Okay. To what extent does that sort of thing correlate with accurate prediction of SQ?
That would depend on the preceding signal stages however as simulated the output is linear for more
- see https://www.onsemi.com/pdf/datasheet/mjl21193-d.pdf graphs for various including THD for pairs.
There should be less IMD which is important for good sound clarity. I could simulate these if I can find the time. I'm happy to make room for others to take on that task??
- see https://www.onsemi.com/pdf/datasheet/mjl21193-d.pdf graphs for various including THD for pairs.
There should be less IMD which is important for good sound clarity. I could simulate these if I can find the time. I'm happy to make room for others to take on that task??
Okay. How much IMD is bad and how much is reasonably okay? And is IMD the only thing that matters for prediction of accurate reproduction? IOW, are we focusing on everything that is important, or focusing on only one factor?
I ask in the context of considering the focusing illusion, which is a ubiquitous human bias: https://www.sciencedirect.com/topics/biochemistry-genetics-and-molecular-biology/focalism#:~:text=Focalism-,Focalism, also known as the focusing illusion, is a prototypical,when making judgments or predictions.
The front end of your amplifier is not RC decoupled from the supply any disturbance on these due to switching effects of the output stage will impact on the front end. That is not your fault the blame for this rests with Rod Elliott. I note that his list of component details is non existent. I don't know what level of current he expects in the green LED which has a 22k Ohm connection to ground. The LED could be flickering for all he knows. There are 6 standard types in the Jaycar catalogue with different operating currents and forward voltages. If other builders have had more success than you maybe they could tell you what their picks were.
How to look for problems (or excuses) when there are none.
A classical green led with a forward voltage of 2.0V and a forward current of 20mA will be fine.
It does not flicker.
G'day Guys,
I have collated all of the suggestions so far and come up with this:
Original
Modified.
C8 added across Q9
R11/R12 decreased to 100R
Rdeg1/Rdeg2 added to Q1/Q2
Rdeg3/Rdeg4 added to Q5/Q6
Some other suggestions were:
Decreasing R13/R14 to 0.22R or 0.1R
R11/R12 to 150R
R9/R10 decreased to 2.7K
R7 increase to 820R
Decoupling for Q7/Q8
Decoupling for Q1/Q2
Changing Q4 to BF469
Resistor changes aren't difficult.
I'm not certain that DC offset was an issue. I didn't write down my measurements but I recall something like 5mV on the good channel and 15mV on the bad channel during my initial tests.
I don't think I measured again after biasing.
Extra decoupling wouldn't be too difficult to do:
If I have the idea right here.
At this point I will do some experiments with the good channel and implement whatever is necessary on a new board.
I have collated all of the suggestions so far and come up with this:
Original
Modified.
C8 added across Q9
R11/R12 decreased to 100R
Rdeg1/Rdeg2 added to Q1/Q2
Rdeg3/Rdeg4 added to Q5/Q6
Some other suggestions were:
Decreasing R13/R14 to 0.22R or 0.1R
R11/R12 to 150R
R9/R10 decreased to 2.7K
R7 increase to 820R
Decoupling for Q7/Q8
Decoupling for Q1/Q2
Changing Q4 to BF469
Resistor changes aren't difficult.
I'm not certain that DC offset was an issue. I didn't write down my measurements but I recall something like 5mV on the good channel and 15mV on the bad channel during my initial tests.
I don't think I measured again after biasing.
Extra decoupling wouldn't be too difficult to do:
If I have the idea right here.
At this point I will do some experiments with the good channel and implement whatever is necessary on a new board.
Q7 has emitter to output; should be to positive rail.
If you use CFP in Class AB you may have difficulty ensuring stability and setting accurate bias.
2EF is not so sexy but they are very stable and can be trimmed for very good sound. The CFP is a very elegant engineering topology which attracts many DIYers but rarely works with repeatable stability.
You could increase the stage current from 2.3mA to maybe 3.4mA with R7 390R rather than 560R and decrease the LTP degen from 100R to 33R each.
Then, in the last, add an emitter follower between the LTP and the VAS, running at maybe 2mA. This increases a lot of transconductance on the LTP and adds more loop gain. Then lag comp can straddle base of emitter follower to base of VAS - two devices, and normally reduces the cap to about 33pF.
Some fiddling with driver base stoppers and output stoppers can give better sound quality. And I'd drop the CDECs on the legs of the LTP. Finesse the lag comp to achieve stability; loading down the LTP is a bad idea.
Getting there........
HD
If you use CFP in Class AB you may have difficulty ensuring stability and setting accurate bias.
2EF is not so sexy but they are very stable and can be trimmed for very good sound. The CFP is a very elegant engineering topology which attracts many DIYers but rarely works with repeatable stability.
You could increase the stage current from 2.3mA to maybe 3.4mA with R7 390R rather than 560R and decrease the LTP degen from 100R to 33R each.
Then, in the last, add an emitter follower between the LTP and the VAS, running at maybe 2mA. This increases a lot of transconductance on the LTP and adds more loop gain. Then lag comp can straddle base of emitter follower to base of VAS - two devices, and normally reduces the cap to about 33pF.
Some fiddling with driver base stoppers and output stoppers can give better sound quality. And I'd drop the CDECs on the legs of the LTP. Finesse the lag comp to achieve stability; loading down the LTP is a bad idea.
Getting there........
HD
Hi Sadface, Be carefull where you place your Cdec's. Only place them on the supply voltages. Not on the other end of resistors. Drawing mistake?
Good decoupling can be crucial for stability. I build a cfp amp that required much less hf compensation combined with a superlow impedance powersupply.
Cheers!
Ruben
Good decoupling can be crucial for stability. I build a cfp amp that required much less hf compensation combined with a superlow impedance powersupply.
Cheers!
Ruben
have you got a copy of something like "bob cordell designing audio power amplifiers" ?
The subjective "this sounds better than that" ... is just subjective. a perfect amp , you are just hearing the 'grunt" of a overkill
power supply . The only exception to this is a CFA , which will add the woofer's EMF to the FB signal.
CF (complimentary feedback) does not add any special sound to a output stage , if anything .. it adds a bit of HF oscillation
to it. A good EF3 is faster still. CFP might get you a few volts closer to the rail , but be far less stable , as well.
My EF3 can hit almost 400V/us with my fastest CFA , "shaming" any P3A !
OS
power supply . The only exception to this is a CFA , which will add the woofer's EMF to the FB signal.
CF (complimentary feedback) does not add any special sound to a output stage , if anything .. it adds a bit of HF oscillation
to it. A good EF3 is faster still. CFP might get you a few volts closer to the rail , but be far less stable , as well.
My EF3 can hit almost 400V/us with my fastest CFA , "shaming" any P3A !
OS
Slew rate is proportional to the error signal between positive and negative input of the amp. 400V/us only happens when you drive it with square wave. Thus, don’t pay too much attention on the slew rate figures.
95% of the people will not notice any audible difference and 5% might even prefer the softer P3A
The P3a has never been described as a low distortion amplifier. It does add color to the sound which some will prefer. The wolverine will actually reproduce without adding anything which others will prefer. I don't understand these debates about getting to "the best". Its like picking a favorite wine, beer, car, painting, etc. "Best" is a personal choice.
Furthermore, a person's preference can change from day to day. Nothing wrong with owning a P3A & Wolverine and liking them both!
For the Ef3/3 vs CFP debate, what's wrong with wanting to build and succeed at both. This is a hobby for most of us here. You should enjoy the process, and if building a CFP amp makes you happy, then go for it! The CFP route may be harder and have setbacks, but that doesn't make it pointless.
Furthermore, a person's preference can change from day to day. Nothing wrong with owning a P3A & Wolverine and liking them both!
For the Ef3/3 vs CFP debate, what's wrong with wanting to build and succeed at both. This is a hobby for most of us here. You should enjoy the process, and if building a CFP amp makes you happy, then go for it! The CFP route may be harder and have setbacks, but that doesn't make it pointless.
Better =Reliability and the ability to still be quite stable when you diverge from the BOM.
Better = no one starts threads complaining about said stability.
Best = circuit can fail "gracefully" (wrong parts + screwups).
Self clamping , rounded clip , thermal considerations make up a design that is future proof.
A design that has enough "fudge factor" to not be affected by future faster/improved semi's.
BTW , I really don't like the wolverine. Cascoded VAS's have that "softer sound" with less THD.
I do like to see a failure , current mirror helper semi failed on a wolverine ... did not take out many
other parts. That semi was most likely a factory "dud".
P3A would be much better with few more strategically placed components. A few more semi's .... would actually
approach "hi-fi".
For the soft sound of H2 , just add it to the source .... instead of being "stuck" with it globally.
The "leach amp" is a 70's design that will adapt and perform perfectly with 2023 parts - but it's designer
WAS a professor. Much forethought went into the design.
OS
Better = no one starts threads complaining about said stability.
Best = circuit can fail "gracefully" (wrong parts + screwups).
Self clamping , rounded clip , thermal considerations make up a design that is future proof.
A design that has enough "fudge factor" to not be affected by future faster/improved semi's.
BTW , I really don't like the wolverine. Cascoded VAS's have that "softer sound" with less THD.
I do like to see a failure , current mirror helper semi failed on a wolverine ... did not take out many
other parts. That semi was most likely a factory "dud".
P3A would be much better with few more strategically placed components. A few more semi's .... would actually
approach "hi-fi".
For the soft sound of H2 , just add it to the source .... instead of being "stuck" with it globally.
The "leach amp" is a 70's design that will adapt and perform perfectly with 2023 parts - but it's designer
WAS a professor. Much forethought went into the design.
OS
I built the original version of the Leach it ran cool and I liked the sound. I used single pairs of 2SB645 2SD665 pairs in the output stage (10 Mhz unity gain types) and MJE340 and MJE 350 drivers. This was in 1980 and I could not source the Leach approved devices. One could not be sure of getting these from overseas outlets.
I ordered some plans for a Leach premplifier with FET inputs in answer to a magazine.advertisement. I had overseas subscriptions to Audio Amateur and Speaker builder and later to Audio Xpress. There was an airmail chargefor these but it does not take 4 or more weeks for post to arrive. I believe these were shipped in bulk to Gordon and Gotch the NZ distribution contractor for these and other overseas periodicals
The plans I send money for never arrived.
Thank goodness for the internet and Diyaudio forums.
I ordered some plans for a Leach premplifier with FET inputs in answer to a magazine.advertisement. I had overseas subscriptions to Audio Amateur and Speaker builder and later to Audio Xpress. There was an airmail chargefor these but it does not take 4 or more weeks for post to arrive. I believe these were shipped in bulk to Gordon and Gotch the NZ distribution contractor for these and other overseas periodicals
The plans I send money for never arrived.
Thank goodness for the internet and Diyaudio forums.