Heathkit AA-100 iron with 6L6GC

Is this amp going to be driven from an outboard preamp, or are you going for an "integrated?"
 
I had originally planned my standard configuration -- single set of RCA inputs with a volume pot that's normally turned all the way up ... with volume typically controlled from the (AirPlay) signal source, and the volume pot there to set the maximum volume. Also then has the flexibility to use an separate pre-amp.

I suspect I know where you're heading -- the Pilot topology with 2 12AU7 input stages with the volume control between them?
 
Yup, that was one of the options I could see working. To list them all that I'm thinking of:

1. The Pilot circuit (with GordonW mods).
2. The standard voltage amp coupled to a split load inverter like many of the Fisher amp sections employ. I've done this one a few times and it works extremely well and is a minimalist design. Even (barely) has enough gain when cathode is partially bypassed to run integrated mode.
3. Variation on #2 is the Harmon Kardon A50K/A500 circuit, which uses positive feedback to bootstrap the voltage amp for more gain. Never built this one, but looks interesting.
4. Voltage amp coupled to a cathode coupled differential pair. This is what I did on my last EL84 amp--I used the 6CG7 for both voltage amp and differential pair. It has oodles of quiet gain, even enough to work as an integrated after 18 dB feedback applied.
5. Differential pair on the input, silicon CCS tail. The differential pair drives the EL84s. This is also a minimalist design. One example of this one is the "Baby Huey" that is described over at diyaudio.
 
Y'know what else would actually work well is Dave's 9302 topology with the floating paraphase front end. It's a fairly minimalist design, too.
 
@kward, another option with a cathodyne inverter would be to use a pentode AF amp stage, like an EF86. Lots of flexibility in the operating point, making it easy to direct-couple to the inverter, and LOTS of gain (say, 40dB or more), allowing for quite a bit of NFB. Downside, of course, is that a pentode is noisy compared to a tube like the 6CG7. What did the input sensitivity of your EL84 amp end up being? I figure I'm going to need ~30Vpp of voltage swing for max output power of (ballpark) 21-23W with the operating point I'm looking at.
 
Open loop, my EL84 amp (all 6CG7 triode frontend) has about 42 dB gain between the first two stages (and with no cathode bypass on the first stage). I ended up applying 18.5 dB global feedback which left me with 0.4V sensitivity. With that much gain, and using AC heaters, I was a little worried that it would be noisy, but no. It's one of the quietest amps I've ever built.

The quietest amp I've built to date uses a pentode frontend. I used to worry about "partition noise" but I can't hear or measure a thing (granted I don't have an FFT analyzer). I think in part, the quietness of it goes to the excellent tubes I used (RCA 6U8's) in addition to DC heaters on the 6U8s. Looking back at my notes, I measured open loop gain (including the slight loss from the split load inverter) at right around 38 dB. With 15.5 dB global feedback, sensitivity ended up being 0.8V, but that was driving a set of 7868 tubes.

I've never tried scratch building a paraphase frontend. One day maybe...
 
Hm, I'm going to need a little bit more open loop gain than that, if my calculations based on your schematic for that amp are correct. For full power on your amp, it looks like you need 8.1Vrms to the finals. The math tells me that your first 6CG7 stage has an open loop gain of 14.5 and that the differential pair inverter has a gain of 7.5. With the -18.5dB feedback, the first stage gain becomes 1.7... 8.1Vrms / 12.75 = .635Vrms ... so my calculations seem at least close to your measured input sensitivity (rounding errors).

I'm going to need somewhere around 10.5Vrms to drive the finals to max power, but using your front-end as is, it looks like I could get the additional gain I need to maintain a .4Vrms input sensitivity (.4Vrms - .5Vrms is the sweet spot I like to hit) by either adding the cathode bypass cap to the first 6CG7 stage or by biasing with a red LED. I'm tempted to try the latter simply because I've always wanted to try it, and the 6CG7 draws enough current to make it practical. But I suppose I'd need a different strategy for applying global NFB if I were to bias with an LED.
 
The voltage gain stage on my amp has gain of 15.8x (open loop), and the differential pair has gain of 8.5x, each output to ground.. I've never tried LED biasing, but I would think it acts similar to a bypassed resistor, so it seems it would be possible to string a diode and a small value resistor (like 47 ohms) together in series and attach the top side of the diode to the cathode and the bottom side of the resistor to ground, and then attach the feedback to the junction between diode and resistor? But like I said I have never tried that.
 
Spent a little time toying with this today. I concluded that I needed a vacuum rectifier to get the voltages into the acceptable range for EL84Ms. As currently drawn, the EL84M plates should get 400Vak and 300Vg2k.

I decided to run the PI and AF amp stages off the same regulated 315V supply that the screens will use. That supply is good for 100mA, which is more than enough. An EFB cathode regulator is pegged to the 315V regulated screen supply. I used the cathode regulator approach rather than applying negative bias voltage to the output stage grids in order to help bring the output stage voltages in line.

The screen / PI / AF amp supply regulator uses a zener string to clamp the regulator input down to 372V. An LR8N3 is used to drive a TIP50. 1% resistors are used in the regulator to prevent thermal drift while in operation.

I goosed the gain a little by adjusting the operating point of the AF amp and PI stages.

Still don't know when I'm going to get a chance to build it, but I figure it's good for a thread bump.
 

Attachments

  • ST-40 Special schematics v0_3.pdf
    59.1 KB · Views: 27
Your original thought of a triode input stage driving a cathodyne inverter describes almost exactly the circuit I published in my thread of some years ago now for the build of my Fisher SA-100 Clone, using the AA-100's OPT. The project turned out to be excellent, and has been built by other AKers as well. Just one more direction you can take your project. Good luck with it!

Dave
 
Your original thought of a triode input stage driving a cathodyne inverter describes almost exactly the circuit I published in my thread of some years ago now for the build of my Fisher SA-100 Clone, using the AA-100's OPT. The project turned out to be excellent, and has been built by other AKers as well. Just one more direction you can take your project. Good luck with it!

Dave

Yah, seems like it's difficult to get the input sensitivity I'm after using that topology. Even using a fully-bypassed 12AX7 as the AF amp stage, I'm going to be looking at 600mVrms input sensitivity with only -12dB feedback, whereas I'm closer to my preferred target of 440mVrms with the Mullard 5-20 style front end (and could goose that up even more by using a 12AX7 instead of a 6CG7 in the first stage). I suppose that's not much of a difference, though.

I suppose another option would be to use a pentode AF amp stage (like EF86) with cathodyne inverters ... doesn't save tube count, but avoids having to build the CCS board :)
 
I take it then you are trying to make a simple integrated (selector switch/volume control) out of the unit? Otherwise, 440 mv would be entirely too sensitive for use with an active preamp. For more sensitivity, I would use an EF86 as you suggested, as that not only would give you the gain, but also allow the use of a volume control with little regard for Miller. It would also steer you clear of the heater/cathode leakage hum that is so prevalent with the high performance pentode-triode tubes made famous in so many Dynaco designs.

Sounds like a great project!

Dave
 
Not really "integrated" with an input selector, etc... but yah, sort of... I generally connect one source ... an Apple Airport Express, and then stream content from any number of devices the family has handy. I usually include a volume control, though... for a small amp in e.g. a bedroom or office, the control is intended to mostly be set < 50%, so I pay more attention to the grid stopper value on the first stage and think a lot about Miller. But on the 6L6-UL amp I built for my living room, the volume control is squirreled away in the back and intended to be turned up to 100% or just shy of... in that case, Miller is less of an issue. That 6L6-UL amp (which has an input sensivity of 440mVrms for 28.5W/ch) is driven by an active pre-amp (a Parasound 2100), and the volume control is really just a component matching attenuator, if you get my meaning.

This 20-ish W/ch amp that's the focus of this project has two goals: Develop a good circuit that can be used for future AA-100 scavenging activities, and build a nice amp that I can sell to my house remodel architect who went ga-ga over my 6L6-UL amp ... unless I end up really liking it, in which case I will of course keep it for myself :)

With that in mind, I want to have a reasonably sensitive input so that it can be used in a standalone fashion without a separate pre-amp. But it should also be easily incorporated into a system with an active pre-amp. Make sense?

Anyway, I'm going to spend some quality time with the EF86 plate curves and see what I can come up with (the Philips data sheet gives a lot of canned configurations, but they all seem to have high distortion figures, so I want to do the computation myself as a sanity check). Gain > 120 for the first stage is easily attainable with EF86, which would give me adequate sensitivity with a cathodyne PI and still allow for considerable NFB.
 
Ok, I toyed around with the EF86 idea while I ate my lunch today and here's what I came up with. Ended up using one of the canned voltage amplifier stages from the Mullard data sheet, mostly because I found the data sheets lacking on critical details that are often found in power pentode data sheets (like screen current curves). Not wanting to solve for 2 variables over a bowl of beef chili, I decided to cut myself some slack. I took a stab at DC-coupling to the cathodyne, but this one awfully tricky even under the best circumstances. I then realized that the canned configurations in the EF86 data sheet rely on the AC loading of the next stage to keep the load line oriented correctly with respect to the plate curves. So, I went with AC-coupling. Not super happy about the extra capacitor in the signal path, so we'll see.

I spent a bit of time this evening tweaking the PI stage. I only actually need 30 or 40Vpp of swing, so I can bias the PI fairly hot to minimize the imbalance on the lower leg. The 47K load resistor in parallel with the 220K grid leak of the lower EL84M adds ~38K of to the specified-by-Mullard 330K, but I don't think it changes the loading conditions of the EF86 all that much. In doing this tweaking, I came to see that I got within a couple of volts (on paper, anyway) of the correct voltages to DC-couple the EF86 to the PI, so I'm kind of toying with the idea. Wondering if the operating point will self-adjust to a certain degree.

Also toying with the idea of adding a 1.2K resistor in series with the PI's 47K plate resistor, and then taking the output from the lower leg of the PI from the cathode rather than the junction of the cathode load (47K) and bias (1.2K) resistors. Not sure how much it matters since I'm not going to be approaching the saturation or cut-off points of the PI.

Thanks for all of your collective input so far!
 

Attachments

  • ST-40 Special schematics v0_5.pdf
    58.8 KB · Views: 16
Yah, not too pleased with that aspect. I may go back to the 5-20 style front end. Still thinking about it.
 
Five phase shifts (counting output transformer). Fun.
One reason why I like the Marantz take on the Mullard, and the dc cathodyne splitter config.
BTW, anything wrong with letting Miller in a triode voltage gain input stage limit the HF input?
 
Sometimes HF rolloff caused by Miller capacitance can't be helped when coupling between stages given the tube choices and operating points chosen. However, and especially for the input, I try not to let Miller do the heavy lifting of HF tuning. It's especially at issue when you have a high gain triode frontend with a high value volume pot attached to its grid. Not only does that induce HF rolloff that can creep down into the octave above 20KHz where you need that area to do HF tuning, but also because the rolloff frequency changes with volume pot setting.
 
Back
Top Bottom