All rights of photos and text reserved. Usage of photos or text from my blog on other websites or for any other purpose only with prior permission. If you want to use any material from my blog please contact me by email.

Tuesday, January 4, 2011

Gain, Headroom and Power


Thanks to all who participated in the poll so far! There seems to be a clear trend for a SE amplifier concept on a budget, followed by LCR phonos. Next is cost no object SE amps. I'll let the poll continue for a while. Let's see how it develops. If the trend stays like this. I will show the SE amp platform concept which can be scaled from very simple and low budget to all fancy, all DHT, cost no object.

In the meantime, I will write some random ramblings about some topics. These will outline some of my basic design philosophies. I don't claim these are absolute truths. There are several ways to reach good sound. What I will show is my way to good sound. At least what means good sound for me. People listen very differently and focus on different aspects of what is the 'sound' of a system. But that might be covered in another post. So let's get to the topic of this post. Gain, headroom and power. I feel there exisits some confusion about these terms, they are often interpreted wrongly or even mixed up. So here are my thoughts:


The gain of a system or component describes the amount by which it amplifies the signal from it's input to the output. So it is the ratio output voltage divided by input voltage. This is usually expressed in dB but can also be expressed as a factor. Ok, so far, so simple, what is there to misunderstand?  I have often seen that in power amps gain is sometimes confused with max. output power. People compare two amps in a system at the same setting of the volume control. Almost each amp design has a specific gain which is different from others. So in this case, one will result in a louder volume than the other. But this has nothing to do with the actual power an amp can deliver. Some amps need more input signal (are less sensitive) then others to reach full power. It is important to match the gain of a power amp to the system. That means it's sensitivity should be such that the preamp can be operated in a sensible and useful range of volume control settings. And of course the preamp needs to be able to deliver the required voltage level for full power, with some headroom (more about that later). I've seen many systems which can only be listend at volume control settings of 9 o'clock max. These systems have too much gain.

This brings me to my design philosophy about gain:

The gain of a system should be just as high as necessary, not more

What does this mean? Ideally you should be able to listen to records which have a lower than average recording level, at the loudest volumes you intend to use. To listen to these at the loudest levels you intend, the volume control should be fully or almost fully turned up. In this ideal situation, there is not a tad of gain wasted in the whole signal chain. More gain typically requires more stages, more components, let's use as few as necessary and these of the highest quality we can afford.

Why worry so much about this you might ask. Let's just stick a resistive divider somewhere if the gain is too much, or a range up to 12 o'clock on my volume control is still enough for me. Well the resistive divider will do the trick of course and might be the ticket if you have too much gain and need to reduce it somewhere. But it's just patch work and it's additional components in the signal path. An argument might also be that you just use higher mu tubes. This way you get a lot of gain without increasing the number of amplification stages. True, but higher mu tubes typically have high plate resistances which means they cannot drive a lot. No long cables for example. If you have a gain stage somewhere which really doesn't need to drive a difficult load, you can use the highest possible gain tube there. Then use that additional gain to reduce the gain requirements somewhere else, maybe by using a lower gain line out tube which will yield a lower output impedance. Or a lower gain, lower rp driver tube in the power amp, which will provide more headroom (more about that below).

I have heard claims from people saying that they generally prefer higher gain power amps in their system even if that means they cannot use the full rotation on the volume control and don't really need all that gain. In such a situation I would have a close look at the preamplifier. Resistive volume controls tend to sound quite different at certain settings. So it could well be the case that the higher gain amp just allows the preamp to be operated in the best sounding settings. It could also mean that the preamp does not have enough headroom, or has rising distortion artifacts at the levels which the amp with less gain needs to be driven to a certain power level. So that doesn't mean that the higher gain power amp is better than the others. Maybe this indicates the preamp should be changed and not the power amp.

Basic message: The overall gain of a system should be well matched to the listening habits. Each component should be chosen such that their gain characteristics match to the overall system.


First let's clarify what I mean by headroom. The headroom of a component or gain stage is the difference between the maximum output signal it can deliver to the maximum signal which is required to drive the following signal chain to full power. Here my philosophy is quite different as for gain:

The headroom of each stage in the signal chain should be as high as possible

Why this? Why being stingy with gain and generous with headroom? Quite simple: Each tube no matter how linear it is will have steeply rising distortion as you get close to the max. possible output level. This can be easily seen on the plate curves. While they might be evenly spaced around the operating point, the distances get closer and closer as you reach cut off and typically wider as you reach saturation. So let's stay far away from those corners under all conditions. Another reason is grid current. As the grid voltage get's closer to 0V grid current will set in, causing the grid to become a non linear impedance. Grid current starts not suddenly as the grid reaches 0V, but long before. Small signal tubes typically start to show grid current at voltages between -1 and -0,5V. So again let's stay away from that region, possibly by a good margin.

So how much headroom shall we choose? I consider 6dB as a good headroom. In the context of a driver stage in a power amp this means: If the output tube needs 100V peak-peak for full power out, the driver stage should be designed such that it can deliver 200V peak-peak or more. In preamps I even shoot for 12dB and more. In low budget designs, especially power amps, this might be difficult to reach. So I would be ok to compromise this aspect if on a budget.

For the complete signal chain this means that when driven to full power, the output tube will start to clip long before any other gain stages upstream are near their limit. So the distortion spectrum is largly determined by the output stage characteristics.


For power I have no straight philosophy to determine how much you should aim for. This is too dependent on the speakers you use and your listeng habits. Still this point needs some discussion. I've already written above how power sometimes gets mixed up with gain. Often people ask me for advice which of my power amp designs they should use. Of course my first question is about the power they think they need. I often get answers like 20, 30W minimum, sometimes even higher. Beeing into single ended, things get difficult with increasing power. While 20-30W are still feasable, it already gets costly in this range if you want to have a certain quality level. While more than 30W are still possible in SE, the cost rises exponentially. So my next question if I get requirements of 20W or more is about the sensitivity of the speakers used. Surprisingly often I get replies of 96, 98 or even 100dB, yet people think they need 20 or 30W of power per channel. Even if they are not half deaf hard rock head bangers or have 100 square meter lofts to fill with large orchestral music.

So where does this come from? While most often it is simple misinformation where people just don't know how much power they really need, I did experience other cases. People who have highly sensitive speakers (or at least speakers which claim to be highly sensitive) and who have tried several amps and claim that they get better sound from the higher power ones. Quite often, what they really need is a better damping factor, not higher power. Higher power amps typically (not necessarily always) have higher damping factors (lower output impedance) than the average flea power SE amp. I have actually seen SE amps which have output impedances of 5 Ohms and higher on their 8 Ohm outputs. While I don't think that huge damping factors are really necessary, a factor of under 2 ist quite low. Such amps will sound quite different on different speakers. If the speaker has variations in it's impedance curve (and most have), this will result in colorations.

While it is difficult to build a SE amp without negative feedback which yields very high damping factors, they can be done such to have at least a reasonable damping factor of say, 3 to 5. In my experience such levels are sufficient to be compatible with a wider range of even 'conventional' speakers. To reach these, higher impedance output transformers are needed. Shoot for primary impedance of 4-5 times the rp of the tube or even higher. Of course there are many other factors which determine the output impedance, like copper resistances of the windings. But the ratio between the internal resistance of the tube and the primary impedance of the output transformer is the main parameter which determines the damping factor of the amp.

A very easy test to check if your speaker needs lower output impedance can be done if the amp has different output taps. Quite often it is beneficial to connect a 8 ohm speaker to the 4 Ohm tap. More often than not, the lower output impedance of the 4 Ohm tap outweighs the disadvantage of the lower output power which results from the mismatch.

So, basic message here: Clarify how much power you really need and if your speaker needs some 'control' verify if that really means power or just low output impedance. I experienced that people often are very reluctant to consider a change in speakers. But lets face it, there are many speakers out there which are just not compatible with SE amps. And to my taste these are not the best sounding speakers anyways. So if you want to dive into the world of SE tube amplification, make sure you have the right speaker with sufficient efficiency and linear impedance curve.

This was my first post about some of my basic philosophies more will follow. If you think they make sense or if you have similar experiences, you might also like some of my circuits and concepts.

Best regards



  1. Thomas,
    I like your methods a lot; I use some of them and they work.
    I saw a reference to a separate rectifier for a driver stage PSU. Myself, I use separate PSUs for power stage and driver stage in my SET amps. I'd suggest it is the most important single thing to do in a SET to get great sound. The more usual shared PSU allows between output stage and driver stage through the PSU, and this messes up the leading edges and transients.

    I also like to have plenty of headroom. For the driver stage for a 300B, I like 200V P-P very linearly, and more like 300V P-P at the limit with a bit of distortion. As I like an active loaded driver stage for good drive capability, this needs a *very* high B+!

    I also agree about gain. I have recently made a new preamp with a lower gain valve, otherwise very similar to the old preamp, and the improvement is very substantial.

    Good words about power and speaker matching. A speaker must work well with the high output impedance of a SET amp if it is to sound good. But this is not black art, it is engineering, well understood.
    Get this balance right and bass can be fast, tight, tuneful, dynamic, rhythmic ... all the things that some people say SETs can't do!
    Myself, I designed the speakers to work with my amp, and tuned them by measurement and by ear; a tricky process but so worth it.

    I also found the 4 ohm output worked best. Well worth trying!

    So it takes some effort to get the best from a SET based system; but it is *well* worth it, IMHO.
    Mozart is playing on BBC R3 FM, with a soprano singing:
    Nothing gives orchestral colour and tonality like a SET.
    Nothing gives transparency like a SET.
    Nothing plays Mozart like a SET.
    And above all, nothing plays female voice remotely like as well as a SET!
    Music so beautiful and emotional, it brings tears to my eyes. Isn't this why we do it! :-)

    Mike C
    Tube DIY Asylum

  2. Tha gain should be sufficiant for max power designed / input standard value... A few db neded for lower level recordings and eventual uncompensated HDCD standard, should be on the preamp side...
    So I can agree with your notion about the needed gain in general.
    We are owrflowed with devices that have very very high sensitivitu inputs like few hundred mili volts. And that is awful to cope with, pot is always at the minimum, and driver stages are suplyed with hig output resistanse for mthe previos input stage of power amplifier...
    Zoran from diyaudio

  3. Thomas,

    Do I understand your comments correctly:

    Good headroom design would result in +6db input and output design ? For example:

    - First stage delivers 8Vpk-pk, than the second stage should have ideally -9V Ug (assuming tube starts to draw grid current at -1; normally -5V would be sufficient, but with +6db we end up with -9V min)

    - Next stage maybe a driver which wants to see 40Vpk-pk. So, our stage should as well deliver 80Vpk-pk (+6db)

    ...the driver than should be able to cope withthis, so Ug needs to be -40V at least...

    correct ?

    Best Regards

    1. yes correct, although this is just a simple rule of thumb and the actual results depend on the linearity of the tubes used