Midgard, a low distortion Lateral FET output Headphone Amplifier

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Simple design but with good performance

1.4mA through each LTP transistor, 10 mA through the VAS, Output biased at 150mA.

Input and feedback has its own little ground island that goes back to star ground, rest of the board has a ground plane with the current sources as well as the supply caps and DC servo connected to this plane, it also acts as a shield.

Simulated THD20 with +/- 15VDC supplies :

THD20 at 1VRMS into a 32 Ohm load is 0.001169% rising to 0.011811% at 7VRMS

THD20 at 1VRMS into a 250 Ohm load is 0.000955% rising to 0.007606% at 7VRMS

THD20 at 1VRMS into a 600 Ohm load is 0.000929% rising to 0.007663% at 7VRMS

Still need to do a little tweaking and adding a DC servo to the layout.

Simple to build with no matching of transistors required and a relatively low component count.
 

Attachments

  • Midgard_LTSpice.png
    Midgard_LTSpice.png
    51.1 KB · Views: 890
  • Midgard_Layout.png
    Midgard_Layout.png
    111.9 KB · Views: 855
Last edited:
This is now my third headphone amplifier design.

What I have done with Midgard compared with Thor and Mjølner is a case of simplifying the build, with fewer parts and no matching of parts required.

Thor was my first project and it had great performance into 32-600 Ohm loads but the need for matched input JFET's and the multistage topology made it unessecarily expensive and complicated to build.

Mjølner was my second project, slightly worse performance than Thor and it was only really good for 150-600 Ohm loads, but I removed a few stages and it was therefore easier to build. However, it still used input JFET's that needed to be matched.

Midgard, now this could be the best of the bunch. It does not have the same 100% symmertry as Thor and Mjølner but it is much simpler to build with no matching of parts required. Gone is the JFET's in the input, a few stages have been removed but it retains(improves really) the same performance as Thor into 32-600 Ohm loads.
 
With a low gain of X2.5 simulation shows it capable to drive up to 5nF loads directly without stability issues. This SHOULD be good enough for majority of cables out there, however, if instabilty shows up there are 3 solutions :

1. Increase the Miller compensation cap. This gives a little higher THD as well as lower slewrate. Slewrate is directly proportional to size
of the Miller cap.

2. Increase gain. Many amps do not like it the closer to unity gain they get, increasing gain improves stabilty into capacitive loads.

3. Add a 2uH//2R filter in series with the output signal. This isolates the output from any capacitive loads seen on the other side of the 2uH//2R filter.

Personally, adding 2uH//2R at the output is my favourite since it does not affect THD or slewrate. This means you can have a low distortion, high slewing amp and still retain the ability to drive highly capacitive loads.

It might not be an issue at all, but until I actually build the amp I won't really know.
 
Last edited:
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.