Floating Regulated PSU Question

[thermionic]

Hi,

I know this is an inefficient way to make a PSU, but I wonder what the potential benefits / disadvantages might be. Say you have a dual-rail PSU board, with separate vregs for each rail. What are the disadvantages (or advantages?) to using it as a single rail PSU by only using +ve and -ve, with the Ground rail floating? Obviously, one can connect either rail to chassis as 'Gnd' if the PSU is floating.

One disadvantage is that any reservoir capacitance, be it pre or post vreg, will be halved because the caps are effectively now in series. If the capacitance is halved, am I correct to assume the output impedance of the PSU is made higher? Empirical experiments, watching the DC waveform under load, suggest that it can have less triangular corruption with the twin-regulator approach than if the load were being driven purely by one vreg. If capacitance is decreased, hence impedance made higher, how can this be?

Aside from being wasteful, where are the disadvantages to the approach?

Thanks in advance.
Justin

 

[abbey road d enfer]

One disadvantage is that any reservoir capacitance, be it pre or post vreg, will be halved because the caps are effectively now in series. If the capacitance is halved, am I correct to assume the output impedance of the PSU is made higher?

This is correct. Being in series, the voltages will add, as well as the impedances.

Empirical experiments, watching the DC waveform under load, suggest that it can have less triangular corruption with the twin-regulator approach than if the load were being driven purely by one vreg.

This is not supported by any theoretical analysis. I can only imagine data corruption due to an error in the modus operandi.

Aside from being wasteful, where are the disadvantages to the approach?

There is no particular disadvantage other than the added complexity. In some cases, this could be a more cost effective solution; let's say you want to build a 48V 5Amp+ PSU. There is no 3pin reg that can do the job safely, so you would need a discrete approach, which is generally more complex and more difficult to tune. You may then want to use two 24V PSU's, each one using a TO3 3pin reg.

 

[thermionic]

This is not supported by any theoretical analysis. I can only imagine data corruption due to an error in the modus operandi.

Maybe I wasn't comparing apples with apples. The comparison was into a 100mA load using identical PSUs @ 30v. I found that the waveform looked much cleaner when using the LM317/337 together, than an LM317HVT on its own. I've never been able to get ripple as low with the HV version as I have with the standard 317/337 vregs, so maybe this was the issue? The fact that the HVT has never looked as clean under load on my projects as the 317/337 combo was what prompted me to ask the intial question.

Thanks.

Justin

 

[abbey road d enfer]

Nothing indicates that the HVT is different of the normal version. But it is very difficult to compare two different implementations of the same basic design. The datasheet attracts the attention of designers on some aspects of the layout that can lead to degradation of performance.