How do you determine a driver stage?

[bluegti]

There is a local friendly competition to build an amp using the 6V6 as the output stage. This is the only requirement, so please do not suggest some other tube, for the output stage.

I know there are a number of existing designs I could use, but I'm trying to come up with my own design as a way of learning. I've read a number of tutorials including the one at Boozhound Labs, but I haven't been able to find a good explanation of the driver stage.

No so much "what is it" but more along the lines of what factors determine what circuit to use, what tube to use, etc?

The Boozhound example uses a simple 6SN7 grounded-cathode for the driver stage. Why is this a good tube (other than its supposed to sound good)? I'm guessing it has to do with the gain, but I'm not sure. If it does have to do with gain, how do I determine how much gain I need?

I'm guessing there isn't a simple answer to these questions, but if you can point me in the right direction, I will be grateful.

[aardvarkash10]

Two words: Morgan Jones

Cheaper but possibly a little less accessible, Fun with Tubes

[Loren42]

Originally Posted by aardvarkash10

Two words: Morgan Jones

Cheaper but possibly a little less accessible, Fun with Tubes

+1 on the Morgan Jones!

[godfrey]

As I see it ...

Step 1: how much voltage?
If you're doing a single-ended triode-connected output stage as per the Boozhound example, the driver needs to provide +-12.5 volts at it's output (about 9 volts rms).

Hmm ... Perfect job for an op-amp! Maybe a nice vintage 741 to go with with whole "retro" vibe? (OK, I'm kidding, please don't do that)

Step 2: how much gain?
For a zero feedback design (like the Boozhound), a voltage gain of at least about 10 is enough. In their example, the input stage has a voltage gain of 15, so the amp will require an input of 600mV rms for full output. (600mV * 15 = 9V) Perfect!

If you wanted to use overall negative feedback, then you'd need higher gain - maybe 50 or so, allowing for reduction by a factor of 4 or 5 when feedback is applied.

Step 3: what topology?
That's easy - if you want voltage gain, you need to use a grounded cathode arrangement. Exception: if you want to use a balanced input, then a long-tailed-pair input stage is needed.

So far, so easy - almost any tube can give 9V rms out and have enough gain.
Now for the hard part:

Step 4: choosing the tube, operating point etc
There's at least 3 options here:

Plan A: Aim for least possible distortion from the input stage.
Plan B: Try to get the right amount of distortion from the input stage to cancel out the distortion from the output stage as best as possible.
Plan C: Worry less about distortion and find something that sounds good.

Plan B sounds smart but there's a slight complication - output stage distortion depends on the load, and even the same speaker will load the output stage differently at different frequencies, so it's a bit like trying to hit a moving target.

Other random thoughts:

For the input stage, you might want to consider noise and microphony, although that's more an issue with low-level signals (e.g. in phono pre-amps.). Shouldn't be a problem here.

Too much voltage gain is no problem - you can always use cathode degeneration to reduce gain (and reduce distortion at the same time).

[Miles Prower]

Originally Posted by bluegti

I'm guessing there isn't a simple answer to these questions, but if you can point me in the right direction, I will be grateful.

There isn't a simple answer here. Although 6V6s aren't exactly a hard load, so that does simplify things a bit.

First of all, you didn't mention how you intended to use those 6V6 finals. Are you going to make pseudotriodes, or use 'em pentodes? This makes a big difference. As a pentode, the reverse transfer capacitance, and hence the Cmiller, will be less. This consideration is an important one since the combined Cmiller + Cgk + Cstray needs to be charged up fast enough to avoid slew limiting at the high end.

6V6 Specs:

Crt= 0.7pF
Cgk= 9.0pF

Rl= 10K (P-2-P)
Po= 10W
Vi=30V(p-p)

Given that, the output at the plate will be: 158.1Vrms or 223.6Vp. The gain is therefore: 15.

Cmiller= 0.7(1 + 15)= 11.2pF

Ci= 11.2 + 9.0 + 10= 30.2pF (Guesstimate Cstray= 10pF)

This gives: Bc= 2pi(30E3)(30.2E-12)= 5.7E-6 (A/V) At 30KHz.

The current that will demand: I= 15 X 5.7E-6= 0.1mA

Your driver should have a minimum current of: Ipq= 5 X 0.1= 0.5mA (By the "Rule of Five" borrowed from solid state practice).

The 6V6 really isn't a formidable load by any means. Of course, this doesn't include the possibility of grid current, which will surely flow during over drive. Still, coming up with a driver isn't that big of a problem. A 6SN7, or one of its relatives, will do nicely here.

Then there is the question of whether this design is going to be single ended or push-pull? Use gNFB or not? That, too, makes a big difference.

[godfrey]

On second thoughts, if you want more than half a flea-power peak output from the 6V6, you'll have to drive it into grid-conduction.

Driving it with a DC-coupled cathode-follower could do the trick.

Something like shown below should be able to kick the **** out of the 6V6 quite comprehensively and give you an extra 5 or 6 dB headroom. (especially with fixed bias)

An input stage with more voltage gain might help too.

[Bandersnatch]

hey-Hey!!!,
I've been looking at that same contest/requirement myself... The thought process went something like this:
How much output voltage do I need? Seems that +/- 20V should cover it nicely. Getting that much from a 250V( give or take a little ) is easy with nearly every single possible choice of tube so on to the next...

How much gain? Given likely linestage availability, I figured 10 would be a reasonable minimum, with something like 50 ( in terms of volts/volts ) would be acceptable. Still within capability of *MANY* good options.

What sort of loading will the power stage present? I'm looking at U-L( or partial triode ) rigging, so it will be between half and a quarter of the triode-rigged input capacitance. Not so large a load that a 10kOhm output impedance from the driver would be working hard. To go one step further, it would want to be idling at between 3 and 10 mA so as to be unpreturbed by the current demands of the load. Still no issues with any likely choices.

So...what will the topology be? Simplicity given the parts bill is a large concern. I settled on the E-Linear circuit. Likely running an EF184 pentode. This simplifies the B+; only a single well-filtered delivery to the amp is made to the OPT. The EF184's plate load is supplied through the U-L tap on the output. Type 12HL7 is another option( and at $2/tube for a known good performer it will certainly get tried ).

Then there is the power supply. B+ around 250V at around 100 mA for a stereo SE amp is fairly easy. PSUDii is used to determine component values. I choose a doubler supply rectified with SiC Schottky diodes and good low-esr 'ytics. A choke in between two sections( the doubler stack and the final capacitor will reduce ripple quite well.
cheers,
Douglas

[bluegti]

Originally Posted by aardvarkash10

Two words: Morgan Jones

I'm in trouble then... I have the Morgan Jones book. :-)

I'm the type of learner that needs to asks questions, get answers, get confirmation using another example. Looks like the answers below will give me a place to start.

[SY]

As long as you have Morgan's book, the input stage for the Bevois Valley amp will work fine with 6V6. And he analyzes it in gory detail.

[bluegti]

Originally Posted by godfrey

As I see it ...

Step 1: how much voltage?
If you're doing a single-ended triode-connected output stage as per the Boozhound example, the driver needs to provide +-12.5 volts at it's output (about 9 volts rms).

Like the Boozhound example, I am planning on a single-ended, triode-connected output stage (at least for now).

I'm assuming that is how you determined "the driver needs to provide +-12.5 volts at its output..." The Boozhound example specifies an operating point of 250V/49.5mA. When I plot this point on the plate curves I see that it lands between the -10 and -15 curves so the driver needs to swing 12.5 volts. (see image below).



Here's where I like to repeat back what I think I understand and get confirmation.

If I were to be using an EL84 operating at 290V/49.5mA then the the driver needs to provide 10 volts (see image below).



Is this correct? If so, this brings me to step 2.

First, did you determine 12.5 volts is about 9 volts RMS by simply multiplying 12.5 * 0.707 = 8.8375V and rounding up to 9V?

Second, I assume the 9 volts you are referring to is being used in the forumla, 600mV * 15 = 9V.

Third, how do you know how much gain you need (keep it simple and assume no negative feedback for now)? You say that "about 10" is enough, but the Boozhound example uses 15 (assuming this is the 15 in the formula above). If I were to use a gain of 10, is the following formula correct? 900mV * 10 = 9V.

Fourth, I see where 15 comes from on the chart in his tutorial (the value in the gain column), but what if I were using a different tube that doesn't have a chart? How do I determine the gain?

Finally, what do I do with the 600mV or 900mV values from the forumlas above?

I hope you don't mind all the questions. Thanks!