Cubie2

Taking of from here (post #485), I adopted this lovely gain stage as the part of a power amp - Cubie sounds great but Cubie2 manages to sound a bit better. :)

So Cubie2 is an extension of BAF 2013 preamp idea, with current mirror (Q3/Q5 and Q4/Q8) instead of resistor/CCS and introduction of negative feedback loop (R7/R16) which sets the gain at about 19dB. The LATFETs in the output stage (Q9/Q10) work in source-follower mode, without source resistors, providing some extended class A regime due to the square law transfer characteristic of MOSFETs.

The input JFET pair can be of any flavor (GR, BL or V), it's just important to dimension R11/R13 so to get about 10-12 mA flowing through the folded cascode branch (Q6/Q7/Rbias). Values on the sch. are based on JFETs' Id of about 6.5 mA. With such JFETs I got about 10mA through Rbias=330R which sets the LATFETs Vgs_Q10 + Vgs_Q9 = 3.3V and the output pair's Id at about 0.7A, but you can use the 1k pot instead of Rbias to set the output stage current to different value or to accommodate for LATFETs' differentiating Vgs/Id characteristic (my current batch of k1058s need 1.70-1.75V for Id of 0.7A and J162s need 1.55-1.60V for the same Id). P1 sets the 0V DC offset at the output, and it's very stable: +/-10mV from cold to hot and +/-2mV when hot. It's also very stable into reactive loads - testing it, I didn't succeed to make it misbehave.
R25 is there to load the folded cascode and determine the OLG (about 45dB) while the C3 shorts the gates of LATFETs for AC signal.
Full power before clipping is 34W_peak at 4R load with 1.3V_peak input signal and about 19W_peak at 8R load with 1.4V_peak at input.

The PCB dimensions are 70 x 25 mm, copper side view. The GND connection for decoupling caps (C5/C6) should be made with separate wires to main GND point, independent from signal GND on the PCB.

The PSU is same as here (post #62).

I tested the amp without the feedback loop (R7 and R25 go out, Rbias is split into 2 x 160R and from their middle point 1k connects to GND) and it sounds great too but with feedback loop it has a bit more definition and bass firmness. Those who enjoy "velvety" quality of BAF2013 preamp might like it more without the feedback loop.
 

Attachments

  • Cubie2JFET-SCH.GIF
    Cubie2JFET-SCH.GIF
    8.7 KB · Views: 7,353
  • Cubie2JFET-PCB_Layout.GIF
    Cubie2JFET-PCB_Layout.GIF
    17.7 KB · Views: 7,218
  • Cubie2-PCB.jpg
    Cubie2-PCB.jpg
    264.5 KB · Views: 6,367
  • Cubie2-modules.jpg
    Cubie2-modules.jpg
    285 KB · Views: 6,051
  • Cubie2-test1.jpg
    Cubie2-test1.jpg
    500 KB · Views: 5,876
  • Cubie2-boxed.jpg
    Cubie2-boxed.jpg
    388.1 KB · Views: 3,128
Nice as usual Juma, I use similar gain stage with a few twists in my amp for over 3 years and it sounds very good.
I think Ayre uses also similar gain stage for their amps and preamps but without feedback loop, just shunt volume control on the output for I/V conversion.
Anyway from my experiments this kind of gain stage sound better with very low local current feedback 5-6dB in my case, then without feedback loop, the difference I heard is in the Hi tones, they are more transparent and soft with low feedback loop.
C7 for compensation in my case was problematic, I`d use just low value shunt cap over R25, but I use different OPS.
Cheers, Borko .
 
Yes it'll fit in the same box - the PCB is about the same size (the new one is 5mm wider), the same output devices are used, PSU and current/voltage/dissipation requirements are also the same.

I drew the PCB so that placement of output devices is the same as in the first Cubie so it's easy to take the old PCB out and solder the new one in.

That LU1014 circlotron is something that's been cooking for a long time but somehow I never managed to bring it to the end (too many versions and ideas and too many other projects - this is not my job, just a hobby so the time it's allowed to consume is limited...)
 

fab

Member
Joined 2004
Paid Member
Any reason why it does not include the double feedback arrangement like the F5 or your variations of the F5.
Also, in your build you seem to have an input cap with few caps in parallel. Can you share your thoughts and/or experience on what type and combination and values are best.

Have you considered lowering vas degeneration resistors value. I have found that it can sometimes improve perceived dynamics.

Thanks
Fab
 
Last edited:
Any reason why it does not include the double feedback arrangement like the F5 or your variations of the F5.
F5-like circuit has a single input pair with fixed (let's neglect the P3 addition) degeneration that connects to ground so sources/emitters come naturally as symmetrical feedback points. It's not exactly the same case here where we follow the idea of symmetrical folded cascode gain cell (BAF2013 preamp) where sources' circuit is "loaded" to GND with a single, current-swing/gain-determining resistor that presents itself as a feedback point. Of course, it can be done in a F5-like manner but it would raise the complexity without a benefit and the DC offset would have to be dealt with in a less elegant manner.

Also, in your build you seem to have an input cap with few caps in parallel. Can you share your thoughts and/or experience on what type and combination and values are best.
Yes, it's a precaution measure since all kinds of gadgets are used as signal sources all the time and some of them leak DC. The combination of Philips MKC and WIMA MKP10 caps I find very neutral. Caps are no big deal by themselves, it's how they are used - take a stage with high output Z , couple it to a stage with low input Z, use the filter-like cable (I had a chance to test very expensive ICs with more than 1nf capacitance per meter) and you are in trouble - every coupling cap will leave a hearable signature.

Please don't turn this thread into futile, never ending discussion about caps, cables and other trivia...
 

fab

Member
Joined 2004
Paid Member
.....

Yes, it's a precaution measure since all kinds of gadgets are used as signal sources all the time and some of them leak DC. The combination of Philips MKC and WIMA MKP10 caps I find very neutral. Caps are no big deal by themselves, it's how they are used - take a stage with high output Z , couple it to a stage with low input Z, use the filter-like cable (I had a chance to test very expensive ICs with more than 1nf capacitance per meter) and you are in trouble - every coupling cap will leave a hearable signature.

Please don't turn this thread into futile, never ending discussion about caps, cables and other trivia...

It has never been my intention to hijack your thread.

Thanks
Fab
 
Last edited:
  • Like
Reactions: 1 user
Sure. Change R14 to 15k/0.5W to dislodge few mA from Q6,7 i.e. to keep their Pd at no more than 200mW. Of course, Rbias needs changing too since less current means lower voltage drop across it and that lowers the current through output stage, but that will depend on your Laterals' Vgs/Id characteristic anyway...