Bob Cordell said:
Bob,
I agree with your reasoning. What am not sure about is how to apportion the capacitance between the reservoir caps and the final caps and the resistance between them. Unless you get it just right, the load-induced ripple on the final caps can be higher then it would be if you had just a single big cap, because of the limited possibility of the reservoir cap to come to the timely rescue of the final cap, as it were.
Is there some rule of thumb for this that you know of? Probably depends on load impedance?
Jan Didden
Hi,
The RCRC style of smoothing in a PSU seems very attractive.
I think the priority must be to ensure the first C can cope with the ripple in that first stage of smoothing.
The second priority is ensuring the second stage can cope with the output peak current demand.
The third priority is then the quality of the second C and whether it will/can respond to tweaking (added //C and/or RC snubbers).
Finally, the total C times the Zout >=160mS for good bass.
I have a suspicion that when all the above are attended to adequately that the first C will be about half the second C and that the second C will be about +-2mF/Apk to +-3mF/Apk of output (into your chosen/design load impedance). i.e 1mF + 2mF/Apk
The RCRC style of smoothing in a PSU seems very attractive.
I think the priority must be to ensure the first C can cope with the ripple in that first stage of smoothing.
The second priority is ensuring the second stage can cope with the output peak current demand.
The third priority is then the quality of the second C and whether it will/can respond to tweaking (added //C and/or RC snubbers).
Finally, the total C times the Zout >=160mS for good bass.
I have a suspicion that when all the above are attended to adequately that the first C will be about half the second C and that the second C will be about +-2mF/Apk to +-3mF/Apk of output (into your chosen/design load impedance). i.e 1mF + 2mF/Apk
AndrewT said:
I tend to agree. The only caveat is that there tends to be a physical size limitation on the electrolytic that is placed close to the output transistors. The bigger the better, but 1000 uF with very good ESR is in the neighborhood. The biggest job of the local capacitor is to force the half-wave Class-AB currents to circulate and get resolved locally, and 1000 uF in parallel with some smaller capacitance should do that job quite well. The really long-time-constant power reserve for bass must come from the main reservoir capacitors which will be in the range of 10,000 to 100,000 uF per rail.
Bob
Has anyone talked about a seperate higher voltage supply to the front end?
In A/B amps this can be quite important, particualar in high power designs that are going to be drive hard into a low impediance.
On the other hand designs with high PSSR and high negative feedback may be less effected by PS anomalies.
Another issue is audible noise. What is the point of a well filtered PS when the background noise comes from the transformer itself and that low distortion design? The transformer can be designed for low audible noise and with toroidals this is very important.
Doug Self has some great guidelines on PS earthing
Ian
In A/B amps this can be quite important, particualar in high power designs that are going to be drive hard into a low impediance.
On the other hand designs with high PSSR and high negative feedback may be less effected by PS anomalies.
Another issue is audible noise. What is the point of a well filtered PS when the background noise comes from the transformer itself and that low distortion design? The transformer can be designed for low audible noise and with toroidals this is very important.
Doug Self has some great guidelines on PS earthing
Ian
pooge said:
The X capacitor is just an additional, more direct path from the positive supply rail to the negative supply rail. At frequencies where this path has zero impedance (of course, in reality, we never quite get there), the Class-AB half-wave currents developed by the upper and lower output transistors will theoretically sum perfectly back to a linear current representation of the current going into the speaker. In that case, those highly non-linear currents would not flow in the wiring back to the main reservoir capacitors and would not radiate nonlinear magnetic fields, and these nonlinear currents would also not get partly dumped into the ground line.
Bob
macka said:Has anyone talked about a seperate higher voltage supply to the front end?
In A/B amps this can be quite important, particualar in high power designs that are going to be drive hard into a low impediance.
On the other hand designs with high PSSR and high negative feedback may be less effected by PS anomalies.
Another issue is audible noise. What is the point of a well filtered PS when the background noise comes from the transformer itself and that low distortion design? The transformer can be designed for low audible noise and with toroidals this is very important.
Doug Self has some great guidelines on PS earthing
Ian
Yes, I virtually always use boosted supplies for the input stage and VAS stage. This is especially important in MOSFET designs where you need substantial forward gate drive voltage; if you don't do this, you waste headroom of the main high current supply and dissipate more power than you need to for a given output power capability. The increased headroom also make the VAS work better and allows headroom for cascoding, etc. It also make it easier in some cases to keep the VAS out of saturation in clipping situations, greatly mitigating or even eliminating sticking. For that reason, I even recommend it for amplifiers with bipolar output stages. Of course, the boosted supplies add cost, but the key is that their current requirements are fairly low.
Bob
Bob Cordell said:
Simply put, the filters are connected between the outputs of the bridge and the transformer ground. The supply outputs are taken from the rectifiers and the speaker ground is returned to the transformer ground as well. (There's a bigger picture here but this hopefully gets the essence)
This arrangement keeps the charging/discharging currents confined to the filter loop while allowing the power to be drawn from the rectifiers. It requires short transformer leads to work best.
The problem with distributed capacitance is the large loop required for the returns.
Regards, Mike.
MikeBettinger said:
Noting that the devil is in the details of course, I have had several cases of oscillatory tendencies with low-esr/esl local output stage supply bypassing; at the time I attributed it to large return loops. With distributed capacitance I find it difficult to select a good central return point. Probably the best solution is a compact construction where the whole supply shebang is very close to the output stage.
Jan Didden
janneman said:
Bob Cordell said:
My take is that there is one central ground and it is at the centertap of the transformer. This is where all returns need to return independently to. As Bob indicated, local loops need to be implemented to minimize the loop area at the point where the work is being done and a lo-z return provided from the central ground.
The input stage (signal and feedback) is using ground as a reference and anything connected to it needs to have essentually the same potential, electrically.
A satellite supply regulator needs it own loop, once again, tight at the point where the work is being done, and the return needs to be part of the loop that supplys it's power; it's output return needs to be thought of separately.
I see this as more of a tree structure than as a star.
It all works if you split ground up into more defined discriptions based on it's function: Power returns, signal return, signal reference, and bypass returns. Each loop needs to be dealt with independently and they should not interact.
To have this approach work requires, as Jan suggested, that the powers supply be located close to the output stage and that both the power and return loops be as compact as possible. Additionally the ground references need to be as clean as possible (which goes back to layout)
Gotta go, Thanksgiving responsibilities are looming over me.
Regards, Mike.
Hi,
I think you should close the rectifier and primary smoothing cap loop to the centre tap. That keeps the charging pulses in the "local loop". Then take a lead from there to the next cleaner stage:- the local loop enclosing the power users (amp PCB),decoupling and Zobel/Thiel network. Finally the local loop enclosing all the signal level returns, signal at PCB, input RCA, NFB.
But I have found that once all the local loops are closed one can directly couple them together (the audio ground) without further deterioration in the quality of the reference. At least to the level that my meagre resources can measure and hear.
The order of connecting them together is important and MUST preserve each local loop.
This brings me back to the first statement.
Local loop for the charging pulses is closed then couple the next cleanest loop directly to it. That is almost the same as making the centre tap coincide with the audio ground BUT subtley different.
I don't agree,MikeBettinger said:
I think you should close the rectifier and primary smoothing cap loop to the centre tap. That keeps the charging pulses in the "local loop". Then take a lead from there to the next cleaner stage:- the local loop enclosing the power users (amp PCB),decoupling and Zobel/Thiel network. Finally the local loop enclosing all the signal level returns, signal at PCB, input RCA, NFB.
I have been preaching this one for a while. Call them by their function and treat them separately, NOT as a common ground.MikeBettinger said:
Again I agree.MikeBettinger said:
But I have found that once all the local loops are closed one can directly couple them together (the audio ground) without further deterioration in the quality of the reference. At least to the level that my meagre resources can measure and hear.
The order of connecting them together is important and MUST preserve each local loop.
This brings me back to the first statement.
Local loop for the charging pulses is closed then couple the next cleanest loop directly to it. That is almost the same as making the centre tap coincide with the audio ground BUT subtley different.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.