Page 1 of 8

HF AC DHT heating

Posted: Fri May 23, 2014 1:10 pm
by Paul Barker
I have split this off from Alex's RH2a3 thread which has become the most diverse thread.

I was uncomfortable with the 40khz frequency and wanted to unwind the choke which sets oscillator frequency.

But I broke it instead. These boards are tightly packed.

So I took out the output transformer at the same time.

Now I have to buy a new electronic transformer. I think the choke could be unwound by cutting one end of each leg of the CMC at the lesser number of turns you decide upon, and bearing back, connecting to it, without having to remove the torroid which doesn't come out unbroken.

The OPT has 95 primary turns so we are looking originally at 110 primary volts but now 154 primary volts.

I am a lttle conscerned we are stressing components by raising the operating voltage. Perhaps where we have inserted a C filter we could change that to an RC filter. This will have to be experimented with but I could envisage a 200 ohm + 20 watt resistor on a heatsink.

The over voltage has affected the frequency, possibly because of saturation in the choke. It would then also affect the quality of waveform and the OPT would probably be saturated.

So we need to take steps to get the operating voltage back to where it is intended.

I would like the frequency more up to 70khz or above if at all possible.

I plan to use this out put torroid to set me up for heating the PX4 amp. I thought I would wind the primary for 2 turns per volt, then I can set my voltages to the accuracy of 1/2 a volt increments. I have to find 2.5v 4v and 7.5v once each for each channel. I am thinking that the single grounding cap on the first output torroid should take be sufficient.
So i don't need to ground each secondary of my second torroid.

At this rate it takes two supplies to make each one. But they are so cheap it isn't worth buying seperate torroids, which near enough cost as much as the whole kit and caboodle.

there will be losses, but the transformers I have are 20 to 50 watt and I am using them on the 20 watt end of that window.

Ex perry mints to follow.

A picture tells a thousand words.

The heavy handed Barker treatment left a void. ( i am better with pipework)

After delicately trying to remove a small RF choke for surgery on it.

The disected OPT.

Posted: Fri May 23, 2014 8:10 pm
by Paul Barker
Killed another attempting to change frequency.

2 down.

So we'll leave frequency alone.

Next comes the filtering without voltage rise. strange thing, EDIT, the following in brackets was garbage, it was the result of me not grounding the cap. (I made a pi filter with 120 ohm resistor and the output voltage now drops to 10.5. This new transformers was 12v unaltered. There remains 2.5v AC ripple.)

So next I shall try a Danbury choke instead of resistor in the pi filter. I'd like to see the back of the 100hz component. (Edit I tried the choke but the current overwhelmed it. On the 50 watt lamp it was taking over double current the choke is rated for so it was just acting as a resistor.)

At the moment the HF frequency is settling at 46khz. I shall just have to live with it. But I am not happy to live with the 100hz pulses.

I now have two spare torroids which I can wind exactly to the voltages I require, multiple secondaries for isolation of each valve. Thye plan remains to wind a primary at a higher turns count to obtain 2tpv so that I can increment in 0.5v steps on the secondaries. If I make the primary tapped I can refine the accuracy on test. This is because without knowing the behaviour under the specific load of this application I cannot be sure what the start voltage is to be. First hook up will be on the longest tap of the primary, then I can claw it back depending on findings. The taps could be in one turn steps. Once the tests are complete the excess (perhaps leaving one over voltage step and one under for different mains voltage scenarios.).

The listening test will be slow coming because suddenly this whole project is starting to look like in needs a piece of wood breadboard hook up.

The pretty in a case PX4 design is going back into obscurity.

I am much happier not trying to make it fit a box and look pretty. In truth I should have gone for a bigger box.

Posted: Fri May 23, 2014 10:43 pm
by Paul Barker
OK so I have edited the above but left the story as it was writen for a history.

Stupid mistake didn't ground the caps.

So with caps in place and 240 ohm resistor on the 50 watt lamp the results are:


The dissipation across the resistor was 14 Watts.

But as I want to heat the valves that amount to 20 watts I need to see what the voltages will be at 20 watts. So I put two lamps in series. and got this. Interesting the frequency varies with power. But to get it up to 50khz pleases.


Now the dissipation in the resistor is 6.6 Watts.

The next part of the process is to raise the resistance so that the output transformer settles to 12v which it is designed for. The whole gadget then should be happier. i am very pleased that at 25 watts the square wave looks much better.

I am now more inclined to trust the RMS voltage prediction of the scope, though the Hameg reads it as 13.48v

Andrew has offered to check my results on his equipment at Owsten so we'll then know which of my two gadgets is right. For a pure SW the RMS is approximately the Peak. As these SW's are not perfect you would expect a different result.

Maybe over the bank holiday weekend I will try heating the valves.

Posted: Sat May 24, 2014 9:37 am
by Alex Kitic
Paul, you really did a lot of work here!

I will not post my comments and opinions as usual, because it would be too much work to cut and paste as needed, so here it comes, taxatively as possible for easier following.

1) Choke and frequency

This is something I was not even considering to tamper with. I am quite confident that 40kHz is good enough, since in theory we are cancelling rather effectively this fundamental frequency in the hummer resistors (or pot) just like we cancel the 50Hz mains. What we are left is mostly 2nd and upper harmonics - and at 80kHz I guess we are safe?

As your example shows, trying to tamper with the choke is basically beyond our capability, and the best thing would probably be installing a different choke with different windings... and beside availability or feasibility, costs would levitate beyond reasonable.

The other reason why I think that it is not necessary to tamper with the choke is the fact that I cannot hear any disturbance whatsoever in the high frequency domain or otherwise. Once ripple was filtered out (mostly), what is left gets totally drained in tube hiss (I guess that is the correct term) which is audible only if you treat your tweeter as a headphone.

With the RH813, this means literally total silence - the only problem that is really important is 100Hz hum/buzz and getting the voltage right. I believe this would basically apply to most other SE DH tube amps, regardless of RH feedback and stuff, because the disturbance is so low that it must reach inaudibility.

2) Voltage increase and its cause

You can simulate easily with PSUDII what happens with a 165nF cap and what changes adding a higher valued cap. I added a 330uF/400V cap because that is what I had at hand. From 208V DC RMS (and huge ripple, basically 329 or so volts), we get to 320 or so DC RMS where ripple is "just" 7.1V. This is a 46x (or more) decrease, basically 33dB.

I have attached two illustrations for this. The RMS values basically correspond to what I measure with my DMM.

You can clearly recognize the pattern "imprinted" on the output waveform of an electronic transformer - the wave that represents the envelope inside which oscillation occurs with burst of (40kHz?) high frequency. The wave is what we hear and what is considered as hum, and the reason why I call it buzz is probably the fact that it is not 100Hz (2nd harmonic for 50Hz) but 200Hz (2nd harmonic for 100Hz) and as such is more irritant (to me, at least).

Thus improving filtering we cannot avoid a rise in DC voltage, because there is no relevant rise in "peak" voltage - just a decrease in ripple of difference, which makes for the RMS reading increase!

Posted: Sat May 24, 2014 10:25 am
by Alex Kitic
3) Filtering out the ripple

Since I consider the ripple as the main enemy in this case, we could try to filter it better. Obviously, this would require separating the source of the voltage (i.e. + pole of the bridge) from the board because there are already some parts there that would interfere with the result, etc.

One way would be to increase the cap - which ultimately could decrese the reliability of the diodes (current surge). Another way could be to add a resistor and a cap: adding another 330uF cap behind a 100 ohm resistor would reduce the ripple to 269mV, which is more than 100x in respect to the untampered with electronic transformer. We are talking about 40dB here.

This solution nevertheless requires two rather large caps and a resistor that would dissipate some current. Maybe applying smaller caps, but an LM317 or eventually TL783 (just in case - but the input-output differential can be made such as not to present a problem to the LM317, which has a slightly better ripple rejection ratio) would greatly reduce the ripple, while peak voltage decrease could remain slight. A calculation could be made for this, and I imagine an LM317 with a couple of resistors and smaller caps could represent a good bargain...

4) DC voltage after rectification and issues

What worries me is not the increase in DC RMS caused by a decrease in ripple, but rather the fact that filtering causes a decrease in peak output voltage, which might be easily correlated to the effect of lower input voltage (mains AC) for an untampered-with unit.

Because, what seems to be one problem with those units, they require a minimum current draw (power) to start the oscillation - and if no-load protected can be powered on without a load -- but in case the mains input is inadequate this could and probably would damage the choke that induces oscillation (if I got this right). Thus, what I am afraid, is that if there is not enough peak (not RMS) voltage after the bridge, the unit might be permanently damaged.

Getting back to a RMS voltage reading after the bridge (with a cap) similar to the reading without the added cap - would equate to a condition possible with much lower input mains AC, like 170V or less. Apparently, 170V AC mains would damage the unit - and I imagine that is due to not enough peak voltage to start oscillation?

4) Higher primary voltage on the output transformer

This might be an issue, but is not something that cannot be calculated. The increase in primary voltage is almost 50%, but depending on wire gauge etc. a calculation could be made to check whether this is permissible (I have not done that yet, and could do it, but I guess others may be more into transformer calculations than I am and could do that better and faster). The fact that the transformer is operating on high frequency could also represent a relaxation factor?

While changing primary turns ratio might be rather easy on your toroid unit, it would be virtually impossible to do on the one installed in my units (picture attached one more time). Furthermore, to do this one needs to add turns, thus more wire is needed - and you need to get some adequate wire (for transformer winding purpose) which might not be feasible for everyone.

My units are powering the output tubes as described - i.e. after adding the capacitor and discarding 3 out of 8 turns (keeping 5 out of 8 turns). This will allow us to get an idea on whether the output transformer suffers from over-voltage in the primary or overheating. So far so good, I've had no problems with that, output voltages keep quite stable at levels described.

Just as I expected, I feel that the units are heating less than before the change of secondary turns, and my explanation for that is simple: after adding the cap the output voltage was probably 17V and at 5A = 85W (the unit is certified as 105W maximum, which is probably what the transistors can do in terms of current and temperature without heatsinks). After decreasing the output voltage in the secondary to probably 10.5V, at 5A this is probably 52.5W, which is self-explanatory.

Now the heat comes mainly from the "back" side where the transistors are, while the "front" side remains rather cool to touch (the side where the output transformer occupies most of the place). From this I deduct that my units are safe to go ahead with...

5) Multiple secondaries and power concerns with low powered tubes

Your PX4 is extremely interesting - I was considering buying some "simpler" unit with toroid transformer, and modifying it for use with the RH300B. One 300B tube consumes 5V 1.2A i.e. 6W, two are just 12W, if I add one winding for the drivers I might squeeze another 4 or 5W... hardly getting to the minimum 20W. Still, my units were operating even with one 10W automotive bulb (I have to check the bulbs to make sure). On the other hand, going for 6.3V secondaries would increase the power and allow for correction (burning the excess voltage with resistors).

I guess each secondary needs a bypass cap on one of the ends. This is obviously some ceramic type... can anyone suggest a value? Since mine had no visible markings...

6) Precise measurement

Not having Fluke 87 units, we cannot do a precise (?) measurement. Still, I believe that the most important is the difference factor between the actual waveform (imperfect square on yours as well) and the perfect square. Because a simple DMM that measures average (even analog!) should measure across a thermal resistance the average peak to peak voltage drop, which is 2x for square and should be just divided by some factor - like the 2.29 which I have adopted after seeing the video you posted. This factor is something you should try to check on your unit, I have the feeling that it is rather similar if not identical on most units!

Posted: Sat May 24, 2014 10:35 am
by Paul Barker

For my filter I used an 83uF thenfor now a 240 ohm resistor and then now have 300uF final capacitance. My ripple is now 500mv. My dc output from the pi filter on a 25 watt load is 284v. Hence the raised output voltage.

I have ordered 470 ohm 10 watt resistors. Then after paying I realised they are from China so that will take a month to come if it turns up at all..

But anyway, that will be a little over sized but the final output voltage is probably better adjusted this way, though it takes up more space heat and cost the Sq W shape is looking better and the DC is cleaner.

I did contemplate puting a CMC and cap on the 12v AC output tuned as a low pass filter to modify the output waveform from Sq towards sign, though it obviously wouldn't be pure sine. Just because the HF harmonics might cause RF interference which might affect our equipment. But this will involave a lot of parts to experiment and adjust the wave form at the right frequency. It would be a big adjust on test scenario.

Right now I'll stick with simple.

Yes I don't recall any sound issues when I used this method previously and I don't think any others who tried it suffered any loss of quality. I think the whole thing died a death because of all this fiddling around getting it to work, and because more of fears about all this HF hash rather than actual facts about the sound.

It reminds me of Thomases amp which showed up real bad on Andrews spectrum analyser but sounded great. With his SMPS DC filaments.

Phil and I both used SMPS DC on the 833b and both those amps sounded great. Mine was an cooperation with Nick.

So alot of these things I would like to address may not neem addressing.

I do suspect though that to boos the mains side with larger CLC filter to suplement the weedy internal one say out of an old desktop power supply microwave over or washing machine to better filter the hash going back down the AC into other equipment would pay dividends.

Posted: Sat May 24, 2014 11:00 am
by Nick
Yes I don't recall any sound issues when I used this method previously and I don't think any others who tried it suffered any loss of quality. I think the whole thing died a death because of all this fiddling around getting it to work, and because more of fears about all this HF hash rather than actual facts about the sound.
In the interest of transparency, I stopped because of the sonics, I had it working fine other than that. But don't let me stop you, just stating what I found.

Posted: Sat May 24, 2014 12:30 pm
by Alex Kitic
I gave it some taught and I think that there is no need to worry about the turns per V ratio of the primary. I think that there is no relevant difference in peak voltage, while the difference in RMS voltage is due to the elimination of waveform constraints caused by ripple after the rectifier bridge.

Maybe the only elements under stress are the transistors due to the change in oscillation amplitude (larger signal all of the time)? Anyway, if dissipation is not too high I guess that is not a problem.

Measurements: your scope shots show a factor of 2.42 for 50W and 2.25 for 20W. This basically confirms 2.29 as a good starting point in calculating the RMS voltage across the load. The lower the load, the higher the "efficiency" as approximation of square wave. What could be interesting is measuring this voltage simultaneously with a garden variety DMM of the average type to check if the peak to peak voltage corresponds to the reading on your scope.

I am not sure whether HF AC can be considered SMPS. After all, what we have is just an oscillator across a transformer, while the real switching action in rectifying this voltage is missing. Maybe the first part is similar, but the more important second part of the SMPS is missing (and the switching artifacts that are caused by it that vary in frequency and hit filter poles and bounce...)?

I did not get whether Nick was using a SMPS DC solution, or HF AC?

I guess HF AC should sound better than SMPS DC because the cathode is at the same potential throughout its length (main AC advantage in general).

Posted: Sat May 24, 2014 3:11 pm
by Paul Barker
Yes Nick tried the HF AC, that is he means.

I'll press on for now, I take on board what you say about the differences between this type of oscilator and a full on DC SMPS. That's a relief.

Now I am homing in to a place from which I can build heater supplies next.

I put 500ohm resistor in the pi filter. The DC voltage is now 255v the AC 100hz element is 100mV the output of the transformer is either 12v or 12.7v depending which piece of equipment is right. Actually I'll just pop back to the workshop and test a pure signwave and a pure Square wave on the scope and on the HAMEG and see if they can agree on that. I'll be back. Also I think you want me to check the AC on a commonal garden voltmeter, one which isn't rue RMS? I think you'll get garbage. The problem is that I don't do commonal garden equipment. I work with my voltmeter so it has to be a good one. This Fluke is all I have right now. All my tools face a lot of prefessional use I attend to 8 heating appliances a day. The kit has to work day in and day out. Anything amateur lasts less than a week in my hands.


The frequency is higher and is it just my imagination or is the square wave getting squarer?

Posted: Sat May 24, 2014 3:55 pm
by Paul Barker
On the 1khz callibration square wave of the scope the scope reads 2.02v which it converts to 0.994 mV RMS the Fluke 0.957 mV RMS the Hameg reads it as 0.9752, so the Hameg and the scope are within 2% of each other the Fluke at 1khz is just under 4% from the scope.

The theory is that a perfect square wave RMS value is the Peak. So that should be 1v. This means at 1khz on a pure square wave the scope is most correct at 0.6% acuracy.

We could extrapolate that for the present we had better use the Scope as our voltage reference. This would put my output right now at 12.7v

We have to factor in that the waveform isn't pure square, but the scope has arrived at it's approximation, which I shall have to accept even if only in the interests of safety of the valves. So in the extreme condition that the hameg is right the 7.5v for the 801a measured on the scope might actually be 7v. The 2.5v of the #27 might be 2.36v and the 4v of the PX4 may be 3.78v.

Nono f those factors are going to make a nat's foreskin of difference, best is, keep on the side of safety.

I am happy to trust the scope.

Regarding the Fluke 116 it reads a signwave correctly at 1khz but at 5khz it falls apart at 50khz you have no chance. I must assume it is true RMS up to 1khz. I didn't sweep it I just went through the decades. It is enough to tell me it is not for use on switching equipment which runs at 30 to 50khz

Posted: Sat May 24, 2014 3:58 pm
by Alex Kitic
Paul Barker wrote: Also I think you want me to check the AC on a commonal garden voltmeter, one which isn't rue RMS? I think you'll get garbage. The problem is that I don't do commonal garden equipment. I work with my voltmeter so it has to be a good one. This Fluke is all I have right now. All my tools face a lot of prefessional use I attend to 8 heating appliances a day. The kit has to work day in and day out. Anything amateur lasts less than a week in my hands.
You might be surprised - I am confident that my DMM is measuring the average peak to peak voltage rather correctly - the problem is converting it to RMS (by division with a waveform difference factor).

On the other hand, your Fluke (117, was it?) is obviously unable to get a reading at all? But the scope and the HAMEG are getting a reading, and the difference comes from the waveform difference factor: I tend to have more confidence in what can be seen on the scope, but it still does not mean anything. A plain measurement of peak to peak voltage across the load would tell you whether the scope is getting it right... but you need a non-TrueRMS meter for that, probably even some analog type would do.

I guess this would be more a case of need for ruggedness (pro) than the lack of precision (amateur).
Paul Barker wrote: The frequency is higher and is it just my imagination or is the square wave getting squarer?
The square wave is not getting squarer, the factor is 2.26 because you are measuring with 20W load - with the 50W load your P-P to RMS ratio will decrease again, and the frequency will decrease back to where it was. You have, probably, gained a better reading thanks to the ripple reduction (simpler waveform).

If you ask me - 40k or 50k, doesn't matter - the 2nd harmonic would be either 80k or 100k, and we are not bats, and these are heaters - not grid inputs... you should try it on an amp and check whether you can hear anything, and whether you actually need anything more than a simple cap at the bridge.

Posted: Sat May 24, 2014 4:08 pm
by Paul Barker
Alex Kitic wrote: you should try it on an amp and check whether you can hear anything, and whether you actually need anything more than a simple cap at the bridge.
Yes that is a plan.

Unfortunately I have to go to my wife's parents tomorrow.... least said soonest mended as we say in Yorkshire. It's aplace where you are in trouble anyway and the butt of the sharpest tongue in Yorkshire from the arrival to leaving. Your role in the operation is to facilitate the wife's pleasure and take all abuse thrown at you with thanks and a smile.

But might hook up a sound on Monday!

Posted: Sat May 24, 2014 5:44 pm
by Alex Kitic
It's the same all over the world. The most important phrase I learned is that "those who have no son in law have no donkey either", to paraphrase the local saying.

Monday looks soon enough! But to get some sound out of it first you need to wind and fiddle with a pair of secondaries? I guess that would take less than one hour...

Repeated question: each secondary needs its own cap. On mine the value cannot be read: what should be used (it's for RF filtering if I'm correct?!

Posted: Sat May 24, 2014 7:11 pm
by Paul Barker
I'll measure the cap and get back to you.

I tried to rewind the torroid and failed. Too small.

At present I am in progress winding a small EI transformer. I just hope the coupling at 50khz will be adequate. there may be greater losses which would put me under target voltage. But the first one will inform me of the necessary ratio.

It occurs to me that to completely rewind the secondary of the original would have been a better idea. Now I no longer look to be saving space.

My amp is looking a 100 miles away from a nice neat little box, but I am doing my mad professor bit. You can't have both.

Posted: Sat May 24, 2014 7:53 pm
by Paul Barker
Oh testicles. (you might need to look that word up Alex)

I'll try to split and tap off the original secondary. This EI transformer is so big it's pointless. Probably be -3 db to -6 db at 50khz so would be a total guessing game deciding on the turns. At least with the original secondary I know the number of turns per volt.

Just a matter of how many of these units I have to kill to get one finished. Present rate is two dead, one knocking it's knees together.

Only cost £5.50. I can afford to kill a few in the name of science.