Quick Question about Smoothing Capacitor Sizing

[E93AFAN]

I am going to use a solid state PSU for my latest valve project (a 7 valve octal GP receiver) as the cost of transformers with separate rectifier windings is becoming silly, not to mention the very high price of rectifier valves. My previous solid state experience in building 60s valve sets with diode rectifiers often specified 8uF - 16uF with a resistor interposed, the results were pretty dire IMHO with lots of ripple, so I am hoping to do better.

The transformer I have to hand is rated 100ma @ 250v and the rectifier I will use is a GBP208 rated 2A @ 800v (because I have one to hand), I'm guessing the set will probably draw about 50ma - 60ma in normal use.

I read somewhere a good rule of thumb for smoothing capacitors is 1,000 uF per 1 amp load, so on that basis the 100uF + 100uF @ 500v can electrolytic I have with a 10 Henry 100ma rated choke interposed would seem to fit the bill.

Does this seem sensible and will it offer reasonably low ripple ?

I did toy with the idea of a valve regulated HT supply but previous experience tells me this is OTT for an am receiver, not to mention the cost of suitable components. So I have decided to throw lots of uFs at the problem and keep my fingers crossed.

All answers gratefully received, as always.

[G8UWM-MildMartin]

Quick thoughts (I'll leave it to the experts to do a proper analysis!) are that it will provide excellent smoothing, but rather high HT (estimated at around 300V, without knowing the choke's resistance) as long as it's not mutually coupled to the output transformer.
And you could gild the lily with a regulated supply to the local oscillator.

[Terry_VK5TM]

Vague recollection of having read something about not overdoing it with the size of capacitors - I think it had something to do with the surge current charging the capacitors doing nasty things either to the transformer or the rectifier valve.

I do know some of the rectifier valve data sheets contain recommended values of smoothing capacitors

The more knowledgeable here will be able to enlighten us if this is true or not.

[Radio Wrangler]

With a solid state rectifier, HT will come up immediately, before your heaters start to warm, so at this time there is no load and your HT supply will peak-rectify.

250 times root(2) = 353v. Reckon on possibly 10% more in places and times of high mains. So >400v capacitor ratings would be prudent. Not just reservoir, but all on HT lines (remember, no drops on HT dropper resistors until current is taken)

A shunt voltage regulator and series R for LO and BFO will reduce drift. Some playing around may be needed to find a resistor value if you use a voltage reg valve like a VR150/30, OA2 etc because they can form a neon relaxation oscillator.

!00u does seem large. It's more usual to de-ripple the feeds to sensitive stages with series-R, shunt-C sections. That series R magnifies the benefits of the shunt C. This will give better protection of the stages which can apply AM due to ripple.

David

[See_Mos]

I agree with all of the above. The amount of tolerable HT ripple depends on the frequency range of the audio output stage and frequency response of the speaker. From the output stages backward filtering is done by R-C networks as David said. If the output stage is push-pull then a little ripple in the HT is self cancelling.

16 plus 32uF was adequate for many years even without the advantage of a choke. Using solid state rectifier you will need a resistor between the rectifier and the first capacitor to reduce the inrush current and this will also help reduce the ripple. Don’t forget that most of the stuff you read about high capacitor value versus ripple relates to low voltage stuff.

[G6Tanuki]

Old valve rectifiers were intolerant of high Peak currents so it was normal to limit the size of capacitor facing the rectifier, and often to add some series resistance too.

8, 16 or 32uF would be typical back in history, with a 10H choke.

By the Sixties thst sort of setup got replaced by semiconductor diodes and the choke - big, heavy and expensive - went, the capacitor got bigger - something like 160+100uF was typically used in TV sets of the Sixties.

Capacitance is cheap and compact these days,in your application I would use a couple of the sort of electrolytics used in SMPS, a few hundred uF, with a low value resistor, maybe 100 Ohms, between them.

[Radio Wrangler]

The thing about surge current and capacitor value is universal, except for choke-input filters.

With a capacitor-input filter, the reservoir initially charges to peak voltage and stays there if no current is drawn. Take current and the reservoir voltage ramps linearly down between the voltage peaks. If it's a little current, or if the capacitance is large, this won't bring the voltage down much by the time it intersects the rising voltage from the transformer and a diode turns on. Sounds innocent so far? This leaves only a very short time, though, before the tranformer voltage peaks and starts to fall and the diode turns off. So the diode spends little time turned on. ALL of the charge taken by the set in a half-cycle of mains frequency has to be replaced in this even shorter period. Consequently the average current in the diode on time has to be several times higher than the average HT current. So the diodes in the rectifier get a harder life if you go larger in reservoir capacitance. This scales up their losses. The same goes for transformer losses. If the rectifier is valved, current surges can damage the cathode. This is where a rectifier valve's max capacitive load rating comes from. With a solid state rectifier diode, notice that it's having to switch much faster than you'd guess from 50Hz. This is where the 1N4007 group of diodes fail to work well. They're not fast enough to work well in terms of what goes on in a 50Hz supply.

Choke input rectification builds up energy in the choke and when the diode would have cut off, the field reduction in the choke creates voltage added to that from the transformer driving the diode to stay on, dumping energy relatively slowly into the reservoir capacitor. When the transformer voltage comes up again and the diode turns on, the choke limits the charging current, robbing energy to charge up its magnetic field again. Good design in a choke input filter is to have the choke large enough that the diodes never all turn off at once. Conduction transfers from one to another, The current into the reservoir ripples, but never goes to zero. But this implies a minimum current to be taken at DC. This usually demands a large choke and some control of the minimum current. The rectifier-reservoir now produces the mean of the transformer voltage waveform. The no-choke option produces the peak voltage. With a choke, run the DC demand current down below the rectifier cutoff value and the DC voltage goes up, reaching the peak rectification value with no current taken at DC.

So both sorts of filters need different design calculations. Both have their (different) problems and compromises. Notice the irony that the lower the HT current can go, the bigger the choke you need in Henries, but the choke in saturation current has to be massive enough to not saturate at the MAX HT current. So for choke dimensions and weight it really is the worst of all worlds, combined. Choke filters are nice, low stress arrangements if you can afford them and handle their limitations.

A reservoir - choke - filter capacitor arrangement is a capacitor input circuit with an L-C lowpass filter stage appended. So you designa as per C-input and treat the second L-C as a ripple reducing filter.

David

[kalee20]

Originally Posted by E93AFAN

I read somewhere a good rule of thumb for smoothing capacitors is 1,000 uF per 1 amp load, so on that basis the 100uF + 100uF @ 500v can electrolytic I have with a 10 Henry 100ma rated choke interposed would seem to fit the bill.

It's a fair rule of thumb for reservoir capacitors... With a full-wave rectifier and 50Hz mains frequency, you'll get at most 10V of ripple.

Larger values of reservoir capacitor result in shorter, higher, current peaks through the rectifier (up to a limit, as determined by the transformer winding resistance), which is why valve rectifiers put a limit on the value of reservoir capacitance. Larger values also take a longer switch-on surge, if using solid-state rectifiers (or even a valve, if it's switched off and very soon back on, while still warmed-up).

The smoothing capacitor is isolated from the reservoir by the smoothing resistor or choke (unless the value is rather small), so generally the larger the better.

[E93AFAN]

Thank you for your thoughts, something to chew over before I rush in.

You'd think I'd learn but I always stumble over the difference in output voltage between valves rectifiers and solid state rectifiers. After nearly 70 years of constructing stuff I'm so used to using a 250 volt transformer with a 5Y3 or EZ81 etc., and seeing 230-240volts at the output of the choke, clearly teaching old dogs new tricks is an uphill battle

I have a transformer rated 190 volt @ 100ma + 2.5 amp centre tapped @ 6.3 volts, the total heater current is only 1.5 amps as valves are EF39 x 4, EBC33, ECH35, EL32 all have 200ma heaters except the mixer which is 300ma. Think this might be better bet as in a quick test the inrush HT seems to be about 275 ish volts until power is drawn then it settles to around 230 ish under load which is closer to the sets design criteria.

An inrush resistor is a very good thought, thank you I'll play about with some values and see which works best and experiment with a shiny new 32uF + 32uF @400v can electrolytic I have found lurking in my spares box.

Gentlemen thank you once again for your time and patience in the face of my naivety, it is a pleasure to hear from you and your sage advice is always much appreciated.

[Gabe001]

250×1.414×2= 707v p-p
Add 10% mains variability + higher voltage at no load (switch-on) and you're a bit marginal on the bridge rectifier voltage rating. Also the peak ripple current into the reservoir cap depending on various factors may be upwards of 1A.

If you can't afford to drop volts using a resistor after the rectifier (note: it gets hot), I've never had a failure with a kbpc5010, rated 1000v 50A. Youll be a bit safer and have a lower peak current and surge if the transformer you're using has reasonable ESR (equally important with valve rectifiers). The 10H will smooth the ripple quite a lot, so you could probably get away will smaller value capacitors.

[John_BS]

There's a very useful PSU simulator program if you fancy a tinker: mine's running on Windows 10. Called "PSU Designer 2"
John

[Diabolical Artificer]

AFAIK the PSUD2 software hasn't been supported for a few years John, you can no longer dowload it & the requisite support packs. Unless I'm mistaken.

Andy.

[Craig Sawyers]

And there is a groups.io for PSU2D, and Duncan contributes regularly and carries out bug fixes https://groups.io/g/duncanampspsud

And the latest version (build 76) is there, with all the support files, along with a beta version for testing.

Back to the HT supply. Back in the day it was common practice to power up the heaters, and then maybe 20 seconds later apply the HT voltages. Tektronix certainly did that, and did/do some audio manufacturers - Audio Research certainly did.

Simple delay firing a relay to connect HT would do the job.

Craig

[Craig Sawyers]

It looks as if PSU3 is under development by Duncan

Craig

[Silicon]

Presumably they switched the HT AC current rather than the DC current?

[Diabolical Artificer]

Thanks for those links Craig, I'll check those out.

Tek used delay relays to vary HT turn on from 20 to 38 seconds. My two 500 series scopes come on at different times.

I've used delayed HT on a few of my amps either using a darlington switched by an RC ramp up or in the last case using a vintage delayed relay.

I use a big power resistor right after the tfmr shunted out by a big relay. So yes, you limit the AC not the DC, easier on the relay.

Andy.