advertisement


Modifying NCC200 boards

last time one failed it destroyed the outlet stage literally ,this time just a pair of rectifier but it blew the tracs around them so had to replace whole board.
I bet that the latter psu failure was a direct consequence of the previous output stage failure .
 
I think it will only make a difference at high current.
I have other reasons for doing this in the NCC300.
Actually @S-Man I've just re-read some of the above and realise that the switching noise is there and of the same magnitude with just the drivers fitted driving a 'scaled' load of 100R (12x 8R) via 'scaled' emitter resistors of 2R5 (12x 0R22)... so sort of confirms it is the output transistors and is proportional to the drive and the load. I'm thinking of changing the 220R//C to something like 1k in series with 22n across the bases of the output transistors (for a time constant of 20µs, which is as near a dammit the time of the rising edge) as a snubber whilst keeping the 100Rs in place.
 
They were different amps the only common denominator is the Kendiel caps.

I've had 63WV Kendeils "puff up" when running cool in an amp with 58V rails after only a few years. I replaced them before something unpleasant happened, but I am now wary of them.
 
Actually @S-Man I've just re-read some of the above and realise that the switching noise is there and of the same magnitude with just the drivers fitted driving a 'scaled' load of 100R (12x 8R) via 'scaled' emitter resistors of 2R5 (12x 0R22)... so sort of confirms it is the output transistors and is proportional to the drive and the load. I'm thinking of changing the 220R//C to something like 1k in series with 22n across the bases of the output transistors (for a time constant of 20µs, which is as near a dammit the time of the rising edge) as a snubber whilst keeping the 100Rs in place.


Good idea but I'm not sure it will work. Charge suckout needs a fairly low impedance to "suck" the charge out of the transistor that's being turned off, there's only a modest voltage to drive this i.e. the PD between the output and the base of the other o/p device.
Along your line of thinking... I wonder if a charge suckout cap (B-B of output devices) would help, even with the driver emitters connected to the output?
 
I've had 63WV Kendeils "puff up" when running cool in an amp with 58V rails after only a few years
Interesting. My amp runs 37 to 38V per rail. Using 3x4,700/40V BC per rail, per amp which have been in there >10 yrs.
I wonder if a charge suckout cap (B-B of output devices) would help, even with the driver emitters connected to the output
I'll have a play at the weekend, although 220R//470n (instead of pair of 100R) didn't have much effect across output transistor bases, maybe slightly less high harmonics as the tail after the spike was a bit steeper. All it seemed to do was increase the total current use by 5-10mA.
 
Tried a couple of things to reduce switching spikes, but they didn't work.
Anyway, I wanted to try to run it without an input capacitor, but despite the output offset being about 2mV with input capacitor fitted, without it the output offset was about -80mV. So I experimented with tr1 / tr2 Vbe and found that using tr2 with hfe about 25% higher than tr1 gave about -60mV output offset which reduced to +12mV with input (and input cap) shorted, which is about what it should be with a pre amp connected as the pre will provide the tiny base current. I still wanted to improve the distortion / switching noise and found completely by accident that reducing the current source on the front end LTP from 2.7mA (220R) to 1.8mA (330R) reduced the switching spkies by about a further 6dB:
20220704-204536.jpg

Who'd have thought that the increased LTP current would cause increased switching spikes...? This also moved the output offset to -50mV with cap and -5mV with input (and input cap) shorted.
I now need to remove the board I installed in my amp at the weekend to modify it as above. :)
 
I hadn't noticed that you had upped the current in the LTP.
This will cause more output offset, due to any mismatch in the correspondingly higher base currents in TR1 & 2.

The switching spike thing seems odd. All your mods have increased the loop gain of the amp, I wonder if the stability has been compromised?
 
Yes, upped the LTP current in the current source coz that's wot it sez in the 'blameless' documentation for lowering distortion in the LTP, but as you say this causes a concomitant increase in base current... Anyway reduced it back, but not as far as the original NCC circuit that has 620R, just under 1mA.
Square waves is the next thing to try, hopefully tonight. :) Aiming for similar to my post 28 up-thread.
Edit: 1kHz and 10kHz square waves look exactly the same as before, same settings. Sine waves completely flat 20Hz to 20kHz, -1dB at 3.5Hz and 80kHz 1W into 7R5.
 
Last edited:
Might have to play with a buffer on the output of my oscillator as it could be that the higher current into the base of tr1 (with higher standing current) is interacting with the voltage divider/filter on the output of the oscillator in a different way to when running tr1 with a lower base current....?
 
The amp board with different front end standing current appears as though it was loading the relatively high filter output impedance I had on the oscillator. To remedy this, I tried several things including an opamp follower between the osc and the amp but was getting too much noise. In the end I removed the filter and rigged up the pre and this gave best results. Changing the standing current from 1.8mA (330R on the current source) to 2.7mA (220R on the current source) makes no difference to the residual distortion. This also remains pretty constant, UNTIL you exceed 8W output. At that point (13W) you get a MASSIVE switching spike form and the distortion residual goes from about -86dB amplitude to about -54dB amplitude...:eek::( Edit: just realised my mistake. This is due to over driving the input to filter. I need to repeat with the voltage divided down. Might have to try adjusting the bias, although this seems to have an optimum setting (set at 2W), below or above which the distortion residual increases.
Any other suggestions welcome.
 
Last edited:
Interesting reading. As an aside, the best caps I've found so far for DIY are the Samwha HC series (as used by Arcam in thier hi-end gear) - well worth checking out; major upgrade in imaging and clarity. - Forgot to clarify - these are reservoir caps - go for the 63-80V versions.
 
Not surprised by that - Samwha capacitors are a Nichicon operation, but made in S. Korea. Excellent value.

Yes, I've happily used them too.
 
All installed and adjusted for similar distortion residual, pretty much like that shown above. I am however picking up rectifier switching spikes, left channel similar to residual distortion levels and right channel about 3x worse so need to investigate with some tin screening... my layout puts the front-end input LTP of the RH channel about 1/2 the distance of that of the LH channel to the PSUs.
Edit: Tin screen didn't work. It's due to earth loop area between the front end 0V and speaker return. Bodging the front end earth with a choc bloc and wire to make it longer and passing it across the board close to the FB network and winding into the loom reduces it markedly.
 
Last edited:
After much poking about and head scratching I eventually noticed the the 220pF filter cap between the base and neg supply rail of the VAS current source transistor (C7) in one channel was the opposite way round to that in the other channel. Putting the 'scope across each in turn revealed a notably larger spike across this cap on the worse channel. I'll switch it round tomorrow. :)
Edit: Finger prod test on these caps doesn't show any real difference one channel to other so not so sure now...
 
Last edited:
Finished it:
20220711-135631.jpg

Re-routed the front end 0V return on both across the circuit board which vastly improved the noisy board, but this only made the other board (which was pretty good) worse. Had to double this wire back and put it under the board for it to be quietest I could get. :)
 
Yes. About 2.5hrs this afternoon. Not good at the description thing, esp with my non-too-brilliant hearing. However, initial impressions are good. It sounds full and clean with a good top end and the image seems a bit wider/deeper with a bit better focus. Doesn't quite have that engaging 'big bouncy' sound of the NCC200, but it is very early days so I dont want to judge too much just yet. I suspect that that signature comes from the 1k and 22k combo on the collectors of tr1 and tr2 respectively, as many others have pointed out. I was running mine with 1k and 2k4 which was about right for me... so will think about modifying my original boards to something close to Qudos spec
 
Yea agree with you about the 1K,22K resisters, tried different combinations and now back to 1K,22K music just sounds more involving.
 


advertisement


Back
Top