upon this thread thanks to @LesE and @QSilver.
Perhaps others can opine as well, I have a thread going for the resurrection of an A-70 integrated Amp.

One of the power supply circuits utilized 2SK373 JFETs. The JFET on the negative side failed spectacularly some time in the past(The back of the case was blown completely off - pics in thread) prior to me picking up the unit.
I picked up two 373s from Dalbani that were supposedly OEM - alas as you might suspect, they were not - and did not work.

So I decided to go the CRD route. Making a longer story shorter I picked up E-352s (Thanks @QSilver @rcs16 )from Mouser. They seemed to do the trick - after some other work the amp lit up, they relay clicked and I actually got to play music through it - for a few minutes. Those CRDs didn't last long both of them went TU the next time I powered up the amp.

Assuming that something else isn't actually causing the CRDs to smoke, which could very well be the issue.....

Here is the question if you have the time to opine:

When replacing JFETs with CRDs - It was pointed out, the Ip should be in the range of the Idss of the JFET.
In my case, the Idss of those 2SK373s(gr rank) is 2.6 ~ 6.5 mA so I chose the E-352 since it's right in the middle of that range.
Should I be choosing a CRD that has an Ip at or near the top of that Idss range?

The section of that circuit looks like this:

Also,
The datasheet for those CRDs shows that as temp and/or V goes up, the Ip drops. If I read the graph correctly, the E-352s Ip at 10V is 3.5mA, but at operating voltage of 63V is around 2.1mA
Should I be picking CRDs that are towards the upper side if the FETs Idss like an E-452 or E-562 (at 63V Ip is around 3.2mA - assuming I'm reading that graph correctly)

The data sheets for these CRDs as pointed out by QSilver indicates that resistors in parallel can help flatten the Ip drop-off due heat.... So, perhaps also pair it with the appropriate resistor?


Thanks for looking....

 

[8-Tracked]

I do not buy that NTE is garbage, I have proof, I used TO-220 drivers in my Pioneer SX-950 20 years ago and they are still fine.
but NTE parts should be used as a last resort since there are usually many cheaper/better options.
for the jfet match Idss is one of the most important spec's in order to match a jfet. Pinchoff voltage is another. Since this application is basically DC abs max voltages and power dissipation are things to match. If you look at many Sansui high end gear, they used a simple bjt/diode/resistors as a constant current source for this linear regulator supply startup application..
Echowars led/bjt CCS is another option.

NTE garbage? Wow that's a rather unique, lesser common unspecified qualification of NTE! Not that I think Toshiba is garbage! Don't get me wrong!

 

[dcmeigs]

I’d like to revive this long dead thread to ask a simple question. Since Q19 and Q20 are being used to control voltage, why not control the base voltage rather than the base current. A resistor/zener network would do that job well, (as would a CRD Zener network if not for the failure mode). Is noise the issue?

I ask because I attempted a CRD fix on a SX-880 using the E202. 2 ma was too much current. I installed a 33k resistor that was about perfect at idle and measured 250ua. I’m tempted to use a 22k resistor and three 14v diodes Instead.

 

[markthefixer]

CRD is used to servo OUT the hum. for clean regulator power. look how they do that instead with LARGE caps in older models.

put up your sx-880 schematic,

constant current diode dances the voltage in opposition to the hum.
the input voltage varies, from hum.
the regulator reference voltage comes from elsewhere in the circuit.

 

[dcmeigs]

CRD is used to servo OUT the hum. for clean regulator power. look how they do that instead with LARGE caps in older models.

put up your sx-880 schematic,

constant current diode dances the voltage in opposition to the hum.
the input voltage varies, from hum.
the regulator reference voltage comes from elsewhere in the circuit.

That’s good to know. The 33k resistor was helpful to verify the circuit was up and operating correctly And to measure the current. I’m really puzzled why the 2ma current doesn’t work in this application as it should. I’m going to replace the D325R with a ksc2073 and see if that makes a difference.

 

[dcmeigs]

Just for kicks, I moved the working SK34 from the negative circuit to the positive circuit. A sanity check of sorts. The SK34 works fine in the negative circuit but not so in the positive. It acts just like the CRD did. Too much current.

Q30 and Q28 are KSC2073 and KSC945Y respectively.

Consulting the C2073 Datasheet, 2ma looks about right.

The voltages around Q28 are 41.75 collector, 14.2 base, 14.33 emitter. What’s the chance the biasing of D22 is the problem? Q28 should be conducting some of the current away, no? I’m using a 1N4148 in place of the 1S1555. Truth is I’m not really sure the purpose of Q28 in the circuit.

 

[ItsGarand]

BTW - I did go with the E-352 to replace the 2SK373s (GR rank ) the Idss is between
2.6mA and 6.5mA

The characteristics (voltage-current) chart for those CRD shows that the E202 has a current of 2.2mA at 10V. In my case that CRD would be seeing nominally 63V. At 63V that chart shows only about 1.5mA for the E-202.

The E-352 at 10V is 3.5mA, but at operating voltage of 63V is around 2.5mA. The E-352s did the trick.

 

[dcmeigs]

Yeah thanks for the input. The mystery with this Sx-880 is why the 2ma supplied by the CRD won’t work in my circuit. It’s too much rather than too little current. My regulated voltage runs much too high with 2ma. A current limiting resistor of 33k installed in its place that passes only 1/2ma (0.5ma) gets the voltage right on target. I can’t find an explanation for that. All the components in the circuit appear to me to be working as they should. The negative supply works fine with the SK34. I suspect I’ve missed something.

 

[dcmeigs]

the regulator reference voltage comes from elsewhere in the circuit.

THIS! This is the answer. So I’ve been holding the CRD responsible for the high voltage from Q30 and that’s just not right? Please tell me how the reference voltage is set.

Edit: I think the R410/R409 voltage divider with D25 sets the Q28 base and the 14v Zener with the 68R resistor sets the Q30 emitter. What controls the Q30 base?

 

[markthefixer]

I can't, I don't have the schematic here in the hospital.

my memory is that the sx-880 drawing is unnecessarily complex, my untangled one is at home, l may have posted it on another thread.

 

[dcmeigs]

I can't, I don't have the schematic here in the hospital.

my memory is that the sx-880 drawing is unnecessarily complex, my untangled one is at home, l may have posted it on another thread.

the schematic is in post #29. Here’s another that’s a little easier to read. Thanks for your help, I hope you’re well soon.

 

[dcmeigs]

The problem is solved.

Q28 was the problem. The original part is a 2SC945A. The foil side drawing shows the pads as being ECB. The replacement part is a KSC945YBU but it’s an EBC transistor. No bueno.

I installed a KSC2383YTA with a ECB pin out for a win.

Voltages are 42.1 and -41.4. Ripple is 3.6mv top and bottom. The hum is gone!

The CRD E-202 worked just fine.

 

[markthefixer]

that a733-a733c ebc/ecb and the c945-c945c ebc/ecb
has been a PITA for decades
and was responsible for my boycott of putting them in any of my true direct parts lists.
instead I list the a992 & c1845.

 

[dcmeigs]

I’ll make a note about the c945 too. You’d think they would give the pin out charge a new number. I guess the 6L6, 6L6G, 6L6GB and 6L6GC might create some confusion, but at least they have a common pin assignment.