I figured out how to get the driver stage decoupling ground trace from the positive rail to the star separate from the output stage ground trace.
However I can't figure out how to resolve the negative rail with the positive rail separate from the star.
The best I can come up with is to use a jumper like this:
This would allow me to resolve positive and negative grounds from the various decoupling stages separate from each other and still keep all the signal grounds separate too.
I have also indicated in yellow where I will place a heatsink across the 2 drivers. My plan would be to cut some 2mm aluminium with a cutout to go over the top of R13/R14 so that the 2 drivers are on the same heatsink. My plan would also be to drill some holes to either side of Q6 so that a cable tie can still be used to thermally couple Q9 to Q6 as Rod recommends.
The other option would be to do as John Ellis suggests and use a BD139 for Q9 with some flying leads so that it could be bolted directly to the driver heatsink. Or plausibly with a lot of rework on the layout, Q9 and the trimpot could be mounted in between R13/R14 so that it is mounted in the centre of the driver heatsink; equidistant from the 2 drivers.
Oh I also added an extra hole on the ground trace at the bottom as an easily accessible test point. I figure that the holes added for the safety resistors on the fuse rails serve the same purpose as test points.
However I can't figure out how to resolve the negative rail with the positive rail separate from the star.
The best I can come up with is to use a jumper like this:
This would allow me to resolve positive and negative grounds from the various decoupling stages separate from each other and still keep all the signal grounds separate too.
I have also indicated in yellow where I will place a heatsink across the 2 drivers. My plan would be to cut some 2mm aluminium with a cutout to go over the top of R13/R14 so that the 2 drivers are on the same heatsink. My plan would also be to drill some holes to either side of Q6 so that a cable tie can still be used to thermally couple Q9 to Q6 as Rod recommends.
The other option would be to do as John Ellis suggests and use a BD139 for Q9 with some flying leads so that it could be bolted directly to the driver heatsink. Or plausibly with a lot of rework on the layout, Q9 and the trimpot could be mounted in between R13/R14 so that it is mounted in the centre of the driver heatsink; equidistant from the 2 drivers.
Oh I also added an extra hole on the ground trace at the bottom as an easily accessible test point. I figure that the holes added for the safety resistors on the fuse rails serve the same purpose as test points.
Hugh,
Thanks for that. Some time ago I tried to contact Ed through the university but could not find any contact details, so assumed he had retired.
His nested feedback idea was quite novel in audio design. I have found that the major improvement in his design for reducing crossover distortion seemed to be the use of inclusive Miller compensation, and he well understood the idea of using degen in the input stages to prevent slew rate overloading, which Stochino later exploited.
Sorry to hear that he is suffering, cancer is not nice.
Thanks for that. Some time ago I tried to contact Ed through the university but could not find any contact details, so assumed he had retired.
His nested feedback idea was quite novel in audio design. I have found that the major improvement in his design for reducing crossover distortion seemed to be the use of inclusive Miller compensation, and he well understood the idea of using degen in the input stages to prevent slew rate overloading, which Stochino later exploited.
Sorry to hear that he is suffering, cancer is not nice.
Not recommended. You want the junction of the bias transistor as close as possible to the drivers junction to thermally track it with as little thermal lag as possible.
I have revisted the simulation I posted on the other thread about linearising the P3a. I did not look at changesto the LTP.
I have persisted witn MJL21193/21194 where the .models were from Cordell while others were from the LTspice on board library. Replacing these with Cordell models gave different results and one of the problems of setting the output standing current to a low level is feasible. I'm talking about 30 m.a. and THD just over 0.03%@20 kHz into 8R at 28 volts peak output.
For the LTP I used 1N4148 diode refences in lieu of the LED and included a common emitter amplifier in the tail. I added a collector resistor in combination with the common emitter amplifier. This does steady enough job where a current mirror would take up board space. This strategy was used in a Playmaster project published in Electronics Ausralia magazine in the 1970's. That design was my second build.
I managed to do without emitter resistors for the drivers. I see these more as a way to increase the Vbe adjustment range. I sought other ways to do that.
l have a copy of Motorola Application Note AN484a which has a part dealing with thermal stability I think I have scanned this paper before
In one part this discusses a situation where the amplifier heats up at high output levels and then abruptly drops to a lower level. If the diodes cool while the heat in the Power Transistors lingers, because dissipation is not instantaneous as music can be, then the diode bias will increase due to less heat in the vbe. Ideally the cooling rate in the vbe should be instep with the heat sink maybe as in the Class A circuit which is a Class B on steroids.
These problems can be solved by the right size heat sink and emitter resistor values. That some builders have success while others miss the bus could experience this problem.
Last edited:
You should look at the Silicon Chip Hummingbird Amp. PCBs are dirt cheap from the Silicon Chip website. Altronics were offering a kit which may still be available. See Silcon Chip Dec 2021 and Jan 2022 issues for details. P3A is getting a bit long in the tooth. Think it was designed before spice was commonly available. Yep, a few tweaks were sometimes necessary for DC offset and to fix oscillation. No big deal and the amplifier does sound good.
Can't see what is gratuitous or negative about my comment. The design is a compromise. It works in a satisfactory manner for me. I use it nearly everyday.
There was a DC offset problem that I fixed by reducing R7 to 470 ohms. Others who had oscillation issues have increased the value of C6 to fix the problem. Apparently the oscillation is visible on the negative cycle of a test sine where the output amplitude is approaching the rail voltages.
I used the Rev C board I purchased from Rod.
There was a DC offset problem that I fixed by reducing R7 to 470 ohms. Others who had oscillation issues have increased the value of C6 to fix the problem. Apparently the oscillation is visible on the negative cycle of a test sine where the output amplitude is approaching the rail voltages.
I used the Rev C board I purchased from Rod.
Last edited:
Thank you for setting the record straight. Sorry I jumped on you, I am getting tired of other people propagating countless 'hear-say'.
Not sure if this will help resolve the layout questions, but the best tool to use is to analyze where current flows. Keep in mind that the charging currents for the filter caps are rather nasty haversine waves, flowing only at the crest and valley of the AC power cycle. The return trace should be considered as a series of small value resistors. Current flowing across these resistors creates a voltage drop. You don't want the more sensitive circuits positioned such that different components are at different points along the voltage drops in the return path as this induces the voltage drops into these circuits. This is why the ground "star" is helpful, as long as the star itself is not part of a high current path. And yes, it is complicated. The PCB layout is really a "component" in the circuit.
Last edited:
A few months ago I made pcb for this amp out of curiosity (according to the original scheme).
Trafo 2x20V 300VAC, 2SC5200+2SA1943+MJE15035/34 from Mouser. Other elements are standard 1% Beyschlag resistors and caps from Panasonic/Rubycon/Nichicon.
DC offset after +-20min warming was 13 and 15 mV. Thanks to helitrim (2k) I added in parallel to R6 (1k2) i set it up to 0V.
Q5 and Q9 coupled with a piece of copper sheet.
Trafo 2x20V 300VAC, 2SC5200+2SA1943+MJE15035/34 from Mouser. Other elements are standard 1% Beyschlag resistors and caps from Panasonic/Rubycon/Nichicon.
DC offset after +-20min warming was 13 and 15 mV. Thanks to helitrim (2k) I added in parallel to R6 (1k2) i set it up to 0V.
Q5 and Q9 coupled with a piece of copper sheet.
Attachments
At this point were you getting the same volume on the speakers from both channels? Previously you stated that the left channel was resistricted to 3 V RMS.
In post 240 I included an image attachment forthe Silicon Chip Class A which has some differences to P3A such as current mirrors and CCS loads forVas collector load. I would have though the heart of the circuit could be used with changes to accomdate a bootstrap collector load for the VAS remove the current mirror in the LTP and draw connecting lines in place. I can post a component side view with tracks shown underneath and the copper side if that would help.
I'll quote the BlackSheep: If the P3A doesn't sound good, you've done something wrong son.
Ok guys, according to the track and trace, I should have BD139/140 by Friday.
While waiting, I've been fiddling with the layout some more. Since the previous layout seemed to not be preferable, I've gone back to basics.
I've chopped off the the extra decoupling stages and gone back to a plain old Rod Elliot style layout.
This is pretty darn similar to Rod's layout however he biggest difference is the extra degen resistors, the larger input cap, larger diameter decoupling caps.
I understand that Q9 should be mounted as close as possible to the Tj of Q6 which I would take to mean the metal tab side of Q6.
However I of course want the metal tab side of Q6 on a heatsink.
Am I over thinking this, should I just be cable trying Q9 to the 'front' side of Q6 and forget about it?
While waiting, I've been fiddling with the layout some more. Since the previous layout seemed to not be preferable, I've gone back to basics.
I've chopped off the the extra decoupling stages and gone back to a plain old Rod Elliot style layout.
This is pretty darn similar to Rod's layout however he biggest difference is the extra degen resistors, the larger input cap, larger diameter decoupling caps.
I understand that Q9 should be mounted as close as possible to the Tj of Q6 which I would take to mean the metal tab side of Q6.
However I of course want the metal tab side of Q6 on a heatsink.
Am I over thinking this, should I just be cable trying Q9 to the 'front' side of Q6 and forget about it?
