Paul Barker wrote:
did you remember to filter the initial 50hz and 100hz from the first rectification of the mains to dc from which the HF frequency generator is powered? That is probably where your hum is getting in. The reason for the problem is that these electronic transformers are specifically designed for lighting, so there is no reason to filter out the 50hz and 100hz elements. After all they replaced the use of ac mains frequency transformer powered lights. So this is an element of the design you have to change if you want to change the use of these transformers.
We used to put additional capacitance off board to fix it.
Actually, I did not - the electronic transformers are "stock", I have not attempted to modify them in any manner. They are good quality German products, not some Chinese no-name items.
The idea about additional filtering of the first DC (after the diode bridge, i guess?) is a good one, although this means tampering with the device, and as such might be "not reproducible" by DIYers who might want to build the amp some day.
As far as hum from these units goes, it is actually not hum in the usual sense of the word (50Hz and 100Hz hum like mains AC would give), but more "solid state buzz".
I just went to the speaker and stopped the music to be able to hear it properly (!). The "noise" is 99% buzz and maybe some 1% hum, if you are intent on hearing it. The noise can be heard from 5-10cm distance from the speaker (i.e. midrange), and gradually gets lost. I don't think I can hear any at my listening spot, maybe in the quiet of the night, if I concentrate on it.
Thus, it is solid state buzz. The same solid state buzz that I hear from voltage regulated DC, using for instance an LM1084. When using DC, the buzz is greatly reduced and can be heard only with your ear very near the speaker. No hum, just buzz. Same as with high frequency AC.
Maybe adding some capacity behind the diodes might improve operation, but it might as well burn the unit (the diodes... or anything else). How much capacity were you adding, and where were you connecting it?
Paul Barker wrote:
Something else worth changing is the frequency of operation.
We raised the frequency by removing turns from the small torroid choke to raise the frequency of oscillation.
The larger torroid HF output transformer sets the output voltage so that can be played with aswell. But you need to scope the output to find out the true RMS, I am not even sure if a supposed true RMS voltmeter is good enough here. At the very least I would check it against a 50hz AC heater and make sure such things as brightness and cathode current are identical for both heaters on the same valve. Over heating shortens the life of TT filaments 10 fold and underheating robs them of power miserably. With TT it is a very fine window of operation you have to hit...
This might be also a good direction, but again it would mean tampering with the units, and it is not something DIYers might easily replicate.
I actually haven't got a scope, nor any means of measuring this high frequency AC (my DMM is good up to 400Hz). While this makes me a little bit nervous (sensitive TT filaments), tests with DC heating (9V, 9.5V, 9.8V tried) have confirmed to me that based on the heaters light intensity I am hitting the correct voltage, probably 9.8 or 9.9V. I am using one 0.15 ohm resistor on each leg, meaning that at 5A current draw, the total voltage drop is 1.5V; the electronic transformer unit is specified as 11.2V-11.7V (depending on current, obviously), so assuming 11.5V at 5A current draw, the filaments get 10V... probably 9.9 or 9.8, but not more than 10V.
The good thing here is that output voltage is supposed to be current draw sensitive, but rather insensitive of mains fluctuation.
Paul Barker wrote:
I have to be honest though, my 212 amp sounded better at my home that day a gang came over. In that version it was heated with mains frequency AC, but there was no hum.
I was slightly less happy with the amp every form it took after that day. Changing to HF ac heating was one of the changes but by no means the only change... That tends to be the way with this hobby, we change too much at the same time and find ourselves lost without scientifically applying logical steps.
My gut instinct is that the heater supply was not the problem. I am fairly confident in the HF heater supply method.
In your situation there isn't much choice anyway.
Indeed I am quite confident in the HF heater supply method. My previous experiments with 2E22 have shown that the HF heater supply is just as good as mains AC, maybe a little bit clearer sounding (slightly less tubey, if that would be the correct definition).
As for the choice - I might decide to try mains AC (getting rid of the buzz, and inheriting some hum in the process. BTW, I am not quite sure whether the hum nulling schematics proposal is applicable to HF AC? Maybe it could be, if the diodes were replaced by high-speed Schottky?
Last but not least, I could continue with the DC voltage regulated LM1084 alternative, as it is not objectionable whether it decreases spurious noise (but the buzz is not something I like, anyway) - and check it's effect on the sound. I might again use classic mains to power the rectirifer/regulator circuitry - or find some more powerful HF AC units (mine are 105W).
Paul Barker wrote:
One issue we had with it which affected reliability was that somewhere in that design of the original HF transformer is something that senses whether the correct number of lamps for safe operation are functioning...
...So what happens today is you buy individual transformers for every lamp.
But you must make sure that your demand on the HF transformer is well smack bang in the right portion of the window. That may not be in the middle, probably better at the lower level. The reason is, that the TT filament is almost a dead short when cold, the sensing mechanism in the transformer probably thinks there is too much load and shuts down before it ever gets off the ground. We experienced this with some transformers. IF this becomes an issue then a resistor in series which is shorted after 30 seconds is the fix.
Well, actually I have to use separate transformers for each output tube, this being an SE amp. I treat them as if they were separate secondaries. The only real difference is the fact that I cannot bypass directly from the cathode, but must bypass from the "mid point" instead.
As for the right portion of the window, using 105W transformers on approximately 60W consumption (11.5V x 5A = 57.5W) means that there is enough current to start the tubes... nice and slow.
(Since current is limited, they start rather nice and slow, max. current is probably just 9A, which is the reason why it shuts down after a while when supplying the rectifier/regulator circuit, since 5A x 1.6 = 8A, and this probably overheats some part of the HF AC transformer and shuts it down).
Putting it simply, a good quality 105W HF transformer for halogen lighting is the perfect choice for heating the 813 - limited current inrush, correct voltage every time, cold and reliable operation for hours and hours, and protected from all sorts of mishaps... if you do not mind the buzz
For DC, a more powerful unit is necessary, something like 200W per tube.
I find it interesting that there is very little mention of "solid state buzz" in all stories about AC and DC heaters that can be found on the web. Some advocate AC as the best solution (mains AC), others advocate "high-end" solutions like Rod Colemans, or the Tent Labs... and there are some that "shamelessly" use high value capacitors to smooth the DC for the heaters without actively regulating it. But no mention of buzz. It seems that DC is perfectly quiet (???) and the whole story is about whether it sounds better on AC or DC, current regulated or voltage regulated... it does not buzz, and is dead silent? Come on, is it just me?
BTW, while there are theoretical explanations on the merits of current regulation, most filaments are actually supposed to operate at a certain voltage, where they exhibit more or less current draw. Forcing a 300B filament to operate with 1.2A current draw means that the voltage across the filament will probably vary between some values (let us assume 4.8V and 5.2V). Now I guess 5V would be the best option, and the filament should draw the current it naturally does at that voltage for which it was "designed"... if we fix the voltage, current will "follow"... but fixing the voltage and forcing a current... does not seem all that correct to me.
This current vs voltage regulation opinion is obviously theoretical. I could easily use a voltage regulator to set the current instead of setting the voltage, and thus let the voltage develop across the filament accordingly. But that would also involve more power loss and heat dispersal: setting 1.2A of current (on the mentioned 300B) would require approximately an 1 ohm resistor, causing an additional voltage drop of 1.2V besides the voltage drop of the regulator, and the subsequent additional 1.44W dissipation... compare that to a low-drop regulator like the L4940V05 which operates 1.2A current on as little as 0.4V drop, and sets the 5V on its own (fixed regulator)... and is limited to 1.5A of current (slow start)... for a dissipation of probably 1W in most applications. Heresy? This is practical theory, and not my intention to step on anyone's toe
The question is: are DC heating applications "buzz free"?
