• WARNING: Tube/Valve amplifiers use potentially LETHAL HIGH VOLTAGES.
    Building, troubleshooting and testing of these amplifiers should only be
    performed by someone who is thoroughly familiar with
    the safety precautions around high voltages.

Trying to work out the function of tubes for purposes of tube rolling

A tube kit amp I'm working on specifies that the output tubes are a gain tube and a follower to reduce output impedance, and lists the best tube types for each stage. In most commercial amps I see, they use the same tube in all locations, presumably for simplicity? Is the gain tube/follower tube common in most small-signal output stages? Should that affect how I think about tube rolling (ie, expensive NOS tubes in the gain, cheaper tubes for the follower)?

I have two real world examples to share, where I am trying to understand the tube rolls. The first is the Audio Research DAC3, which uses four 6922 tubes in it's output stage (schematic here: https://arcdb.ws/Database/DAC3/ARC_DAC3MKII_schematic_and_parts_list.pdf). I assume V1/V2 are the gain tubes and V3/V4 are the followers, so changing V1/V2 would have the biggest impact on sound?

My second example is the Audio Research SP8 preamp (https://www.arcdb.ws/Database/SP8/ARC_SP8_manual.pdf). It uses three tubes in the line stage and three tubes in the phono stage, as well as two more in the power supply. Notably, they output sections use two 12AX7s plus a single 6922. I don't understand this circuit, and don't really have any assumptions about which tube does what.

Anyway, just trying to get a bit more understanding before I stock up on vintage lightbulbs.
I assume V1/V2 are the gain tubes and V3/V4 are the followers, so changing V1/V2 would have the biggest impact on sound?

I don't understand this circuit, and don't really have any assumptions about which tube does what.
It's just two gain stages, plus a cathode follower, with overall feedback to acheive the RIAA curve. Again, the gain stages will have more impact on the sound than the cathode follower. However, it's not as easy to roll a phono amp because the RIAA curve depends on the gain stagecharacteristics, so you would likely end up throwing off your equalization in the process. Or you can think of it as adding some extra, unspecified EQ, which obviously will have a large impact on the sound.
@Merlinb Thank you, that was exactly what I was looking for! For the phono stages: If I source a closely matched pair or appropriately spec’d tubes, is that not enough? Like, if the amp is currently running a matched pair of Electro-Harmonix 6922 and I replace them with a closely matched pair of NOS Mullard 6922, that could throw off the EQ more than doing the same swap in the line stage would change the sound?
Interested in Tube Rolling?

Start with an EL34 amplifier (push pull Or single ended).
Roll from EL34 to KT77 (if you go from a KT77 amplifier to an EL34, be sure to add a wire from Pin 1 to Pin 8).
EL34 and KT77 specifications are as close as they can get (except the needed jumper for the EL34, and KT77 Pin 1 is not connected internally).
Plug and Play (self bias; fixed bias requires re-adjustment or at least measure to make sure the current is the same as the original tube type).

Start with an EL84 amplifier (push pull Or single ended). Or from a 6BQ5 amplifier to an EL84.
Roll from EL84 to 6BQ5.
EL84 and 6BQ5 specifications are a close as they can get.
Plug and Play (self bias; fixed bias requires re-adjustment or at least measure to make sure the current is the same as the original tube type).
Note: Unfortunately, marketing ruined all this when they decided to dual label those tubes EL84/6BQ5 or 6BQ5/EL84.
A True 6BQ5 is a Beam Power tube with Beam Formers.
A True EL84 is a Pentode tube with a Suppressor Screen, g3.

To be fair, start the test setup by sending the Left channel music to Both channels, use two speakers (one connected in opposite phase) bring the speakers very close together (1 Inch) Facing each other, and adjust the balance control, or separate channel volume controls for a very deep null of the sound (it almost collapses to zero). Now, you have adjusted the gain of the two tube types to be equal.
Now set it up with the switches, and only one speaker. Face the speaker forward.

For a definitive listening test, make one channel EL34 and the other channel KT77 (or one channel EL84 and the other channel 6BQ5).
Then send the Left channel music signal from a CD player. . . first to one channel, and then to the other channel.
Simultaneously switch the Input signal to one channel and then to the other channel; and switch a single loudspeaker from one channel to the other channel.

Bring in a lot of listeners, and have one person switch back and forth many times from one channel to the other.
Give that person a break, put him or her with the other listeners, and bring a different listener to do the switching.

depending on exactly how you administer this listening test, you have a double blindfold test, or at least a Pseudo double blindfold test.
Last edited:
  • Like
  • Thank You
Reactions: 3 users
Aha, OK.

Tube rolling with Only the same Tube number.
6922 Only!
No 6DJ8, or other tube types need to apply.

Get ready to test the quiescent current of all those 6922 tubes in your preamp, one by one.
Then . . . decide if you need to make the currents equal before you listen.
Failing to have the same current might not a fair test.
But at least you can decide which tube sounds better, without regard to the different quiescent tube currents.

The same tube at different currents has different plate impedance, rp. And different rp can affect the RIAA equalization on many preamp circuit topologies.

If you don't set up the different tubes in different channels at the same time, and switch the signal back & forth from channel to the other channel,
You are counting on your "Audio Memory" to know which one sounds better.

Which RIAA recordings will you use?
Which phono cartridge will you use?
What is all the wiring capacitance from the cartridge all the way to the RCA phono jack at the preamp.
All these affect the "RIAA" equalization accuracy.

Just plug in the different 6922 tubes, and decide which ones sound better.
Don't worry about all the above complicated factors; rely on your ears.
Last edited:
Quiescent current is different from triode to triode in the same envelope, different from one tube to another tube, etc.
Much of the current difference is due to the chosen bias methods, the particular cathode resistor RK, plate resistor RL, B+ voltage, grid return resistor, Rg, etc.
Some circuits Magnify the above triode to triode, and different tube to tube current differences.
Some circuits Reduce the above triode to triode, and different tube to tube current differences.
Your preamp circuitry may be one, the other, or in the middle.

When our local audio club was at a listening session, and the level of concentration was way to high (people were listening for technical sound differences, not the actual music sound differences) . . .
I sometimes tried to get them to loosen up and relax a little; so I said: "Hey that clarinet player is using a # 1-1/2 reed, he should be using a # 1 reed".
Last edited:
  • Like
Reactions: 1 user
I edited my Post # 9, after your Post # 10.

Using exactly the same portions of groove of a vinyl recording:
If there are technical differences, then there should be musical differences.
If there are no musical differences, why look to change or worry about the technical differences.
The purpose is to make music sound pleasing, accurate, or some other thing.

Caution: Playing a section of vinyl over and over again, without rest time between, will not allow the groove to restore its original position.
That means the music sound is different from one listening to the next listening test.

Have you started to measure the tube plate voltage, cathode voltage, and if necessary grid voltages?
I do not know your circuit topology, so just measure all three voltages.
Installed tube voltages?
Different tube voltages?
Measure a set of tubes voltages. Then listen.
Measure a second set of tubes voltages Then listen.
Try and reduce the number of insertions and removals from the sockets.
Are the voltages close to each other, or far apart (the answer to what is "far apart" requires knowledge of the circuit, and how immune to differences it is).

Otherwise you only know how it sounds to you.
May be a good test, may not be a good test.

Just my opinions
I do not expect anything.

I am a person who has been making electronic measurements since 1959 (simple measurements at first).
It is just one aspect of my electronics experience.

As an example of what to expect, there is a Mullard 3 Watt amplifier, discussed a number of times on different Tubes/Valves threads.
The most often problem noted is that old Mullard EF86 tubes work, most others do not (in that Mullard circuit).

Measure and know.
Do not measure, and do not know.

Tube rolling also has interesting outcomes:
Plug and Play
Plug and Pray
Last edited:
If you can carefully and safely measure those voltages, write them down.
Then post them here.

If you are not experienced as to how to do that, then do not measure them.

I do not have the schematic of your preamp, I am shooting in the dark.
But even with the schematic I may, or may not, have an idea of how much voltage difference is acceptable.
Here is a link to the schematic: https://www.arcdb.ws/Database/SP8/ARC_SP8_manual.pdf

The tube layout is attached.

For general information: https://www.arcdb.ws/model/SP8

My plan was to install the new tubes, and see if I heard a difference, and if not, probably try to resell them.

I’m all for measurements. I’m just not sure what to do with the result? How should I interpret the measurements I take?
Measuring DC voltages will tell how emission is. Measuring distortion and the rate of various h2 h3 etc will tell more.
Tubes definitely work different from each other, since they are hand made.
They significantly vary, not only by brand, but tube to tube , same model, same brand, even if made one just after another.



A typical glass audio tube is made on an assembly line by people wielding tweezers and small electric spot-welders. They assemble the plate, cathode, grids and other parts inside a set of mica or ceramic spacers, then crimp the whole assembly together. The electrical connections are then spot-welded to the tube's base wiring. This work has to be done in fairly clean conditions, although not as extreme as the "clean rooms" used to make semiconductors. Smocks and caps are worn, and each workstation is equipped with a constant source of filtered airflow to keep dust away from the tube parts.

Again: HAND MADE which means inconsistent.

Imagine the grid to cathode distance: 1 mm or so.
A 0.1 mm error (or more) will significantly alter gain (transconductance), bias,etc.

So you will definitely hear differences, but you cn no assign them to any soecific brand.

"Sovtek is bright", "JJ is warm", etc. is nonsense and not backed by statistics.

Most differences posted online are based on "samples of one" so meaningless.

That said, it applies to single tube stages, used "raw", as-is.

The moment you introduce Negative Feedback those differences are erased

Now response depends on NFB network, not on individualtube idiosincracies.

My point being that in your preamp, which uses NFB, you will hear no differences ...beyond Placebo effect that is.
  • Like
Reactions: 2 users