I do not quite understand your argument, but I will pick up on one phrase and expand on it.Christer said:
"no current returns to the center tap"
this would imply that there would be no voltage difference in the wires at the centre tap. If there is no current and no voltage difference then it would be the ideal place for a zero voltage reference.
However, some current does indeed flow in the two wires that form the centre tap and from/to the centre tap when the amplifier is generating output. There is also a current flow to the centre tap when ever the current in the +ve and -ve supply rails is NOT IDENTICAL. This does occur when the amp draws unequal current during the quiescent (idle) period and when current flows to the load.
Mull over that and come back.
Bob Cordell said:
Bob,
I am intrigued by this rail-to-rail cap, and have tried to understand what it does, and if it does anything. My first reaction was that it should hardly make any difference, since only one half of the output stage conducts at a time and thus there are no return currents directly between the rails (except for the small cross-conduction currents around zero).
However, Spice seemed to think some current should flow through it
, at least if adding some considerable resistance to the PSU leads. I thus tried to figure out where the currents flow, and what happens. It seems to me (based on some sim experiments only) that what the rail-to-rail cap does is to provide a path so both rail-to-ground caps appear in parallel. That is, during a positive half cycle, the upper transistor draws current both from the positive rail-to-ground cap, and from the negative rail-to-ground cap, by discharging the latter "backwards" through the rail-to-rail cap. For small PSU wire resistances there seems to be little difference, but for somewhat higher values it seems to have an effect to lower the magnitude of the currents through the rail PSU wires and also slighty decrease the ripple at the transistors. However, the rail-to-rail cap should then ideally be quite larger than the other caps, since it appears in series with the cap it helps to discharge.It also seems that the cap slightly decreases the harmonics of the PSU wire currents. I do not quite understand why that happens, but as far as I can see, it probably must be some synergy effect between the rails when trying to always utilise both rails in this way.
Does this correlate with your wiev on what goes on when adding such a capacitor, or is there some other way to understand it?
AndrewT said:
No, they are the rectifiers. They are reverse biased whenever the (typ.) 100000 uF capacitor has a larger voltage than the transformer. This is the case for most of the sine wave. In fact, if the capacitor does not loose charge (i.e. unloaded), there will be no current through the diodes after the initial surge at power_on.
Gerhard
Hi,
I have finally bitten the bullet and drawn a diagram. Please excuse the quality, MS Paint & me don't get on too well.
Although the diode symbols are missing you all know where they are.
Follow the number sequence. That is the route the charging pulses take when current flows in one direction. When the phase changes to the opposite half waveform the flow reverses but takes the other route through the diodes 15-14-13-4-5-6-7-8-9-10-11-12-3-2-1-16-15. ( I have called this the figure of 8 route in the past)
Notice that the legs from 16 to A and from 8 to B do not feature in the charging route.
IF the current from the +ve pole and -ve pole is zero then NO CURRENT flows 8 to B and 16 to A. Similarly if the current flowing out of the +ve pole exactly matches the current arriving back at the -ve pole then again NO CURRENT flows from 16 to A and from 8 to B.
However, look at the short stretch between the two secondary windings. This is the centre tap.
The route 15-16-1 passes through the two wires forming the centre tap. Charging pulses pass through the centre tap even when NO CURRENT flows to the load.
Need I go further? Is it falling into place yet?
I have finally bitten the bullet and drawn a diagram. Please excuse the quality, MS Paint & me don't get on too well.
Although the diode symbols are missing you all know where they are.
Follow the number sequence. That is the route the charging pulses take when current flows in one direction. When the phase changes to the opposite half waveform the flow reverses but takes the other route through the diodes 15-14-13-4-5-6-7-8-9-10-11-12-3-2-1-16-15. ( I have called this the figure of 8 route in the past)
Notice that the legs from 16 to A and from 8 to B do not feature in the charging route.
IF the current from the +ve pole and -ve pole is zero then NO CURRENT flows 8 to B and 16 to A. Similarly if the current flowing out of the +ve pole exactly matches the current arriving back at the -ve pole then again NO CURRENT flows from 16 to A and from 8 to B.
However, look at the short stretch between the two secondary windings. This is the centre tap.
The route 15-16-1 passes through the two wires forming the centre tap. Charging pulses pass through the centre tap even when NO CURRENT flows to the load.
Need I go further? Is it falling into place yet?
Waiting for the diagram Andrew.
However, I think I your text is enough to tell me that we are talking about different things. Of course current flows into or out from the center tap during charge pulses. There is no discussion about that. The point is that current flows only during the charge pulses, why it cannot serve as a ground point for anything else but charge pulses.
However, I think I your text is enough to tell me that we are talking about different things. Of course current flows into or out from the center tap during charge pulses. There is no discussion about that. The point is that current flows only during the charge pulses, why it cannot serve as a ground point for anything else but charge pulses.
Hi Christer,
now that you have a diagram.
On it you will see the letters indicating the returns that need a common reference. Imagine all those little circles are solder tags. bolt them together to close off each of the local loops and where possible keep the connections short.
Now bolt each of the local loops together in the order from dirtiest (mains safety earth) through charging loop to power to signal. That is the simple and effective way to create a common reference and preserve the tight local loops.
As an aside, take the wires 1-2-3 and 15-14-13. couple these closely together or even twist them together and the electromagnetic fields due to the flow and return currents cancel leaving little radiation to affect other circuits.
Similarly the wires 4-5-6 and 12-11-10 can be twisted together. 6-5-to PCB and 10-11-to PCB along with the return from C1 &2, this triplet should be twisted together. That brings the audio ground to near 6 & 10.
Finally, each of these twisted sets should be kept short wherever possible and this fits with the recomendation that output stage should be close to the power supply.
Note also that it is no accident showing the smoothing cap common connection (B) right next to the centre tap connection (A). Keep the loops tight and short.
My last request would be "how do we layout amplfier PCBs" to minimise the radiated energy that comes from isolated +ve or -ve or return traces spread eagled across the extremities of the PCB? Pity we have all the most sensitive circuits buried inside this mishmash of radiating traces.
now that you have a diagram.
On it you will see the letters indicating the returns that need a common reference. Imagine all those little circles are solder tags. bolt them together to close off each of the local loops and where possible keep the connections short.
Now bolt each of the local loops together in the order from dirtiest (mains safety earth) through charging loop to power to signal. That is the simple and effective way to create a common reference and preserve the tight local loops.
As an aside, take the wires 1-2-3 and 15-14-13. couple these closely together or even twist them together and the electromagnetic fields due to the flow and return currents cancel leaving little radiation to affect other circuits.
Similarly the wires 4-5-6 and 12-11-10 can be twisted together. 6-5-to PCB and 10-11-to PCB along with the return from C1 &2, this triplet should be twisted together. That brings the audio ground to near 6 & 10.
Finally, each of these twisted sets should be kept short wherever possible and this fits with the recomendation that output stage should be close to the power supply.
Note also that it is no accident showing the smoothing cap common connection (B) right next to the centre tap connection (A). Keep the loops tight and short.
My last request would be "how do we layout amplfier PCBs" to minimise the radiated energy that comes from isolated +ve or -ve or return traces spread eagled across the extremities of the PCB? Pity we have all the most sensitive circuits buried inside this mishmash of radiating traces.
Andrew and Mike Bettinger,
Hm, I think we are still discussing somewhat different issues, and maybe there is some misunderstanding from the start, about what Mike Bettinger really meant. My point is that, yes, of course it is possible to use the center tap as a star ground. Since the midpoint of the caps will also be connected there, any return current goes to the center tap and then either into the windings or directly on to the midpoint of the caps (depending on whether and how much the caps are charging). However, since the current loops through the windings are broken most of the time (by the diodes) the return current will most of the time just continue to the caps. Hence, I don't understand the argument that the center tap should be the ground points since all currents return there anyway. Most of the time the currents return there only if we have already choosen the center tap to be the the ground point, which makes the argument circular.
Hm, I think we are still discussing somewhat different issues, and maybe there is some misunderstanding from the start, about what Mike Bettinger really meant. My point is that, yes, of course it is possible to use the center tap as a star ground. Since the midpoint of the caps will also be connected there, any return current goes to the center tap and then either into the windings or directly on to the midpoint of the caps (depending on whether and how much the caps are charging). However, since the current loops through the windings are broken most of the time (by the diodes) the return current will most of the time just continue to the caps. Hence, I don't understand the argument that the center tap should be the ground points since all currents return there anyway. Most of the time the currents return there only if we have already choosen the center tap to be the the ground point, which makes the argument circular.
Now I am starting like mikeks, answering my own questions.
Maybe this is not so strange after all. Since the rail-to-rail cap achieves that both rail-to-ground caps supply current (to some extent) irrespective of which half of the OPS is supplying load current, the current supplied by these are likely to vary somewhat less over time, and similarly the charge currents from the PSU (and the main caps) will then also smooth out over time.
Christer said:
Maybe this is not so strange after all. Since the rail-to-rail cap achieves that both rail-to-ground caps supply current (to some extent) irrespective of which half of the OPS is supplying load current, the current supplied by these are likely to vary somewhat less over time, and similarly the charge currents from the PSU (and the main caps) will then also smooth out over time.
Christer said:Now I am starting like mikeks, answering my own questions.
Actually, i wind up answering my own questions after my own research efforts because folks tend to ignore them.
For instance, all of my questions in this thread were ignored!
AndrewT said:
A picture is worth a thousand words. Now I believe I know why we have not been on the same page as to the layout and implementation of ground. (this isn't the first try Andrew and I have had at this).
I'm to tired right now and want to re-read this again in the morning, but what is happening on the centertap terminal is the key worth examining.
One other item is that for power amps I don't use diodes or dual bridges in the ground return path. Just a simple bridge with the returns directly connected to the centertap. I see no purpose to placing anything especially a diode in this path. Although I have raised this question, I have not received any clear indication as to why one would chose to-do or not to-do this.
Regards, Mike.
MikeBettinger said:
I fully agree here!
* I have hard time to understand why one wants to "chop up" the earth path with 2 rectifiers, the earth should be a quiet point.
* We also lose another +&- 0,7 Volts of rail efficiency.
* We have doubled up the noise sources, eg. diodes switching on and off, and not necessary at the same time in both rectifiers due to imbalances.
I would whish the double rectifier sheme could be burried ones and for all.
Cheers Michael
By keeping the ground-traces, +ve and -ve close together.
In the few designs i've made they always end up at the center of the pcb, not at the extremities.
This also offers the advantage of having short connections for local smoothing/decoupling caps.
In practice , most of the times this means you have to turn the DRAWN schematic inside-out, not simply "copy" it to the pcb.
Klaas
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.