24 February 2013

Amps: retubing Marshall 1923C 85th Anniversary - the BIAS nightmare


I was deeply testing lots of pre-amp tubes combinations, as I've commented in my previous blog entries.
Once the pre-amp configuration was clear to me, there was the time to try to swap the stock power tubes with a matched pair of Svetlana Winged C EL34s. Tone wise, I think stock tubes were already Svetlana, since I am not finding a tone change, neither a dimensionality change or more or less harmonics or a change in power or a change in the distortion character.
But, I can say this after a real nightmare trying to biasing myself this amp.

As the Marshall 1923C 85th Anniversary is based in the well known JCM2000 DSL50 platform, I started to search for information about how to bias this amp and, http://www.eurotubes.com has a video that explains how to do it yourself and it seemed really easy.

Well, the operation is easy but... you cannot get good results if you don't take into account some other important things when biasing your new tubes. This entry will go in detail about the things that I had to take into account before I've reached a good biasing of my new tubes.

Being my very first biasing, I was very uncomfortable with the overall process. Now, I would like to save this experience in this blog entry to be sure I can recall everything for the next time.
Since I've bought a matched quartet, next pair of tube will not need re-biasing (hallelujah!) but, I will needed it in some time.

Mixed and confusing information

I've started to search Internet and, basically Marshall forums (http://www.marshallamps.com/forums) and, found a lot of apparently contradictory information and, lots of different biasing values recommended by very different people.

What's biasing?

The first thing we need to understand is that biasing a tube is to provide the right amount of current for the tube while it's in IDLE status and, that means that it isn't receiving any signal (our guitar signal) so, be sure to completely roll off gain and volumes when biasing your amp.
In Idle status, the tube is alive, and there is some flow of electrons that we want to avoid to reach the plate, by negativizing higher the grid (the pin that receives our guitar signal).

The plate (anode) can dissipate a certain amount of power (in form of heat) before starting to deteriorate and, this amount of power is being named Maximum Plate Dissipation Power, and it changes for each tube type and each amp class (class A, class A/B...). This is the energy transformed into heat that the tube is releasing or, how hot can go tube components before they start to disintegrate themselves, while in idle status.

The Marshall DSL50 is a class A/B amp loading EL34 power tubes. The Maximum Plate Dissipation Power for an EL34 in class A/B is 25 Watts.

It is recommended to bias the tube to achieve a Plate Dissipation Power less or equal to a 70% of its Maximum Plate Dissipation Power, for class A/B amps. And a 90% is the maximum for Class A amps.
That means that when biasing, we can get a maximum of (25 W * 70% )  17.5 W of Plate Dissipation Power and, beyond that value, our tube will start to worn really fast.
But, most of people will tell you that a 70% means that the tube works really HOT and, that the recommended ratio is around 60%, being 50% a value that is often seen as COLD.
So, our biasing should achieve a Plate Dissipation Power between the 60% (15 W) and 70% (17.5 W).

The formula is as follows:

Plate Dissipation Power / Plate Voltage = biasing current

So, if our plate voltage was 475 V, our biasing values would be between

(60%) 15 W / 475 V = 0.0316 A = 31.6 mA
(70%) 17.5 W / 475 V = 0.0368 A = 36.8 mA

But, if our plate voltage was 450 V, our biasing values would be between

(60%) 15 W / 450 V = 0.0333 A = 33.3 mA
(70%) 17.5 W / 450 V = 0.0388 A = 38.8 mA

So, we can see that the biasing values are HIGHLY DEPENDENT on the Plate Voltage and, this is just when issues begin.

Plate and Cathode method

If you want to read to someone that knows something about this topic, please, read this thread in Marshall forums: http://www.marshallforum.com/workbench/10890-biasing-70-vs-90-vs-plate-current-vs-cathode-current.html?referrerid=0.

To make it shorter, from the beginning amps where biasing using the Plate measuring method. This is a very dangerous operation (since voltage inside amps are higher than 300V and can go around 500V) and, it's how amp techs did it all the life and, still do. This is the accurate way of biasing an amp and, why it's highly recommended to let a qualified tech to do it.

The second method was issued by some amplifier makers, as Marshall, to provide an easier way for the common guitarist to do its own biasing. This is an "easy" way, not so dangerous and more inaccurate but, the only way we, the guitarists without proper electronics equipment and knowledge, can use (if the amp brings us such a possibility).

The rules about the Plate Dissipation Power still apply to both methods but, we are measuring different things in each method.

In the Plate measuring way, we are accurately measuring the Plate voltage and, therefore, the exact value that allows an accurate biasing procedure.
In the Cathode measuring way, we are summing together two values: the Plate and the Grid currents so, we need to adapt our equations above to do it in the right way.

The maximum power dissipation of the Grid is 8 W and, since we are in class A/B, same 70% rule applies for this component of the tube. So, we have to consider both dissipation powers in the equation, when measuring via cathode. That is:

Maximum Dissipation Power = Max. Plate Dissipation Power + Max. Grid Dissipation Power
33 W = 25 W (Plate) + 8W (Grid)

70% of dissipation will be now: 33 W * 70% = 23.1 W
60% of dissipation will be now: 33 W * 60% = 19.8 W

And applying those values to the equation mentioned above, we will have these values for 475V:

(60%) 19.8 W / 475 V = 0.0416 A = 41.6 mA
(70%) 23.1 W / 475 V = 0.0486 A = 48.6 mA

And, for 450V:

(60%) 19.8 W / 450 V = 0.0440 A = 44.0 mA
(70%) 23.1 W / 450 V = 0.0513 A = 51.3 mA

So, we see that the biasing values are VERY DIFFERENT in the Cathode method than in the Plate method, for same Plate Voltage.

Measuring Plate Voltage?

It seems a very dangerous operation, according to the warnings of most of people very done to amp works and, something I wanted to avoid, since I am not confident on handling voltages of such a high value.
The operation seems to be related to measure the output of the power transformer and, it's not being described in detail, to avoid people to try something like that without the needed knowledge and expertise handling lethal voltages.

So, ok, reading the different threads around biasing the DLS50, I've also found the official biasing procedure by Marshall here: http://www.drtube.com/schematics/marshall/JCM2000%20Bias%20Instructions.gif, which is recommending to bias to 45 mV (or 45 mA, since the bias reading is being done thru a 1 Ohm resistor and, therefore V = A).

While lots of manuals are recommending maximum 38 mV for EL34, this Marshall's value seems to be out of scope but, is it?.
Not really!. If we take into account that the biasing method is cathode reading and, we also consider grid dissipation, a "good value" would be 41.6 mA (@475 V) and, maximum would be 48.6 mA (@475 V) so, this value falls in between both.

Plate voltage varies depending on the source voltage (your mains voltage) but, it also varies depending on the tolerance values of all the components of the amp in the path, as well as it depends on the Power Transformer capabilities.
As per comments of most techs, the usual Plate Voltages are around 470V, but, there are variations between 450V and 500V.

In the ideal world, as the amp was designed, for your mains theoretical value (220V in Europe), the Plate Voltage would take the value of 475V.
In this scenario, the recommended biasing value of 45 mV would determine the following power dissipation:

power dissipation / plate voltage = bias value => power dissipation = bias value * plate voltage
power dissipation (45 mV @ 475V) = 0.045 A * 475 V = 21.38 W

That, represents a 64.77% of the maximum dissipation value so, between the safe 60% and the maximum 70%. It's a reasonable value, indeed.

I will explain later how I've got my amp biased without directly measuring the Plate voltage but, using an alternative approach, measuring the mains voltage. But, first, let me talk about my first wrong attempts. That's how I've learned and reached to that approach.

First attempts (Wrong)

Well, thanks to the probes and the measuring device that comes with the TAD BiasMaster, the operation is easy (if you have the right values) and safe. You haven't to remove the chassis and, you haven't to use your multimeter probes in the narrow space of the three pins used for biasing the power tubes.

My first try went really bad. In first place, the current values measured by the BiasMaster were varying very much between a second and the following one, the amp seemed to be in a constant instability.
That was my first error. I made the biasing WITHOUT turning off volume and gain controls and, therefore, the tubes weren't at its idle status. So take note of this: turn off your volume and gain pots before biasing!. Remember, we want to measure the tube on its idle status. The tone stack was left to their noon positions, also, in following attempts.

Second try, with the volume and gain controls off, I had similar situation than before and, I've seen that the left tube was giving very crazy readings, sometimes even zero. I realized that the biasing I did in the first attempt (that was for 37 mV) was giving me a reading of about 44 mV, probably because the controls were open during the first attempt.

Anyway, this issue with the left tube and the highly oscillating values reading in both tubes made me think if I had some issue with my mains current.
I've used my multimeter (set to 600V AC) and plugged its probes in a wall socket to measure the voltage and, in the space of 1 minute or 2, I've got readings ranging from 215V to 235V, with the numbers oscillating really quickly.

So, I thought: "there is no way I can consistently bias the amp in this situation, because I don't have stable current values". I've decided to leave the biasing for one day later, expecting a more stable voltage source.

Next day, voltage was way more stable, oscillating between 234V and 235V. Very high!. Way over the theoretical 220V that I should have!. And then, my thought was: "if the ratio of disipation varies with the plate voltage and the plate voltage can vary so much, mabye 45 mV is too much for my actually voltage. What can I do?".

So, I've decided to follow an indirect approach, to determine the biasing value depending only on the mains voltage.

The right attempt

Table of biasing values

Seeing that my mains voltages were over the common value, I wanted to make a table calculating the Dissipation Percentages for each Mains Voltage, given a certain biasing value.
My goal was to keep the dissipation percentage just a bit over 60% in the full range of measured voltages (from 215V to 235V) and, a maximum 65% at the top notch voltage value of 240V (which I've seen a few days, some time ago).

The idea was, to have a table of different biasing values to apply depending on the mains voltage I was receiving at time of biasing my amp. Look at this table (click on the image to see it full size):

On top, the data I was using for calculations, the EL34 and grid dissipation values and, the percentages that correspond to maximum dissipation (70%), Marshall value (65%) and the safe value (60%).

The table below left calculates the BIAS values for each mains voltage, based in the theoretical plate voltage. To do that, I've supposed a "lab" plate voltage of 475V for a "lab" stable mains value of 220V so, I can establish a direct relationship between the expected plate voltage depending on the mains voltage

The columns BIAS, correspond to the bias values for a 70% of dissipation (column Max) a 65% of dissipation (column Marshall) and a 60% of dissipation (column Safe).
Rows highlighted in green stand for the standard values that the company should be delivering in my mains wall sockets (between 220V and 230V).

Rows highlighted in yellow (including the green ones), are the values measured in my mains.
240 is the maximum Voltage my power conditioner will support before switching off the unit.

At the end, there are three rows. The first one calculates an average value taking into account the whole range of voltages. The second one just the values highlighted in green and, the third one, the values highlighted in yellow (including green).

This would give me an idea about which are the right values for each disipation ratio depending on the mains voltage. The average values should give me the rigth value to cover all voltage variations while remaining under the target disipation percentage.
You can see that the average values are very close in the three cases and, it seems that 44 mV should be a safe value in my crazy electrical environment, instead of the 45 mV recommended by Marshall.
But, I've read that at 60%, the amp has better dynamics and punch and, the tubes are under way less stress (and last longer). That average 40 mV (safe 60%) are very consistent with the values that Eurotubes is recommending in it's video.

The second table calculates the dissipation percentage for each "plate voltage" (theoretical value calculated based on the mains input), given a certain biasing value. So, we can see how the plate disipation ratio varies with mains voltage variation. Since my readings were of 234V to 235V, I wanted to achieve the 60% on this spot and, avoid going over the 65% at maximum mains input (240V).

If you look to both tables, which rows are related, you will see that the "theoretical" 220V (in dark green), corresponds to a "theoretical" plate current of 475V and that, with a biasing value of 41 mA, this means a dissipation of 59.02% (see last column of table 2).

At the measured voltage, during biasing, 234V, the related Plate Current should be around 505V and, the dissipation corresponding to a biasing value of 41 mA would be of a 62.77%, what is a very comfortable value.

You can see that, biasing at 41 mA, the maximum percentage doesn't reaches the 65%. At 240V, the dissipation would be of a 64.38%, still safe.

I would probably go up in the biasing value, maybe up to the 42.5 mA that some people recommends (instead of the "hot" 45 mA that Marshall recommends) but, I wanted to check first if this would give me a good sound, first.

The damn Phase Inverter tube

Once again, the mixed information drove me to buy tubes for PI position without matching triodes, because some people stands that is even better to the sound to have unbalanced power tubes. So every tube I've bought wasn't checked for balanced triodes. My fault.
I've read (after my purchases), that this is damn wrong. The amp maker is already designing the power stage slightly unbalanced, to give the right sound in the notch positions (when a tube goes down and the other goes up in a pull/push design) and, therefore, we need a very balanced PI tube to allow those power tubes to remain just as unbalanced as they were designed, to get the best tone.

Some time ago, I've read the reports of Myles Rose about all the tests he performed about several batches of tubes of each maker and model. The Sovtek 12AX7-WC was identified as a good tube, with values very close to 12AX7 specification so, I've decided to source a few. Later, I've read in the New Sensor's page, that the 12AX7-WC is highly recommended for Cathode-Follower positions, as well as an awesome PI tube.

Even than my preferred PI tube is the 12AX7-LPS, since I had some reliability issues in one of my amps (the Koch Studiotone Combo), I decided to throw there a Sovtek WC, instead of the LPS.
Both tubes were bought to TAD and, they come with the basic test (oh... it seems to work!) but, without checking for matching triodes (highly recommended for PI tubes).

So, my suspect was that I was taking that crazy readings for the left power tube because I had a PI tube with a bad triode. So, for my next biasing try, I swaped that WC tube (that never suit my bill) with an LPS.


Clearly, the WC tube had very unbalanced triodes (contrary to what Sovtek says, and contrary to what Myles Rose found in his tests). Once the LPS was in place, the readings for both tubes were very consistent and, it was a child's game to bias the tubes to my target value (41mV, 41mA).

I also removed the WC in V3 (cathode-follower) and plugged there a TAD 12AX7-Cz (a verified JJ ECC83S), since I wasn't confident of the WC tube type anymore. I think I am gonna buy tubes only with balanced triodes, seen how inconsistent were the readings with that tube.

Testing the amp, biased at 41 mA

Yes, yes, yes!.

The amp started to sound really good and, today, I've enjoyed playing for first time in the last month.
I don't know what will happen when my mains voltage will drop up to the low 215 V I've measured, maybe it will go very cold, I dunno.

If this is the case, I will rise the bias maybe up to 42.5 mA and I will test again and, maybe, I end biasing it to 45 mA, as per Marshall recommendation but, I don't want to go so high with a mains power that behaves so crazy.
I would probably go straight to 45 mA if I had a Power Stabilizer, that delivered a constant flow of 220 V AC, independently of the mains voltage.

But, look at this table, calculating the disipation percentage for a biasing value of 45 mA:

At 45 mA, the Plate Dissipation percentage can go over the maximum 70%, for mains inputs over 237 V and, I am actually very close (around 235 V !!!).

The safest 44 mV that I've calculated as the average value (see table 1), will bring me a safer solution, since no mains voltage up to 240 V would cause a dissipation over the 70%. See next table:

And, probably, the safest value that will give me a good tone along the whole range of voltages would be 42.5 mV (as read in some forum, casually).

You can see that, this will ensure me a good value (60.06%) in the lower side of the table (215V mains / 466V plate) and that, at the actual mains value (234V) I am getting more or less the same dissipation ratio (65.07%) as if I was applying Marshall's suggestion for "standard" values.
Maximum dissipation ratio would be of a 66.74% for the highest mains voltage (240V).

So, I would test a biasing value of 42.5 mV, just to check that I am achieving a better tone during the whole range of mains voltages, if the currently set 41 mV seems weak when the voltage drops to 220 V to 215 V.

I am fully conscious that this is NOT an accurate method, but it's an approach that seemed to work in this particular case so, I wanted to log it and to share with all you. Anyway, I am under the recommended Marshall's value so, I still have some room!.

I think, I could take same approach with other amps if I would know which is the "standard" plate voltage expected at "standard" mains power, and, what's the maximum dissipation power of plate and grid of each related tube (6V6, 6L6, EL34, EL84, etc).

What I can certify is that a bias of 41 mV, works awesome at my actual mains voltage (234-235 V) and delivers a great tone. That means a disipation value between 62.77% and 63.04%.
Let see what happens when the mains voltage drops seriously down!.

Final Update 25/02/2013

I've re-biased it to 42.5 A (or 42.5 V) and, I am getting a consistent sound along the whole range of mains voltages. Today, voltage was oscilating around 224 to 226 V, what is a very good value and, everything sounded well. If the voltage goes up, I will never reach the 70% of Maximum Disipation Power so, I am safe and, if it goes to 235 V as yesterday, I will be in at 63% (still below Marshall's value). Today, I was around a 62%.

So, I am expecting the amp going a bit hotter when the mains voltage rises (but not so hot to achieve a bad tone) and a bit colder when voltage rises (but not so cold to achieve a sterile tone).
I think is a good compromise value for my crazy mains power.

I am very happy now that my amp was bringed back to life.

No comments:

Post a Comment

Please, feel free to add your comments.