Valve audio circuit design questions.

[Binaural]

Hello all,
I have done a good amount of searching (the net as well as this forum and others) and have not been able to track down the answers to these questions. Sorry for posting so many at once, but it'll be nice to have one thread that has all the answer! Some are general and some are a little more specific, although not confined to one specific project or design. Most of the questions deals with interpreting and working with the data sheets of the valves.

What is the difference (other than quiescent current/voltages) between different bias voltages and also what reasons are there not to center bias a circuit? For example, I am using a 12SN7 for an input tube with a B+ of 400V and Ra is 33k. This gives a value of around 12mA (400/33000) and from looking at the load line, this presents many different biasing points. Center biasing would be around -10V which would give 20V of theoretical headroom. With a max input voltage of 3V, this allows for 17Vs of perspective points of bias. Finally, are there any ever circumstances not to center bias with RC coupling?

Is there an optimal place along the bias lines to operate? It seems for most valves, the farther up the bias line, the straighter it gets. Is this what they mean when they say valves are "linear?" It seems the higher up the line, the lower the 2nd harmonic distortion (more symmetrical amplification) but lower on the line would yield more gain (line is more horizontal) but would increase distortion. I realize this is probably an application dependent question, but if possible, a general answer is needed.

How are load lines calculated for transformer coupling? Is the primary impedance used in place of Ra? If this is the case, for push pull is this value halved to find the operating points for 1 tube or is the full value used and the chart would then represent both tubes? So then to find Ia, the value given on the chart would be for both tubes and to find one tube's Ia, the value would be 1/2 Ia?

When dealing with a gain stage that uses a CCS on the anode such as the 10M45, how is the load line calculated since in this situation the Ra is very high. Wouldn't this put the operating point down in the bottom part of the graph where the lines get bunched up?

While on the topic of CCS's, what is the difference between using them on the cathode or on the anode? What's the advantages/disadvantages of each. It probably isn't possible to use one on both the cathode and anode, but if so, why? Also are CCS ever used in the cathode of a DHT? I've never seen it done and was wondering why this was the case.

Finally in a gain stage that uses fixed bias, is the grid leak always connected to the bias source? If so why? If not does it matter which one should be connected closer to the grid?

I hope all this is comprehensible. Also if there are any books or websites that deal with or have information on transformer coupled circuits, could you let me know? It seems like transformers have been given a bad name in the audiophile community, but I think otherwise. There seems to be a shortage of info about them and any info anyone can pass along is greatly appreciated. This goes for CCS's as well. I'm interested in learning as much as I can about them.

I look forward to your wisdom!

Joe

[Gordy]

Joe, you asked too much for me to answer it all, but here is some of it:

Originally Posted by Binaural

What is the difference (other than quiescent current/voltages) between different bias voltages and also what reasons are there not to center bias a circuit? For example, I am using a 12SN7 for an input tube with a B+ of 400V and Ra is 33k. This gives a value of around 12mA (400/33000) and from looking at the load line, this presents many different biasing points. Center biasing would be around -10V which would give 20V of theoretical headroom. With a max input voltage of 3V, this allows for 17Vs of perspective points of bias. Finally, are there any ever circumstances not to center bias with RC coupling?

Choose the bias point that gives you the results that you are looking for. For most folk that means lower distortion which in turn means ‘as far left as practical’.

Some tubes start to draw grid current at -0.5 or -1 V, so it’s worth keeping away from that region at maximum input swing. So with your example of 3V peak input it would be sensible to start looking at a bias point of around -4.5 V more-or-less.

Originally Posted by Binaural

Is there an optimal place along the bias lines to operate? It seems for most valves, the farther up the bias line, the straighter it gets. Is this what they mean when they say valves are "linear?" It seems the higher up the line, the lower the 2nd harmonic distortion (more symmetrical amplification) but lower on the line would yield more gain (line is more horizontal) but would increase distortion. I realize this is probably an application dependent question, but if possible, a general answer is needed.

‘Optimum’ is defined by the designer (you in this case!). Generally more left = more consistent, but also it means that you may be able to configure the circuit with a lower voltage power supply, which is always a bonus.

When the lines run quite parallel (after the initial upwards curve), such as in a 300B DHT, that is what most folk refer to as ‘linearity’. However don’t talk too loud because many vacuum tube fans don’t like to recognise the very, very parallel (horizontal) lines of some semiconductors! (Oh no, I can hear them sharpening the knife and chanting “Tubes are the most linear device known to man; get Gordy, get Gordy…”)

Originally Posted by Binaural

How are load lines calculated for transformer coupling?

I’ve always done parallel feed so can not advise you.

Originally Posted by Binaural

When dealing with a gain stage that uses a CCS on the anode such as the 10M45, how is the load line calculated since in this situation the Ra is very high. Wouldn't this put the operating point down in the bottom part of the graph where the lines get bunched up?

No, it puts a theoretical line horizontal across the graph at the level / setting of the CCS. When you look at that you will see that (for a great many tubes) the bias point increments are spaced far more evenly, and hence distortion is likely to be reduced. However a portion of current is still diverted to the load and hence not through the tube. This means that the higher the resistance of the load (and hence the lower current transferred to the load) the more horizontal the actual load line will remain, and hence the lower the distortion will be.

Originally Posted by Binaural

While on the topic of CCS's, what is the difference between using them on the cathode or on the anode?

I believe that they can be used on the anode of a common cathode amp, but I’m not sure of the details.

The normal use of a CCS on the cathode is as the ‘preload’ (I don’t know the correct term) of a cathode follower to replace the cathode resistor.

Originally Posted by Binaural

Finally in a gain stage that uses fixed bias, is the grid leak always connected to the bias source? If so why? If not does it matter which one should be connected closer to the grid?

Yes, because the bias will be at a negative voltage and hence there is no other place to connect it. (Or maybe I misunderstood the question). It provides bias voltage and a path for grid current.

Originally Posted by Binaural

…if there are any books or websites that deal with or have information on transformer coupled circuits, could you let me know?

These sites might help:
Tubecad: John Broskie's Guide to Tube Circuit Analysis & Design
Valve Wizard (hello merlinb):How to design valve guitar amplifiers
Aiken Amps (Tech Info): Aiken Amplification

Books:
Valve Amplifiers by Morgan Jones (available new)
Radiotron Designer’s Handbook by Langford Smith (second hand, but may have been republished recently)

[bob91343]

Asking these questions is like asking what sort of gasoline to use. It's entirely dependent on what you are doing with it.

Study the first sections of an RCA receiving tube manual for a good beginning. Terman's handbook is good but may raise more questions than it answers because it's a profound text. The ARRL Radio Amateur's Handbook from a few decades ago has good information.

Transformer coupling is a great idea but the implementation is often wanting due to limitations on transformers. After all, even a perfect transformer will present a very low impedance at low frequencies. The considerations are many, and you need to understand the fundamentals before getting into these questions you ask.

[kevinkr]

Originally Posted by bob91343

Asking these questions is like asking what sort of gasoline to use. It's entirely dependent on what you are doing with it.

Study the first sections of an RCA receiving tube manual for a good beginning. Terman's handbook is good but may raise more questions than it answers because it's a profound text. The ARRL Radio Amateur's Handbook from a few decades ago has good information.

Transformer coupling is a great idea but the implementation is often wanting due to limitations on transformers. After all, even a perfect transformer will present a very low impedance at low frequencies. The considerations are many, and you need to understand the fundamentals before getting into these questions you ask.

Actually take a look at the curvature of the characteristic curves that intersect your load lines and try to pick a bias point where the slopes of the grid voltage curves vs plate current are relatively constant across the range of plate current you propose to swing.

Incidentally in the case of a 12SN7 plate currents of 8 - 10mA typically provide just about the best linearity this type is capable of. (I typically run them in this region.)

Morgan Jones Valve Amplifiers 3rd Edition has a lot of useful information and is also very fond of the 6J5/6SN7/12SN7 family of tubes for good reason. I recommend the book strongly.

[boywonder]

Originally Posted by Binaural

Hello all,




How are load lines calculated for transformer coupling? Is the primary impedance used in place of Ra? If this is the case, for push pull is this value halved to find the operating points for 1 tube or is the full value used and the chart would then represent both tubes? So then to find Ia, the value given on the chart would be for both tubes and to find one tube's Ia, the value would be 1/2 Ia?

I hope all this is comprehensible. Also if there are any books or websites that deal with or have information on transformer coupled circuits, could you let me know? It seems like transformers have been given a bad name in the audiophile community, but I think otherwise. There seems to be a shortage of info about them and any info anyone can pass along is greatly appreciated. This goes for CCS's as well. I'm interested in learning as much as I can about them.

I look forward to your wisdom!

Joe

See Miles Prower's very informative posts #11 & #12 here:

PPP OPT clarification

See the load matching sections 1-4 here:

education+diy

[Miles Prower]

Originally Posted by Binaural

What is the difference (other than quiescent current/voltages) between different bias voltages...

None.

...and also what reasons are there not to center bias a circuit? For example, I am using a 12SN7 for an input tube with a B+ of 400V and Ra is 33k. This gives a value of around 12mA (400/33000) and from looking at the load line, this presents many different biasing points. Center biasing would be around -10V which would give 20V of theoretical headroom. With a max input voltage of 3V, this allows for 17Vs of perspective points of bias. Finally, are there any ever circumstances not to center bias with RC coupling?

There are any number of reasons not to pick a Q-Point right in the middle: you don't need that much output swing, you can move up the loadline and find a Q-Point that gives a better THD estimate.

Is there an optimal place along the bias lines to operate?

Some would say it's where the Rp and Gm change at nearly the same rate. That may, or may not, be "optimal" due to other considerations.

It seems for most valves, the farther up the bias line, the straighter it gets. Is this what they mean when they say valves are "linear?" It seems the higher up the line, the lower the 2nd harmonic distortion (more symmetrical amplification) but lower on the line would yield more gain (line is more horizontal) but would increase distortion.

Not with triodes. The closer the loadline approaches the horizontal (constant plate current) the higher the gain and the lower the distortion. Sometimes you just can't hit those straighter lines since that would put you deep into red plate territory.

How are load lines calculated for transformer coupling? Is the primary impedance used in place of Ra? If this is the case, for push pull is this value halved to find the operating points for 1 tube or is the full value used and the chart would then represent both tubes? So then to find Ia, the value given on the chart would be for both tubes and to find one tube's Ia, the value would be 1/2 Ia?

What kind of xfmr coupling? Output or interstage?

When dealing with a gain stage that uses a CCS on the anode such as the 10M45, how is the load line calculated since in this situation the Ra is very high. Wouldn't this put the operating point down in the bottom part of the graph where the lines get bunched up?

An active plate load is effectively a horizontal loadline.

While on the topic of CCS's, what is the difference between using them on the cathode or on the anode?

At the anode, a CCS develops maximum gain and output swing. At the cathode, it kills gain since it now looks like a huge, unbypassed cathode resistor. Active tail loading can, and has been, used for cathode followers.

Also are CCS ever used in the cathode of a DHT? I've never seen it done and was wondering why this was the case.

It can be done, and has been done.

Finally in a gain stage that uses fixed bias, is the grid leak always connected to the bias source? If so why? If not does it matter which one should be connected closer to the grid?

No. You can return the DC grid return resistor to ground if you apply a positive voltage to the cathode. This is what you do when you use LEDs to bias the VT. Since the forward voltage is stable, and the AC impedance of a forward biased LED is quite low, it's effectively fixed bias with the grid return connected to DC ground.

I hope all this is comprehensible. Also if there are any books or websites that deal with or have information on transformer coupled circuits, could you let me know?

Pete's Site (Go here and get a copy of the Radiotron Designers Handbook, fourth edition. Definitely start there, and you can pick up lots of other helpful books and white papers over there as well.)

Steve Bench's Site

Tube CAD Journal

Frank's -- Esssssss-loads of vacuum tube spec sheets.

[Binaural]

Ok WOW! Didn't expect so much so fast!

Miles, your explination in the posts you linked to was great. It'll take me a day or 2 to digest all of what you wrote, and I appreciate you being so thorough. Thanks for that Boywonder.

As far as the transformer coupling question, I was mainly concerned with basic ol' push pull in either the output or interstage positions. Would the only difference between these is what is connected to the secondary as well as the impedances? I.E. a plate to a grid (interstage) with much less of a step up/down ratio such as the OPT going to a speaker.

Are there any graphic examples of the horizontal loadline with CCS that you were talking about? It'd help me connect the dots a bit easier.

Gordy, hope you sleep safely tonight... don't want them crazy toooob peoples coming after ya!

This is enough info to keep me busy for a while. I actually spent the last couple days on Steve Bench's site. He's got a lot of great stuff on there, inverting a tube?? Never would have thought of it. I'll be back to this thread periodically with new questions. Thanks to all who responded, I've got a lot to digest now. Exactly what I was looking for.


Joe

[Arnulf]

Originally Posted by Binaural

Are there any graphic examples of the horizontal loadline with CCS that you were talking about? It'd help me connect the dots a bit easier.

Can you draw a straight horizontal line at Ia = Ia(q) ? Shortly before you get as far right as B+, it changes slope (it plummets into Ia = 0), because CCS swing its output past its supply rail.

FWIW, no CCS has infinite impedance and no load (which appears in parallel with CCS) is ever infinite so the loadline is never horizontal. Usually it is the grid resistor of the following stage that determines the slope of the loadline when CCS is used.

[Miles Prower]

Originally Posted by Binaural

As far as the transformer coupling question, I was mainly concerned with basic ol' push pull in either the output or interstage positions. Would the only difference between these is what is connected to the secondary as well as the impedances? I.E. a plate to a grid (interstage) with much less of a step up/down ratio such as the OPT going to a speaker

The output xfmr operates as an impedance matching device: to match the Lo-Z speaker to the higher load impedance that the plates operate into. In the push-pull case, how you match depends on the class of operation. For Class AB*, one half will be cut off, and out of the circuit while the other half is approaching its max current. In that case, you're mainly concerned with half the primary. If your load is the usual 8R, and you decide that an Rl= 2K / phase looks good, then the impedance transformation is: 2K : 8R= 250 : 1. This corresponds to a voltage ratio of sqrt(250)= 15.81 : 1. Since you need a balanced two phase system, the entire primary will have a voltage transformation of 31.62 : 1. This will give an impedance transformation ratio of 31.62^2= 999.82 : 1. Therefore, 8R at the secondary will look like 7K998 across the whole primary. Call that 8K, plate-to-plate, and that's the transformation ratio you'd specify when ordering the OPT.

Let's say that you like Class A operation instead, with Rl= 3K. In this case, both halves will be conducting for 360deg, and the whole primary will be involved at all times. For a balanced two-phase system, the two halves are effectively in series. So what you'd want is a transformation ratio of 6K : 8R across the entire primary. So that's what you'd order: an OPT with 6K (P-2-P) : 8R. (That xfmr would also be suitable for Class AB* finals operating into 1K5 / phase). OPT matching xfmrs also like to have a large Xl. The larger the primary Xl, the more closely the actual loadline approaches the straight line approximation, and the less phase shift you have at the low end of the audio band. This is important in that it keeps the loadline away from the high distortion region around plate current cutoff, and it makes it less likely that you'll see low frequency instability under gNFB.

As for the case of interstage coupling, this is a bit different. Since there is effectively no load connected across the secondary, the whole secondary electronically disappears since its load (infinite load impedance) is way above the (wM) transfer impedance. The primary side essentially operates as if it were a choke-loaded stage. For the best results, the inductive reactance of the primary should be at least equal to the tube's Rp at the lowest frequency to be passed. Ideally, the Xl of the primary should be 4 -- 5 times greater than Rp. This is easier to meet with RF circuits than it is at audio frequencies. The best tubes to use on the primary side are triodes with the lowest Rp's, such as the low-u vertical deflection triodes, operated at a stiff plate current to get that Rp down. This operation is quite "CCS-like", and the primary side will see a gain ~u. Secondary voltage step-up can increase the gain even more, but you shouldn't over do that since the driven tube's Ci + Cmiller will appear to have been increased at the primary side, increasing the current demand. Usually, you won't see more than 2 : 1 step up ratios for that very reason.

The other consideration is to operate the driver with fixed bias. You don't want to bias a triode with an unbypassed cathode resistor since this will negatively impact the low frequency performance since that resistor will increase the effective plate resistance. You also don't want a bypass capacitor there either since, with triodes, capacitive currents at the cathode means capacitive currents in the plate. This can interact with the xfmr primary inductance to cause resonances that will result in hellacious distortion. Your best bet is to use fixed bias, and return the cathode directly to ground.

If done right, interstage xfmr coupling can give some excellent results. Of course, that means decent ISTs, and these don't come cheap.

[Arnulf]

Originally Posted by Arnulf

because CCS swing its output past its supply rail.

Only just noticed a typo: the word cannot is missing, "... CCS cannot swing its output past its supply rail." Edit option is gone now.