GM70 Amplifier

[Paul Barker]

FWIW, I used a current source for the CF load.

id need more voltage then though. What about hybrid with choke providing the ac swing source providing the operating point of the el84?

[Nick]

AHH I assumed you had a split supply. My mistake.

[Paul Barker]

Well, im now thinking “shame to unbuild the 833 output stage.” But hang on, I dont need it at home. Ok I’ll unmount the tributes and give them a more appropriate home.

So what gain do I need to drive the GM70 grid 100v peak so lets make that 120 volt peak as beyond 120v there is no anode voltage, meaning the come crashing down limit is not the voltage amplifier. 120 peak is quite demanding. Might sound nicer aswell in two stages., anyway one stage would limit and compromise the final stage.

Well I like the 5687 I have some . If I use for two va stages and employ the 650v B+ rail I could use 30k resistive load and get plenty of voltage gain, such an amount that I may be able to play with unbypassed cathhode resistor and still have enough gain. I go on my spreadsheets of years ago to see what actual gan A is on two stages . Not yet looked at the el84 cf . Ill start here with the 5687’s.

Reflecting on this it is easy decision to use 5687 first stage against sn7/j5.

But I woke up with a niggling feeling. I didnt run the maths for the 76.

So the maths went horribly wrong. Mainly because unbypassed cathode resistor turns plate resistance of 9500 into 9500 + 37260. For the headroom in a 430v power rail anode load 36k so instead of ratio going 1:10 no not even 1:3 its 1:0.8 and A(ctual) gain is 6. But there is a fix. We dont have the headroom to increase the resistive load. So use CCS load, even if we conservatively allow 500k because the load resistance is parallel with the input resistance of the next stage. The A becomes a respectable 12.62.

So to bring all the A(ctual) gain numbers under the same post first stage A with unbypassed cathode resistor for sn7/j5 = 9.28 for 5687 is 14.8 and 76 (with ccs load) is 12.62. So has anyone got an opinion ? Its between the last two, both have plenty of input head room input capacitance similar except plate to cathode lower on 5687. Might 5687 with its resisor load sound more natural? Or is the ccs ‘trick’ faultless? Anyway Im only going to build it up one way. Remembering my past emotions I can vividly recall the 76 brassy sound. So I’ll ccs that.

Adendum.

Ronald hasnt replied to my email so I’ll be sending some 56 and 76 valves to someone to match a pair please? I dont think I can make another curve tracer. Though theyre cheap enough

[Nick]

Or is the ccs ‘trick’ faultless?

Depends on your definition of faultless. Can be close enough. Of course will sound different to a resistive load as the load line is different. Depends on your frame of mind if its better or worst. Much like your preference to not bipass the cathode resistor.

I should keep out of expressing opinions beyond the electrical as I have long past given up on chasing between differing distortion profiles.

[Paul Barker]

Ive never forgotten what Morgan Jones said in Vry’s flat. “Why not use youre main b+ to load your voltage amplifier with High value resistance.” Well I can use the 650v supply, nock off 50v to decouple another 250v for anode voltage and I get 70k ohm load, which results in gain of 8.27. But thats not enough, so I’ll dice with death and ccs the 76 and see if I like the result. Me neither chasing too much so Im not building an amp to try out different ideas on. Im building an amp with a chance off the bat of a pretty good and powerful sound on my speakers at home.

otoh I wander what gain Mark obtained from the 4p1L pentode mode? (To involve alternative distortions) Just to involve him, hes been quiet so far. I suspect, not enough gain.

Anyway Im enjoying this and everybodies efforts.

Which brings me to mension someone on dyaudio in the past discussed a 811a amp driven by a 4P1l cf which didnt need a necative supply because bias voltage was +7.5. But I thought why use a valve with gm of 7 down hill with the wi d assistance when we have the EL84 which nearly doubles that? Parameter which I would provide the dominant benefit of all reasons to chose the cathode follower.

[Mike H]

May or may not help, but I had considered (perhaps erroneously) 6J5 to be an equivalent for 6SN7, or near enough. In other words I lke the idea of using 6J5, if you've got them. Don't know anything about 76.

For my latest iteration of the 801A SET, I'm considering 2 stages of Voltage gain, military version of EBC33 (single triode), and 12SN7 with anode choke.

Or, the anode choke might be a 1:1 interstage, with cap coupling across the 'live' ends.

[Max N]

Ive never forgotten what Morgan Jones said in Vry’s flat. “Why not use youre main b+ to load your voltage amplifier with High value resistance.”

My 211 push-pull amp used that idea, I had around 750 volts after smoothing. The input/gain stage used 50K loads, and the phase-splitter used about 25K top and bottom. That was the best sounding version of the amp I think.
I didn’t get any problems with it, but I have wondered since whether I was flaunting the maximum anode voltage at start-up. Might be worth sticking a zener across the valve, anode to cathode, for protection?

[Max N]

May or may not help, but I had considered (perhaps erroneously) 6J5 to be an equivalent for 6SN7, or near enough. In other words I lke the idea of using 6J5, if you've got them. Don't know anything about 76.

Yes, 6J5 is the same as one half of a 6sn7 or 12sn7 or 7n7

[Max N]

One last thought - if we use a ccs load, and it’s a perfect ccs, does it matter anymore whether we bypass the cathode resistor? Because the voltage across the resistor won’t change.

[Nick]

Even if it's a perfect CCS there is still the grid resistor of the following stage in parallel with the CCS at AC.

[Max N]

Good point, I’d forgotten that, thanks Nick

[Paul Barker]

Having observed all the posts and Nick’s mention of grid impedance, I may have forgoten to factor that in, but I think I allowed 500k parallel with ccs. I usually allow 1 meg for a ccs so grid resistor would be dominant. I probably calculated gain from 1/2 a meg. The receiving valve 6j5 can cope with one meg but anyway I have grd chokes, so dc resistance doesnt really matter because in dc terms it will be much lower than a meg. So in the event its not something that I need to address.

Im still going to build for 76 because my remembrence of 76 dominates my opinion of what to do. Secondary to that is the consept about distortion profile for ccs which Morgan may have felt could be an issue when he said what he said, but he didnt expand on what he said to give that info.

At the end of the day if the amp sounds great with the ccs loading the 76 I’ll be a happy amp builder. If it doesnt I’ll look for a reason, one of which might be the 76 or the ccs. Immediately more factors than enough to juggle.

So best is build it as chosen and see how it sounds at the first pass.

Still not heard from Ronald, and note he doesnt offer built versions so eventually I shall build another. But not in time for this project. I doubt that matters if a valve is wildly out I’ll know to swap valves. I diubt the gm70’s will vary much.

[Paul Barker]

Oh my calcs are ok I did allow for the parallel nature of anode load. “ So use CCS load, even if we conservatively allow 500k because the load resistance is parallel with the input resistance of the next stage. The A becomes a respectable 12.62.” so 12.62 is correct.

[Mike H]

[Paul Barker]

So finding 6J5 equivalents Ive found a nos boxed pair with the micky Mouse ears. But also used L63 grey glass. Iv ev put two copper gm70’s in the bases! The lenvo 6.5 A filament supplies are ready and waiting.

I’ll try get to lockup tomorrow and fumble around for 56’s and 76’s.

Ive ordered ixys 90’s 45’s unobtainium.