RH2A3

[Paul Barker]

By the way another diy er called Kurt back in the day was planning a commercial project by making an oscilator in a box which was planned to cost about $100 US but soon afterwards this King chap discovered the use of lighting transformers and Kurt's plans died a natural death.

[Alex Kitic]

First things first, Paul was right about the conversion factor. As tested this morning with several transformers and loads, it turns out to be 1.356 - with normal AC source.

Thus the tubes are probably getting 10.2V, which might or might not be on the higher side. We must take into account DMM precision. It is difficult to shave off 0.2V since it would require a total recombination of resistors.

Nick is also correct, although we do not agree on all terms

In theory, AC RMS x 1.41 less drop across diode results in DC RMS. As Paul cautioned, the result is going to be way lower than 1.41, but once we have established a reference, we are ready to go ahead. After all, that is how an analog DC meter can be used to measure AC, and the rest is obviously calibration and standardization if it leads to manufacturing.

Now we get back to the nature of the waveform, because what we need to know is it's heating power, it needs to deliver 50W of power to the heater, 2 ohms wire, basically. If the waveform was triangular, the same peak value would result in less energy, to put it that way.

A 10 Vrms triangle wave applied across a 2 Ω resistor also produces 50 W of heat. That 10 Vrms triangle wave has a peak voltage of √3 V (≈17.32 V), and a peak-to-peak voltage of 2√3 V (≈34.64 v).

I am trying to say that it is my opinion that the approximation measurement via diode plus cap with a tested conversion factor leads to a good approximation of the voltage across the filament. Maybe even better would be a test of current draw, which I might try with my DMM in series with the filament? If it draws 5A, we should be at the correct spot?

I believe that the waveform is more similar to sine than triangle, thus at the same peak value it has maybe slightly less heating power than the sine wave. This translates that aiming at 10V we actually get slightly less due to the erratic form of the wave. But the ultimate measure should be current draw: imagine that as a current regulated DC supply where one has omitted the voltage reference part and lets the voltage settle at a value relative to the current draw of the filament, which is not perfect, but some would prefer it to voltage regulated.

Last but not least, I understand how the additional cap has increased output voltage: by "cancellation" of the 100Hz format ripple. What we get this way is much more similar to (assuming) 40kHz wave, perhaps still loosing some energy in respect to a pure sine wave.
While this explains the perceived increase in brightness and increase in measured peaks, it also should mean that the result is a waveform close enough to a sine to be considered as such, maybe having slightly less heating power (thus slightly less voltage) than measured. The 10.2V I am getting are probably just perfect, if we are trying to stay as close to but not exceed the 10V RMS mark?

[Alex Kitic]

I need to correct myself on DMM accuracy, mine is 0.5% on DC up to 200V, and 0.8% on AC up to 200V. Basically 10x more accurate than I wrote...

AC response is average, calibrated in RMS of sine wave.

AC average should be the same as RMS for a sine wave (this is where the problem lies, waveform shape).

I somehow missed Paul's extremely interesting post. Nevertheless, I knew how the standard untweaked waveform looks like. This is the change that gets introduced by the tweak: the "dips" get filled with signal. Without the added cap, ripple is 230V at 100Hz, the 330uF cap brings that to maybe 10V or less.

The only remaining "issue", without a true RMS DMM, is the waveform. If square wave, it should have more heating energy than a sine wave for the same peak reading. Still, this is an imperfect square wave, thus the energy is lower than for a perfect one. Nevertheless, 3.19, if that is your estimate, is less energy than 2.828, but more than 3.464.

What I am trying to say is that my proposed procedure to get a reading actually errs by giving a result higher than it actually is. If 3.19, the average reading across the filament as measured by my DMM as 29V becomes just 9.25, and based on brightness of filament, that is not correct.

I will try to measure current across my DMM to get a better idea, but it should result in just one iteration of the voltage drop result across a resistance...

[Alex Kitic]

This might help...

He also showed the RF suppression on the input. This is fundamental in hifi to make sure you don't put back into the same line you are using for your sources phone stage preamp ect RF noise from this SMPS. I would suggest that the standards of the domestic power industry which these electronic lighting transformers have to adhere to are not adequate for good hih sound.

So you definately need to supplement the RF supression. Without a scope, one approach would be to assume they have correctly designed it for the frequency of operation, and duplicate it. So you could buy two of these cheap transformers to use one, and rob the rf suppression components from the spare to put in series with the existing one.

I think it has been made easier today by the frequency they are now using. I believe the main problem you will encounter is RF noise affecting everything else plugged into the same main power source unless you supplement the RF filtering.

I have experienced no problems whatsoever with RF. Either it is the fact that my preamp power supply already has an RF filter at the input, or because my electronic transformer unit looks high endish in respect to those shown... or the fact that I am using the amp on a separate "special" plug derived from a voltage boost transformer (my mains is 223V as average, so I use a box with a 6.3V filament transformer of the EI type inside wired to add 6.3V an d slightly more to my special 230V plug). Probably does a good job at filtering what would be transferred to the line?

And yes, your posting is very helpful

[Alex Kitic]

This might help.

There is a fairly reliable fact, that the manufacturer of the transformer has to meet standards. So in original form it produces a little under 12v.

googling the problem I found a very informative guy who shows his true rms voltage reading at 11.8v so that makes perfect sense.

Besides, if I got this right, all one needs to be able to measure the RMS output of an electronic transformer is a Fluke "true RMS" DMM?!

This means I can buy one as soon as I get a job
In a local shop they start at 160EUR approximately. The only reason I don't have one of these is being unemployed now, and not needing one when I still had some means of sustainance. A true RMS meter is not strictly necessary to measure your B+ or your resistors... or your mains supply voltage.

But it can be assumed that a DIYer building an smp with TT tubes like the 845 or GM70 or 211, or 813 - where these electronic transformers find primary application (in my view), can be expected to own or buy one. After all, I do my soldering with cheap irons bought on the flea market, but DIYers who build my designs own "soldering stations" with temperature control, which I cannot afford (and probably wouldn't know how to use).

Thus, someone building an RH813 with electronic transformers for heaters supply could easily fine tune his heaters voltage by simple measurement, provided his/her DMM is not some cheap rubbish but a Fluke... easy.

[Paul Barker]

Thinking about it the RF filter built in is probably adequate because many people with hifi will have a number of these transformers on the same supply. We use them one for each lamp these days so a kitchen could have 6, 8, 10 or more.

But a nutcase on power supply design would definately want to see more protection.

I have a number of stages of RF filtering on my power supply to my hifi. Yes it makes a considerable difference to the sound. It might possibly be why my newly discovered love of solid state has been given freedom of expression. Just might affect solid state amps more than valve amps.

[Paul Barker]

I happen to have a 20 - 60 watt transformer from my work.

So I'll also do some tests and well see where we can go.

I find that the output torroid is wound with 10 turns. It is rated at 11.6v at 5 Amps.

So I should be able to find a suitable turns ratio to give you your correct voltage without resistors, for one. This makes it a more flexible device, because you can power for example 2.5v 6.3v or any other filaments by just winding your own secondary. More lower powered valves can share the one transformer because the valves would have individual tailor made secondaries.

I am pretty sure for the 212 this is how I did it, I altered the transformer.

So with 1.16 volts per turn (in the case of the transformer I have) 9 turns gives you 10.44 Volts at 5 amps. the tolerance of the valve is 5% either side, so you are at the edge of tolerance, perhaps a 0.1 ohm 5 watt resistor would be expedient. Or you could perhaps loose the voltage in the leads, just make sure they don't heat up too much.

Any way I'll do some tests.

[Alex Kitic]

Yes, that would be the perfect way to go - instead of burning voltage across resistors like I do. And would open it for other applications, like 2E22 at 6.3V 1.5A, eventually 300B (not in my case, but might be useful for people with higher efficiency speakers who might hear some hum)... and maybe increasing the ratio for GM70 (20V).

Still, my high-endish units actually do not have these simple toroids at the output. I might have to check again inside, but they look rather different than most units shown on the net.

Most important: have you got a true RMS DMM? I guess the important thing is maximum frequency measurable. Some units are specified to measure up to 40kHz, others go up to 500kHz as measurement range. Besides my not-totally-precise method, the true RMS DMM would be paramount in fine tuning. A specified turns ratio like 8 turns for 10.4V might still prove imprecise enough to require measuring to chek the result.

Do not forget to measure with and without added cap! Particularly if you are going to check it with a scope (and see the dips disappear). The resulting RMS must consequently be higher.

[Alex Kitic]

More lower powered valves can share the one transformer because the valves would have individual tailor made secondaries.

I am not sure about that. If the secondary goes directly out towards the output terminals, than yes, provided that total power is not exceeded (current). 2x 300B is 2x 5V secondary, total current 2.4A... but only if it goes directly towards the output. In that case it saves money, space... you name it.

[Paul Barker]

Yes Alex the torroid is the largest one on opposite end to mains input. It is the last think on the far end. The ground ref Cap is across it though. That must be kept.

[Alex Kitic]

Yes Alex the torroid is the largest one on opposite end to mains input. It is the last think on the far end. The ground ref Cap is across it though. That must be kept.

I have to check my units again, since I do recall having seen two transformers, but neither was a toroid: what I mean to say is that it is not as easy to change turns ratio (unwind a couple of windings) on a toroid and on some other type of output transformer.

Looking at some 3.15V x 5A being burnt (i.e. 15.45W each channel) in the voltage dropping resistors make me very interested in changing the ratio of the secondary

I usually do my best to keep warming at bay - cool running toroid transformers, avoid burning too much heat in the regulators... and with an amp of this kind heat is plentiful just from the output tubes (70W anode dissipation, some g2 dissipation, and 50W heater dissipation, that easily makes for 250W heat, without the additional heat burnt by such voltage dropping resistors).

Yes, the cap is what I had in mind: if we add another secondary winding, I guess it should have a reference cap as well?


Additional note:
BTW, the amp sounds soooo good. Everything I was looking for - but the heat. Bass is amazing actually - it was just the sag caused by the inadequate power transformer used for first runs. Now that there is no hint of any noise or hum everything has cleared up to leaves previously achieveable only with 2A3 or 1619... but with enormous power reserves.

[Alex Kitic]

I found a reference similar to what we are doing (or trying to do), illustrating the waveform before and after the tweak:



I guess clicking it should link to the page where I found it. While there are no photographs illustrating how it should be done, we actually do not need any (should be fairly easy to unwind a couple of turns, or wind again a few turns on a small toroid).

While cautioning about getting the voltage correctly after the tweak (which is what I am trying to stress), there are several statements in the text that are not all that accurate - like the harmful effect on heating the tubes, or using wire-wound resistors (I guess those I am using to burn excess voltage are exactly that - wire-wound resistors)...

Anyway, I am not interested in following this guys path towards a HV supply (I prefer my tube rectifier stuff), but maybe some of the basics could be interesting when it comes to winding two or more secondaries for tube heaters...

[Paul Barker]

I found a reference similar to what we are doing (or trying to do), illustrating the waveform before and after the tweak:



I guess clicking it should link to the page where I found it. .

No it doesn't so here it is.

[Alex Kitic]

I found a reference similar to what we are doing (or trying to do), illustrating the waveform before and after the tweak:

I guess clicking it should link to the page where I found it. .

No it doesn't so here it is.

Thanks - when I insert the pictures from my blog they usually represent a link...

The attached picture is a cut-out from the interesting video you linked. The math shown on the picture is 28.4/12.35=2.29 (peak-peak / RMS)- now if we can relate the average reading (i.e. 28-29V across the load) with this RMS we could get a result from that.

If my average reading is equal to the RMS reading (I hope not), this would mean that I am having 12.35V RMS across the heater?! But it is lit approximately just right, almost the same as with 9.8-9.9V DC?!

There must be some correlation between RMS and average on this signal... the best I have yet is the AC to DC ratio across the "concoction" diode plus cap.

EDIT - picture attached

[Steptoe]

Dear Alex, hotmail is still denying my existence but Pre65 has kindly given me your contact details and I will get the 813s in the mail now. Best wishes, Steptoe.