The discussion is not where to put zobels - this is well known.
There is distortion introduced AFTER coil. Maybe the coil is suboptimal dimensioned. If we take a look at latest simulation file you can see a zobel 0.µ/10R at amplifier output followed by 0.9µH coil with parallel 10R then 10 inch cable to the rear side of amplifier followed by an output relais and second zobel directly at output terminals. Relais seems to be not the reason as input and output distortion is the same...
My question to the community simply was
"Is it possible that at high frequency (> 10 kHz) and high output levels (~ 40V rms) the output coil is adding distortion?"
BR, Toni
Absolutely. However this is not the full picture. I am currently struggling with an LFET output stage that needs a very specific snubbing scheme around the outputs. What matters is the total loop inductance from collector to emitter. In other words, collector inductance is in series with RC inductance and both need to be minimized. If your collector decoupling is not close to the transistors, then minimizing RC length to ground won't help much. The RC return should go directly to the collector decoupling.
This is already fulfilled by the current pcb. The distortion rising starts directly after output coil.
I think the 10R resistor value is too high, try in the range from 2R and 5R.
You shouldnt underestimate the importance of not only where the zobel is located but also the direction it is placed with regards to the pcb, output transistors and even the transformer. As an example having it at 90 degrees with regards to transformer can be detrimental in some cases compaired to having it at 180 degrees. Using flat copper wire for the coil also helps minimize magnetic pickup but is hard to find.
I think Self has some writings about coil characteristics and effects.
... found one major cable routing problem.
I have erroneously routed the speaker output cable from pcb parallel to the power supply wires about 4 inches. The power bjt switching spikes (~ 1Vpp / 20kHz) from power rails have had a chance to crosstalk into speaker cable ...
At those very low distortion levels at those high output power levels the cable routing inside the case is extremely sensitive.
The speaker output cable is now routed under the mounting plate - see picture.
BTW coil and resistor are parallel but resistor is NOT inside coil.
Thank you all for your tips - every small step can help us DIYers to get the most out of our amplifier.
Now have to measure again and again and again ...
I have erroneously routed the speaker output cable from pcb parallel to the power supply wires about 4 inches. The power bjt switching spikes (~ 1Vpp / 20kHz) from power rails have had a chance to crosstalk into speaker cable ...
At those very low distortion levels at those high output power levels the cable routing inside the case is extremely sensitive.
The speaker output cable is now routed under the mounting plate - see picture.
BTW coil and resistor are parallel but resistor is NOT inside coil.
Thank you all for your tips - every small step can help us DIYers to get the most out of our amplifier.
Now have to measure again and again and again ...
Attachments
no !!!!
a single Flow or Return cable passing through a hole is not allowed. Does it make a current transformer? Or is it some other effect?
Both the FLOW and Return must pass through the same hole if it must through a hole at all.
But why have you got a single wire. It should be a twisted pair !!!!!!
That pair is both resistant to picking up interference and is good at not transmitting interference.
Always close couple Flow and Return Pairs. Always. That is what makes Planes good. The close coupling of the Flow and Return currents. In wire connections twisting achieve similar.
Rod Elliot has written briefly about higher distortion after the output coil. Check the article about a classic 60 Watt amplifier on his site.
As someone has already pointed out, 10R is too high. IIRC Self advocated 1R.
You can always have one Zobel on the amp board, Coil with parallel snubber R off the amp board, close to the output terminals, followed with a second Zobel right on the output terminals. I am not able to recall the author, month/year etc., but there was an article in Electronics World about how to correctly dimension the second Zobel.
As pointed out by Andrew, the signal and return cables should run together in some geometrical arrangement which rejects Common Mode artifacts and preferably low inductance type. If you are familiar with weaving CAT5 wire as Speaker cable, you could try it. I am sure you will have much better results.
As someone has already pointed out, 10R is too high. IIRC Self advocated 1R.
You can always have one Zobel on the amp board, Coil with parallel snubber R off the amp board, close to the output terminals, followed with a second Zobel right on the output terminals. I am not able to recall the author, month/year etc., but there was an article in Electronics World about how to correctly dimension the second Zobel.
As pointed out by Andrew, the signal and return cables should run together in some geometrical arrangement which rejects Common Mode artifacts and preferably low inductance type. If you are familiar with weaving CAT5 wire as Speaker cable, you could try it. I am sure you will have much better results.
Hear! Hear!
Get your coil and its parallel resistor off the output PCB and as far as possible away from things like power-supply cables and anything made of iron or steel.
What are you using as a load when testing? Any non-linearity in the load will cause a non-linear current to flow which will then drop a non-linear voltage across the output coil and wiring, causing higher distortion readings after the coil.
How about a toroidal output coil? It shouldn't be hard to take the existing coil and bend it into a circle and glue the end together or use a zip tie. That might fix all of this.
I found a 16-foot piece of severed RG6 coax lying on the ground one day. Very stiff stuff with a braid+foil shield and thick copper core. I can't solder to the braid or the foil though.
I found a 16-foot piece of severed RG6 coax lying on the ground one day. Very stiff stuff with a braid+foil shield and thick copper core. I can't solder to the braid or the foil though.