layout
There is this hot fix with the gnd wire between the two caps on the PCB. And the note:
"Placing a short thick wire connecting the ground pins of the supply caps reduces THD"
Two questions:
- whats the background, why does it reduce THD ?
- doesnt make this wire a big GND-loop around the PCB ?
Thx !
There is this hot fix with the gnd wire between the two caps on the PCB. And the note:
"Placing a short thick wire connecting the ground pins of the supply caps reduces THD"
Two questions:
- whats the background, why does it reduce THD ?
- doesnt make this wire a big GND-loop around the PCB ?
Thx !
MikeB said:Hi Pierre, i have not forgotten the high power symasym, i did not get into it yet. I know i have promised it...
I can't give any definitive schedule, it's hobby not work...
How much power would you need ?
Thanks for the reply Mike. No problem...I just wanted to know it it was still in your project list . For the power... 200W in 4 Ohm would be nice
Hello Mike,
Thank you for your prompt response.
You are more proffesional than hobyste !
This long Week-End i' will make the modification you suggest.
Coil change by resistor and add emiter resistor in input!
The output is mjl 3281/1302 the drivers mje15030 /15031 according your web site.
JC BALTHAZARD
Thank you for your prompt response.
You are more proffesional than hobyste !
This long Week-End i' will make the modification you suggest.
Coil change by resistor and add emiter resistor in input!
The output is mjl 3281/1302 the drivers mje15030 /15031 according your web site.
JC BALTHAZARD
Hi Cortez, there are many more ways to increase THD than by unlinearities in the amplifier itself, one is modulated ground, caused by large currents flowing in the gnd. This "thick wire" simply shorts out the current ripple introduced by the 1000uf caps. This mechanism can create easily >0.1%thd, getting really bad for higher freqs.
In theory this wire creates a nice loop which could act as antenna, but it seems that the signals received by this are very low compared to the currents flowing, at least it did not show problems.
With the revised layout (5.3) a single trace goes to this connection, nearly nulling out these currents.
Hi Pierre, btw, symasym in its present state already delivers >100watt into 4 ohms without problems, just be sure that drivers and output-bjts have hfe > 100, and that the PSU is strong enough...
It's unlikely that you really notice the difference between 120 and 200 watts. I thought you would like to have 500watts...
Balthazard, with the MJLs symasym should not show instabilities, this problem only appeared with the Toshibas, and with c14=10pf you should have enough safety margin... But, i did not check with Iq=500ma. How big was the oscillation ? Few mv's ore more ?
Have you changed the feedbacknetwork to lower gain ?
Mike
In theory this wire creates a nice loop which could act as antenna, but it seems that the signals received by this are very low compared to the currents flowing, at least it did not show problems.
With the revised layout (5.3) a single trace goes to this connection, nearly nulling out these currents.
Hi Pierre, btw, symasym in its present state already delivers >100watt into 4 ohms without problems, just be sure that drivers and output-bjts have hfe > 100, and that the PSU is strong enough...
It's unlikely that you really notice the difference between 120 and 200 watts. I thought you would like to have 500watts...
Balthazard, with the MJLs symasym should not show instabilities, this problem only appeared with the Toshibas, and with c14=10pf you should have enough safety margin... But, i did not check with Iq=500ma. How big was the oscillation ? Few mv's ore more ?
Have you changed the feedbacknetwork to lower gain ?
Mike
Hi Mike
> This "thick wire" simply shorts out the current ripple introduced by the 1000uf caps.
But this wire operates between the two caps, there is no big current there in theory.
> This mechanism can create easily >0.1%thd, getting really bad for higher freqs.
And whats the base about this frequency dependence ?
Grounding is "just" wiring, which means mainly ohmic resistance, in theory.
> In theory this wire creates a nice loop which could act as antenna,
> but it seems that the signals received by this are very low compared
> to the currents flowing, at least it did not show problems.
I meant opposite, this GND-loop is the "radiator", and the other
parts inside the loop are the receivers, cause current loops are
coupled to eachother, arent they ? -> another GND current- modulation ?
> With the revised layout (5.3) a single trace goes to this
> connection, nearly nulling out these currents.
It wouldnt be better if there was a separate wire from the non-signal GNDs to the main PS star-GND ?
(Under non-signal GNDs I mean: the local filter caps, CCSs, etc...)
And an other GND-wire would go from the signal GND to the same main PS star-GND.
(main PS starGND = right at the main caps, not on the PCB)
> This "thick wire" simply shorts out the current ripple introduced by the 1000uf caps.
But this wire operates between the two caps, there is no big current there in theory.
> This mechanism can create easily >0.1%thd, getting really bad for higher freqs.
And whats the base about this frequency dependence ?
Grounding is "just" wiring, which means mainly ohmic resistance, in theory.
> In theory this wire creates a nice loop which could act as antenna,
> but it seems that the signals received by this are very low compared
> to the currents flowing, at least it did not show problems.
I meant opposite, this GND-loop is the "radiator", and the other
parts inside the loop are the receivers, cause current loops are
coupled to eachother, arent they ? -> another GND current- modulation ?
> With the revised layout (5.3) a single trace goes to this
> connection, nearly nulling out these currents.
It wouldnt be better if there was a separate wire from the non-signal GNDs to the main PS star-GND ?
(Under non-signal GNDs I mean: the local filter caps, CCSs, etc...)
And an other GND-wire would go from the signal GND to the same main PS star-GND.
(main PS starGND = right at the main caps, not on the PCB)
Still in slow progress. Tomorrow may be I can bring them to life. 
Cheers
Karel
An externally hosted image should be here but it was not working when we last tested it.
Cheers
Karel
Cortez, sorry, i am not in the best mood for techtalk these days...
About the grounding wires beeing ohmic resistance only, their inductance seems high enough to cause troubles for higher frequencies...
And with larger supplycaps onboard, the typical star grounding could defeat their benefit.
You're right, this groundloop does induce the halfwaves flowing there into the inputstage, that's why i changed it to single trace in v5.3.
These half waves are visible in the THD measuring as even harmonics.
As i had no problems with hum, i did not research grounding further...
The PSU-ripple in the output measured with -110db for a 2.5v signal, that's <2uv.
Grounding is difficult as it is THE reference voltage throughout the system, but has to deal with large currents.
Mike
About the grounding wires beeing ohmic resistance only, their inductance seems high enough to cause troubles for higher frequencies...
And with larger supplycaps onboard, the typical star grounding could defeat their benefit.
You're right, this groundloop does induce the halfwaves flowing there into the inputstage, that's why i changed it to single trace in v5.3.
These half waves are visible in the THD measuring as even harmonics.
As i had no problems with hum, i did not research grounding further...
The PSU-ripple in the output measured with -110db for a 2.5v signal, that's <2uv.
Grounding is difficult as it is THE reference voltage throughout the system, but has to deal with large currents.
Mike
> Cortez, sorry, i am not in the best mood for techtalk these days...
No problem of course, and I wish you the bests !
> About the grounding wires beeing ohmic resistance only, their
> inductance seems high enough to cause troubles for higher frequencies...
I see, really, so low ohmic and inductance GNDing is the ideal.
And what about capacitive GND problems, how they occur ?
> And with larger supplycaps onboard, the typical star grounding could defeat their benefit.
Why? I dont see this. The point is, just not to share different GND-currents on the same GND wire, isnt ?
So if these onboard caps have a different GND-wire to the main caps, it wouldnt be a problem.
> These half waves are visible in the THD measuring as even harmonics.
Sometimes unbeliveable how much depends on wiring, GNDing, layout....
Otherwise if I put my main caps near to the output transistors
(5-10cm) it also recommended to use these 1000uF local caps on board ?
> As i had no problems with hum, i did not research grounding further...
I would be very interedted in this researches, maybe it counts a lot...
> Grounding is difficult as it is THE reference voltage throughout the system, but has to deal with large currents.
Grounds should be connected to the main GND on caps as late as possible, i guess.
And with a thick wire, to minimize the resistance too.
Thx for reply, cheers !
No problem of course, and I wish you the bests !
> About the grounding wires beeing ohmic resistance only, their
> inductance seems high enough to cause troubles for higher frequencies...
I see, really, so low ohmic and inductance GNDing is the ideal.
And what about capacitive GND problems, how they occur ?
> And with larger supplycaps onboard, the typical star grounding could defeat their benefit.
Why? I dont see this. The point is, just not to share different GND-currents on the same GND wire, isnt ?
So if these onboard caps have a different GND-wire to the main caps, it wouldnt be a problem.
> These half waves are visible in the THD measuring as even harmonics.
Sometimes unbeliveable how much depends on wiring, GNDing, layout....
Otherwise if I put my main caps near to the output transistors
(5-10cm) it also recommended to use these 1000uF local caps on board ?
> As i had no problems with hum, i did not research grounding further...
I would be very interedted in this researches, maybe it counts a lot...
> Grounding is difficult as it is THE reference voltage throughout the system, but has to deal with large currents.
Grounds should be connected to the main GND on caps as late as possible, i guess.
And with a thick wire, to minimize the resistance too.
Thx for reply, cheers !
Maybe this help will be helpful for anyone.
The length of the wire must be inferior at lambda / 10 where lambda is C / Fclock
So, in calculuus :
l < ( C / Fclock ) / 10
Where
l was the lenght of wire in meters.
C , i'm not sure at all but that's must be the capacitance.
Fclock, the frequency of the clock, where in our case are from 3.5 to 145 KHz.
To my mind, the lenght of the wire can not exceed 15 cm for the 145 KHz frequency and 1000 uF capacitance.
The length of the wire must be inferior at lambda / 10 where lambda is C / Fclock
So, in calculuus :
l < ( C / Fclock ) / 10
Where
l was the lenght of wire in meters.
C , i'm not sure at all but that's must be the capacitance.
Fclock, the frequency of the clock, where in our case are from 3.5 to 145 KHz.
To my mind, the lenght of the wire can not exceed 15 cm for the 145 KHz frequency and 1000 uF capacitance.
Hello:
I am working on a DIY Headphone amp similar in topology to what you guys are doing hear I posted about it in another thread since my question related to a subject they were talking about. I really never got answered my question so again I want to make my DIY Headphone amp stable without using compensation capacitors. I think by adding local feedback resistive around the voltage gain stage will solve for my desired result. http://www.diyaudio.com/forums/attachment.php?postid=893692&stamp=1145084241
I am working on a DIY Headphone amp similar in topology to what you guys are doing hear I posted about it in another thread since my question related to a subject they were talking about. I really never got answered my question so again I want to make my DIY Headphone amp stable without using compensation capacitors. I think by adding local feedback resistive around the voltage gain stage will solve for my desired result. http://www.diyaudio.com/forums/attachment.php?postid=893692&stamp=1145084241
Hi ppl, with this topology you have way too much gain to get away without compensation caps, you should go for folded cascode like shown in post #10 in this thread (you don't need the tripledarlington). For headphone amp you don't need large amounts of feedback or low zout and you can use full ClassA operation.
With a double darlington output and carefully chosen openloop gain you can get away without comp caps, or minimal compensation like a 10pf or smaller from vas to negative input...
I once built the amp shown without any comp cap, it was not bad. Here is a thread about this circuit.
Mike
With a double darlington output and carefully chosen openloop gain you can get away without comp caps, or minimal compensation like a 10pf or smaller from vas to negative input...
I once built the amp shown without any comp cap, it was not bad. Here is a thread about this circuit.
Mike
MikeB said:
Thanks mike for your suggestions and the link. This is a real interesting discussion on your Design. I came to the same conclusion as you did regarding the large amount of open Loop gain; I decided to go back to the folded Cascode method I started out with, however using refinements from the KISS amp. Have you considered returning them Vas load 100K resistors to the inverting input and not ground? This would give you a inner feedback loop around the vas. Regarding not requiring a triple Darlington for headphone use, Well from my findings with low impedance headphones in the 32-100 ohm range that the measured distortion under load when tested with a real headphone Load. I have seen Headphone Amps measure 0.001% THD with a 50 ohm Dummy resistor however slap on a pair of Grado Headphones and that number changes to something like 0.06% with a Dual stage Diamond Buffer by adding another stage with that addition being the output stage using larger transistors than previously used as output transistors driven from an op amp (AD8610) this changes the THD under load a lot the 0.001% with 50 ohm Resistor to 0.008% with the Grado headphones. This was at the same Bias setting for the output devices. This is but one of hundreds of different test on different output stage Topologies used to drive headphones. I also should not this was at 1 volt output no where near the output capability of the Amplifier so I would assume the performance difference gets worse at higher levels.
As for Class A this amp started out with an ICq of 500 Ma on the output stage and 10 mA on the input with big old MPS4392 JFETs on the input. That version was most of the time stable. It was sensitive to the type of headphones used some Brands would make compensation capacitors required. I will post my updated Schematic when finished.
Oh I found this statement by you totally true and I can see you are like myself and Love micro dynamics
“As i switched to doubledarlington these caps got very big, this has
the advantage that the vas follows exactly the voltage inside the
drivers, beeing syncron with the outputstage. This result in perfect
behaviour for fast transients, means squarewaves are reproduced
perfect, not a single sign of overshooting/ringing in the whole circuit”
Thanks for your input and we will see how the next version of my Folded Cascode Headphone Amp works out, Wow I don’t have to change the file name because it will have reverted back to the Folded Cascode.