Hi Valery,
Yes, I think that is Ed's well-known loading paper. I think I was probably at the Convention when he delivered it. Thanks for bringing that to my attention.
Cheers,
Bob
Yes, that is the version that was being referred to as the one where each emitter (in the BJT case) had its own Zobel. I think that is a nice idea for distributed Zobels, but I have not taken it that far, just putting multiple Zobels on the output rail. With the speed of MOSFETs, your approach might be quite advantageous. This would be especially the case if there was some inductance in the source resistors. I also lean toward paralleled (if necessary) metal oxide 3W resistors for source or emitter resistors (rather than non-inductive wirewound resistors).
Cheers,
Bob
The reason I first proposed the connection point before the emitter resistor was that it looks very attractive to use SMT thick film resistors directly to the transistor emitter leads.
Not only does it avoid the emitter resistor inductance but lowers the trace inductance about as low as possible, and then halves it from the paralleled connection with two resistors per pair.
And doubles the number of resistors to spread the power dissipation even more.
Carries your distributed resistor idea to limit.
Best wishes
David
Last edited:
Well, easy for me to say it doesn't seem hard, do you have any concerns?
Best wishes
David
Yes, how to layout the board to have the leads soldered directly to the emitter leads with the transistor under the PC board. Maybe I am having trouble visualizing this for myself. Do you have an example? or picture?
Yes, how to layout the board to have the leads soldered directly to the emitter leads with the transistor under the PC board. Maybe I am having trouble visualizing this for myself. Do you have an example? or picture?
No example yet, I am at work on the layout process myself and think I understand the physics but have to learn the practicalities, so very helpful to discuss possible issues.
I definitely assume a double sided board.
A 2512 resistor is a convenient width to connect more or less directly to the wider part of a TO-264 output transistor lead closest to the die.
I plan to simply have a pad and hand solder the resistor.
My initial idea was to have the board horizontal and OPS transistors vertical down one side to connect directly to the vertical heat-sink face.
That is simple to visualize but does limit the space around the OPS transistor leads, OK for most transistors but a bit of a constraint for the extra leads of the Thermal Trak devices I have.
The other common style is to have the transistors bent over parallel to and under the board.
Then it is not necessary to have them down one side, I think this is what you envision?
That leaves more space for connection flexibility but I haven't really worked out that option.
Best wishes
David
Last edited:
Stack 5 of these on top of each other...
500pcs 5 1watt Wattage 2512 Chip SMD Resistors 1 Ohm Resistance | eBay
500pcs 5 1watt Wattage 2512 Chip SMD Resistors 1 Ohm Resistance | eBay
Ahh, it's always assumptions that we don't mention that lead to confusion.
I just visualized one 2512 and that makes it easy, they are available rated up to 3W.
That's 6 W dissipation per output pair, that should be more than adequate.
Best wishes
David
Last edited:
Magic, if you have more accurate models for bc550c, please post them here.
The one I remember predates his well known AES one. Maybe in one of the Australian journals, comics or even an internal AWA magazine. It's very old cos I remember adverts (schematics?) of valves (tubes to yus Yanks) on the same page.
And da Aboriginal elder's name was Neville. Dick Small was only an adopted Aboriginal elder.
I think Guru Cordell's
pretty much sums up my present thinking.
The only new bit is that the Zobel at the speaker terminals can be replaced with a single 22n ceramic with advantage.
This lets me use loadsa 'real life' RFI proofing work I've done on mikes & preamps.
____________________
Duu.uh! I thought Guru Cordell was the first to suggest this.
I've done a tiny amount of SPICE work which suggests this might be a good thing for the more flighty OP devices.
But there are possible issues with stability as each half switches off.
Could we have a straw poll as to who has tried this in 'real life'?
Though I've great respect for the theoretical pontificating of various gurus here, I've even greater respect for 'real life' stuff they've tried.
http://www.diyaudio.com/forums/solid-state/240712-cfa-topology-audio-amplifiers-644.html#post4109825
Damir, is this your invention?
_____________________
.. just a caveat that the power rating of stuff is critically dependent on their mounting. 5x1W stacked will have MUCH less than 5W power handling.krisfr said:
Also, the higher power handling of modern resistors in small packages is obtained by running them at higher temperature.
An Immutable Law of Nature. Heat dissipation/removal is proportional to Area x Temperature. Smaller area (device) needs higher temperature to move the same amount of heat.
Be careful they don't unsolder themselves in use or mild overload.
It's common to mount thru lead power resistors some distance off the PCB on ceramic standoffs so their heat doesn't damage or discolour the PCB.
Usual for most competent commercial amps of the late 20th century. It sacrifices low inductance for better thermals.
Last edited:
As far as I know, it was me. Some links are >HERE< and there was an another one I can't find where Edmond Stuart and Andrew T responded.
But it is sufficiently obvious that I can't believe I was the first to think of it.
I have asked repeatedly but had little response to previous enquiries, perhaps this time someone will have an earlier reference.
The same possible issues concerned me too and I haven't tried it in practice yet.
I planned to experiment with it for my next amp, now I know Damir has the same idea too, I am even more confident to try it.
Best wishes
David
The ex-physics student in me can't let this pass. This is not an law of nature, just an old approximation for typical losses at room-ish temperature. Radiative losses increase much faster with temperature, for instance.
Last edited:
Not my idea, I think nattawa suggested it here http://www.diyaudio.com/forums/solid-state/243481-200w-mosfet-cfa-amp-21.html#post3690559, but more as the compensation.
Damir
I hadn't thought of replacing the terminal zobel, but yeah that would be clever to let it double as RFI shunt as well. I'll have to sim this, and model the speaker leads more accurately to see how this holds up. Like .01R, 100nH series in both leads.
I grovel at Guru Zan's pedantic feet
.. and his championing of the Stefan-Boltzmann conditions.However, for our poor resistor, I beg to suggest plain old Fourier & Newton is more applicable.
Of course I should have said "Heat dissipation/removal is proportional to Area x Temperature DIFFERENCE"
There's a number of subtleties .. eg raising the resistor above the PCB increases Convection so isn't quite described by Fourier & Newton.
But the point I was trying to make was that if you have two 5W resistors, one much smaller than the other, the smaller one has to run hotter for its rating.
This holds whether Fourier, Newton, Zan or Stefan-Boltzmann are the most applicable eqns. That's the bit that's Immutable. The eminent gentlemen only tell you how MUCH hotter.
Four thoughts:
1) I am always paranoid about resonances and the like at high frequencies. Just a straight 0.022 cap to ground at the output terminals makes some sense, but if that is all there is I worry about opportunities for resonance against the wiring back to the board or the speaker cable - or it causing some "tuning" of the speaker cable transmission line. For that reason, it still might be wise to parallel that capacitor with a Zobel to assure good damping at that node. I am assuming that the cap and Zobel are right across the output terminals.
2) Apart from differential-mode effects at the output terminals and speaker cable, we need to remember that the speaker cable and the speaker itself comprise a huge common-mode antenna. How pick-up from this is handled may also be a concern.
3) With regard to individual Zobels on PNP and NPN emitters, I don't think this will cause any stability problems when one side switches off. Bear in mind that the off-side emitter is still connected to the common output rail through a very small-value RE that should be non-inductive.
4) In thinking of output stage physical design and inductances, we need to keep in mind that the length of the output rail may not be insignificant, so distributed damping by Zobels along it may help suppress resonances. In amplifiers that use multiple output pairs, how they are arranged may also make a difference; i.e., where all the NPNs are grouped together or where pairs of NPN and corresponding PNPs are grouped together. I like the latter because the highly non-linear half-wave-rectified class AB currents can be locally resolved (circulated) through a smaller loop, as with a rail-rail capacitor.
Cheers,
Bob
1) I am always paranoid about resonances and the like at high frequencies. Just a straight 0.022 cap to ground at the output terminals makes some sense, but if that is all there is I worry about opportunities for resonance against the wiring back to the board or the speaker cable - or it causing some "tuning" of the speaker cable transmission line. For that reason, it still might be wise to parallel that capacitor with a Zobel to assure good damping at that node. I am assuming that the cap and Zobel are right across the output terminals.
2) Apart from differential-mode effects at the output terminals and speaker cable, we need to remember that the speaker cable and the speaker itself comprise a huge common-mode antenna. How pick-up from this is handled may also be a concern.
3) With regard to individual Zobels on PNP and NPN emitters, I don't think this will cause any stability problems when one side switches off. Bear in mind that the off-side emitter is still connected to the common output rail through a very small-value RE that should be non-inductive.
4) In thinking of output stage physical design and inductances, we need to keep in mind that the length of the output rail may not be insignificant, so distributed damping by Zobels along it may help suppress resonances. In amplifiers that use multiple output pairs, how they are arranged may also make a difference; i.e., where all the NPNs are grouped together or where pairs of NPN and corresponding PNPs are grouped together. I like the latter because the highly non-linear half-wave-rectified class AB currents can be locally resolved (circulated) through a smaller loop, as with a rail-rail capacitor.
Cheers,
Bob