Unregulated PSU are simple and they work, but I would like to understand more. I read here and there and it looks like different people have very different approaches. I understand that a compromise is needed, and I seek your help with the calculations to find the best one.
From what I understood the optimal PSU should have Q=0.5 and Z as low as possible. First question: Z is a function of frequency. Should I try to minimize Z in all the audio range ?
This is wonderfully simple, but if I try to decrease Q, I increase Z.
Let's start with a CLC filter, tube rectified. the goal is to have 400 V @ 200mA.
Here are my problems: in order to decrease Q I can:
1) increase the DCR of L, but like this I do not have a good regulation
2) increase L, but there are not many inductors more than 10-12H for this current.
3) increase the last C
4) add RC in parallel to the filter cap for damping. If I add this I do not know how to calculate Q. Additionally I found formulas to calculate good damping. and the first requirement is that the damping cap should be 16 times the value of the filter cap. Now I have 200 uF filter cap 500 V. If I have to multiply by 16 if will get VERY big and expensive.
To decrease Z I have to:
1) choose low DCR transformer, but the rectifier does not like it ! Additionally this will cause ringing and spikes depending on the reservoir cap value.
2) choose low DCR inductance.
I made some simulation also with LC filters, but a very high voltage is needed in input, and further filtering is needed to have decent ripple.
I could not find a document where these topics are treated all together. some people focus on the ripple, other on the q and others on the Z. It looks like sleeping with a short blanket, you have either feet or head cold.
Thanks,
Davide
From what I understood the optimal PSU should have Q=0.5 and Z as low as possible. First question: Z is a function of frequency. Should I try to minimize Z in all the audio range ?
This is wonderfully simple, but if I try to decrease Q, I increase Z.
Let's start with a CLC filter, tube rectified. the goal is to have 400 V @ 200mA.
Here are my problems: in order to decrease Q I can:
1) increase the DCR of L, but like this I do not have a good regulation
2) increase L, but there are not many inductors more than 10-12H for this current.
3) increase the last C
4) add RC in parallel to the filter cap for damping. If I add this I do not know how to calculate Q. Additionally I found formulas to calculate good damping. and the first requirement is that the damping cap should be 16 times the value of the filter cap. Now I have 200 uF filter cap 500 V. If I have to multiply by 16 if will get VERY big and expensive.
To decrease Z I have to:
1) choose low DCR transformer, but the rectifier does not like it ! Additionally this will cause ringing and spikes depending on the reservoir cap value.
2) choose low DCR inductance.
I made some simulation also with LC filters, but a very high voltage is needed in input, and further filtering is needed to have decent ripple.
I could not find a document where these topics are treated all together. some people focus on the ripple, other on the q and others on the Z. It looks like sleeping with a short blanket, you have either feet or head cold.
Thanks,
Davide
Davide,
Your analogy is correct, real life gets in the way of ideal. I like your words regading the short blanket syndrom. IIRC, the ideal PS fiilter is that of a butterworth.
The requirements of the circuit being supplied will dictate how far one must go to have optimum performance. Optimum almost always falls short of ideal. For class A the variation in current draw is small, but PS noise rejection is typically poor so most tend to optimize for the lowest ripple practical at the expense of Z. (again $$ rule, to have lower Z, the components tend to become heftier and more expensive)
For Class AB, current swings quite a bit, here many choose for lower Z and will let ripple float. This is really apparent in SS.
In your example above, the tube rectifier introduces many of the limitations to the PS design. If a CLC type you may even have to introduce series resistors prior to the rectifier tube to keep it alive. The rectifier tube itself suffers from a "high" impedance. The LC is better in reducing the load on the rectifier, but as you pointed out will require a higher voltage ($$) and a high quality choke ($$). An LC filter within reason, also has better voltage regulation.
There is some good reading to download on PMillet's site that goes into even better detail.
Your analogy is correct, real life gets in the way of ideal. I like your words regading the short blanket syndrom. IIRC, the ideal PS fiilter is that of a butterworth.
The requirements of the circuit being supplied will dictate how far one must go to have optimum performance. Optimum almost always falls short of ideal. For class A the variation in current draw is small, but PS noise rejection is typically poor so most tend to optimize for the lowest ripple practical at the expense of Z. (again $$ rule, to have lower Z, the components tend to become heftier and more expensive)
For Class AB, current swings quite a bit, here many choose for lower Z and will let ripple float. This is really apparent in SS.
In your example above, the tube rectifier introduces many of the limitations to the PS design. If a CLC type you may even have to introduce series resistors prior to the rectifier tube to keep it alive. The rectifier tube itself suffers from a "high" impedance. The LC is better in reducing the load on the rectifier, but as you pointed out will require a higher voltage ($$) and a high quality choke ($$). An LC filter within reason, also has better voltage regulation.
There is some good reading to download on PMillet's site that goes into even better detail.
Which book are you referring to ? I downloaded and had a look to almost everything in Millett web site.
Thanks,
Davide
Thanks,
Davide
Z is a function of frequency.
I really worry about Z only in the 40Hz to 100KHz range.
A reasonable sized inductor will look very high Z at any reasonable frequency.
So the transformer, diodes, first cap and DCR of the inductors seem to be at best second order issues.
The ESR of the last capacitor is crucial. This single parameter controls the Z at audio frequencies. Above that, the inductance may take over, which is why bypass caps are sometimes used.
HTH
Doug
I really worry about Z only in the 40Hz to 100KHz range.
A reasonable sized inductor will look very high Z at any reasonable frequency.
So the transformer, diodes, first cap and DCR of the inductors seem to be at best second order issues.
The ESR of the last capacitor is crucial. This single parameter controls the Z at audio frequencies. Above that, the inductance may take over, which is why bypass caps are sometimes used.
HTH
Doug
I am using film capacitors (shizuki) as last capacitors. On paper their ESR is close to zero.
What about damping ? What is the effect of a parallel RC damping on Z ?
From the load point of view, this should be seen as an impedence in parallel with the last cap, so decreasing the overall Z. Is it correct ?
Thanks,
Davide
What about damping ? What is the effect of a parallel RC damping on Z ?
From the load point of view, this should be seen as an impedence in parallel with the last cap, so decreasing the overall Z. Is it correct ?
Thanks,
Davide
- Status
- Not open for further replies.