opinions on williamsonish build?

[Lingwendil]

opinions on sort of Williamson build?

So I recently grabbed a pair of affordable 3x7x12" BUD industries aluminum chassis, and have decided to finally build a pair of monoblocks to use some extra parts that I've been hanging onto.

Plan so far was to go for something of a Williamson based design, but with pentode connected (150 volts on G2) 12AV5GA as outputs, since I have a bunch and they don't require herculean drive voltage for decent power. I was planning on using 6FQ7/6CG7 (cleartop, BTW, for cool points, since I have a bunch) for the front end. I have a couple older Radio Shack 12.6 volt 3A filament transformers, and a couple RIB industrial control transformers I was planning to use for power. I have a couple no-name 6K:8 push-pull transformers to use as well, rated for 30 watts. This is a sweep the floor build, intended to use only extras as a cleanup of the parts bin

The power transformers were discussed in another thread a long while back, but I had planned on using the 240 or 277 volt output (bridge rectified) of these for my power supply, so I would be able to make up for a bit of loss in the windings when run this way to get somewhere 300-ish volts., problem is I only get about ~270 or so rectified and at the input to my mosfet ripple filter, which ideally should drop 15 volts or so to give a nice smooth output. That leaves me with maybe 250~260 or so volts supply to my driver stage, which is a little lower than I would prefer. Fortunately, I load tested the transformers all yesterday evening, and found that they run barely warm with 120mA DC draw, so they can handle the build just fine power wise.

With the way that my outputs will be run, I will be at ~20-25 volts per grid, so I don't need a tone of voltage swing, but would this be adequate voltage to run the 6CG7 as intended, with some adjustment of parts values, of course? I'm thinking that I should probably be fine, because the driver will not be asked to swing much voltage in the first stage (assuming that I have an unbypassed cathode, I may get a gain of ~10 or so, and for a 1.5 volt peak onput this should be well within the capabilities of a 6CG7 running with ~160 volts on the plate, and linearity doesn't look to be an issue around this area. I'll likely run a DN2450 cascode CCS to maximize the load impedance up here, and then use a step network (as in my flea amplifier) or self-bias for the concertina section. Sounds good, right?

For the driver tubes I was thinking that the 6CG7 types wouldn't be a great for this position at this supply voltage, but I do have plenty of the russian 6N1P on hand, that I think would be a better fit given the limitations of a good load impedance at this voltage. MY only real question here, is should I go with the original Williamson scheme of having unbypassed cathode resistors here, or would I be better off bypassing them, ar even running something like an LM334Z CCS tail under each pair? I have several of the 334Z on hand, and when built as a temperature compensated CCS they work well enough for an LTP that has both grids driven, and I think would probably be a better choice than running the drivers separately. If there's a "gotcha" here that I'm missing, I'd love to hear some ideas.

So, in summary, the amp is tentatively 6CG7-->6N1P-->6AV5GA Williamson, with solid state rectification and smoothing, fixed bias, and let's say ~260 volt supply.

Feel free to shoot me ideas (no, I'm not going for triode connected ) and If you have any suggestions let me know.

 

[Lingwendil]

I've attached the basic idea of how I'll do the chassis layout as well.

 

[kodabmx]

I'd switch it to 6N1P > 6CG7 > 12AV5... Or 6F12P: If you only need 30V of swing, why not just go 6F12P into 12AV5 directly? Like the circuit I sent you for PCB making

I'd also triode strap it, and run it at 300V/60mA but that's just me

I also read your chassis size as 17x12x3 and wondered why they looked small LOL. I did my monoblocs in 17x10x2 Hammond boxes.

 

[Lingwendil]

I need to go back to that PCB design! That would be very useful here! I have practically zero access to an actual computer these days, so haven't been able to do much PCB stuff at all

If I had 6F12P on hand I might like to use it, but this (as usual) is trying to use what is already in the pantry, so to speak. Other than the chassis boxes and maybe an odd resistor or two it should be able to be done for zero additional money or supplies.

Do you think the 6N1P is linear enough to go up front? I've used them in a flea amp build experiment a while back, I liked it as a differential but didn't quite like it as a basic voltage amp up front. I couldn't put my finger on why, but it did sound different enough that I preferred the other tubes (6N2P, 6SL7, 6CG7) that I tried for that idea that eventually became the front end for the flea.


You and your triodes I'll be converting the EL86 amp to 6P43P triode connected once I'm done with these

Do you think individual cathode resistors is fine, or should I gild the lily with a CCS under them?

 

[kodabmx]

I sent you a PM...

I think the 6N1P is pretty damned linear. I basically look at it as a 6SN7 evolution. Where the west had the 6CG7 and then went to the 12AU7, Soviet developed the 6N1P. Higher mu/GM but similar OP to 6SN7. 6N5P is basically the same. 6N14P is an upgraded 6N3P which I think is also quite nice. In reality, I tend to prefer the Soviet tubes to the US ones, at least on a "bang for the buck" level. I say if you need a 6DJ8 get a 6N24P for $1 and rewire the socket

Yup. Triodes make more sense to me. Basically a voltage controlled resistance instead of voltage controlled current source. I just find that with the sweep tubes, the added peak current handling let you run like 300V with a 1k - 2k Ra-a so you still get power. I'm getting 10W RMSout of a pair of 6P43P triode strapped. While I might get 25W in pentode, I'd rather use 6P36S for that and retain the linearity and simplicity of triode operation

Triode is what makes tubes stand out from MOSFETs IMHO. If Sony had continued their SIT designs, I'd probably be using them, too.

I'd go fixed bias instead of cathode bias. In triode, you get a lot more power from the extra voltage available and there's less waste heat but if you go cathode, resistors should be fine. Bypass with a larger cap than you think you need. I usually go for a 1.6Hz corner so like 1000uF maybe?

Seriously though, I've built the 6F12P driving the 12AV5GA PP and it was pretty nice. You'll get about 20W in triode, probably like 50W in tetrode I guess and you probably won't need two rails. 300V B+, 60mA, ~-50V bias. As George found out, 12AV5GA is basically a 6B4G with a cathode if you strap it.

Here's my take on it in triode. You know how to change it to pentode, and to change it to cathode bias... If you do go pentode, you won't need the drivers. I picked 320V because it's what you get from a voltage doubler off an isolation transformer Cheaper that way This design also uses the VPT12 for ~3k:8R. VPT18 will give you 1k3:8R which is nice for bigger tubes like 6P45S.

 

[Lingwendil]

Yeah, I'm definitely going fixed bias. I was going to do the bias scheme on valve wizards page that lets you do it without an additional winding, although I suppose a small transformer running off my filament tranny would be a good fit too, I'll have plenty of room inside.

You know what? You win. I'll go triode this time. It's simpler, and might be a better fit for my no-name output transformers anyway. You've been giving my arm the triode twist for a while, so it'll be a nice change of pace.

 

[trobbins]

At some point you are going to have to manage feedback stability. Do you have any low and high frequency equipment that could be used to identify performance of the OT by itself beforehand, or of the amp after construction? You may want to add in shelf networks for both low and high frequency roll-off from the start, and allow for terminals to swap part values.

Symmetry of part location may be better if the drive dual triode is centreline with the output stage valves, and how the physically large coupling caps underneath the chassis influence the valve socket layout, and that the caps are sufficiently 'up in the air'. You may need to preempt some under-chassis and top of chassis screening between the PTs and input stage, and even to the OT, or at least an input valve socket with screening.

 

[Lingwendil]

I have some limited access to some measurement gear at work but the availability depends on me finding someone willing to give me a hand (I'm on the security team at a certain silicon valley tech company, and have a few connections in the testing labs) but I don't have open access. At home these days I'm limited to a couple DMMs and my ears pretty much.

From previous experiments with these transformers back when I still had more gear I didn't have any issues using them in a 6AS7 Williamson build, and I remember just some slight tweaks to parts values to stagger the capacitor time constants, and a small capacitor across the feedback resistor was all it took to correct a smidge of overshoot. I figure this time I can use the capacitor in the step network to help set a pole as well, but worst case I'll just run it open loop until I can test it formally.

Do you think that a long-ways layout, with the power transformers on the narrow end, then the output transformers, output tubes, and then the small tubes on the opposite end would be better? I could bring input and feedback up to the front tubes with a shielded cable to have some extra insurance.

 

[trobbins]

Mocking up a couple of layouts using known part sizes can be well worth the effort, as it pushes you to work out the orientation of valve bases to make the most direct lead paths, and to appreciate where any compromises may be and the options to avoid inadvertent coupling.

I am a fan of using valve base terminals, or custom located terminal strips, rather than a 'tag board for all parts' approach for circuitry that relies on symmetry and where parasitic capacitances between parts, and from parts to chassis, could be an influence up in to the 100''s of kHz (where phase and gain margins may still be a concern).

At least parts are not as large as back in the 1940's, and your DMM's are likely to allow part matching to better than 1% for resistors and caps, and valve bases are unlikely to show leakage.

 

[Lingwendil]

I'm going to be going point to point, with most components done off of the socket leads, with maybe a 4 position tag strip nearby each pair of sockets for larger stuff. Fortunately the layout is fairly simple, so I only plan to use a PCB for the PSU, bias, and maybe the CCS I end up using. I was also considering having a steel shield inside the chassis to isolate the AC wiring from the rest of the internals.

I've never been a fan of tag boards for some reason. Last one I used was for a Marshall 18 watt lite build, and I would rather have had a nice PCB.

 

[kodabmx]

When I built my monoblocs, I put the power transformers on the left, the OPTs across the rear, and the tubes in the front center/right.

 

[kodabmx]

Oh. And here's the power supply for B+ See how easy? L1 is 5$ at a1parts,ca (Coil-58).

You can use these instead of the two in series I show in the drawing of course... B43541B8476M000 EPCOS (TDK) | Capacitors | DigiKey

For the heaters, I use a VPT18 transformer, and use a full wave DC setup using one of these STPS60L30CW STMicroelectronics | Discrete Semiconductor Products | DigiKey mounted with one of these 4180G Aavid, Thermal Division of Boyd Corporation | Fans, Thermal Management | DigiKey directly to the chassis.

I get 11.6V which is within the spec (and I like to run the heaters 5-10% low as it seems to make them last longer).

 

[kodabmx]

The inside of my monoblock looks like this. This one uses boost converters running from the heater supply instead of the quadrupler. One of the boost converters does the -110V bias, the other does 560V through a doubler. The automatic bias board in the middle, the 12V smps for the touch-power switch, and a small 6V coil for the bias module. Also, those 2W cathode resistors blew up time ago, and were replaced with 5W MO type.

Could I have made it a little neater? Sure. But it works exceptionally well anyway

 

[kodabmx]

BTW what's the Ra-a of the OPTs? If you give me some measurements (PriR and SecR and Zratio) I'll run a sim with ppcalc and see how it looks and what power you can expect...

 

[Francois G]

Lingwendil said in post #1: I have a couple no-name 6K:8 push-pull transformers to use as well, rated for 30 watts.

 

[kodabmx]

Missed that, thanks.

So I'd use the 16R tap to get 3kra-a if it (hopefully) has them. Should get 10-20W depending on the resistance of the primary/secondary.

 

[Lingwendil]

I'll try to measure the winding resistance tonight, but yes, 6K:8, I figured this would be a reasonable choice here. And if I use them with 4 ohm speakers it would still be a reasonable choice.

 

[Lingwendil]

Resistance readings on the primary are 200 end to end, with a 90/110 split. Secondary is around 2.5 ohms.I have no way to measure inductance unless there's an arrangement with a DMM I'm unaware of.

Looks like fixed bias of -50 volts or so, about 250 volts plate to cathode, and 6k plate to plate will be my output operating point. I can't go higher voltage with these transformers so I'll take what this gets me

 

[kodabmx]

With the voltage that low and 6000 Ra-a, you'll only get about 3W in triode out of it. You might just go pentode...

 

[trobbins]

What is the lowest AC current you can measure with your DMM's?

Back in the 1940's, the original WW articles said to aim for less than 150uA on the AVO 7 10mA scale, with the full primary across a 5V 50Hz heater winding. You would have 60Hz, but same ballpark.