Kay 703 - neutralizing hum - humdinger on 60FX5?!

[mxrshiver]

it's actually a Paris PC-1, but essentially the early Kay 703 design with the oddball radio tubes. the attached original schematic is like, 85% accurate to the original circuit.

this amp's being rebuilt and thoroughly redesigned (all metal film resistors, safe primary w fuse & 3 wire power cord & iso transformer, tone pot isolated from DC, cathode bias for 18GD6, etc) but keeping the 18GD6 and 60FX5 tubes and basic gain staging, to have some unique character.

the goal is, for the output from the Line Out i'm installing off the speaker winding, to have recording-quality signal to noise ratio, even for as small an input as perhaps 10mV for soft vocals. anticipating excess hum at the very least, from the large magnetic fields around such high voltage heaters, i plan to use the maximum amount of hum-mitigating techniques in the initial design as possible, and hopefully have some time at the end to try scaling back or removing a few of them, to see what makes the most difference. perhaps my goal is a fool's errand, but i'm using the process as an opportunity to study exactly how much hum and heater noise it's even possible to eliminate from an amp like this, given the large heater voltages with larger magnetic fields... and, more importantly, which techniques make the biggest difference, for the smallest cost and effort.



here's the details of the new heater supply...

i'll be using 3 separate power transformers, one for the high voltage supply, one for the 60FX5 heater (Bel 241-5-48, 12VA, 48V secondary), and one for the 18GD6 heater (Bel 241-4-16, 6VA, 16V secondary). i'll be feeding both heaters with very well filtered DC from these separate supplies; each secondary will have: thermistor > small dropping resistor > humdinger with 30VDC reference to CT > fuse > bridge rectifier > at least 3,300uF filter > another small dropping resistor > heater.

not too big of a deal for the 18GD6. but the 60FX5 is a little nuts... luckily i've found 80V 1,200uF caps for pretty cheap and am planning on using 3 in parallel. and the 48V winding should actually put me no more than about 5V above the 60VDC target, so i won't need to use very large dropping resistors. but my main sticking point is implementing a humdinger on a freakin 48V transformer winding...



i'm really not sure how much resistance to use for the humdinger on the 60FX5. i'd love to use a low resistance value of around 500R so transformer leakage current can be easily shunted to ground, but that would dissipate almost 5W and draw about 100mA... so i'm trying to figure out how high resistance i can get away with, in order to draw little enough current to stay within the 12VA power transformer rating, which i believe is already being pushed a bit here... and stay within Rhk max as well, especially considering i'm returning the humdinger center tap to a DC elevation reference.

i'm thinking of using 2x 1W 4.3K's on each side of the heater winding, and a 1K 1/2W pot in between, for the humdinger, making about 10K total resistance. this still allows 5mA to flow, more than i'd like, but only dissipates about 1/4W total, which makes for more relaxed component ratings.

assuming Rhk max is similar to the 20K of the EL84 (my only reference for an actual stated Rhk max for a pentode on a datasheet), that leaves me about 10K left for the lower arm of the DC reference voltage divider, allowing me to obtain a 31VDC reference voltage from a 33K/10K divider off the 135V screen node. this draws about 3mA, also more than i'd like, but not a huge deal.

but i don't know if a 10K humdinger would cause issues, or wouldn't work the same as a lower resistance one. if there would be issues, perhaps the answer is that i just have to find a different power transformer rated for more than 12VA, in order to accomodate the extra current the low resistance humdinger will draw... or perhaps i'm not as close to that 12VA rating as i think? based off of the 100mA heater current, and assumed 0.5 power factor due to the large 3,300uF filter, i've been thinking the VA would come out to 60VDC x (100mA/0.1) = 12VA... but to be honest, i'm not great on power factor equations... and i've been wondering if the power factor is better than the terrible 0.5 on a similar 6.3V circuit, because the 60V heater provides much more load resistance... so perhaps i have a few VA to spare?



thanks in advance for any thoughts or advice!

[danhei]

Are you willing to add a cap and choke before the plate supply?

[mxrshiver]

Are you willing to add a cap and choke before the plate supply?

finally got some time to return to this project... thanks for the suggestion, it was pretty silly of me not to include a choke from the start given my goals!

schematic and final bench pics attached, the blue edits are final tweaks made during testing.

i haven't written in the voltage measurements yet, but they're largely right where i want them to be - filaments about 5% below rated voltage, 60FX5 is at about 4.5W/145V on the plate and 0.75W/120V on the screen, 18GD6A is nice and cool as are all transformers and power dropping resistors - and it turned out sounding really lovely too!! nothing like a simple pentode driving a pentode... and especially with the line out, it's the sort of amp that could function as a pedal in the front of another amp, a preamp driving another power amp, or really sound like a full bodied amp on its own. the CTS speaker goes a really long way towards that... the magnet is so tiny, but the bass response somehow stays so clear even when it's overdriven! is there a typical way folks share audio files on here?

the noise floor is pretty good! but i'd like to get it even better, and i have a lead as to how i might do so, and i was hoping someone might have context for what i've observed.

there's 1.1mV on the output regardless of settings, and i believe it's largely coming from the 60FX5 filaments. adjusting both humdingers made a minor difference, but eliminating the 100R 5W power dropper from the high voltage rectifier made no impact, neither did adding either filament voltage dropping resistor, and neither does wiring position as far as i can tell, or position of mic/guitar to amp, etc. i thought perhaps i needed a lower impedance path to ground for the filament leakage current, so i placed another 4K3 1W in parallel with each existing one, but that actually DOUBLED the size of the noise waveform, to my surprise!

my lead on eliminating the noise, came simply from touching either 60FX5 filament pin with my multimeter probe while in voltage measurement mode - the audible noise dropped by about a third! even better was when i then connected the other multimeter lead to ground - the noise was barely audible even with my ear on the speaker, and the measurement dropped to 0.6mV! results were identical whether i was in AC or DC mode. you can see in the attached scope pictures - the first is with nothing attached, the second is with one multimeter probe attached to one filament pin and you can see a nasty spike being eliminated, and the last is with the other multimeter lead attached to ground and the whole waveform is reduced a bit in amplitude.

i'm pretty puzzled by this phenomenon... it seems perhaps the DC filament will couple less noise into the audio signal, if it has some sort of ground reference attached after the rectifier, rather than before the rectifier?

one possibility is that the DC elevation is messing with the other aspects of the filament circuit, as the roughly 24VDC of elevation is putting one of the filament pins very close to ground, as opposed to being balanced around ground with the other pin. the voltage measurements between the pins, also don't quite match up with the voltage measurements from the pins to ground, so i think the multimeter impedance is messing with the readings... which tracks with it making such a difference in the noise in the first place. here's some relevant voltage measurements:

60FX5 rectifier (+) to (-): 56.8VDC, 2.5VAC
rectifier (+) to ground: 47.9VDC, 6.9VAC
rectifier (-) to ground: 0.6VDC, 6.9VAC
humdinger CT to ground: 23.8VDC

here are my ideas for next things to try:
- reducing DC elevation voltage
- removing humdinger and adding post-rectifier balancing resistors referenced to DC elevation voltage
- removing humdinger and adding post-rectifier balancing resistors referenced to ground
- removing humdinger and attaching negative rectifier output to ground
- removing humdinger and attaching positive rectifier output to DC elevation voltage

[mxrshiver]

waveform pics...