LuFo Lite - a 1 Transistor SE Class A Headphone Amp

Founder of XSA-Labs
Joined 2012
Paid Member
This is an outshoot from the LuFo amp described in the Pass Forum here. Someone asked for something really simple and so I removed the upper cascode and dropped the voltage and bias current. What happens when you drop a transistor from a 2 transistor amp? You get perhaps the simplest amp in the world. Paired with a choke load (and I am using inexpensive microwave oven transformers) you can get some amazing sound. The PSU requires low noise and slow ramp up in order to prevent the SMPS from going into hiccup mode. So you do need a capacitance multiplier unless you are using a car battery as your power source.

Here is the circuit:
967135d1626325893-lufo-amp-39w-se-class-28v-rail-lufo-lite-schematic-15v-32ohms-hpa-jpg


The predicted performance for 15mW into 32ohm headphones is 0.0085% THD and with a superb SE Class A harmonic profile.

The amp was very simple and can be dead bugged on the 3 pins of LU1014D TO247 IMS aluminum PCB adapter board (Gerbers and PCBs available here).

Here is the LU1014D mounted on the IMS adapter board which makes it easy to use on a normal heatsink:
957971d1623053194-assembling-lu1014d-ims-247-adapter-ims-to247-assembly-closeup-03-jpg


In practice, there is something off with the model of the LU1014D in LTSpice and I needed to use a 2.5ohm resistor below the MOT to get the bias current low enough (1.5A). The circuit has been verified by myself and Jwjarch so far. The sound IMO, is very dynamic, natural, organic and transparent. Like nothing there but the muscle. It is very quiet - cannot tell it’s on. If you use the Cap Mx with a slow 20 second ramp up, there is no turn on pop sound either.

Proof of concept P2P build:
967129d1626324352-lufo-amp-39w-se-class-28v-rail-6edcb877-4b4c-48ca-9cae-02a7847ce580-jpeg


One channel up and running:
967130d1626324353-lufo-amp-39w-se-class-28v-rail-b34efa2c-e985-4658-bf33-963695f679ed-jpeg


And btw, it will power an 8ohm speaker (as shown in photo) to 1W with respectable distortion.

I was able to use the output of my DAP source without a preamp and got the headphones as loud as I would care to ever listen to. But if you need more output, any decent preamp should work.
If you want to have higher output voltage swing - just build the LuFo but with a single cascode MOSFET and keep bias current circa 1.5A. That will let you get up to 25Vpp swing from a 15v PSU.

For reference, here is what a full size LuFo looks like with a linear cap Mx PSU and a big enough heatsink to make 40w into 8ohms with a 28v rail. Still using a single LU1014D JFET to make the music.
961586d1624332091-bench-tonight-obt-7a850f46-9d48-493b-8521-32a5d2da0194-jpeg


Measurements on full size LuFo show 0.01% THD for 2.83Vrms into 10ohms with a superb harmonic profile:
954528d1621898005-lufo-amp-39w-se-class-28v-rail-lufo-2-83vrms-10ohms-fft-jpg


A great preamp for the full size LuFo is the OPA454 that I recently made:
962412d1624593694-opa454-front-lufo-amp-b543d4c7-ca91-470c-b157-e4a4c1d9598b-jpeg


Built in +/-35v PSU on the sub mezzanine:
962208d1624516193-bench-tonight-obt-ee3dfdb2-632b-409f-bf99-333f429a64c5-jpeg


962413d1624593696-opa454-front-lufo-amp-d1fe8f7e-8858-42ce-b119-2c4312aef82d-jpeg


Of course, for HPA duties, only +/-12v would be required for an opamp stage.

For your reference, I used this cap Mx design by Juma. Use a 1000uF cap instead of the 220uF to get a very slow ramp up to prevent SMPS hiccup mode from happening. I also used a TO247 N channel MOSFET like IRP240 to handle more current easily.
36990363054_f4dbe6e8e2_z.jpg


Edit Oct 28, 2021: the manufacturing data and Gerber files are located here.

The LuFoLite was demonstrated to work well with a 12v 20w halogen lightbulb as the resistor R3 in the schematic above and cam be used with a 24v SMPS for 1.63A bias current and 20v rail after the cap multiplier.
993307d1635383484-lufo-lite-1-transistor-se-class-headphone-amp-d6dbff60-3418-4619-9780-e4b7b3d7eb58-jpeg
 
Last edited:
X,

I was just commenting to a colleague of mine that you are a never ending machine; you never sleep, even though you say you do and you are probably healthier than most of the diyaudio members!

Thanks for the offshoot LuFo Lite! I have a pair of Lundahl 2773/3.5A on order, we'll see how it works out in a regular LuFo. I'm planning on wiring the coils in series for maximum inductance, and DCR should be 0.4 ohms or so.

On my MOFO, I have the Lundahl 2733 [running 2.5A bias, and both coils wired in parallel], at 10W RMS using Bode Plot analysis, the amp is flat (0db)at 10Hz. At 14W RMS (clipping on my design since my 'pre' stage has a max output of 31Vpp), it is -3dB at 20Hz. I'm sure there is some core saturation of the choke at that point but it's good to know the little Lundahl choke can handle it.

Great experiments as always...

Best,
Anand.
 
Last edited:
Founder of XSA-Labs
Joined 2012
Paid Member
Thank you for the kind words, Anand. For the record, I got 7 hours of sleep last night! I do run 5-6 miles a day though, everyday, rain or shine. I pay my sleep dues now, you already paid yours years ago in your precession when we were both young men. :)

Thanks for sharing with us that the Lundhals work well. To get 10Hz performance is very respectable, so it seems they are less prone to saturating than the MOTs. I suspect for headphone duties we won’t have any issue with saturation of the MOTs.

Btw, I just got word that samples of my custom EI balanced dual primary choke has been built and will ship out tomorrow. That will be interesting to have the balanced drive LuFo with 100w into 8ohms from a 28v rail.
 
How good is your spice model of LU1014D ?
I have this:
Code:
.model LU1014D NJF(Beta=28 Rs=4.6m Rd=4.6m Betatce=-.5 Lambda=.9 Vto=-1.15 
+ Vtotc=-2.5m Cgd=363p M=.3805 Pb=.4746 Fc=.5 Cgs=784p Isr=84.77p Nr=2 
+ Is=8.477p N=1 Xti=3 Alpha=10u Vk=100 Kf=111.3E-18 Af=1)
And you have this:
Code:
.model LU1014D NJF(VTO=-1.8 BETA=29000M LAMBDA=0.18 RD=0.0065 CGD=363P CGS=784P VTOTC=-2.6M)
Which one is closer to reality?
 
Founder of XSA-Labs
Joined 2012
Paid Member
I finally assembled the LuFo Lite in stereo. I settled on 1.5ohm 2% 10w Dayton speaker crossover resistor below the MOT. This gave 1.3A bias current and using matched JFETs from the GB the bias currents are almost perfectly matched as you can see the voltage readings across the 1.5ohm resistors. It’s powered by a 12v SMPS from an SMSL Class D amp that I had lying around. After the cap Mx, the voltage at the drain of the LU1014D is about 8.0v. I think this is all that’s needed for most headphones as the output swing is about 85% of 2x Vsupply or about 13.6Vpp for this case. That’s good enough for about 700mW into 32ohms.

attachment.php


attachment.php


attachment.php


I connected it to my Khadas Tone Board DAC and playing Amazon Ultra HD tracks - I am immediately struck by how good it sounds. Very clear and transparent. Very dynamic and powerful. Being a zero feedback amp, it images very well. It is also absolutely silent. Cannot tell it is on if music is not playing.

First track was Lithium by Nirvana - nice! Then Back in Black by AC/DC - Super!
The next up was Us and Them by Pink Floyd - amazing.

I can immediately tell that this amp is special. Perhaps one of the best headphone amps I have ever heard, no joke. And I have heard a lot of headphone amps.
 

Attachments

  • EA33C8CC-E28A-48DB-9C8A-FE27D2A9E1CA.jpeg
    EA33C8CC-E28A-48DB-9C8A-FE27D2A9E1CA.jpeg
    556.1 KB · Views: 3,955
  • ABFFC6A7-F9AD-4B0C-8BDD-0A61E4E27B40.jpeg
    ABFFC6A7-F9AD-4B0C-8BDD-0A61E4E27B40.jpeg
    492 KB · Views: 3,691
  • 74BFEE5C-937C-4651-BF83-6584DA36A363.jpeg
    74BFEE5C-937C-4651-BF83-6584DA36A363.jpeg
    546.1 KB · Views: 2,797
Founder of XSA-Labs
Joined 2012
Paid Member
How good is your spice model of LU1014D ?
I have this:
Code:
.model LU1014D NJF(Beta=28 Rs=4.6m Rd=4.6m Betatce=-.5 Lambda=.9 Vto=-1.15 
+ Vtotc=-2.5m Cgd=363p M=.3805 Pb=.4746 Fc=.5 Cgs=784p Isr=84.77p Nr=2 
+ Is=8.477p N=1 Xti=3 Alpha=10u Vk=100 Kf=111.3E-18 Af=1)
And you have this:
Code:
.model LU1014D NJF(VTO=-1.8 BETA=29000M LAMBDA=0.18 RD=0.0065 CGD=363P CGS=784P VTOTC=-2.6M)
Which one is closer to reality?

Hi Lineup,
Your model seems to give lower THD by about 2x less, and the bias current decreased from 1.7A to 1.1A. For the 8ohm case, I got 0.02% THD vs 0.04% from my model. I think previous measurements of THD with the model I orginally had was pretty accurate. Your model may be undersestimating the THD. Thanks for the model though.
 
I run your model now.
Possibly it gives closer to reality values.

May I ask why you run so much current, 1.72A ?
It is possible to increase the resistor values and get lower current and lower THD.
See image.
 

Attachments

  • LU1014D-xrk971-3.png
    LU1014D-xrk971-3.png
    47.1 KB · Views: 464
Founder of XSA-Labs
Joined 2012
Paid Member
Hi Lineup,
I am running 1.3A in reality and it’s high in order to be able to use as a fleawatt amp for 8ohm speakers. Several people expressed interest in an application as such. For headphone applications there is a lot less power required unless we are talking about the handful of cases such as the HE6 which requires up to 3w at 32ohms. But that will be voltage limited and not current limited.

Thanks for running the sim at lower bias current.