DIY Audio Projects Forum

With the idea for a new amp rattling around in my head, I thought that once again I would let everyone see my process for the design and build of a new amplifier. I haven't done this all in one place since I documented the 300B SET Design Project starting back in September of 2010. I thought some of the newer forum members might find the whole process interesting.

The spark that set this off was my completion of the 6L6 SE-UL optimization study documented here. This provided ample data for the operation of the 6L6 in the SE-UL topology and it looked like a good candidate for a new "Office Amp". Additionally, I've been wanting to tackle a more integrated (i.e. more complicated) main amplifier for a while and this seemed like the right opportunity. As I walk through the design and build I can explain the various choices I'm making, solicit comments and inputs, and answer question from forum members. So, without further eloquence, let's be off.

As always, I wand to start with some requirements to keep the whole process anchored. The list is not extensive, but it does capture my main desires for the new amplifier. Here they are:

1. Multiple selectable inputs
2. Individual left/right volume controls for balance control
3. Tone stack with bass/treble controls
4. Single master volume control
5. All tube / “no sand” design
6. Natural warmth of even harmonics as the amp goes into overdrive.
7. Optimized SE-UL 6L6 output.

The first requirement, "Multiple selectable inputs", is self explanatory and supports my desire to have multiple input sources for my office system without jockeying equipment or cables. This is easy to accomplish with the various selector switches currently available.

The second requirement, "Individual left/right volume controls for balance control", is all about the environment in which the amp will be used. My office is configured for working not as a dedicated listening area. This means that achieving the right channel-to-channel balance is not always easy depending on where I'm sitting or working. Having individual controls will allow me to set the balance and the front end gain of the amplifier.

The third requirement, "Tone stack with bass/treble controls", is also all about the environment. The acoustics of my office are far from optimal and this will allow me to correct for some of that.

The fourth requirement, "Single master volume control", actually works with the previous two. Once I achieved a balance and tonal coloring to my liking, I want to be able to use a single control to vary the overall volume. This single control accomplishes this objective.

The fifth requirement, "All tube / “no sand” design", was brought about by my looking at some old photographs. As I was looking at them I began to think back to my "electronic shop of horrors" from my misspent youth and some of the very old equipment I had there. I wanted to create something that would remind me of that time. So, this amp will be all tube; silicon need not apply. That means no diodes, no regulators, and no LEDs. This one needs to harken back to the golden age of tubes.

The sixth requirement, "Natural warmth of even harmonics as the amp goes into overdrive", is about the sound. As some of you know, I recently boxed up a little class-D amp for my workshop. Now don't get me wrong, I love that little amp. But while working on a project the other day, for some reason I noticed the very "transistor like" sound of the amp. And, whereas this is more than acceptable for my dusty and noisy workshop, I decided it just wouldn't do for my quiet office. So this is something I will have to track as I choose gains and bias points for all the stages.

The seventh requirement, "Optimized SE-UL 6L6 output", goes without saying. It's time to put all that data which I collected to work.

So with all that being said, I have penciled out a basic signal chain design for my new amp. This is shown in the figure below.
In my next post I'll discuss what drove the decisions for this particular order of elements and I'll talk about the overall gain and margin study that went into the decisions.

So what do people think? As always, comments and questions are more than welcome.

_________________
Matt
It's all about the Glass!
http://www.CascadeTubes.com
Cascade Tubes Blog

Hmmm... Strike one? Let's try another post.

So, let’s talk about the order of the elements in the signal chain. When dealing with a multistage signal chain with lots of elements, there are a few driving considerations. The first is gain control and signal swing. The second is distortion. And the third is noise.

Overview:

Because I have four active fixed gain stages (buffer, gain recovery, driver, & power stage) and three variable gain stages (volume, tone stack, & master volume) in this amp I have to be especially mindful of gains and signal swings. Too large a signal swing in certain places can drive following stages into cutoff or compression before the power stage can reach full power. In general, smaller signal swings through the variable stages is preferred.

There is also a complication in the operation of the tone stack. Although the tone stack I’ve chosen (James-Baxandall passive) has a nominal design gain of ≈ -21dB, the gain can vary by as much as 38dB dependent on frequency and control settings. To prevent the inadvertent overdrive of later stages, the tone stack needs to be in the front end of the signal chain where voltage swings are small in the absolute sense.

The solution is to virtually divide the amplifier into two pieces. First a small signal section where gains are set, tonal variations are made, and overall output level set. And second, a traditional power stage which is voltage driven to produce the desired SPL at the speaker. The dividing line can really be on either side of the master volume control in my last post. Hence we are left with the low signal swing portion (i.e. the “preamplifier”) and the large signal swing portion (i.e. the “power amplifier”).

This approach also helps with distortion. Smaller swings in active stages in general means lower distortion. This means that the preamplifier can be made relative low distortion and the overall distortion can be controlled by the design of the driver and power stages.

Now the overall noise figure of the amplifier is largely set by the first gain stage which is the “gain recovery” stage. Before this, each of the stages represents a loss which hurts noise figure. This means we need to pay special attention to the first four stages so we can guarantee a nice quiet amplifier.

Design Decisions:

As I said above, the goal is to have relatively low signal swings in the preamp sections. As such, I was willing to take a hit on noise figure and put the lossy elements (Gain controls, buffer, & ton stack) in front of the first amp to control that swing. During the build I will take extra care to keep the noise low prior to the first gain stage.

The gain recovery stage is just our old friend the 4S. I did this for two reasons. First, the buffer driven tone stack followed by a 4S has been proven to have excellent tracking and performance. Second, there are situations where it may be desirable to increase the gain a little and this topology also handles this eventuality particularly well as evidenced by these response plots.

The only other real choice was the positioning of the master volume control. This could go between the gain recovery and the driver stage, or between the driver and the power stage. The advantages of the former are lower distortion and greater control of the power stage operations. The latter would allow some tonal shaping of the output by the driver stage. I chose the former as the goal in this amp is tonal neutrality of the basic design and this placement of the master volume better supports that goal.

Gain Study:

The next step was to look at the end-to-end gains and see if there are any problems. This required me to make some design decisions about the individual stages. For the preamp I decided to use the stages designed for the Baxandall circuit here.

For the power stage I needed something that would cleanly drive my 6L6 SE-UL stage to its full bias of 32 volts. While investigating various drivers, I was perusing the back of my RCA RC-30 Receiving Tube Manual and noticed a recommendation in the “Resistance-Coupled Amplifiers” section using a 6SL7 with a 3.2kΩ self bias and a 220kΩ load. This as about a 33dB gain stage and has the added benefit of being biased closer to cutoff than conduction and hence is even order distortion biased at high drive levels.

So here is a gain summary spreadsheet with the assumed stage gains and stage bias numbers inserted for each stage as appropriate.

In this example the input to the amplifier is assumed to be 2v-rms (2.218v peak). The gain controls are backed of by -10dB (≈ 2 o’clock) and the master volume is backed of by -20dB (≈12 o’clock). The tone stack controls are set flat with a nominal gain of -21dB.

In this example, the output level is about 16dB down from maximum output (still fairly loud) and there are no overages in the signal chain. However, looking at the tone stack plots above, it is possible to boost the high or low frequencies by ≈19dB above the nominal value. This means that if one was to turn up the bass or treble frequencies at the gain settings, somewhere near the highest settings, the power stage could be pushed into overdrive at those frequencies.

Here is a modification of the spread sheet showing the bass control turned fully clockwise.

Here we see the power stage driven ≈2.7dB into overdrive. However, this assumes first, that the bass frequencies are at the same level as the rest of the music and second, that the gain controls are unmodified. However, by backing off either the gain controls or the master volume by only 3dB the problem is eliminated.

This also demonstrates that with both the gain controls and the master volume set at “half volume” (i.e. 12 o’clock) there remains ample margin for even maximum bass or treble boost with out overdriving the power stage. This is shown below with the gains set as indicated and the tone controls flat.

Here there is ample room for both volume increase and tonal shaping without overdriving anything in the signal chain.

Overall this is a very well behaved and balanced signal chain design. In the next post we’ll solidify the design and discuss some tube rolling options for this amp. As always, questions and comments are more than welcome.

_________________
Matt
It's all about the Glass!
http://www.CascadeTubes.com
Cascade Tubes Blog

Looks like a professional diy project Matt which I will read with great interest and hopefully learn more of this hobby. I am still absorbing the wealth of information in the 300b set project I started reading at the beginning of last year for my little 300b project.

_________________
John

Hi Matt, I love the 6/12SL7s. Especially the old ones with round anodes. Sylvania, RCA and I have a batch of National Union ones (lovely sound). For the 6L6 you might consider the Sovtek 6L6WXT+. It has a little more output and overdrives nicely. When in linear mode it has a clean sound in U/L.

EDIT: I was up fairly close to you about two weeks ago. Visited Gold Beach Oregon. I could get to like that place......even if I am a desert rat.

Good listening
Bruce

_________________
Some of my DIY Tube Amplifier Projects:

• PoddWatt: 5751 / EL84 Push-Pull DIY Tube Amplifier Project
• OddBlocks: KT88 Push-Pull DIY Tube Amplifiers (Mono Blocks)
• ForeWatt: ECC802S (12AU7 / ECC82) DIY Tube Preamplifier Project
• GrooveWatt: Tube (Valve) RIAA Phono Preamplifier Project

Looks great!
I always enjoy these write ups.

Before we get to the detailed design there is one more topic I’d like to discuss: stage coupling and blocking. Now anyone who has followed my design discussions over the years knows that I despise blocking distortion.

Blocking (or bias excursion) takes place when the input voltage to a stage briefly exceeds the bias of the stage and the grid momentarily draws current. When this happens the DC voltage on the coupling capacitor increases and the stage momentarily has lower or no sensitivity. This situation persists until the excess voltage on the coupling capacitor can discharge back to steady state.

The normal way to handle this is to limit the bias excursion recovery time constant to a predetermined short time. But here is the hitch; bias excursion recovery time is inversely proportional to low frequency roll off point. So the shorter the bias excursion recovery time, the higher the low frequency roll off. Normally, in a single stage this is not an issue because the designer can choose a design point that is satisfactory for both. However, if the design cascades a series of stages, then the low frequency roll off losses from each coupling are additive and the overall low frequency response of the amplifier can be seriously compromised. The solution to this predicament is to find out when bias excursions set in and control them only where required. The other couplings can be set at much lower frequencies without fear of bias excursions except during periods of serious signal chain overdrive.

Let’s go back to to the signal chain diagram with some additions:
Here I have added indications of all the places where there is capacitive coupling within the signal chain. These are labeled A, B, C, and D. At each one of these places there will be a low frequency roll off. All of the losses added together will limit the overall low frequency response of the amplifier. As such, we will want to be careful how these points are set. So let’s examine each point in turn.

‘A’: This capacitor is between the input volume control and the cathode follower buffer in front of the tone stack. The input buffer has a maximum AC input voltage 104v peak or 73.5 v-rms. It will be virtually impossible for this to cause a problem so the capacitor can be sized for a very low frequency rolloff. Blocking is not an issue for this coupling capacitor.

‘B’: This is the isolation capacitor at the output of the buffer. The tone stack is directly coupled to the input of the gain recovery stage. In order for this capacitor to contribute to block, the output of the buffer would have to exceed the bias of the gain recovery stage plus the minimum loss of the tone stack (≈ -2.0dB). The following is a gain setup showing the limits.
In this example I’ve used a 2 v-rms line level input voltage. With the initial gain settings at maximum, and the tone controls turned up to maximum boost, there is still about 0.7dB of margin before any grid conduction might occur at this point. And as a point of interest; with these settings the master volume has to be turned down to essentially -34dB to prevent overload in the output stage. This is the amp at essentially full volume.

These setting are very extreme. And with these settings the amp will sound terrible. As such, I think it’s reasonable to assume that blocking should not be much of a problem for this coupling capacitor either.

‘C’: This capacitor is between the gain recovery stage and the master volume control. Given the low bias of the driver (≈2.3v) it might seem that this capacitor could be prone to bias excursions and blocking. However, it is important to see what is happening in the whole signal chain. For this example I have again set the tone controls to maximum boost. However, I have also set the master volume to maximum (i.e. zero loss) and controlled the signal chain with the gain control. Here is the gain spreadsheet.
In this example we see that the driver is just on the edge of overdrive. However, at this drive level, the power stage is already over 10dB into overdrive conditions. So even though it is possible to have problems with capacitor C, it is highly unlikely that this would happen on any regular basis. The only opportunity for a bias excursion here would be the sudden overdrive peak which is in excess of 10dB over the level required to drive the power stage into overdrive. And in this case, the level of driver stage overdrive would likely be rather benign.

So the solution for this stage would be to watch the bias recovery time constant, but not be too concerned if it becomes too long. I will likely make bias recovery time compromises at this coupling point to help preserve the overall low frequency response of the amplifier.

‘D’: This coupling is the DC blocking capacitor between the driver and the power stage. Since this likely the first place where overdrive will occur (i.e. power stage overdrive) this will be the coupling where it is most important to control the bias excursion recovery time. Here, I will likely apply my 400 beats per minute rule and limit the bias excursion time constant to no longer than 30mS.

So the whole reason that I walked through this long explanation is to show how I will be managing the low frequency response of the signal chain. It is however important to remember that this amplifier also has the ability to apply almost 19dB of bass boost over nominal. This capability gives me a little leeway in my overall design process.

As always, questions and comments are more than welcome. Next time I think I’ll dive in to the power stage design.

_________________
Matt
It's all about the Glass!
http://www.CascadeTubes.com
Cascade Tubes Blog

"hese are labeled A, B, C, and D. At each one of these places there will be a low frequency roll off."
I generally use cut off freq of 9hz or lower for couppling caps and 0.5-5hz cut off when I want very good base response

Matt, haven't seen a post on the 6L6 project for a while. Hope all is well.

_________________
John

All is well, I've just been very busy for the last 6 months and had not had time to devote to the project.

As winter sets in, I am hoping to have more time to devote to the design project and associated writeup. My next step is the UL Rp investigation and measurements. This will help in the choice of output transformer.

_________________
Matt
It's all about the Glass!
http://www.CascadeTubes.com
Cascade Tubes Blog

Good to hear Matt. Understand the need to prioritise.

_________________
John