my take on the szekeres

[geekysuavo]

alright, parts are starting to trickle in from mouser and digikey (the 2sk214's i ordered will probably be backordered through the new year, knowing my luck) and my custom footprints all work out, so i'm gonna test a few small parts of the circuit and then its off to the fabs for my board.

one small note: the sims show 5mV on the output of the amp, which is probably fine, but it means the first thing im measuring after the magic smoke test will be dc offset...

~ brad.

[AndrewT]

the simulator is using the models and calculates from them to find all the predictions.

Output offset is probably one of the most inaccurate predictions after distortion.

Expect the offset to be anywhere when you use real transistors and real tolerance resistors. Try not to unbalance any LTP to remove offset.

[geekysuavo]

just so nobody starts worrying the project was canned...

i've received all the components and the boards, and - even though i'm extremely anxious and ready to get back to some electronics - i'm waiting until after the holiday to solder it all up and test it out. i'll have updates here at that time, as well as at geekysuavo.org/buf.

~ brad.

[geekysuavo]

ok, i finally got the board wired up and of course, i have problems.

1.) the opamps are getting wayyy too hot. i fear oscillation. would increasing the capacitance at the opamp outputs help solve this instability problem?

2.) tons of mains hum. i dont have any wires crossed where hum could be getting injected. this hum sounds like a ground loop, nice and loud.

does anyone have any ideas as to why this circuit would hum like crazy and heat up the opamps?

thanks,
~ brad.

[geekysuavo]

furthermore, it seems like the whole darn unit is extremely sensitive to capacitive coupling to ground, specifically via my hand or head. i fear i'm missing some kind of grounding issue.

ideas?

thanks,
~ brad.

[geekysuavo]

ok, i have a few ideas for future steps to take:

first, i'm pretty convinced that a major problem with the current layout is the grounding scheme, specifically the fact that the ground rail extends all the way back to the start of the unregulated filter section. this has to be creating multiple non-equipotential points in the ground return, causing induced currents. i think i'll split the ground into two rails (one per supply) until absolutely necessary, and then join everything at a single point at the end of the regulated supplies.

again, please feel free to chime in with suggestions or criticisms.

~ brad.

[geekysuavo]

i've taken a few measurements of the amp, to clarify the issues i'm having.

a.) the scope is set to 1us/div and .1v/div here, showing the oscillation on the output (after the mosfet buffer, 20R load) with the inputs grounded.

b.) ripple voltage at the end of the unregulated supply showing a ~1vpp ripple at 120hz. you can clearly see the high-frequency oscillation riding on top of the ripple waveform.

c.) a closer view of the ripple with added hf oscillation.

d.) finally, the voltage at the output of the regulator V+ rail, showing 15.7vdc with essentially no noise other than the hf oscillation.

it seems i basically invited murphy into my home by paying little attention to layout and grounding in a design using a high-bandwidth, high slew rate opamp.

so i have a few ideas:

1.) swap the ad843 for an ad797, which has significantly better noise performance in the audio band and slightly better THD figures. the lower slew rate and bandwidth of the '797 would ease design requirements. does anyone have any opinions about differences in audio quality between these two chips?

2.) add a low-pass filter and high-pass filter to the inputs of the opamps to reduce the possibility of incoming signals pushing DC to the headphones or inducing hf oscillation. is this really necessary, considering the phase-lead capacitor in the feedback network?

3.) redesign the layout with a ground plane or star ground. it's easy to see that the circuit would greatly benefit from better grounding and layout, so i plan on experimenting first with this aspect of the circuit. soon, i'll finally have a nice dual-rail lab bench power supply to test bits and pieces of the design. i'll continue to post layouts with questions as i go.

~ brad.

[AndrewT]

phase lead cap in the NFB loop. Try reducing it, in stages, to about 1/16 of present value.

Do all three proposals 1.) to 3.)

[geekysuavo]

hi andrew, thanks for the comments.

Originally Posted by AndrewT

phase lead cap in the NFB loop. Try reducing it, in stages, to about 1/16 of present value.

decreasing the phase lead capacitor extends the bandwidth of the amplifier even more, right? the only reason i can understand doing that is in the context of also doing proposal 2.)... have i missed something?

also, caps that small are probably ceramic. is a C0G-type ceramic suited for duty inside the signal path?

Originally Posted by AndrewT

Do all three proposals 1.) to 3.)

thanks. a new one-channel-only-no-psu pcb layout is in the works to that end.

~ brad.

[geraldfryjr]

This a very interesting thread and I just found it.
A few years ago I started experimenting with such circuits and I had designed them for High Voltages for esl's and es headphone use by stacking the fet's.

And I had just recently figured out how to build a fet ccs to replace the load resistor aswell as stacking the fet's in the ccs for HV use.

Excelent! Keep up the good work guy's. jer