Jo Sharp:
Careful with this modification. C3 is there for a reason. It is to balance the DC bias currents on the input pair of transistors (through the 27K=24K+3K). As these are not JFETs this will be a significant voltage.
Firstly it will add >>100mV of DC on the speaker outputs. This may not be a big issue (crackling when connecting speaker cables, not suitable for active speakers, switch on thump). However anything plugged into it which uses muting relays which open for muting will cause a Huge crack out of the speakers (amp gain x DC, so could be several volts). I guess the MDAC does this, from the conversations about output/headphones relays.
I'm surprised to see an audio designer with John's following on here recommending this. He appears too careful about his appearance here online to give such advice. Maybe power amps are not his area.
There are ways of designing this in, but don't go ripping out parts without knowing how the design works. Try different capacitors of varying type and quality by all means, or bypass them with polypropelene ones. But don't short out C3 unless you like unpleasent noises...
As I said each to there own – but I’m pretty blasé about details like this, I don’t care about a few little pops and thumps when changing inputs if the end result is better sound quality.
Being easily distracted while working on the BDAC’s PCB layout I’ve just spent 15 minutes making a spice model of the NCC series amps – interesting as I have recently been designing our MAMP so “Spice Model” comparisons can be informative.
For a nice change I had all the transistors models in the spice library and used “Real” device models – not “generic” spice models – so the models should be very close to reality.
The NCC topology is a basic “text book” amplifier circuit, the final SQ will be strongly determined by its construction and component choices rather then design – that’s where the mysteries and the skill of the Black Art come into play
DC Offset results with “perfect” matched input transistors:
NCC with coupling capacitor (so in standard form) -27.2mV
NCC with input coupling capacitor bypassed, with no Input device connected -19.4mV (Lower due to the extra 100K Pre-Cap now also paralleled to Ground).
NCC with input coupling capacitor bypassed, MDAC powered off (MDAC internally terminates to 10K) 1.9mV
NCC with input coupling capacitor bypassed, MDAC powered on (Modeled as an effective 5 ohms source impedance which should take into consideration MDAC output impedance, connector and cable resistances) 12.9mV
A significant portion of the NCC inherent DC offset is caused by the imbalance in Diff pair tail currents, 649uA (TR1) over 494uA (TR2), while the input bias current is 1.6uA - the resultant output offset levels are still small IMO and of no real concern.
So, in fact the DC offset (without taking into consideration MDAC’s offset which may still causing a slight Pop sound when un-muting) is LOWER when the NCC input cap is removed while connected to MDAC, with only an 11mV difference in output offset at the speaker terminals between muted and unmated states.
NCC standard (with Cap fitted) -27.2mV
Offset at speaker when MDAC powered off: 1.9mV (Cap Bypassed)
Offset at speakers with MDAC powered on 12.9mV (Cap Bypassed)
Pls. note that these results do not take into consideration the amplifiers “inherent” offset due to imperfect input device matching. Unbalanced input device will magnify the differences – it would have been nice to see a trimmer across the Emitters of TR1 / TR2 to null the inherent offset.
In all cases the input offset is low as not to be a concern, in my experience and now confirmed by simulation results the input bias current is low so as not to pose a concern – in fact DC offset levels are LOWER in the case of NCC with the MDAC as an input source impedance in both Muted and Unmuted states...
As I say each to there own...
John