Recommend an input stage triode please

[zigzagflux]

Attached schematic. Basic ST-70 iron, variant of El Cheapo, but I wanted a little additional gain, hence the input stage. Don't want gobs of feedback, just enough to give a decent damping factor. Amp will be 6dB HP filtered at about 60Hz.

So which tube is recommended for the gain stage? Tubes I have on hand are few: 6SN7, 12AT7, 6N6P, 7119. This amp is not worth the 7119, IMO, and the SN7 might be considered a waste of a good tube given the ST-70 iron, but they are easy to get. 6N6P is a nice little tube, but likes a lot of current/lower voltages. Tough to get from my 420V B+.

Any dual octal or 9 pin triodes are welcome. No AX7 or AU7 please. THX.

 

[SY]

ECC81 or 6SL7. Go with the classics.

 

[G]

How about a 6C8G? Nah. Too rare. It seems like a 5670 would be a good candidate. Never mind. Not enough gain from either. Sy's suggestion sounds like the way to go.

 

[Eli Duttman]

Attached schematic. Basic ST-70 iron, variant of El Cheapo, but I wanted a little additional gain, hence the input stage. Don't want gobs of feedback, just enough to give a decent damping factor. Amp will be 6dB HP filtered at about 60Hz.

So which tube is recommended for the gain stage? Tubes I have on hand are few: 6SN7, 12AT7, 6N6P, 7119. This amp is not worth the 7119, IMO, and the SN7 might be considered a waste of a good tube given the ST-70 iron, but they are easy to get. 6N6P is a nice little tube, but likes a lot of current/lower voltages. Tough to get from my 420V B+.

Any dual octal or 9 pin triodes are welcome. No AX7 or AU7 please. THX.

Dude,

That's a Mullard style circuit in your drawing. It's "classic" and certainly is a reasonable option.

You can use "El Cheapo" topology, with the NFB connected to the non-inverting triode of the 'T7 splitter by "fronting" it with a common cathode 6SN7 section whose cathode resistor is unbypassed. Such a 6SN7 section is highly linear and is just fine, without loop NFB being applied. As the cap. coupling the voltage gain triode to the splitter is outside the NFB loop, a Soviet surplus PIO part could work out well.

 

[zigzagflux]

Eli:

Yes, I recall you mentioning something like that in a previous thread of mine. How would I compensate the amplifier for high frequency response? Would I keep the paralleled RC across the 6SN7 plate resistor?

Also, wouldn't this method require a negative supply?

 

[Sheldon]

I had a nice commercial EL34 UL amp with 12AT7 on the input and 12BH7 for the splitter/driver. That was before I started building amps, so I don't know how the feedback was applied, but the distortion profile was very good.

Sheldon

 

[Eli Duttman]

Eli:

Yes, I recall you mentioning something like that in a previous thread of mine. How would I compensate the amplifier for high frequency response? Would I keep the paralleled RC across the 6SN7 plate resistor?

Also, wouldn't this method require a negative supply?

Yes, "El Cheapo" topology requires a negative rail. You can get that without any additional "iron" by SS rectifying the B+ and voltage multiplying the 5 VAC winding of the ST70 power trafo.

The "textbook" way to obtain phase compensation uses a cap. in parallel with the NFB resistor. An o'scope is used to tweak the value of the compensating cap., for good square wave performance.

My technique for phase compensation is brute force, but it works reasonably well. Use inductive WW resistors to load the phase splitter anodes. That peaks HF gain. Short the NFB loop out above 80 KHz. by connecting a mica cap. of approx. 150 pF. between the "hot" wire of the O/P trafo and ground. This method relies on the natural HF roll off of the O/P trafo to bring things into alignment.

High gm small signal types are indicated in an amp with loop NFB. High gm provides resistance to HF error correction signal induced slew limiting. Both the 6SN7 and the 12AT7 are reasonably high gm.

 

[SY]

The "textbook" way to obtain phase compensation uses a cap. in parallel with the NFB resistor. An o'scope is used to tweak the value of the compensating cap., for good square wave performance.

At least it was until Crowhurst demonstrated why that might not be such a good approach.

"Understanding HiFi Circuits" has a clinical dissection of that means of compensation.

 

[GordonW]

I was just looking around... and was checking out the 5842/417A for this.

Looks like it could be the "mack daddy" of input tubes. Will work great on low B+ (150V or less), which can work well with the direct-coupled Mullard style phase splitter.

Cgp is only .55pf! Cg is about 9pf... but that part DOES NOT MULTIPLY with gain! For a gain of say, 40, in the input, the 5842 would only have about 30pf effective input capacitance, while a ECC81 would be over 60pf, and the 6SL7 would be over 110pf! This means the 5842/417A would have TWICE the bandwidth of the ECC81, and close to FOUR TIMES the bandwidth of the 6SL7 at the same input driving impedance!

Also, with a plate resistance of only about 1.8K (!) and a transconductance of 24K micromhos (!!), it could drive the crap out of ANYTHING.

As long as a gain of no more than30-40 would work on the input stage, this could be a REAL contender. These tubes aren't cheap-cheap... but I note they can be had for only about $25 each, from some common places...

Regards,
Gordon.

 

[kevinkr]

I was just looking around... and was checking out the 5842/417A for this.

Looks like it could be the "mack daddy" of input tubes. Will work great on low B+ (150V or less), which can work well with the direct-coupled Mullard style phase splitter.

Cgp is only .55pf! Cg is about 9pf... but that part DOES NOT MULTIPLY with gain! For a gain of say, 40, in the input, the 5842 would only have about 30pf effective input capacitance, while a ECC81 would be over 60pf, and the 6SL7 would be over 110pf! This means the 5842/417A would have TWICE the bandwidth of the ECC81, and close to FOUR TIMES the bandwidth of the 6SL7 at the same input driving impedance!

Also, with a plate resistance of only about 1.8K (!) and a transconductance of 24K micromhos (!!), it could drive the crap out of ANYTHING.

As long as a gain of no more than30-40 would work on the input stage, this could be a REAL contender. These tubes aren't cheap-cheap... but I note they can be had for only about $25 each, from some common places...

Regards,
Gordon.

Hi Gordon,
I like and use the 5842 a lot, so much so in fact that I now have pretty close to a life time supply in my stash. The one potential caveat is they like to operate at very high plate currents for the best linearity - I generally run them at 10mA and beyond which could be a problem in an ST-70 type design unless allowances are made for the current. I would say using the 5842 would call for the use of the 5687 as a splitter, again at relatively high plate currents.. Great bandwidth and drive capability - would drive quite a few EL34 in parallel with ease. Borrow the 5-20 topology as the basis for this design. Oops I guess I have drifted away from the original intent of the thread..

 

[zigzagflux]

Deviations are always welcome. I learn about more tube varieties and options.

The design is actually for my brother; his first foray into glow-in-the-dark bliss. I was going to run the 6n6p as the LTP, as it has lots of oomph at low voltages; the high current demands, as Kevin mentioned, steered me away from it. Although, he did buy the new Triode Electronics 'upgraded' power tranny, and I run the EL34's at around 45 mA instead of the stock 50, so there is a little leeway here.

The 5842 really is a nice performer by the datasheet; but one triode per glass, errr.... not for my brother. Me, sure. Him? nah.

Really don't think I need tons of gain at the input stage, though. I'd have to run through the numbers (again). With direct coupling, I would agree, Gordon, the input stage wants a tube capable of operating with low plate voltages. This has frustrated me, trying to pick operating points that still work with the direct coupling.

The AT7 thus far looks to be in the lead I guess. SY, that operating point has me a little confused. Read your RLD article at Bas' site, where you mentioned great linearity at 100V, 2mA. I just don't see linearity on the curves with your 133K resistor. What am I missing? Lots of 2nd harmonic, gets filtered by the PP stage, sure. Was it measurements alone that confirmed this operating point?

Honestly, if I'll be using the AT7 LTP at 3mA per triode, I have plenty of current to work with. Maybe the 6n6p would be a good option here ? 150V B+, 16mA ? Lot of current just to feed a LTP....

 

[GordonW]

I would agree, Gordon, the input stage wants a tube capable of operating with low plate voltages. This has frustrated me, trying to pick operating points that still work with the direct coupling.

The 6DJ8/6922 can probably do what you want, at low plate voltages. It, in fact, can't go over 135V, period, for DC plate voltage. Pretty low plate impedance, too. Gain is a bit low (33 for the tube, lower than that for a stage with cathode feedback), but if that isn't a problem, it should work. It likes decently high plate currents... but not as much as the 6n6, IIRC...

Regards,
Gordon.

 

[SY]

zzf, the 2nd harmonic doesn't get filtered by the output stage- as for as that stage is concerned, it's part of the signal. But at the full swing of the input stage (11V peak or 22V p-p), the 12AT7 with a 133k load run at 2mA showed less than 0.1% second.

 

[ray_moth]

An alternative to direct coupling, which both avoids problems with voltages and preserves LF stability right down to DC, is a step network. A resistive potential divider is used, with the upper resistor bypassed by a capacitor so that DC is attenuated but not AC. This article explains it to some extent.

For example, a 1Meg resistor connected to the plate of the first stage as the upper arm of the potential divider, bypassed by 0.22uF, and a 500k resistor to ground as the lower arm. The output to the grid of the second stage is taken from junction of the two resistors. This drops the voltage at the second stage grid to 1/3 of the plate voltage of the first stage but does not attenuate the signal. E.g. if there is 180v on the plate of the first stage, there will be only 60v on the grid of the second stage in this example. (Other ratios are possible.)

This is a particularly useful thing to do when coupling one differential stage to the next, because it also helps to minimize the effect of DC mismatching in the first stage.

 

[JoshK]

Thanks Ray for the link. I wasn't familiar with this type of circuit. I hadn't seen that link before.

 

[zigzagflux]

zzf, the 2nd harmonic doesn't get filtered by the output stage- as for as that stage is concerned, it's part of the signal.

Maybe I wasn't semantically correct; I was referring to Morgan Jones, p. 388: "Additionally, even harmonic distortion, caused by unequal gain on positive and negative half cycles, is cancelled, whilst odd harmonic distortion is summed."

This is accomplished at the transformer, not output stage, correct?

An alternative to direct coupling, which both avoids problems with voltages and preserves LF stability right down to DC, is a step network. A resistive potential divider is used, with the upper resistor bypassed by a capacitor so that DC is attenuated but not AC. This article explains it to some extent.

I'll have to chew on that one for a while, Ray, but I do recall that article from a while back. At least for this project, I will try to stick with the topology I have, and not experiment too much. Hate to mess up brother's amp, and I'm superior at messing things up- just read my historical posts !

Originally posted by SY
But at the full swing of the input stage (11V peak or 22V p-p), the 12AT7 with a 133k load run at 2mA showed less than 0.1% second.

Me little brain confused. Attached is my loadline for the AT7 near your operating points, O/P about 2V. This doesn't look linear to me ??

 

[SY]

"Additionally, even harmonic distortion, caused by unequal gain on positive and negative half cycles, is cancelled, whilst odd harmonic distortion is summed."

Morgan is referring to the even harmonic distortion generated by the output stage, not distortion from earlier stages. After all, once a harmonic is added in, it's an inextricable part of the signal. Here's a way to think about it: let's say the input signal consists of a 1kHz sine wave mixed with a 2kHz sine wave at some arbitrary ratio. I pass this through a perfect input stage. What will the output of that stage be? The mixed wave, of course. And that's what the output stage will be fed and will be expected to reproduce.

Now suppose the input stage is not perfect, it has 2nd HD, and I feed it a 1kHz sine wave. The output stage is again fed a mix of 1 and 2 kHz, and it will reproduce that, since it doesn't "know" that the 2kHz component is distortion.

 

[SY]

This doesn't look linear to me ??

Actually, for a 10 or 11 volt swing on either side of the idle point, it looks pretty good. And on the bench, the actual results with actual tubes confirms that.

 

[zigzagflux]

Got it. Then 12AT7 it is. Post results when complete.

Thanks all !

 

[zigzagflux]

How's this look so far? For free, you really can't beat ExpressPCB software. So easy to use !