FET replacement - any of these worth trying

[yestertech]

I'm working on a Pioneer amp which has had a melt down in the PSU area. The main rails are fine, but the regulated lower voltage +ve rails are missing.
Ref to the attached diagram, I have found subs. for Q1 and Q3, but Q5 is also shorted. This is specified as a 2SK34 N channel FET. I don't have any of the recommended subs, two of which are given as BF 244/5
However, I have unearthed 2N 4858, 2N5457 and 2N5486 and I'm wondering how critical the device is in this circuit ?
I also have a trusty old 2N3819 or two, equivalents for which are stated as BF 244/5 ?

Any thoughts on whether any of these subs would be suitable.


Thanks

Andy

[MrBungle]

JFETs are usually incredibly poorly specified. A lot of the time they don't even work in circuit as specified and need to be hand selected as I just found out with a batch of J310 ones.

I'd start with something with about the same Idss range and see if it works. I'd start with the 2n5457 myself looking at a 2SK34 datasheet. Looks about right. Not sure if pinout is the same however. 2n3819 and BF245 will probably be fine as well.

[Mooly]

Are you sure the FET is short? If checked out of circuit its possible for the gate to float to a voltage that will cause the FET to 'conduct' and show a low resistance across the D-S channel when it is in fact fine.

One quick thought is to swap Q5 with Q6 and then fit (say) a 3k3 resistor in place of Q6.

Q6 is just used as a current source and can be replaced with a resistor, at least for faultfinding whereas Q5 looks a little enigmatic its configuration and I can't at first glance see why its needed.

Is that part of the circuit drawn correctly?

[Maarten]

A JFET will show a short when the gate isn't charged. It will only turn off if the gate has a negative charge. VFETs in Sony amplifiers were sometimes condemned because of this, while they were perfectly fine.

[Radio Wrangler]

You have a huge range in not only Idss to contend with but also Vp.

One of the FETs in the diagram is being used as a constant current source as well

It is quite likely that the FETs in the original were selected. The bounds on normal types are far too wide for most circuit design techniques.

Are you really, really sure those transistors are dead? JFETs are usually quite robust.

David

[G0HZU_JMR]

Originally Posted by Mooly

whereas Q5 looks a little enigmatic its configuration and I can't at first glance see why its needed.

Is that part of the circuit drawn correctly?

It does look a bit odd... I'm guessing that the design aim for this part of the circuit is to provide 52V at 11mA with a very low source impedance over most/all of the audio bandwidth.

R3 and C5 set the bandwidth of the feedback and Q5 is there for startup and also as the variable resistor (controlled by the fast feedback) that controls the bias current of the BJTs such that the loop stays firmly closed and this gives a low source impedance at about 52V. So I'd expect the 52V to be very stiff and clean even if current is gulped from this circuit at an audio rate. This is because of the wide bandwidth of the fast feedback loop.

So I'd expect this circuit to start quickly but also to have a very low source impedance and presumably this will also be obtained with low noise.

The worry for me would be finding a part for Q5 if it has failed. During startup it could very briefly see a high voltage across the D-S channel. So I'd expect that the 2SK34 is going to be rated at a higher voltage than a jellybean P50 JFET like the 2N3819 for example. I'm not sure what happens if the Vds ever gets higher than the figure on the datasheet. It might just leak a bit if the voltage isn't too excessive. As Q5 is meant to pass current during startup this might be OK. Once it's up and running at 52V at the output I'd expect Q5 to be running near pinchoff with a very small current passing through the channel and Vds will be down around 15V. One other worry is about how robust this circuit is if the 52V output was briefly shorted with a test probe. It might not survive this?

I'm not sure that device spreads in Idss or Vpinchoff are that critical in this circuit as there is plenty of voltage range available for Q7 to find Vpinchoff in Q5 and Q5 will normally only be operating at a few uA, so it is well away from Idss except during startup.

[G0HZU_JMR]

I guess a lot depends on how quickly the 68V supply rises at startup. If this 52V circuit can charge C8 (47uF) quickly compared to the rise time of the 68V supply then the JFET Q5 might not ever see a high voltage across the D-S channel at startup as the 68V and 52V fails will both rise together gracefully. I'm not sure how it all behaves on shutdown, but it might be possible to use a regular P50 JFET in this circuit with little risk of exceeding the Vds rating of the JFET. The JFET mainly needs to operate near pinchoff and it needs to be able to pinch itself down to very small channel currents in the uA range.

[G0HZU_JMR]

I have found subs. for Q1 and Q3, but Q5 is also shorted.

I think that in this circuit, the JFET would be the least likely to fail out of all these transistors so I think Mooly and David's suggestion that the JFET might actually be OK is a good one.

If the original JFET turns out to be OK when tested then hopefully this section of the PSU will run OK once Q1 and Q3 are replaced. However, I'd recommend having a close look at the original dead Q1 and Q3 to see if there is a gain group letter on them. Presumably the 2SA733A has a lowish current gain (just a guess) but the 2SC1735 datasheet suggests there are a few gain groups for this device. As the current gain of each BJT affects the gain around the feedback loop in the PSU it might be wise to try and keep this in the same ballpark as the original devices. If there is too much loop gain the phase margin may become compromised and instability may creep in. But hopefully this would require a really large increase in loop gain.

[Julesomega]

I had big problems with the voltage regs in another Pioneer tuner-amp, the SX-1990* so I was hoping I had some answers, but comparing schemas they are quite different configurations. In mine the regulated supplies were ±80V and the FETs were acting as current sources to give 2mA as the collector loads for the comparator transistors. In yours only Q6 on the -ve side is used this way, with 10k considered adequate for the +ve side. If you need to replace Q6 I suggest take a FET of your choice, add a source resistor and vary that to give a suitable source current, but I can't work out what that would be (not much help)

I had to replace several bipolars on the +ve side, they kept failing. I used an MJE13005 (700V Vceo) and 2 × 2SC2240 (120V) but they needed 100pF b-e to cure oscillating ~2MHz

Have to go out now but will see what more I can find later today

*I wrote some observations in the thread Pioneer SX-1250 and SX-1980 Flagship receivers

[yestertech]

My thanks to everyone for the valuable input. I'm not sure I quite fathom the 'FET testing short' idea.
If I put one of the subs on the Peak Atlas tester it tests fine ( although the readout is unable to correctly identify D and S ) . If I do the same with the 2SK, it shows 'short red to blue' where one of them is the gate.
I will reinsert it anyway and see what happens, using the bench 60-0-60 supply with current limit on, fed into C2/C3 I think , just in case. This will also show whether the vacuum display and preamp sections are, in fact , operating. The Power amp section is known to be working, as it takes it + and - rails directly from the power transformers, via a separate bridge rectifier.


Andy

[G0HZU_JMR]

It does look like your 2SK34 is dead if it shows a short at the gate on your tester. So I think it is not worth risking it again.

I've never seen a PSU regulator like this before with a JFET at Q5. I suspect that the JFET helps to speed up the startup process and to minimise the stress on the fairly feeble pass transistor Q1 during startup. Once the supply is regulating at 52V I think the JFET runs very close to pinchoff with maybe 1uA or less through its channel. The channel current is controlled by Q7 to keep it all regulated at 52V and it will be typically 11mA divided by the combined current gain of Q1 and Q3. So the JFET channel current could be 11mA/10,000 which is just under 1uA when the PSU is regulated. Looking quickly at the rest of the circuit it does look like the PSU is designed with some degree of 'sequencing'. The 52V (and -52V) supply is needed to start up the other PSUs for example.

Once repaired, it might be wise to be careful how you turn the whole amp on and off under normal use. I'd advise against turning it on/off/on in a quick sequence. Also, avoid being a bit vague with the on/off switch, don't ever hold the main power switch such that it vaguely wiggles in between on and off.

[Julesomega]

Thanks to Jeremy for a better analysis than mine! Two further points I wanted to make are about heatsinking and rebuilding the board.

Heatsinking for the series pass devices is woefully inadequate in the SX-1990, with the +80V device melting itself off the board and taking other parts with it. There was little space to increase the heatsink size and I ended up putting in the smallest possible fan running at reduce volts/speed - the noise was hardly noticeable

My replacement for the original 2SC945A was a 2SC2240 (ft 100MHz) which destroyed the board again due to self-osc. In order not to damage anything further I veroboarded the same circuit while I tried replacement components and optimised values to suit the new devices. It seemd a safe approach and I now have a handy little 70-90V prototype bench PSU!

[G0HZU_JMR]

I will reinsert it anyway and see what happens, using the bench 60-0-60 supply with current limit on, fed into C2/C3 I think

I think that your 60-0-60 PSU needs to have a reasonably slow risetime in order to prevent stressing the Pioneer PSU components as they start up.

Whatever you do, don't hard switch the 60-0-60 PSU direct to the Pioneer circuit with external toggle switches. This will have an instant risetime and this could cause a lot of stress. At a guess, I think a 60-0-60 PSU risetime of >10milliseconds would be quite safe.

[Mooly]

Originally Posted by yestertech

My thanks to everyone for the valuable input. I'm not sure I quite fathom the 'FET testing short' idea.
If I put one of the subs on the Peak Atlas tester it tests fine ( although the readout is unable to correctly identify D and S ) . If I do the same with the 2SK, it shows 'short red to blue' where one of them is the gate...............

Most small signal J-FET's are totally symmetrical and so the notion of which is Drain and which is Source has little meaning and the device operates the same which ever way round it connected.

If you take a J-FET and connect a meter on its ohm range across D and S then you could get a reading from anywhere from a few ohms to open circuit depending on the charge on the gate pin. Keep touching just the gate pin and you'll alter the charge and should see the reading D to S change as the D-S channel resistance alters. Testing this way is a little meter dpendent (test voltages used). An AVO on the high ohms range (with its 15 volt battery) is pretty good for this.

[Julesomega]

Sounds like a MOSFET, not JFET?

[kalee20]

Yes, MOSFETs and JFETs are quite different in circuit behaviour, and internal operation.

As Radio Wrangler says, JFETs are quite robust devices, no more susceptible to electrostatic damage than bipolar transistors. It's just a shame that production techniques haven't reduced the spread in key characteristics for a given type number!