Introduction

The Ella is an integrated amplifier kit supplied by Brian Cherry of DIYhifiSupply based on EL34 (or 6550/KT88) and 6922, 6SN7 tubes. I bought mine in early 2005 because I was interested in trying to build one rather than because I needed a tube amplifier. I already had an EL34 based, Chinese amp that I thought was quite good but once I completed the Ella I realised it was much better. So I ended up keeping it and experimenting with changes to components to see how much difference they would make.
One of the good things about the Ella is that it is almost completely point to point wiring rather than using printed circuit boards so it is easy to change. I've now rebuilt almost all of it, and although the original was very good the current version is much better.
The kit I bought had the added options of the Choke, PEC potentiometer, Obbligato capacitors and Kiwame resistors, with Valve Art EL34 valves.
I changed the wiring layout on the 6922 tubes so that both channels were the same. The original kit had the triodes reversed on one channel but as the 6922 is a dual triode it makes no difference to the sound having them the same and does make it much easier to avoid any wiring errors during assembly.
The wire colours were also changed from the manual. For example yellow wires for ground, blue for the B+, green for the 6.3VAC and so on. This probably increased the build time by about 20% as each time I cut a wire I had to make sure I would still have enough of the same colour for other places I wanted it!
My B&W Matrix speakers preferred more power so the EL34 (35-40W) tubes were swapped with Electro-Harmonix KT88 (50W) tubes. This made a big difference, particularly to the bass. This meant all of the tubes were EH (6922 gold pin, 6SN7 gold pin, KT88).
A lot of the wiring was changed from the original copper to the silver in teflon. I used the 0.5mm wire for the signal in (up to the 6922) then 1mm for the other wires except for the ground bus which was 2mm. There was a noticeable change in detail but it would have been much easier to build it with silver wire in the first place. Wires around the 6922/6SN7 and some ground wires weren't changed though.
The Django Transformer Volume Control (Amorphous core, silver wire and silver RCAs) was added and the Ella rewired to bypass the PEC volume control. A very big difference in sound quality.
My ceramic 6922 sockets weren't holding the tubes very tightly and if the tubes were moved could even result in some shorting. Heatshrink around the pins had fixed this but I decided to change the 6922 sockets with the new teflon ones. This was much more secure. I also took the opportunity to change some more of the copper wire to silver.
BF
Arrow
AF
I used the Cree 5A 1200V Schottky diodes (CSD-05120A from Parts ConneXion) to replace the original ones. Certainly an improvement as the sound became more focused and clearer, with voices and individual instruments easier to follow. Percussion had more impact too.

Note that on the Crees the metal plate on the cover is live so I decided it was safer to heatshrink them.

BF
Arrow
AF
The Chinese RX21 10R and 220R 4W resistors were replaced with Mills 10R and 220R 5W (wire wound non-inductive). The sound became a bit warmer and silences much quieter but there was a loss in the highest frequencies. So I changed the power tube signal resistors from Kiwame 1K 2W (carbon film) to Vishay 1K 1W (VTA52, bulk metal foil). This made a huge difference as it became much clearer, but now a lot of the warmth of the music had gone. Some high frequencies sounded quite sharp and the upper midrange had a thin, metallic feel.
After a few days I changed the Vishays for Audio Note 1K 1W tantalum resistors. These were much better, with the right level of detail and no harshness. A rounder, fuller sound than the Vishays and far more enjoyable while still being more detailed than the Kiwames.
BF
Arrow
AF
Brian provided some new black test jacks and upgraded bias pots (0-22K in 5 full turns instead of less than one turn on the old pots). The screws were different sizes to the old ones and there were no pins for the bias PCB so I replaced the PCB with an offset tag board attached to the transformer cover screw (if you do this be careful not to attach any wires to the mounting leg tag!) and the position of the pots and jacks were swapped as they fit the holes better that way. The chassis needed a bit of filing for the plastic sleeve of the jacks to fit. I also wired it so that turning the adjustment screw clockwise increases the measured voltage, which I find more intuitive than the normal Ella's decreasing voltage. (See the Diagrams page for the new layout).
BF
Arrow
AF
Using information provided by Dennis Grimwood I used my old power supply IN5407 diodes on the power tubes circuit. These go between the output transformer and the screen grid resistor and block out any AC current on the screen. Rather than try soldering the diodes to the tube sockets I used another pair of offset boards attached to the transformer cover screw. The sound was now a bit smoother and had more "presence", with perhaps more midrange detail.
Here is a summary of how they work: Message from Dennis. The Diagrams page shows how to fit them in the Ella.
Since the volume control hadn't been in use since I put in the Django I removed it completely and changed the faceplate. The new one is still 10mm aluminium but anodised in a glossier black finish.
I swapped some more resistors, ending up with:
NFB: Shinkoh 30K 1W  Much smoother than Kiwames but a bit "fat" with Kiwame signal resistors, so:
Input: Audio Note 1K 1W  The AN tantalums increased the range without becoming too bright. I also tried the RikenOhm 1K 2W here but found they had too much high frequency, especially with the Kiwames still on the NFB.
6SN7: Audio Note 1M 0.5W
To decrease any background noise I also changed a lot of the non-signal resistors:
Power Supply PCB: Wirewound 1.5K 7W
Power Supply PCB: Kiwame 220R 2W
Power Supply PCB: Kiwame 4.7K 2W
Zobel: Kiwame 24R 2W
The result of all of these changes was that the sound became fuller, smoother (less grain) and warmer. There was more bass than before and both bass and mid range seemed clearer.
BF
Arrow
AF
After months of thinking about it I decided to try changing the coupling capacitors. It took months because this wasn't a cheap change - I replaced my Obbligato 0.1uF 630V caps for 0.1uF 600V V-Caps from VH Audio, then at $40 each. The sound improved, mainly in the midrange (which I found clearer) and the dynamics (far more impact). The bass became a bit rounder and perhaps the treble was more controlled too. They are better than the Obbligatos but do seem to take a long time to burn in, and of course are far more expensive.
BF
Arrow
AF
I realised there was a single diode (IN4007) on the negative bias circuit that was performing half wave rectification, when full wave rectification should make the current in this circuit smoother. I tried using a cheap bridge rectifer (DB107) and found that the music sounded more powerful but with a bit of added harshness in higher frequencies, probably caused by diode switching noise. So I replaced the DB107 with four Cree 1A 600V Schottky diodes (CSD-01060A). Brian did a simulation and found the DC ripple reduced even further if the 220uF capacitors were replaced with 470uF, so I did this too using the same type as the originals (Rubycon YK, 100V). I think this was one of the better changes I've done - vocals are smoother, there is more detail in the high frequencies and the sound seems much "fuller".
BF
Arrow
AF
Dennis saw my bridge rectifier change and suggested that I also add bypass capacitors on the negative bias circuit. This circuit is supposed to be DC only, providing a steady bias to the power tube grids. But AC can get into the circuit too, for example the signal itself via the 100K resistor and 22K potentiometers. The rectifier (or diode) will then convert that AC into DC and add it back to the bias supply, causing instability. By adding non-polarised capacitors at the output of the bias supply at the PCB any AC can be drained to ground without affecting the DC at all. I used Solen 0.47uF 630V caps, although anything up to 1uF and rated for 100V is fine.
I had seen some encouraging reports on the Gold Lion KT88 Genalex reissues so thought I'd try them. They do sound much better than the EH tubes I had been using and are worth the extra cost. I'll stay with the EH 6922 and 6SN7 though as the best alternative to those seem to be expensive NOS tubes.
BF
Arrow
AF
The newer Ella uses improved output transformers (Japanese "perfect lay" 5K, 75W, Z11 core) so I decided to try them out on my Ella. Wiring them in was easy enough; the problem was physically fitting them. The cores are much bigger so I had to do quite a bit of filing of the brackets, particularly rounding off the top back section so the transformer covers would fit. And I also had to drill some holes in the chassis. The result is a clearer sound, especially in the lower frequencies and also improvements in the dynamics.
The original Ella had a Zobel network (a resistor and capacitor in series across the speaker terminals) that Brian said was there to remove a slight ringing in the older transformers. With the new transformers he didn't think it was necessary so I removed it. This didn't make a huge change to the sound - maybe a bit more edge in the higher frequencies.
BF
Arrow
AF
I had some doubts about adding a circuit board that I couldn't fix myself if it ever failed, but it's easy to install so would be easy to swap back to the original manual bias system if necessary. And this module is completely analog so at least there is no digital noise added.
First I changed the resistor on the boards from the supplied 3K3 to 7K5 and wired it in parallel to one of the existing 22K bias pots. This gave me a variable bias up to 60mA per tube rather than the fixed bias current it defaults to. (A better option is described here). Then I had to move the bridge rectifier I had added so that the boards would fit. After that installing them was quite easy as I could use a bracket to mount them off the tube sockets. The only problem now is that I have a hole where the removed 22K bias pot used to be, but at least that lets me see one of the blue LEDs on each board.
Having already improved the negative bias supply with the rectifier and capacitors the additional improvement in sound wasn't that huge, but there did seem to be a little more detail and improved bass. The module certainly does what it's supposed to - the bias goes to the level I have chosen and stays there regardless of any change in amp temperature. Before, the bias could change quite a lot from when it was first turned on to after a few hours use, as well as changes from day to day, presumably due to fluctuating mains power supply voltages.
BF
Arrow
AF
After all the previous changes the wiring around the power tubes had become quite messy. Also since the socket pins could move slightly I found that after changing tubes I sometimes broke a connection. So I decided to replace the sockets with the teflon ones where the pins are fixed and to take the opportunity to rearrange the wiring. In particular I used pin 6 for the 100K negative bias resistor so that I could remove the tag board I had added in June 2006.
While doing this change I also upgraded the 0.47uF capacitors on the 6SN7 sockets to the Obbligato premium film ones.
AF
I was introduced to one of Mark's friends in Hong Kong called Kenneth Wong. He winds transformers by hand to a very high quality so I asked him to make some for my Ella. I had already heard his transformers in other amps so I decided to try replacing the input transformer as well as both outputs. As I was arranging my move from Hong Kong to New Zealand at the time I left my Ella with him to install them. He did a great job with the sound improving a lot and I would certainly recommend him.