Out with the old, in with the new
After experimenting a little with the first chassis, I found out that choke loading the 26 was a great improvement in sound. It sounded even more open and natural than before. It is really amazing how much this simple modification can change. The choke allows the 26 to get it's maximum gain (just a bit lower than the mu). Another benefit of the choke as a load is the fact that the DC resistance is very low, which lets you design a B+ supply with a lot less voltage compared to standard RC loading. The lower voltage drop over the choke (only 5-10V depending on DC resistance of the choke) against up to 150-200V for resistance loading. This means a higher efficiency since your anode resistor isn't being used to heat your home as well as acting as a load for your tube.
My old chassis was already showing signs of ending up in the trash bin. The hum was still there and I had no room to put the chokes onto. This left me with only one alternative: make a new chassis... Ugh, this was not one of my favorite parts of DIY tube audio, but it would have to be done. So here is a handy lesson.. never build a chassis unless you know for 100% that there won't be any problems with it. The problems I had were in the home-potted power transformers, but I didn't find that out until I had put them into my new chassis. I had to learn the hard way :))
A few weeks later I made a breadboard of the choke loaded 26. Testing showed that this time there was no hum at all on the B+. I had a special power transformer made by Bartolucci in Italy which had the exact windings I needed. This thing was just huge! I've seen 50W power amps with smaller transformers on it. Hooking up was easy, but took me a few days since my free time is limited these days. So... the breadboard worked fine... now let's make a nice chassis.
This time I didn't go overboard with brass plates an housings, just a simple 2mm aluminum plate with a wooden frame around it. The Bartolucci iron was beautiful enough to be placed on top anyway. I made a separate chassis for the power supply and one for the actual signal part so that no AC would be present near the signal path. This would also allow for more room might I want to change something on the inside or whatever. What took me four years of tinkering to develop and improve was now finally realized. The chassis is not yet polished or the wood yet oiled/waxed, but it already looks very appealing.
Bottom of the signal chassis
B+ supply under construction
Bottom view (yuck!)
My experience with the 26 has been sort of a dilemma. It is the most superb sounding tube I have ever heard, yet it is also the most hummy, microfonic tube I've ever heard. The con's of the 26 cancel out almost all the pro's, but in the end it still finishes with a grade of B+ (sorry for the pun). I have been struggling to remove the heaviest con's against this tube, but with no luck... until now...!!!
The 26 is known to most people, but most fear using it because of the extreme sensitivity to hum, noise, microphonics and downright nastiness of it. Add these problems to the hum induced by moving your hand over the tube and your average Joe will dump it in the next RAT ad or E-Bay auction. Only this week have I found the solution to almost all the 26 related kinks... FINALLY!!!!!!
I've tried everything to get the hum down. The sources for this hum are numerous. It can come from a high B+ ripple, a wrong grounding, a magnetic field, filament ripple, electrical fields; you name it. The B+ ripple can be solved by improving the power supply (more capacity, choke input, higher choke value etc), but I have found B+ is almost never a real issue unless you really do something wrong. Magnetic field interference can be solved by moving the 26 away from any AC voltage supplies... but filament hum?!?!?!?
The low voltage of the 26 tube (1.5V at 1.05A) is extremely low for the current it is pulling. Most tubes with a one ampere filament are all in the 5V region. These filaments are easier to drive since most US tubes use this voltage, so there are more transformers available for it. Also 5V is an ideal value for voltage regulators. Anyway, because the voltage is "so" low and the accompanying current is so high (relatively), the hum is not caused by the 1.5V but by the current itself. A passive filament supply (rectifier and capacitors only) can do a lot, but you need mucho micro-Farads to get hum down to a reasonable level. The best option for passive filament supplies is using a choke filter much like that of common B+ supplies (20mH or up). This will actually work great with "normal" tube filaments. The chokes smoothen out the currents as capacitors do for voltage. The lowered ripple due to the choke means you can lower the capacity a bit.
You will find that this supply will not really remove the hum, but only reduce it enough to make it tolerable. A better solution spec-wise is a voltage regulator. This will make sure that the output voltage is really stable. It also really affects the sound quality, so much so that I don't consider this an option unless you have absolutely no other way to remove the hum properly. Tom Ronan employed a constant current source (CCS) using a normal voltage regulator. What he does is stabilize the current to the filament, or actually make a current source that is set to the required 1A for the 26. This option is probably the best way to go since it has less effect on the sound quality than voltage regulation.
My own solution really came from the experimenting of a friend, he used batteries to supply the filaments of some of his tubes (KC1). The 26 is by no means a battery operated tube with it's 1A filament, far from it really. Still, this seemed like a shot worth taking. I have a few 2V/8Ah lead batteries that could do the trick, just add a 0.47 ohm resistor and I end up with 1.5V for the 26 filament.
At first I only hooked up a signal generator and a B+ power supply to see what happened. I still had the voltage regulated supply I normally use, self made with lot's of capacitors. The measured ripple at the output was in the area of 150mV. This is a lot of hum if you consider a 2V input signal and 15V output signal.
I hooked up the battery with good hopes and connected the scope probe to the output. I was kind of confused since I was not measuring any kind of output signal anymore. Hmm, the filament was lit, the B+ was there, the cathode resistor had 13V over it which suggested about 6mA current. This was all as it should have been. AHA.... the input from the sound generator wasn't hooked up yet, so that was the problem.... but... HEY!!!! What's this?!?!? No signal on the B+ before... my scope was on the 0.02V setting and I was only picking up some noise. I had a 150mV 100Hz hum here just a minute ago... WOW... the battery really works; no hum at all. I now measure about 2mV of noise on the output, could this be right?! There is even no hum when I move my hand over the tube...
The battery filament supply really is the solution to 90% of the problems associated with the 26, in my experience so far anyway.
The numero-uno problem of the 26 tube! I've had nightmares about the hum problems I have had with this tube, I even lost a few years of my life I think :) It is an oh-so-lovely tube, but extremely difficult to get quiet. If you do succeed in removing all the hum, then the result will be well worth the effort. There is not a more lovely, musical and airy tube out there!!!
Here are some tips if you, like me, are experiencing heavy hum:
- The filament supply must be DC, there is no getting around this in a line-level preamp application unless you will tolerate the hum.
- Make sure the ripple on the filament is next to nothing, I mean zilch !!!
- Capacity in the filament supply has to be at least 47000uF and preferably in two RC or LC filter stages if you are using a passive supply.
- Use a 6V winding and rectify. The extra voltage allows you to use a hefty RC section to filter noise/hum.
- If capacity doesn't help, try using two 2.2mH air-core (!) chokes from a speaker cross-over (0.8mm diameter wire or more) in the filament supply, this will dramatically decrease hum. The air-core will prevent saturation due to the high dc current.
- Shield the outside of the tube (glass), it really picks up hum sometimes. Move your hand along the 26 when it is playing and hear what I mean.
- Check that the B+ supply has little or no ripple, although this is not the main cause of 26 hum.
- If the filament keeps inducing hum in the 26, try regulators (yeah yeah, I know!) or Tom Ronan's idea for current sourcing.
Now as to sound... Going from the first SRPP 6SN7 design to what is now standing before me, a choke loaded 26 preamp, there were some considerable changes in sound and character. I've already sworn never to use anything but a DHT unless really really really impossible. I think this will never happen though, since even RIAA can be accomplished with an all DHT setup. But anyway...
The sound of the choke loaded version is sooooooooohooooo smooth, unlike anything I've heard up to now. Most people will want to hear all the gorey details on what it sounds like, but I really think that it is too subjective. The best way to find out what it sounds like is to build it yourself. The only thing I will say is that you will NOT be disappointed. A few tips though:
Component quality really is important in this preamp. Use only paper-in-oil or MP type capacitors and try to keep the capacity value as low as possible. This will give a quicker response which you will hear in the detail of the sound. Use a good grade iron, especially for the plate choke. Bartolucci makes specially designed chokes for this purpose which are layered and wound differently from their normal supply chokes. Other brands should have these kinds of products as well, so ask for it before buying a supply choke. The difference is that a supply choke was designed around 50-100Hz, but your audio signal is a little broader than that... Anything below 2x2x2 inches is rubbish as a plate load choke, the core has to be larger to handle the lower frequencies...
Always underrate your power supply... If you need a B+ of 400V/10mA, make a transformer that can handle 400V/50mA, you will really hear the difference. The same goes for filament, for 4V/1A use 4V/3A or something.
This is the best damned preamp I've ever built. It certainly took me long enough, but it was worth it. The trip during this preamp was filled with knowledge and experience which is crucial to making good amps...
When using rectified DC for the filament add a 470uF cathode bypass capacitor!
Ug = -9V
Ia = 8A
Rk = 1125 ohm = 1K1
Rp = 7600 ohm
L = 200H
f0 = 6.05Hz = 6Hz @ -3dB
R = 470K//470K = 235K (see text)
C = 220nF
f = 3Hz @ -3dB
This is the -3dB frequency of the filter at the output. Since the connected power amp has an input impedance (resistance) of 470K, the load resistor of the preamp will be paralleled with this 470K. Therefore to calculate the output frequency of this filter we need to take this into consideration. The 470K of the preamp paralleled with the 470K of the power amp gives a load of around 235K. The resulting -3dB frequency would be 3Hz.
f = 1kHz (as example)
L = 200H
Xl = 1.256.637 ohm
The impedance of the choke at 1kHz would be about 1.26M ohm, this is high enough to make the gain approach the value of mu.
Rp = 7600 ohm
Xl = 1256600
Zout = 7600 ohm = Rp
The output impedance would be 7600 ohm at 1kHz, which is the same as the Rp of the tube itself. Would I fill in a lower frequency, the impedance would drop a bit further. For anything over 1kHz, the impedance of the choke is of little or no influence and Z would be equal to Rp. We do have to take into account the grid resistor in the output filter. The output impedance of the preamp would be Rp paralleled to Rg.
Zout = (7K6 * 470K) / (7K6 + 470K) = 7479 ohm
In conclusion; for all cases the output impedance of the preamp will be equal to, or just lower than the plate impedance of the tube; 7K6. Would I use RC loading, this would always be higher than the Rp, actually almost equal to Ra. Choke loading avoids a large (DC) resistance value in the signal path, the drawback is roll-off in the low end due to induction (or lack there-of) and possible roll-off in the high region due to capacitance between the choke windings. The roll-off in the low frequencies is calculated above (6Hz).
Rl = Xl = 1256600
Rp = 7600
mu = 8.3
A = 8.25x
Power supply section
B+ supply (for one channel)
See my battery charger project for the appropriate chargers...