100W Ultimate Fidelity Amplifier

Founder of XSA-Labs
Joined 2012
Paid Member
That cap sees <1V

Ok - my mistake then as it was the only one rated 63v on the last board Inkade before transistor pins got flipped. Btw, that is real good because I was thinking the same thing. You can mount this on a 50mm square CPU heat sink with small fan and add a 7812 on board to power the fan. Could be a 2in "cube" amp.

Speaking of caps - is there a reason we need a 10uF electrolytic input cap? I would rather not have a 'lytic in the signal path. I swapped it out for a 1uF film cap and two 1uF ceramic caps in parallel and listening to it now. Will it suffer lack of bass extension with just 3uF on the input? Sounds nice to me so far. Overall sound quality seems more accurate and textures are clearer. Forgot to mention I also added a 47pF ceramic disc bypass across the feedback resistor per earlier suggestion to get rid of overshoot on square wave. I wonder if that is also changing perception of clear sound?
 
Last edited:
layout of my fx8

I don't like how the feedback is referenced to that ground point (along high current draw of CCS from input ground). The red path of the current source contains a quite high current.

In the past I found I prefer taking the current from positive supply instead of ground. I have a FX8 PCB where I installed resistors in 3 positions (input ground, PCB star ground, positive supply) ready for switching connection, to see which one is a better point to source current for the LTP-ccs (But I haven't gone that far as I have some other parallel projects)
 

Attachments

  • groundccsfb.png
    groundccsfb.png
    370.1 KB · Views: 1,101
I don't like how the feedback is referenced to that ground point (along high current draw of CCS from input ground). The red path of the current source contains a quite high current.

In the past I found I prefer taking the current from positive supply instead of ground. I have a FX8 PCB where I installed resistors in 3 positions (input ground, PCB star ground, positive supply) ready for switching connection, to see which one is a better point to source current for the LTP-ccs (But I haven't gone that far as I have some other parallel projects)

CCS ground can be reconnected.
 

Attachments

  • groundccsfb.jpg
    groundccsfb.jpg
    124 KB · Views: 1,017
That 22k (or 20k) resistor does indeed see a lot of current. It feels fairly warm almost hot. When I had the bias improperly set by accident on startup of the second unit due to bad pot position - it got smoked and had to be replaced. Should it be 1/2watt?

If you're talking about resistor that biases constant current source transistor, it is dissipating less than 0.1W in the worst case with +/-45V rails.
 
Speaking of caps - is there a reason we need a 10uF electrolytic input cap? I would rather not have a 'lytic in the signal path. I swapped it out for a 1uF film cap and two 1uF ceramic caps in parallel and listening to it now. Will it suffer lack of bass extension with just 3uF on the input? Sounds nice to me so far. Overall sound quality seems more accurate and textures are clearer. Forgot to mention I also added a 47pF ceramic disc bypass across the feedback resistor per earlier suggestion to get rid of overshoot on square wave. I wonder if that is also changing perception of clear sound?

1uF ceramic caps that you used for AC coupling are probably made using X7R of Y5V dielectric, which is not temperature stable. This results in considerable fluctuation of capacitance and has impact on the cut off frequency and associated phase shift. If you really insist on not using electrolytic capacitor for AC coupling, use physically smallest 4.7uF film capacitor that you can find.

47pF capacitor across feedback resistor is way too much. For this amplifier only a very small capacitor in the range of 2.2pF-4.7pF can be used for lead compensation.
 
Hi xrk! Did you measure the VAS current in FX8 (on 47 Ohms)? Simulated value is 3,6mA only, which (theoretically) is too small current for driving Laterals' input properly, i.e. less able to charge-discharge the input capacitance of FETs. Nonetheless can not be detected any slowing of squarewaves. In VSSA I measured 13,8mA in VAS.

Meanwhile I simulated many of similar circuit (Sansui AU-317, 417, 517, Pioneer A30) and all of them have a small overshot on the top of their squarewave beyond 2 kHz. It seems impossible to remove these spikes. Another interesting observation is that these circuit don't show rounded squarewaves in higher frequencies rather trapezoids with very sharp edges. Very interesting... Sansui Au-317 an up have nearly perfectly flat squarewave at 5Hz! :eek:

All in all this is a very mysterious topology. I can't wait to arrive the ordered parts.
 
Last edited:
Hi xrk! Did you measure the VAS current in FX8 (on 47 Ohms)? Simulated value is 3,6mA only, which (theoretically) is too small current for driving Laterals' input properly, i.e. less able to charge-discharge the input capacitance of FETs. Nonetheless can not be detected any slowing of squarewaves. In VSSA I measured 13,8mA in VAS.

Meanwhile I simulated many of similar circuit (Sansui AU-317, 417, 517, Pioneer A30) and all of them have a small overshot on the top of their squarewave beyond 2 kHz. It seems impossible to remove these spikes. Another interesting observation is that these circuit don't show rounded squarewaves in higher frequencies rather trapezoids with very sharp edges. Very interesting... Sansui Au-317 an up have nearly perfectly flat squarewave at 5Hz! :eek:

All in all this is a very mysterious topology. I can't wait to arrive the ordered parts.

It is depend on the compensation. My simulation did not show an overshoot on 100kHz square wave. Very high slew rate :cool:
 

Attachments

  • FX8 mod.asc
    10.7 KB · Views: 374
Speaking of caps - is there a reason we need a 10uF electrolytic input cap? I would rather not have a 'lytic in the signal path. I swapped it out for a 1uF film cap and two 1uF ceramic caps in parallel and listening to it now. Will it suffer lack of bass extension with just 3uF on the input? Sounds nice to me so far. Overall sound quality seems more accurate and textures are clearer.

LF cutoff can be calculated (to find the minimum capacitance), but I found that bigger capacitance ("theoretically" bigger than necessary) brings better bass or short term dynamics.

My standard is 4.7uF minimum. Polarized cap is very bad, but with bipolar (BP) I found that capacitance is not an issue. I mean, even 47uF is fine when it is a BP. And BP has less footprint than MKP (if interference is an issue). MKP might be "clearer" but also "harsh" compared to cheap BP. Good BP (such as the Black Gate N or something from Nichicon) is my preference.

I also added a 47pF ceramic disc bypass across the feedback resistor per earlier suggestion to get rid of overshoot on square wave. I wonder if that is also changing perception of clear sound?

As mrmax said, that's too high. It should be done with oscilloscope or simulator, to remove UHF bump. You risk lowering the HF extension, which might result in "clearer sound" when the HF is exaggerated or contains HF garbage (or the tweeter is incapable of producing HF cleanly).
 
44V/22k=2mA, 44Vx2mA=0,088W...

Yes, I think it is not the issue with power (may be noise?), but I haven't done my HW checking the issue, but I doubt it will make any audible difference so when it is already done... But I will always use 1/2W here before I have the chance to conduct the experiment...

My FB series resistor is also 1/2W. I don't know if 1/4W will make audible difference or not. Just a piece of mind knowing that the Johnson noise is lower, beside, there is long distance from output to input transistor.