For a high frequency transient the inductance of the longer traces and the cables prevents much current coming from the MF decoupling or from the PSU.
The majority of the HF transient current comes from the HF decoupling capacitor.
Treat the capacitor as the sole supplier of current for that very short period transient. The HF transient is not music/audio, it is interference and ultra fast corrections to the audio, that's why we need to use multi-megaHertz transistors.
So starting with the idea that the capacitor is the sole source of current, draw in the route that takes current from one lead of the cap through the processor to the load and back to the other lead of the capacitor.
Once you have drawn that route, decide whether the total length and shape of the route fits with a requirement to pass an HF transient, i.e. ultra low inductance.
The majority of the HF transient current comes from the HF decoupling capacitor.
Treat the capacitor as the sole supplier of current for that very short period transient. The HF transient is not music/audio, it is interference and ultra fast corrections to the audio, that's why we need to use multi-megaHertz transistors.
So starting with the idea that the capacitor is the sole source of current, draw in the route that takes current from one lead of the cap through the processor to the load and back to the other lead of the capacitor.
Once you have drawn that route, decide whether the total length and shape of the route fits with a requirement to pass an HF transient, i.e. ultra low inductance.
Last edited:
The different current sources behave different if we look at the thermal stability. The green led ensures a nearly constant current from (winter-) cold amplifier to full power driven (summer-)temperature ranges.
Using the zener would kill us more as 4V rail voltage headroom compared to dual diode or green led variant.
My first decision was to use a simple current source where we loose less rail voltages. The bad thermal behavior implied the use of green leds.
I didn't know about cutworks.com. Just a pitty that their website is only German.
I find Schaeffer a bit too expensive to create bottom and top panels. Cutworks looks interesting but I need to learn German
Thank you for the tip.
Could briefly explain how cutworks works; I mean I guess I have to create a design with a CAD program. What are the next steps; how do I get a quote and order panels?
On this side of the world laser cutting becomes more cost effective when you are getting a complete 4' x 8' sheet of metal done. We pay more for the operator to load the machine, and less for the laser to cut. That might be the case there too.
Most prefer CAD drawings. Laser just requires 2D drawings.
Most prefer CAD drawings. Laser just requires 2D drawings.
New cutworks shop is available mostly also in english.
The standard home page seems to be only in german - google translate can help here...
create/register a new account then:
BR, Toni
The standard home page seems to be only in german - google translate can help here...
create/register a new account then:
- create your part with a cad program to generate a dxf file. QCAD, LibreCAD or similar.
- login to the shop and upload your dxf file using "create part" and follow the wizard to "next step choose material"
- chose steel, stainless steel or aluminium
- button "next to justify"
- button "next to define cut"
- button "next to summary"
- select number of copies and button "save and goto ..."
BR, Toni
Thanks Toni.
Schaeffer's Front Panel Designer is able to export DXF files. Have you ever tried using that to order panels at Cutworks? Do you think it would work?
Hi. I don't want to clutter Toni's thread anymore with this. Perhaps we should transfer this to the thread dealing with my implementation of Self's circuit. Link above. But if we think of decoupling caps as reservoirs or supply depots for fast transients, I see the current flow from cap to op amp supply pin, through the op amp and its load, returning to the op amp negative supply pin and back to the aux PSU. Surely only leakage currents or currents at frequencies for which the decoupling cap presents a low impedance path actually flow through the cap to its other pin (connected to ground or whatever the case may be). In any case, it seems to me that a flow of any current via signal ground is not ideal. Signal ground is only very indirectly connected to the aux supply ground and we don't want to pollute the signal reference with HF supply artefacts. Next best would be for any bypass currents to flow directly to the aux supply ground. And surely best of all is for such currents to pass directly to the lowest potential of the supply i.e. from positive to negative aux supply rail. Do you disagree?
Thx. I suspect in the end we are just saying the same thing, no? The green LED variant is simply a better current source because of its higher output impedance (resulting from higher emitter degeneration). A zener, or other form of voltage reference, might allow higher emitter degeneration = higher output impedance = a better current source, but at a cost of voltage headroom. As zener dioxides are just average voltage references, I was wondering if performance could be improved further with a lower noise, better performing, reference - even if such reference did not consume as much headroom as the 6.2V zener in Bob's example. Say, a high performance 5V or 2.5v voltage reference. (SPX431 or such like.)
(I seem to remember warnings being given in the thread dedicated to Bob's book with respect to his seemingly favoured "feedback current source".)
Thanks Andrew. Let me chew on it. I believe the below is the point in the conversation when Toni switched the decoupling of the supply to the input stage from signal ground to power ground. (C3/4 equate to C14/15 in the final schematic). In my implementation the 10R resistor referenced by kgrlee (it's actually R3 in the final schematic) is removed because signal ground on my input board is directly connected to the "silent" star ground board with thick wire. Effectively I have signal ground 'spanning' two boards and make the connection to silent star ground via the balanced to single-ended input board rather than Toni's path from IPS/VAS to OPS "dirty" ground via R3 and onward from there.
Nice!
Does the output stage typology of SA2014 have a particular name? I note it is like the Triple in that the drivers are connected emitter to emitter and not the output rail. Is it "the Double"? I note you were steered away from exploring the Triple or Darlington OPS typologies but I am wondering what led you to your typology choice in the first place. I did not see it discussed by Cordell but see his Fig 15.14 shows a similar configuration (without the paralleling). Was it solely to have the drivers operating in class A?
Also, what is the purpose of R46, R47, R49, R50 given all the driver emitters are connected via R48? Is it to isolate each of the transistors in a parallel pair from each other?
Does the output stage typology of SA2014 have a particular name? I note it is like the Triple in that the drivers are connected emitter to emitter and not the output rail. Is it "the Double"?
I note you were steered away from exploring the Triple or Darlington OPS typologies but I am wondering what led you to your typology choice in the first place. I did not see it discussed by Cordell but see his Fig 15.14 shows a similar configuration (without the paralleling). Was it solely to have the drivers operating in class A?Also, what is the purpose of R46, R47, R49, R50 given all the driver emitters are connected via R48? Is it to isolate each of the transistors in a parallel pair from each other?
Last edited:
Never tested a xxx431 based voltage ref. for ccs. Need to sim this variant.
Indeed there was a discussion of HF oscillation if base resistors are missing or wrong dimensioned using different transistors. The thermal behaviour is also bad. This was another reason to decide to use the simple but very efficient CCS variant using green LEDs.
