coppatone GM70 /1

OK, let’s get started to adapt jc morrison’s recent architecture—specifically the one used in the silbatone RI-25—for a GM70 amplifier. This design is primarily for my friend Thomas Schick (yes, ‘Tonearm’ Schick), but also my friend Daan van Egmond got interested so I am going to also cover a variant for his situation.

Long ago I promised Thomas a GM70 amplifier design. At the recent ETF he got inspired by jc’s amp design and said to me ‘how about that, with a GM70 and 2A3?’ Sure, I thought, that sounds reasonable. Over on the GM70 yahoo group this driver is quite popular. Daan was interested but had some 300Bs laying around that he wanted to use as a driver. We’ll do that too.

In honour of the silbatone brand and the fact that GM70s are available with copper (plated) plates, I decided to call this design the coppatone GM70. Don’t take that too serious however, I am not the type of guy to name an amp after a piece of Bach music…

On purpose, I am not going to include the principle schematic in this post. For this round of designing, it suffices to know the following:

• 3 stages of amplification: input, driver, output;
• input and output transformers are used;
• GM70 output tube;
• choke loaded driver, choke or current sink loaded input tube—full gain is available from both.

Where it comes to the top-level design points we are going to check off today, the grass has been mown before my feet by that young, upstart blogger—and friend—Thomas Mayer: gain, headroom and power. Thomas and I seem to have come to exactly the same conclusions, so I let him do the explaining.

Let’s pick up the GM70 datasheet and the Russian one too. Amplifiers are designed from the speaker terminals towards the input. The first design choices concern the output stage and they are intuitive choices: set the tubes up for 100W dissipation, run them in class A2 to get 40 Watts of power out and use a 7k output transformer. Rationalising that:

• 7k output transformers have been used quite a bit before with the GM70; I can see how using higher impedance (10K or more) drives transformer winders nuts and using lower (5K or less) leaves no damping factor—more on that in a minute;
• 100W dissipation where the datasheet says 125W max? Apart from the question if copper plates can take the 125W, the effort is quite futile: to have 25% more power (50W instead of 40) means the ability to play almost 1dB louder through your speakers. Wow;
• in comparison, having 40 Watts of A2 power is a good deal—7.5dB—more than the sort-of 7W one gets out of a 300B amp. That does not sound like 5 times louder, more a good step louder. But it is really valuable as power headroom.

The operating point I pick should come as no surprise: V = √(100 × 7000) = 837V (call it 840), I = √(100/7000) = 120mA.

The grey line sets up our load line by drawing it from twice the current to twice the voltage of the operating point. The orange stretch is the part that has to be exercised to get 40W output. The dotted lines show the operating point smack in the middle—the point of choosing the operating point my way. We can see the GM70s will be biased with around -70V and it takes 130Vpeak on their grids to get to max power. The first (70/130)² × 40 = 11.6 Watts will be in virgin class A1, for those afraid of A2.

Checking the plate resistance naively, the numbers given for µ and gm in the datasheet would lead us to believe it to be 1k1. Looking at the slope of both -60 and -80 grid lines at 120mA show it to be 1k67. With a 7k output transformer we get a damping factor of 4. A 5k OPT would not have been enough.

Grid current? There is a graph for that in the Russian datasheet:

The dotted lines on the right show that for the +60V we will push the grid at maximum and the plate being at 90V at that point, the grid current will be 30mA at that point. That sounds scary, but since driving a grid is driving a diode, a good deal of that 30mA is the DC component and the plate choke of the driver will take care of that.

That is enough on power, now on to headroom. I also set as a mandatory requirement 6dB headroom for the input and driver stages for a power amp. This means that everything before the GM70 shall be designed to deliver 260Vpeak to its grid. And it turns out we just get away with it using a 2A3:

The red line shows the 300V plate limit and the yellow load line shows the 25mA operating point. 25mA is where the zero Volt gridline crossed 40 plate volts—300V minus our 260V design goal. Is that linear enough? let’s have a look at local level, with some simple image scaling:

That looks pretty damn linear to me. So the 2A3 driver will be run at 300V, 25mA. Further checking the 2A3 datasheet shows the rp to be about 1k1 and the µ to be 4. Sizing the plate choke for 1k1 and 5Hz gives us 35H. A quality choke has no more than 10 times the Henries in DCR, so 350 Ohm max. Half of that would be quite excellent because it is the low DCR that deals with the DC grid current of the GM70.

Now for the alternative driver, 300B. With a 400V plate limit we have more choice, but I must say that things look not that linear at viable currents:

I think it is grotesque to run a driver as high as 80mA, but on the other hand linearity does not look good below 40mA. Right now I am thinking 50mA and that would give us 325V on the plate—for a given current, a lower plate voltage gives more linear operation. The grid will be at -70V, rp 900 Ohm and µ 3.85. Choke: 29H, 290 Ohm max.

What threw me off for a moment is that both these drivers have a µ of (about) 4, still they need to be biased at (about) the same grid Volts as the GM70. Well, of that factor 4, half of it goes into the fact that the GM70 grid needs to be driven with about twice its bias volts (130Vpeak) for A2 and the other half into the 6dB headroom. Mystery solved.

The last thing we have to look at today is gain. How much is enough? This directly controls the required µ of the input tube. Thinking of what speakers Thomas has and my own experience of what kind of maximum power is required for louder-than-I-can-take volumes on 89dB/W/m speaks in 65m² room—three whole Watts—I say: set gain so that with 1Vrms input, 4W comes out. This means that even with a standard CD player hooked up without a preamp, one can fully turn it up and get 16W out. If really more is needed, an active preamp that can put out 3.2Vrms will drive the coppatone GM70 into clipping.

So what should be the gain of the input tube? I am going to combine both driver variants here because the 300B µ is so close (96%) to that of the 2A3. Four Watts is 10dB below 40W, so the peak voltage on the GM70 grid for that is 130V/√10 = 41V; that requires 41/4 = 10.25V on the driver grid; 1Vrms is √2 Vpeak; required input gain is 10.25/√2 = 7-and-a-bit. What a nice surprise, instead of (make-me-)nervous computer-age tubes, we end up with superior DHTs being a good match, like 26, 27, 205D and my personal favourite, the 10Y. Already a 6SN7 with a µ of 20 is too much for this amplifier.

Next time we will put all 3 stages together in a schematic and will see that with the design of the input stage, additional requirements will appear for the input tube that will thin out the field of contestants. See you then.