Cubie - small F5 variant with GR grade JFETs and LatFETs

In need for a nice 10-Watt amp I went through my inventory and found enough spare parts to make this little F5 variant.
Transformer is a 200VA, 4 x 12V AC (was saving it for a small circlotron but too lazy to obtain enough same-sex output parts).
Input JFETs are k170/j74 GR grade (Idss about 3mA) leftovers that come very handy here because with Rs=10R they run at Id of about 2.5mA which is high enough to drive properly Lateral MOSFETs and low enough to provide enough OLG (Rd is about 650R for LatFET's Vgs of 1.6V i.e. Id of 0.7A).
Laterals are k1058/j162, used without Source resistors so using the square law benefit to provide more class A headroom which is really noticeable when driving the low Z speakers.

The Cubie (kids are merciless) was extremely fast so I had to tame it a bit (that's why C19 was installed). The oscillograms show 24Vpeak-to-peak output (the clipping starts at about 26Vpeak-t0-peak) at 90kHz with 3R3 and 3R3||1uF loads and the amp's behavior is exemplary.
Thermal behavior is also very nice - heatsink size is 23 x 18 x 4 cm and after two hours they reach a temperature of 18 degrees Cels. above ambient (MOSFETs are 3 degrees hotter). DC offset drifts about 15mV from cold to hot.

It's worth mentioning: output MOSFETs' bias (Id) will depend on Idss of input JFETs and theirs Drain resistors (R16+R18 i.e. R17+R19) so wide range of values can be used but it's essential to measure JFETs Idss first.

The PCB doesn't show C19 because I added it later (prototype was tamer then the PCB version). I installed C20 on the RCA connectors - it's optional, I use it only because some of my signal sources leak DC. C20 is a combo of 1uF Philips MKC and 100nF WIMA MKP 10 which sounds very neutral to me.

Schottky rectifiers (MBR745) are used in PSU to minimize loses i.e. to avoid diode heatsink - too much work, there's 16 of them... Ripple voltage under the load is about 20mV_rms and the amp is dead silent on my 93dB/W/M speakers.

Finally, the sound surpassed my expectations - on my 4 Ohm speakers (Canton CT-1000 and Epos M5) it's not worse than my "end of the road" amp (F5 variant with k2013/j313).
_____________________________________
Edit: there is a text error on the PCB drawing - the text states dimensions 60 x 25 mm but that's a typo - real dimensions are 65 x 25 mm
 

Attachments

  • Cubie_PSU-AMP-PCB.gif
    Cubie_PSU-AMP-PCB.gif
    81.2 KB · Views: 5,949
  • LookInside1.jpg
    LookInside1.jpg
    207.2 KB · Views: 5,861
  • LookInside2.jpg
    LookInside2.jpg
    256.1 KB · Views: 5,282
  • LookInside3.jpg
    LookInside3.jpg
    169.2 KB · Views: 5,087
  • 90kHz@3R3.jpg
    90kHz@3R3.jpg
    58.4 KB · Views: 4,904
  • 90kHz@3R3+1uF.jpg
    90kHz@3R3+1uF.jpg
    60.9 KB · Views: 1,163
  • Front.jpg
    Front.jpg
    95.3 KB · Views: 1,168
  • Back.jpg
    Back.jpg
    123.2 KB · Views: 1,078
  • PCB.jpg
    PCB.jpg
    66.8 KB · Views: 1,372
  • PSU.jpg
    PSU.jpg
    98.9 KB · Views: 1,681
Last edited:
I like the description, '"velvety" quality. I hope the 'bit more pronounced' will be o.k. with my Fostex168Sigmas?!

Right now I am driving both the speakers connected in series and a single channel ACA. They sound more real that any amp I have had so far, with any speaker, for that matter.

The authoritative part for the Cubie is inviting, but I guess, more than 10 watts would allow for better dynamics.
 
Hi why no power supply decoupling at the supply lines ? IMHO that would be good design practice.

It usually is.
If you look at the schematic you'll see that every channel of the amp has independent PSU so inter-channel coupling through supply lines is practically nonexistant.
Also, schematic shows 1uF film caps (C4, C9...) which are stack-foil type (almost non-inductive) which are soldered beneath the PSU board. Putting them on the amp's PCB introduces the risk of creating a resonant circuit (together with inductance of wires that connect PSU to the amp) because this amp is extremely fast.
So, good design practices are implemented here...
 
Were you using reasonable gate resistors?

I talked to them - they seem to be very reasonable... :D
The schematic is in the post #1 - value of gate resistors in the output stage doesn't make as much difference as at the gates of the input stage JFET's.
So, as Mr. Pass suggested, using the value higher than 1k (3k3-4k7) at the input would eliminate the need to use C19.
 
It usually is. ...............
Also, schematic shows 1uF film caps (C4, C9...) which are stack-foil type (almost non-inductive) which are soldered beneath the PSU board. Putting them on the amp's PCB introduces the risk of creating a resonant circuit (together with inductance of wires that connect PSU to the amp) because this amp is extremely fast.
So, good design practices are implemented here...

Yes, that's the way things are usually done, but in F5-like topologies it can cause instabilities - I experienced it myself...
these two posts seem to be indicating ringing on the supply lines.
That is a reason to sort the ringing, not for omitting the decoupling. That means moving the decoupling to where it has a chance of working. Maybe even adding a bit of damping resistance.
 
Last edited:
these two posts seem to be indicating ringing on the supply lines.
That is a reason to sort the ringing, not for omitting the decoupling. That means moving the decoupling to where it has a chance of working. Maybe even adding a bit of damping resistance.


Buy some tools and parts, spend some time in the workshop and come again when you actually build and test some circuit...
 
Last edited: