Capacitor Selection Pi-Section Smoother.

[Takapuna]

My current project is an RIAA phono amplifier using valves(!).

HT source will be a reasonably standard Transformer-EZ80-C-L-C-L-C combination. I’m looking for c.25mA total.
Heater supplies for the signal valves will be DC from a different transformer for which I require c.2A.

When selecting capacitor values for either side of an inductor (or smoothing resistor) following the rectifier in these “standard” power supply configurations, more often than not the two or more capacitor values appear to be chosen equal in the examples I’ve seen.
I've done a number of internet searches trying to find some wisdom as to whether this is the best solution or not as I can see arguments for unbalancing capacitors one way or the other. I welcome any thoughts and guidance from other members who will have a much clearer view of this please.

As I understand it:
Making the input capacitor smaller reduces the peak currents for the rectifier and transformer, thus giving them an easier ride and dropping fewer volts across their internal resistances. It obviously leaves more ripple for the following LC or RC combination(s) to deal with but if the following capacitor is made large this can be dealt with.
Unbalancing the values in the other direction, making the input capacitor large and the output capacitor smaller has the advantage of further reducing voltage ripple straight away…as it were. This leaves less work for the following LC/RC combination. I appreciate this leads to high peak currents and can stress transformers rectifiers and the input capacitor if they are not rated for the ripple current.

Once past the first smoothing inductor or resistor the following stages are reasonably isolated and large capacitance values at this point are not such an issue.

I’ve used the PSU designer from Duncan’s Amps and it appears that for the low-current high-voltage supply a small input capacitor of, say 10uF, as opposed to the maximum allowed 50uF for the EZ80 leads to results which are every bit as good as with a 50uF input capacitor with the added bonus of reduced stresses and faster start-up time (not that the start-up time matters). OK, with a low current and so much smoothing results are always going to be good but the predicted resultant ripple is similar and negligible in both cases.

Running PSU designer for the 2A heater supply suggests that placing a big capacitor at the input to the smoothing section gives the better result, assuming ripple ratings permit. I’m not sure how smooth the heater supply needs to be, so it’s tempting to use capacitors which are likely to be thought of as “too-large” if such a thing exists. My examples were run with 47,000uF and 10,000uF which I interchanged.

My question boils down to: Is there any “right” or “wrong” way to select the capacitor values when using a Pi-Section for smoothing please?

Thanks.
Phil
-----

[G6Tanuki]

Pragmatically, the value of the capacitor-directly-connected-to-the-rectifier should be what the rectifier's manufacturer recommends [along with the minimum series-resistance, which will be made-up of the transformer's DC-winding resistance plus appropriate series-resistors].

Once you're downstream - when there's a few hundred Ohms of DC resistance and/or inductance separating the downstream C from the rectifier - then to be honest, the more the better.

The underlying thing is to protect the [valve] rectifier from having to handle excessive peak-currents which can damage the cathode's emissive surface. This is dealt with by limiting the peak current that can be delivered into the first-stage smoothing capacitor [by using a low value for this capacitor and/or series resistance]..

With proper design, "Start-up time" should be a couple of cycles of 50Hz. Remember also that the time taken for the filament to come up to full-emission provides a 'slow=start' effect.

Problems can sometimes occur if there's a brief power-outage - short-enough for the load-side capacitors to discharge but not short enough for the rectifier's cathode to cool. Then when power is restored you've got a fully-emissive cathode feeding into the empty capacitors.

That's when your series-resistors really show their worth.

[bikerhifinut]

CLCLC filtering? Blimey that's some amount of filtering.
EZ80 is fine with 50uF on the end of it so use something between32 and 50 uF.
I'll assume you're using something like a 10H choke.
After the choke you can use as large a capacitor as you fancy, there's no logic in your concept of "balancing the values each side of the choke.
I wouldn't personally go any higher than 200uF here and frankly 100uF will be plenty.
Subsequent to the first CLC filter, if you need to fine tune the HT volts then it makes sense to use a resistor to drop the volts followed by another capacitor to form another RC filter.
By now the ripple will be all but immeasurable.
And don't forget to decouple the HT at the actual preamplifier especially if the PSU is separate supplied via Umbilical. This will give another measure of RC filtering.
You mentioned the heater supply and as this is a phono preamp I would strongly suggest a regulated DC heater supply. A basic regulator using a 317 type adjustable reg (other varieties are out there) will be fine, or even the easy option of a 7805 with a couple of silicon diodes in series to earth from the bottom leg of the reg will get you 6.3V for heaters. Or other fixed voltage regs if using different heater voltage.
I honestly do not think you'll get any benefit whatsoever from the expensive use of twice as many chokes as needed.
If you are that concerned about your PSU and it's a good thing to start with a clean power supply, then instead of all this palaver I would just go for a properly regulated HT and LT using silicon regulators for a rock steady supply.
And you aren't going to like this I suspect, but why a valve rectifier? extra load on the transformer, Voltage drop through the valve and at the end of it all you have is a more complicated and heavier PSU. and with silicon diodes you can use a nice big reservoir capacitor.
I've been there.........................

I assume you are also using a properly thought out circuit for the actual RIAA stage itself.

I don't wish to sound negative, please don't take my comments the wrong way.

OOPs crossed with G6Tanuki

Andy.

[Sideband]

With regard to a DC heater supply, I used this on a valve headphone amplifier using a couple of ECC82's. The heaters of these two valves were connected in series so required 12.6V. I used nothing more complicated than a 7512 regulator with the earthy end connected to chassis via a 1N4007 diode (to raise the output of the 12V regulator to 12.6V relative to chassis) and using a 4,700uF capacitor for the main smoothing to the input of the regulator and a 100uF capacitor on the 12.6V output. No detectable hum on the output at all.

You could probably do the same thing with a 317 adjustable regulator and you wouldn't need the diodes in the earthy end.

[Max Power]

I would be inclined to use a Maida regulator on the HT rail, and then a 317 as suggested by Sideband for the heaters.

A lot cheaper than huge chokes !

[Sideband]

Originally Posted by Max Power

A lot cheaper than huge chokes !

Sometimes it's easy to 'overthink' these things. Huge chokes cost money and take up a lot of room whereas a couple of regulators are peanuts in comparison for this sort of job anyway.

I've been using my headphone amp for about 5 years with no problems. The heater regulator runs cool (admittedly using the metal chassis as a heatsink) and is built on a small piece of Veroboard. There is no measurable hum.

[GrimJosef]

There's something to be said for rectifiers and for solid-state HT regulators. But the point of the OP's build is to make a valve phono stage. If he was going to start putting solid-statery in then he might as well pick a quiet op amp for the audio stages and have done with it.

To be honest, low-ripple HT supplies for audio pre-amps aren't very hard to make. I'd support G6Tanuki's suggestion to go with a smallish first capacitor (strictly called the reservoir) and then a larger second one (the first smoother). I'm afraid you'll quickly get into the regime where most of the noise won't come from the HT rail but from internal noise in the valves, pickup in the signal wiring and unwanted effects in the grounds. Sorting those out is where the fun is though !

Cheers,

GJ

[Takapuna]

Thanks to everyone for the interesting and informative discussion. Four replies before the evening ended is quite remarkable along with others earlier today.

As per the observation in the previous post, my aim is to build a phono preamp using valves if possible. I already had, or have been given, salvaged, or brought many of the components already.

No offence taken by the "guidance" from everyone to consider silicon rectifiers. This would remove many of the HT smoothing considerations along with volume weight and watts. I know it makes sense but...

I will look at the use of 3-Terminal regulators for the heater supplies if my proposed transformer has sufficient headroom. It should leave the supply ripple-free and circumvent the problem of absentmindedly pulling one or more valves only to find the others start to glow more brightly!

My proposed design does split the power supply and amplifier into two boxes and, yes, an extra R-C stage for the HT input at the amplifier input is within the design. GrimJoseph is quite right; I guess hum, noise, pick-up and grounding are going to be the fun areas.

As an aside and if you are wondering where I'm coming from, I have a large (20+) collection of vintage radios in various states of repair or completeness along with the usual old VHS/Beta/Video2000 machines in the loft and a considerable collection of Laserdiscs! There are a couple of cassette decks that have been an interesting diversion recently too. Not to mention a Freeplay Radio that shed the teeth on the small spur gear attached to the belt pulley. Similar to everybody else I suspect..?!? I also enjoy music and especially listening to vinyl. Hence I have one foot in the Audiophool camp!

Thanks again to everyone for your comments.

Phil
----

[Craig Sawyers]

Be careful in LC filter design that there is a critical inductance phenomenon to ensure stable operation. If you get it wrong you can end up with the voltage too high and oscillatory behavior.

See for example https://www.diyaudio.com/forums/powe...nductance.html

http://amfone.net/Amforum/index.php?topic=37557.0;wap2

Craig

[Craig Sawyers]

A good explanation of the difference between a CLC and an LC filter follows. Critical inductance applies to an LC filter. But a CLC filter in which the input C is small starts to move towards an LC situation. Oddly the LC filter, if properly designed, gives better regulation

There is a derivation of the formula Lc > Rc/(6pif) and a good general description here

http://www.r-type.org/articles/art-144.htm

[ms660]

For rectification, pages 174 to 193 is a good read:

http://www.nj7p.org/Manuals/PDFs/Books/RCA_1940_VTD.pdf

Lawrence.

[kalee20]

Since microfarads are cheaper than henries, I'd spend the money there.

Craig's comments are bang on, but for the low-current demand (which moreover is constant anyway), there's no point in having a choke-input filter.

Another comment: the EZ80 does have a maximum reservoir capacitance of 50μF, but, rules are made to be broken... If you add sufficient resistance in series with the transformer secondary to limit current to 90mA, the EZ80 can never be over-run, so you can use as many μF's as you like for the reservoir, and the small 10-15mA of load current can still be supplied. And then follow with a couple of RC smoothing stages. You'll soon end up with hum levels at the microvolt level.

The purist in me would indeed keep semiconductors out of an all-valve circuit...

[Craig Sawyers]

That is an excellent reference in general - thanks for that!

The pages you give are Schade's seminal paper on calculating currents and voltages in a whole selection of filters for mains rectifiers.

It is so often used and referenced that Rudolph Moers wrote a paper for Linear Audio (Vol 8, Sept 2014, 83-127 "The Otto Schade Method") in which he built a generic circuit to test Schade's graphs. Some are bang on within small errors, and some are a bit wide of the mark. Worth a read if you like some maths and some concrete rules.

Craig

[ms660]

Here's another one that I always keep on file, (pages 82 to 97) it's by Kauzman and it simplifies in graphs for peak steady state current & max. hot switching current....Anode dissipation.....Total added series resistance dissipation, all more or less in agreement with Schades:

https://worldradiohistory.com/ARCHIV...w-1947-Mar.pdf

Lawrence.

[Craig Sawyers]

Originally Posted by kalee20

Since microfarads are cheaper than henries, I'd spend the money there.

Craig's comments are bang on, but for the low-current demand (which moreover is constant anyway), there's no point in having a choke-input filter.

For stable operation at 25mA you need at least a 13H inductor for an LC filter, and possibly somewhat more for safety.

However it is worth bearing in mind that with just a 50uF reservoir capacitor at 25mA current, you get a ripple voltage of 3.5Vp-p (=It/C with t=7ms)

So you need to do something to get that down. One favorite instead of a CLCLC with expensive and large L's or is CRCRC. That involves some voltage loss, but the filtering can be impressive. If R=1k and C = 50uF, the breakpoint is ~3Hz. With three in cascade you can get 100Hz ripple down to a very low level and near sinusoidal waveform. Of course there is a voltage drop that needs to be taken into account - 3k of series resistance at 25mA is 75V.

Good luck!

Craug

[kalee20]

And if you have 100μF and 470Ω as an RC filter, it reduces 100Hz ripple by a factor of 29. Two such RC filters in series reduce Craig's 3.5V of ripple at the reservoir to 4mV, with under 25V drop; another two will give 5μV of ripple, with total 50V drop.

But of course, you could use bigger than 100μF and smaller R. I've just looked up: Farnell are selling (812-6852) 220μF 400V electrolytics (Rubycon, so they're decent) for just £5.52 each. £22 plus four resistors for 4μV ripple on the HT rail is pretty good value!

[Takapuna]

Thanks to everyone for continuing to take an interest and for all the new ideas.
I'll take some time out to reconsider the design and look at the information in the various links.

However, a couple of years ago I bought 2x 20H/460R/75mA chokes of the "affordable" variety (hence the high resistance?) and recently have acquired 47uF/500V capacitors. I'm tempted to go with these as I've spent the money already. The CRCRC arrangement would be cheaper I agree.

Craig's first post had me a little worried, but the design was not going to be a true choke-input filter. After my discussion regarding making the reservoir small to reduce the strain on the rectifier it's tempting to use 47uF as the EZ80 is happy with this value and be done with it.

Let's see. Whilst a split box design that can easily be carried anywhere by two persons might appeal to me, it does not score so highly with my Wife regardless of how good it may or may not sound!

Regards,

Phil
----

[trobbins]

Originally Posted by Takapuna

As I understand it:
Making the input capacitor smaller reduces the peak currents for the rectifier and transformer, thus giving them an easier ride and dropping fewer volts across their internal resistances. It obviously leaves more ripple for the following LC or RC combination(s) to deal with but if the following capacitor is made large this can be dealt with.
Unbalancing the values in the other direction, making the input capacitor large and the output capacitor smaller has the advantage of further reducing voltage ripple straight away…as it were. This leaves less work for the following LC/RC combination. I appreciate this leads to high peak currents and can stress transformers rectifiers and the input capacitor if they are not rated for the ripple current.

You may want to check PSUD2 again about first capacitor ripple current versus capacitance value. Often in practise the first filter capacitor has to be a certain size to provide a sufficient ripple current rating, and a smaller capacitance value part can't be used.

If you can hear or have the ability to measure power system related noise, then distance and layout (mechanical and electrical) of the power supply are your best friends for reducing power supply related ripple/noise at the amplifier, so starting with an external power supply module is worthwhile if it is shielded and distant. Layout and distance is also important in such a power supply module (rather than cramping all parts together), from orientation of PT and chokes, to how the caps are connected to/from, and how you manage the 0V line and protective earthing of the equipment.

[Takapuna]

Thanks, I'll check with PSDU2 again. I'm assuming capacitor ripple current ratings need to be better than the mean value predicted by PSDU2 and not the peak values because peaks are of course very high. Please correct me if I'm wrong.

I'm planning to star-point the 0V around the reservoir cap negative terminal to which I will run the PT centre-tap directly. I may extend from this point a little with a short length of LARGE gauge solid tinned copper to facilitate easier layout.

Mains protective earth goes down to a stud close to the inlet. I trust it's OK to make a single connection between this stud and my star point...and that this can be a simple flexible tail. That is to say my star point does not have to be constructed at or around the physical chassis connection location?

Phil
----