Slewmaster - CFA vs. VFA "Rumble"

Hope this is helpful those running VSSA like CFA IPS and chasing low digit distortion. Value of the bypass cap really change HD, say this because seen some schematics running low values. Attach example three cap values.
 

Attachments

  • BypassCap.png
    BypassCap.png
    48.2 KB · Views: 475
I change driver using 2SC4793 and 2SA1837. Driver current is ~36mA. Resistor gate is 22 Ohm. THD rise a bit but slew rate increase significantly. This is the result:
PM 79 GM 16
180W/8Ohm, 1kHz -> 0.003693%
180W/8Ohm, 20kHz -> 0.012127%
70W/8Ohm, 1kHz -> 0.000439%
70W/8Ohm, 20kHz -> 0.003299%
317W/4Ohm, 1kHz -> 0.000787%
317W/4Ohm, 20kHz -> 0.007792%
SR ~328

It looks like after using a 10-12 mA EF buffer, THD no longer depends on MOSFET input capacitance, all you can win with more powerful buffer is the slew rate.

By the way, comparing THD of one-MOSFET and three-MOSFET configurations, it is clear that three-MOSFET is better - looks like here we face the other MOSFET special feature - poorer output impedance, than the one of BJTs, so a number of MOSFETs in parallel can drive the high loads better.

I think, with dual EF buffer we can drive 5 pairs of IRFP240/9240 with no problem.

Cheers,
Valery
 
Hi Guys,

Dragging you back to the standard OPS. I am still having problems with one board. I just couldn't get it to stop oscillating so I pulled the MJL21193/4 outputs and replaced them with MJL4281/4302. One board still oscillating. So then I replaced the drivers, (MJW3281/1302) with NJW0281/0302. One board still oscillating. Then I put new MJE340/350 in the pre-drivers and Cap multiplier. Still oscillating so I put 2SC3423-Y in the two VBE multipliers. No change. I tried 33p and 47p in C107/109,(NPO). The 33p was slightly better. I tried 47p, 82p, 100p and 150p in csa/csb(NPO). 150p seemed the best. Both boards have identical parts yet one oscillates and the other doesn't. It is no longer runaway oscillation but the bias is not very stable and I can see fuzz on the scope when music is playing. My other OPS boards have never been a problem. The differences between the other boards and these are this. The old boards use MJL3281/1302 outputs, MJW3281/1302 drivers, MJE340/350 pre-drivers, MJE340/350 CM and KSC3503 VBE. C107/109 are 33p Mica and csa/csb are empty. The only other difference is that for C111/113 I used 56uf in the older boards and 100uf in these.

So now I have replaced every single device and tried several different caps. Any suggestions where to look next? The biggest mystery is why one of them oscillates and the other doesn't. One more clue. The board that oscillates doesn't oscillate unless there is a load on the output and a source plugged into the input.

Thanks, Terry
 

Attachments

  • MJL4381-4302.jpg
    MJL4381-4302.jpg
    464.7 KB · Views: 435
Last edited:
The board that oscillates doesn't oscillate unless there is a load on the output and a source plugged into the input.

Thanks, Terry

Hi Terry,

Given that a load is needed for the board to oscillate suggests that there may be a grounding and/or decoupling problem. Considerably higher currents will be flowing in the output stage when driving a load.

Have you checked the star ground wiring and is it identical to that of the working channel? Have you checked the PCB for damaged, open-circuit and short-circuit traces? Try grounding the main heatsink if it is floating. Are the transistors on all heatsinks fully insulated?
 
...
So now I have replaced every single device and tried several different caps. Any suggestions where to look next? The biggest mystery is why one of them oscillates and the other doesn't. One more clue. The board that oscillates doesn't oscillate unless there is a load on the output and a source plugged into the input.

Thanks, Terry

Can't see the zobel filter usually used between output devices and coil.
If really missing, try to add a 47 - 100nF + 10R zobel directly be4 coil.
 
Can't see the zobel filter usually used between output devices and coil.
If really missing, try to add a 47 - 100nF + 10R zobel directly be4 coil.

It doesn't look like it in the picture, but the Zobel is in fact electrically before the coil, and it is present.

Terry, have you checked to ensure the Zobel components are intact? If the 10R has failed that could open you up to issues.
 
Hi currentflow,

Those are good points. Now that you mention it, it does act like a grounding issue, though there is no hum. The boards have been removed from the heatsinks several times. Maybe it is time for some new kapton tape. I have checked for continuity between the heatsinks and output collectors each time. I will try some different grounding schemes and see if that alleviates it.

Thanks guys. I will double check the zobel.

Thanks, Terry
 
Last edited:
Hi Guys,

Dragging you back to the standard OPS. I am still having problems with one board. I just couldn't get it to stop oscillating so I pulled the MJL21193/4 outputs and replaced them with MJL4281/4302. One board still oscillating. So then I replaced the drivers, (MJW3281/1302) with NJW0281/0302. One board still oscillating. Then I put new MJE340/350 in the pre-drivers and Cap multiplier. Still oscillating so I put 2SC3423-Y in the two VBE multipliers. No change. I tried 33p and 47p in C107/109,(NPO). The 33p was slightly better. I tried 47p, 82p, 100p and 150p in csa/csb(NPO). 150p seemed the best. Both boards have identical parts yet one oscillates and the other doesn't. It is no longer runaway oscillation but the bias is not very stable and I can see fuzz on the scope when music is playing. My other OPS boards have never been a problem. The differences between the other boards and these are this. The old boards use MJL3281/1302 outputs, MJW3281/1302 drivers, MJE340/350 pre-drivers, MJE340/350 CM and KSC3503 VBE. C107/109 are 33p Mica and csa/csb are empty. The only other difference is that for C111/113 I used 56uf in the older boards and 100uf in these.

So now I have replaced every single device and tried several different caps. Any suggestions where to look next? The biggest mystery is why one of them oscillates and the other doesn't. One more clue. The board that oscillates doesn't oscillate unless there is a load on the output and a source plugged into the input.

Thanks, Terry
Do all your boards have the bias transistor hanging on wires? Are you possibly inducing a current into the bias circuit when you load the amp down?
 
Yes, they are sitting side by side, powered from the same PSU. Voltages are very close to the same. The troubled board pulls slightly more current through the CM when playing music. I assume that is due to the oscillation since the bias rises more than the good channel. It is 5am here so I haven't had a chance yet to try any of the suggestions yet. I will in a couple hours.

Thanks, Terry
 
Yes, they are sitting side by side, powered from the same PSU. Voltages are very close to the same. The troubled board pulls slightly more current through the CM when playing music. I assume that is due to the oscillation since the bias rises more than the good channel. It is 5am here so I haven't had a chance yet to try any of the suggestions yet. I will in a couple hours.

Thanks, Terry

Hi Terry, another suggestion - keep an oscilloscope connected to the problematic board's output at all times - this way you will see if there is an oscillation and what "character" it is (sometimes it's "full scale", sometimes it's just some "ringing", ets.). Knowing that may help understanding its nature...

Cheers,
Valery
 
Ok, let's go back to OPS stability with different output devices for a minute.
I did some simulations, actually confirming what's been said above.
Analyzed circuit is my tube-hybrid IPS together with SlewMonster OPS.

OPS creates additional pole on overall frequency response curve.

Picture 1 - with faster NJW3281/1302 transistors this pole is around 4.6 MHz and 5.5 db below the unity gain level. Acceptable - we cross unity gain with 20 db/decade slope.
Picture 2 - with slower MJL21193/21194 the pole is located at 3.2 MHz and actually lays on the edge - roughly at the unity gain. This is a problem - frequency response slope changes from 20 db/decade to 40 db/decade here and we don't want 40 db/decade to be located at the unity gain level.
Picture 3 - this is how we can solve it - increase the value of C107, C109 (pre-drivers shunts) from 33pF to, let say, 150pF. The pole stays at 3.2 MHz, but moves down to -6 db - and we are fine again.

Anyway, the faster devices are the better. SC2922/SA1216 Sankens would be a good choice here as well, with Ft = 50 MHz and Cob = 250 pF (half of any of the above).

Hope this helps.

Cheers,
Valery


Hi Valery

cob = 500pF and 40MHz FT for 2SA1216

http://www.onsemi.com/pub_link/Collateral/NJW0281-D.PDF : NPN/PNP Gain Matching within 10% 400pF 30Mhz
 
Last edited:
Yes, they are sitting side by side, powered from the same PSU. Voltages are very close to the same. The troubled board pulls slightly more current through the CM when playing music. I assume that is due to the oscillation since the bias rises more than the good channel. It is 5am here so I haven't had a chance yet to try any of the suggestions yet. I will in a couple hours.

Thanks, Terry

Swap the VAS PCB and observe if oscillations follows to another output PCB.