Help me make this psu cheaper

[rotation]

Hi guys,

i'm happily using this psu for every tube project i build, it works very well. But i was wondering if i could use smaller caps than suggested (400u ). I thought about using something like 100-220u for the first two filter caps, 10-22u for cap that is in parallel with zener, and 100-220u cap at the output.
This looks like a lot of capacitance for regulated psu and well designed preamp or whatever, and this big caps take lots of space...

My plan is:
-1st cap would be around 100-220u because i calculated ripple attenuation for both and difference is not very big. I also have two more filtering caps..
-2nd cap would be much smaller, about 10-22u. I'm not sure if i really need it if i have big caps at the input and output.
-3rd cap would have 10-22u. I know this cap takes away some of zener noise and i think it acts as a capacitance multiplier. If so, how much is beta of TIP50 at this condition, i would like to calculate by how much is capacitance multiplied (if it is at all)?
-4th cap, i'm not sure if i need big cap here. If i'm correct about cap. multiplier and if TIP50 beta is more than 10, i think i don't even need it. Should i leave it out or put in?

What do you think about my plan? I don't need anything crazy here, just a simple, small and cheap regulated psu.

Miha

 

[tv]

If you used an IRF840 instead of the TIP50, you could scale the 3rd cap and the 10K resistor and thus even magnify the "capacitance multiplication".

If you want to avoid zener noise, you could make a 2nd order "lpf", i.e. include another RC cell (after the first R-zener-C) to gate. This could be useful for preamps, but the exact values should be determined in advance. (obviously this wouldn't be cheaper)

 

[JohnRoberts]

Using a darlington in place of the TIP 50 pass transistor (if available and cheaper than parts savings elsewhere) would need less base current and lower power components in zener and base feed resistor. Also you could use a smaller cap across zener for same ripple spec.

JR

 

[PRR]

> just a simple, small and cheap regulated psu.

SPECS!!

Volts, Amps, Ripple.... 

> regulated

WHY?? Tubes run fine on unregulated B+.

DEFINE your needs. I will guess 5mA-15mA at 250V-300V, with ripple on the first stage B+ node to be lower than tube self-hiss at the plate (say 0.1mV).

A rule of thumb for the first filter cap is 1uFd per mA. Older designs used less, but caps are cheap today. But even when cheap, 200uFd is a BIG cap for a 10mA load.

1uFd per mA will give around 1V of ripple. Going "20 times bigger" with 220uFd is folly: parasitic resistance will not give you 20X lower ripple. Get steady DC with a few volts ripple, then filter it.

While 1uFd per mA for 5mA-15mA says 5uFd-15uFd, it may be convenient to wait until you find the other caps and then use the same size for the first cap.

Note that the negative end of the rectifier (in this case, the PT CT) MUST go directly to this first filter cap. Do NOT run this garbage through the chassis as in your drawing.

Start with "too much" voltage. 20%-40% excess. Your 350V to 300V-250V drop is a good plan.

We "use" the excess DC voltage in an R-C-R-C... filter, to reduce ripple a lot with tolerable DC loss.

Knowing your current (??), figure how much resistance will drop 5%-10% voltage. Use nearest standard value Power resistor. (The running dissipation may be low, but the start-up surge is brutal.) 10mA will drop 20V in 2K resistor.

Use standard R-C formula to find a capacitor which will work against this resistor at a frequency 50-100 times lower than your ripple frequency. Ripple is usually 100/120Hz, so aim at 2Hz-1Hz. 2hz at 2K is 39.8uFd, use 40uFd.

100Hz/2Hz is a factor of 50, so the ripple is reduced by 50 in each R-C stage. We started from around 1V ripple. We want some power at 0.1mV ripple, 10,000 times lower. Two stages 50*50 is 2,500, not enough for the first stage (but likely good enough for other stages). We need 3 sections. 50*50*50 is 125,000 of ripple reduction, apparently 0.01mV ripple, better than we need. We should not count on full theoretical ripple, but still we might try smaller caps to save money/space.

Since amplifier stages after the first stage have higher signal and do not need such low ripple (if there is a ripple problem, it is usually that first stage), we can tap-off power for high-level stages before the last filter stage. Now the last resistor carries less current and may be scaled larger with reasonable (5%-20%) voltage drop, and correspondingly better filtering.

This plan is time-tested. It can be made nearly short-proof, a great benefit in DIY. As shown, a short probably sags your 350V to 300V, each resistor gets 100V, 5W parts will survive many hours under dead-short, and if you do leave it shorted all night they are not expensive to replace.

 

[strangeandbouncy]

Hi PRR,


    many thanks for this! May I congratulate you for answering my question BEFORE I even got round to asking it?  ;D


    As ever, we are SO in your debt . .



      Kindest regards,


        ANdyP

 

[rotation]

> just a simple, small and cheap regulated psu.

SPECS!!

Volts, Amps, Ripple.... 

> regulated

WHY?? Tubes run fine on unregulated B+.

DEFINE your needs. I will guess 5mA-15mA at 250V-300V, with ripple on the first stage B+ node to be lower than tube self-hiss at the plate (say 0.1mV).

A rule of thumb for the first filter cap is 1uFd per mA. Older designs used less, but caps are cheap today. But even when cheap, 200uFd is a BIG cap for a 10mA load.

1uFd per mA will give around 1V of ripple. Going "20 times bigger" with 220uFd is folly: parasitic resistance will not give you 20X lower ripple. Get steady DC with a few volts ripple, then filter it.

While 1uFd per mA for 5mA-15mA says 5uFd-15uFd, it may be convenient to wait until you find the other caps and then use the same size for the first cap.

Note that the negative end of the rectifier (in this case, the PT CT) MUST go directly to this first filter cap. Do NOT run this garbage through the chassis as in your drawing.

Start with "too much" voltage. 20%-40% excess. Your 350V to 300V-250V drop is a good plan.

We "use" the excess DC voltage in an R-C-R-C... filter, to reduce ripple a lot with tolerable DC loss.

Knowing your current (??), figure how much resistance will drop 5%-10% voltage. Use nearest standard value Power resistor. (The running dissipation may be low, but the start-up surge is brutal.) 10mA will drop 20V in 2K resistor.

Use standard R-C formula to find a capacitor which will work against this resistor at a frequency 50-100 times lower than your ripple frequency. Ripple is usually 100/120Hz, so aim at 2Hz-1Hz. 2hz at 2K is 39.8uFd, use 40uFd.

100Hz/2Hz is a factor of 50, so the ripple is reduced by 50 in each R-C stage. We started from around 1V ripple. We want some power at 0.1mV ripple, 10,000 times lower. Two stages 50*50 is 2,500, not enough for the first stage (but likely good enough for other stages). We need 3 sections. 50*50*50 is 125,000 of ripple reduction, apparently 0.01mV ripple, better than we need. We should not count on full theoretical ripple, but still we might try smaller caps to save money/space.

Since amplifier stages after the first stage have higher signal and do not need such low ripple (if there is a ripple problem, it is usually that first stage), we can tap-off power for high-level stages before the last filter stage. Now the last resistor carries less current and may be scaled larger with reasonable (5%-20%) voltage drop, and correspondingly better filtering.

This plan is time-tested. It can be made nearly short-proof, a great benefit in DIY. As shown, a short probably sags your 350V to 300V, each resistor gets 100V, 5W parts will survive many hours under dead-short, and if you do leave it shorted all night they are not expensive to replace.

Hi PRR,

this is excellent, thanks a lot!
I didn't give my current draw because it's different for different projects, normally from 10-20mA at 250-310V. I used regulated psu because it SOUNDED better than the old unregulated psu. The problem were back to back transformers (as on G9), one of them was overheating and i could hear in sound of preamp that something was wrong.

Btw, when i'm building HT, heater and phantom supplies on one board, do i have to ground heater with the other two or leave it floated? I had big hum in my preamp when regulated DC heater was not grounded (someone told me to do so), and when i grounded it to the last filter cap of HT hum completely dissapeared.
Another question about grounding; my input transformer has two shields, one is between the windings and one is case. Output transformer only have wire for grounding it's case. In a project i built i connected this shields directly to circuit ground (input tx near input, output near output) and it didn't hum. Now i'm wondering if i didn't create ground loop, my circuit is now grounded at three points, no?

Miha

 

[tv]

>>> because it SOUNDED better
o rly. this happens with 9-volt circuits, too.

why don't you just use a mosfet? it's only marginally more expensive than tip50.

then, you could get by with 2W or even 1W resistors.
you could loose the CRC before the transistor and use only a single C there and "transfer" the big caps into 2nd order LPF in fet's gate circuit, like I said.

Better regulation (no bjt base-dip when overloaded), better ripple rejection, cheaper/smaller caps and resistors, better everything...

You can "practice" the circuit on 15volts with a BS170...

 

[rotation]

>>> because it SOUNDED better
o rly. this happens with 9-volt circuits, too.

I'm quite sure that unregulated psu had bad sound because one of two back to back transformers was to small, it was overheating and buzzing. This one was powering HT..
This question comes up often, but i never really tested if there is difference in sound between proper reg. and unreg. psus.

You are right, IRF840 costs only 30 cents more than TIP50. Can you explain what i should do to build it or have schematic? When i'm buying material for PRR's advice i can build yours too and test all three.

Miha

 

[tv]

A "generalized" circuit is in atachement.
You will have to calculate the resistor/cap values by yourself.
Each R/C section (there are two of them - R1C2, R2C3) should have "corner" frequency of cca 1-2Hz. The "gate stopper" could be 330R-1K. Add the gate protection zener as well.

You will have to calculate the first RC section with the Zener current in mind, i.e. the R1 will have to pass the proper current to the HV zener. R1 will have to be a 2W type.

In theory, if you use 56K resistors, you could use 2,2uF for the C3 and 22uF for C2 (or possibly C4). C1 and C4 you will have to come up with your values.

here is another example:
http://www.solderingpoint.com/projects/mosfetps/mosfetps.php

 

[tv]

Hey, what happened?