Class A2 Direct MOSFET Coupled SE

[mogliaa]

Originally Posted by eboz

Hi Ale,

I am glad to see your interest in this amplifier, as I always enjoyed reading about your efforts in the 26 preamp thread and your blog. I have a schematic that I used to get a rough idea of bias points and behavior of the amp. The actual schematic I am still working on, as it's quite involved. There is a standby circuit that is turned on by a key switch, so I can be sure others won't be able to operate it if I'm not around. The B+ will only turn on after the heat and bias comes up, and it also enforces a 2min delay. B+ can then turn on through two 10ohm NTC's in series, which are bypassed 3s later. The amp can stay on heat+bias mode as long as you want before hitting the B+ switch. In theory after the MV tubes are settled, I should be able to leave the B+ switch "on" and then just hit the heat+bias switch and the amp will take care of itself, but I've not tried that yet. I am using Rod Coleman's DHT heater and also the SSHV1 for the 4-65a cathode supply. I used shielded coax for signal throughout the amp, and much work to have the d3a or other high mu pentode strapped triode as behaved as possible. It's a CRCLC into a LM317 in CCS mode, which is floated around 55v from the SSHV supply, and ceramic caps right at the heater pins shunting any hf noise to ground. You can see that the topology is a little different in that the cascode mosfets are not folded. At least in simulation, this appeared to look better to me in terms of headroom and distribution of harmonics (lower in 3rd and higher in 2nd). I actually used two 1n60 mosfets, and dn2540 was used only for the CCS. I am not that knowledgable in the SS department, so it's quite possible my choices are less than ideal in the mu-follower. So far I can say that the music is sounding excellent, and the amp is extremely quiet. I did test and scope about every section of the amp, except for B+, before assembling it into this huge two chassis affair. I have multi taps for B+ at 650, 750 and 850v, currently running at 650v. I can certainly say that anyone using Eimac tubes would be wise to get the plate cherry red, because it does sound dull and awful when the plate is dull and grey! I'll have the schematic soon, and get the other channel finished a little down the road.

Cheers,
-MW

Interesting mu-follower arrangement. Haven't tried this before.
Would be keen to get some of your feedback regarding your experience in building this amp. At the moment I'm focused on the LT and HT supply design. I will use +600V, and two stacked +300V (one with salas SSHV version 2). This amp clearly will require a lot of iron! which I have already..
Filament supplies will be JMS transformers (they are superb and not expensive at all here in the UK) with Rod Coleman boards.
Configuration is as per my previous design: 26-46-4-65a. First 26 probably transformer coupled but will breadboard all this before making any decision.
I see this amp will be a challenge from a layout and chassis perspective, so I'm taking my time as I'm not that good with wood and metalworking. Your amp looks astonishingly good by the way!!

I will probably test some 814 wired in triode which I may get hold of shortly. Only few posts and curves I found are quite promising, and probably and easy valve to drop in here. I will trace curves as soon as they arrive....

No rush here, I want this amp to sound (and look) really good

My wife hasn't seen all the iron and has no idea of the size of what is coming down the line. So far everything is hidden up in my attic workshop. Will see if she kicks me out!
cheers,
Ale

[eboz]

Here's pretty close to what I actually implemented. I only brought DC into the signal chassis, and the final caps of the raw supply are in the signal chassis. The regulators are placed as near to the tubes as possible. I have test points on top of the chassis, and trim pots are also accessible from above.

-MW

[Fuling]

The schematic in post 19 is very interesting. I´m messing around with some old RCA 808 triodes for the moment, they´ve been on the shelves for years but earlier this week I found the parts needed to build the required 7,5V 4A filament power supplies.
So far I´ve managed to build a messy cliplead thing that delivers 11-12W through an Edcor 3k/8R 25W OPT (and radiates quite a lot of heat from the bright 30W filament and dull red 35W plate). Open loop bandwith is not very impressive but a fair bit of local plate to grid feedback did a good job there.
I use a 2SK7000 Fet to drive the grid, input stage will probably be a pentode with "Schade" plate to plate feedback.

[Michael Koster]

Originally Posted by eboz

Here's pretty close to what I actually implemented. I only brought DC into the signal chassis, and the final caps of the raw supply are in the signal chassis. The regulators are placed as near to the tubes as possible. I have test points on top of the chassis, and trim pots are also accessible from above.

-MW

Is the grid connection to the 4-65A correct? It should connect to the top of the 560 ohm source resistor for Class A2 operation with grid current.

[Michael Koster]

Originally Posted by Fuling

The schematic in post 19 is very interesting. I´m messing around with some old RCA 808 triodes for the moment, they´ve been on the shelves for years but earlier this week I found the parts needed to build the required 7,5V 4A filament power supplies.
So far I´ve managed to build a messy cliplead thing that delivers 11-12W through an Edcor 3k/8R 25W OPT (and radiates quite a lot of heat from the bright 30W filament and dull red 35W plate). Open loop bandwith is not very impressive but a fair bit of local plate to grid feedback did a good job there.
I use a 2SK7000 Fet to drive the grid, input stage will probably be a pentode with "Schade" plate to plate feedback.

Using a pentode for the input stage is clever with positive grid bias because you can connect the screen grid of the input pentode to the cathode of the output tube.

[eboz]

Originally Posted by Michael Koster

Is the grid connection to the 4-65A correct? It should connect to the top of the 560 ohm source resistor for Class A2 operation with grid current.

Wow.. thanks, I will correct the schematic and re-post it. Luckily I've not done it this way in the amp.

-MW

[Fuling]

Using a pentode for the input stage is clever with positive grid bias because you can connect the screen grid of the input pentode to the cathode of the output tube.

That´s very clever, by doing that the whole thing gets self-adjusting. I will take a closer look at dc coupling, the power transformers I´m using now are too small but I have a hugh beast with a 450-0-450V winding somewhere...

[eboz]

Originally Posted by eboz

Here's pretty close to what I actually implemented. I only brought DC into the signal chassis, and the final caps of the raw supply are in the signal chassis. The regulators are placed as near to the tubes as possible. I have test points on top of the chassis, and trim pots are also accessible from above.

-MW

Corrected schematic is attached.

-MW

[eboz]

I've learned many things about this amplifier since completing the "finished" pair that was not learned in the prototype phase. I'd just like to share some of those experiences for anyone who's interested in the circuit as well. For example which brand and style of 4-65a tubes are preferred, most stable bias voltages and arrangement, temp stability, etc. I at first thought that the Penta-labs tubes with their more substantial base were nice in that they appeared to run cooler, judging by the plate color, and in theory would last longer. This was a "NOS" pair from ebay, and while one tube biased like brand new, the other tube was much lower current at the same bias. The bias also seemed to drift more with heater variation/environmental temp. So I would advise on using the more standard and smaller 4-65a envelope. With the more typical envelope, try to find a pair of tubes that have a closely matched date code, and they should bias up similarly. This is the code that's engraved and written on the internal base structure of the tube.

I was at first using the SSHV (first version) as the power supply to the cathode of the 4-65a, but I ended up frying one of these due to an improperly sized insulator. I swapped these out for the SSHV2 and found that the new version is much much more stable in this role, and less temp dependent. For the 4-65a I am using Rod Coleman's filament supply and while it sounds wonderful, it is rather sensitive to temp. I am using one heatsink per device, and each heatsink is then solidly bolted and coupled to the chassis. I'll either have to settle for resetting these supplies for winter and summer seasons, or implement a cooling fan.

The only changes I've made to the schematic I last posted, is that the d3a is biased with a 1.8v AA LiIon battery (1.8v is the voltage of these guys when new, much higher than the 1.5v AA we know of), and a 1 ohm resistor in the cathode for measuring purposes only. The input transformer is now additionally terminated with a zobel tuned to roughly 35khz (this is in parallel to the nominal 33k ohm load). The zobel removed a slight sense of unease to the overall sound, and was tuned by scope to best damp the ringing. The output tubes sound best to me when biased a bit below max dissipation. It is -100v and 70mA with 735v B+.

One behavior that still troubles me a bit, is that the bias on the output tubes seem to drift higher (higher in voltage, resulting in more current and brighter plates) during sustained HF transients in the music, like a crescendo of cymbals. I need to scope this, but I don't think it's just the output moving into A2 mode, because it takes too long to recover. I thought that it could be the upper mu-follower supply sagging, but that should put the bias lower rather than higher. Another thought would be oscillation, either on the input or output tube? Yet another possibility is the SSHV2 supply sagging, but this really should not be happening since it's a well regulated supply with current headroom, and it's supposed to be isolated from A2 demands.

I'm still trying to convince a coworker to come over to take some higher quality pictures of the amps, but for now I'll show you the pair in action. The connection to the 4-65a anodes can't be seen in this pic, but it's a piece of solid core copper that runs up above the heatsink/cap and then makes a fancy little spiral to the connection point.

Cheers, and Happy New Year!
-MW

[mogliaa]

Hi Michael,
Many thanks for sharing your invaluable experience with this amp and I am looking forward to the high quality pictures of the amp as it looks superb!

Over the Christmas break I worked intensively when I got some time to build my 4-65a amp. I managed to complete the filament supplies and gyrator boards for one channel as well as did some tests on the 46 driver in filament bias. All good for now, and hopefully I can complete the breadboard for one channel soon.



I think what you are experiencing is that your gyrator bias point gets unstable when going into A2 operation since you are feeding the voltage reference CCS from the unregulated power supply that provides the peak current in A2. In my design I use V4 (which is the SSHV2 regulator) instead to feed the LND150

Implementing the 46 in filament bias was a crazy thing to do. It added an additional hefty filament supply and a lot of heat in the filament resistor, but I hope it is worth the effort in terms of sound, which I'm sure it will do.

Nic Wilsher implemented a 4-65A amp many years ago and recommended that is key to run them at their limits in order to keep the getter healthy, otherwise they become gassy after some use.

A final addition to my design was the inclusion of a crowbar circuit (R6, U3, R11, C2 and R12) as suggested by Rod Coleman. This is tuned to blow up the 600V power supply fast fuse in case it doesn't blow when expected. This will protect the OT.

What are you listening impressions so far?

Cheers,
Ale