Bob Cordell Interview: Power Supplies

[KSTR]

Hi James,

this LC-"snubber" is a bit tricky and has only little figure of merit in simple circuits, if any.

It must be very carefully tuned (WRT frequency, Q and impedance), it then effectively shifts ripple to higher harmonics and keeps peak-to-peak ripple voltage lower, compared to a circuit with the same overall filter capacitance but without the inductor (which carries no DC, a nice feature). And it can be cascaded to do even more so.

All of which is of no worth (or even a bad thing) when there is no post filter with a steep slope, because power supply ripple rejection of amps tends to decrease with rising frequency, usually with a first order slope. The post filter might be passive (LC) or an buffered filter (source or emitter follower), with something like a 3rd order Bessel characteristic.

This make it obvious that only few circuits would see a net improvement with this, given the overhead. That's why you won't see it often in the industry, it has too many disadvantages to be considered universally useful.

Klaus

[jameshillj]

Thanks Klaus - interesting.

[Pocoyo]

Hiii

I am newbee in DIYAudio (1 year )
little confuse about grounding

My amp system

---- + -----I---I --- (+,0) -- I-----I --- + I-------- I
---- G? . . . I T I . . . (G?) -- I PSU I --- - I Channel I
---- - ----- I__I --- ( -,0) -- I____I --- G I________I



Maybe i need a simple direction from you
is PSU need grounding from cable ?

If i look some PSU their connect just from transformer
(+, 0, 0 and - ) but if we use cable and plug theres
3 way cable +, 0 and ground

Is this ground need to connect ? (From Cable to PSU)
If it is connect must i add a resistor in there ?

I use 6pcs 10000uf 80V and bridge rectrifier 32a 200v
for 45VDC

really need your help
regards, jeffry

[gootee]

Hi Jeffry,

Sorry. You have not provided enough information.

There are many different types of power supplies. Each type of power supply might connect to the transformer in a different way than some other types connect to it.

Below is a link to a web page that has diagrams of how some types of power supplies and transformers are wired:

http://sound.westhost.com/power-supplies.htm

From your words and diagrams, it looks like you are talking about the "center tap" wire, on the 'secondary' side of the transformer. And you are naming that wire "Ground".

Often the center tap IS used as a 'ground reference' voltage for a PSU [which might be typical for some types of dual (i.e. +/- voltages) power supplies]. But sometimes the center tap is not used at all, or is used differently.

If your PSU makes two equal and opposite + and - DC voltages, then your center-tap wire 'probably' IS used as the PSU's ground reference voltage. But where and how that center-tap wire connects would depend on how you want to design your grounding scheme. (Maybe you should do some searches for 'star ground'.)

The transformer's 'center tap' is not really a 'ground'. It is just the 'middle' voltage, between the voltages of the two outer secondary windings' wires.

For example (NOT for YOUR power supply; only an example), you 'could' pick ANY one of the secondary wires, and call that wire the "ground". You COULD even CONNECT any one of the secondary wires to the real 'earth ground'. ONLY the voltage DIFFERENCE between the wires is important. Example: a "25-0-25" transformer could ALSO be used as a 0-25-50 transformer.

[Pocoyo]

thank you goote

[djmiddelkoop]

In electronics school I learned the equivalent circuit for an active device, like a transistor.
In this equivalent circuit the feeding power supply rail was always considered a shortcut for signals, hence the signal the transistor ( or any active device) handles goes through the supply to ground, i.e. the supply is equal to ground for signals.
I hope I can explain this clearly, but since than I realised that for any kind of stage to work properly, the supply has to be a shortcut for any signal.
So, like mentioned before, the VAS of any amp needs it’s own very well decoupled powerlines, or its powerline currents will interact with the larger currents from the output stage when powerlines are shared.
Same applies for the output stage, signal currents in de powerlines will interfear with the charging current peaks of the (also not so perfect) electrolytic capacitors.
Like said before, splitting these capacitors (CRC) and their grounding points will make things a lot better: the active stage sees a better shortcut to ground for signals, and the C connected to the rectifier handles the charging cycles with less interference with the signals through the other C.

Hence we have to realize that the signals of each active stage has to travel through the powersupply with the smallest resistance possible and without interfering each other.
This also explains why the quality of the used capacitors is more important than the value.

Dick

[GK]

Here is a basic simulation schematic of the regulated +/-45V power supply I have designed for the power output stage of one of the three monster amps I am currently working on.
The amplifier is question is rated at 800Wrms into 1 ohm (40V peak) with a bias current of 5A for class A operation to 200Wrms into 4 ohms. Output current will be hard limited at 100A, providing full amplitude drive into a load resistance of 0.4 ohms.

Only the positive half of the regulator is simulated, the negative half being identical except that polarised components are inverted and the transistors are swapped for their N and P channel complements.

The design is unusual in that, while having a very high loop gain and therefore high (far better than required) DC precision, global negative feedback is not utilised to produce a low output impedance, as massive over compensation is used.

[GK]

The series pass element consists of twenty MJL21193 PNP transistors in a CFP arrangement with one MJL21194 NPN driver transistor.
The CFP provides a lower output impedance than a straight emitter follower due to the local NFB loop.
The 20 parallel series pass devices ensure that the output impedance does not rise significantly at high load currents due to beta droop.
Q17, Q18 and Q19 form a triple emitter follower driver for the CFP.
Transistors Q5 and Q6 form the error sensing long tail pair, comparing a fraction of the output voltage developed across R5 against the reference diode potential.
Q1 and Q2 form a current mirror load for the LTP and Q3 and Q4 form a buffered VAS, resembling a common audio amplifier circuit.
Q21 is a 1mA constant current load for the VAS. C1 is the frequency compensation capacitor and the VAS current is limited to 2mA by Q22.
This results in a maximum push-pull charge/discharge current for C1 of +/-1mA, giving the global loop regulator a response .001/1000e-6 = +/-1volt per second.

This response is so slow that it ensures the regulator loop does not “fight” the audio signal current to keep the output voltage constant and provides unconditional stability into the excess of 20,000 uF of filter capacitance on the output (2 X 10,000uF + 1000uF + 100uF + 10uF + 1uF + 0.1uF).

[GK]

Attached below is an output ripple simulation with a load current alternately switched between 100A and 5A at a rate of 1Hz. Due to the 1V/s time constant of the feedback loop, the regulator takes ~45 seconds to ramp up to the +45V regulated output voltage.

[GK]

Output ripple at 1kHz