• 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.

EL84 Amp - Baby Huey

Ex-Moderator
Joined 2003
Interesting results; the amplifier has a comparable output resistance to a single-ended triode amplifier without any feedback, and you've compared/likened its sound to a single-ended amplifier. The big question has to be whether the reason that single ended amplifiers are considered to sound "nice" is actually something to do with the output resistance and the transition from voltage drive towards current drive? Or is the distribution of distortion harmonics more important? Either way, if we can obtain the same audible result using either SE or PP then the efficiency and cheapness of PP has to be a good thing.
 
EC8010,
With the cross coupled shunt feedback resistor in the 16K to 22K range it does indeed sound very Single Ended Triode-ish. With that resistor pushed up to 47K it cleans up a little more with some added detail but starts to sound more, dare I say it, Solid State -ish (anyway a bit "dry"). I think this gets back to balancing the 2H vs 3H harmonic content rather than trying to squeeze out the lowest possible harmonic distortion level which I'm increasingly convinced is the key to gorgeousness and "emotional content" in the music.

I need to try deleting that 4700uF in the cathode bypass before settling on the final value. I checked what I was using and it is an ordinary commercial type so may be adding some "blurr". When this amp was a "Bevois Valley" I only used the 470uF/50V Blackgate standards across the then 470R Cathode resistors.

Anyway everyone can "tune" this shunt feedback set resistor to their own personal preference and speakers. I think mine is going to be at around 27K or 33K BUT that is with my nominally 6 Ohm VAF DCX Speakers.

Some more swapping between Elke Brooks and Gwen Stefani coming up this weekend - with some Robert Palmer as a sanity check.

Cheers,
Ian
 
ctaudio,
I tried removing the 4700uF bypass cap. Yes it did cleanup and remove a layer of "blurr" and did much as you described BUT I thought it lost too much bass. A search thru' the capacitor bin did'nt find anything to replace it with but did turn up 4 more 470uF/50V Blackgates so I dropped those in. bypasses are now 2 x 470uF/50 Blackgate and 1uF PPS. I'm waiting for the Blackgates to settle in (you can't actually tell what Blackgates are going to finally sound like untill they've been in for about 40 hours, plus and the last bit of settling in usually involves a large lift in bass) but initial impression is that I still need a bit mor e capacitance.

Re-reading your earlier post I was a bit concerned about you replacing the 16R current set resitors with trim pots to "exactly balance " idle currents. This should not be required BUT more importantly I was wondering how you were then determining and balancing the idle currents. The 16 Ohm is the only fixed resistance in the loop, across which you can do a voltage measurement to calculate the currents. I thought I should check that you are actually balancing the idle currents and not setting the current sources for equal cathode voltage or some other scheme which would not guarantee idle current balance.

Cheers,
Ian
 
overload recovery

I just received the Wima and Xicor (low ESR) caps. I’ll try them in the next few days. (Blackgates are not something I have laying around the house.)

I've been reading the Morgan Jones book. Over the weekend, I decided to look at the overload behavior of the amp. I used a USB soundcard to alternately drive the amp into overload and just before visible distortion (@200Hz). I captured the waveforms at the output driving an 8ohm nominal speaker using three different bypasses: (1) 4700uF+470uF+1uF, (2) 470uF+1uF, and (3) 1uF.

http://cheunglai.smugmug.com/gallery/1463716

Between case (1) and (2), I don’t see much difference. There is a slightly longer recovery time with the 4700uF, but not much. Unless we are crossing some magic threshold of the human hearing, the difference in recovery time does not seem to justify the difference in sonics between the two cases. Something else may be going on there. One thing I noticed though is the crossover distortion in overload in both cases.

I then tried removing the 470uF also, leaving only the 1uF in the bypass, mainly for my curiosity (case 3). I noticed a couple of things. The amp goes into overload at a much lower level. The crossover distortion in overload is gone. The recovery is now almost instantaneous. I have not done much AB comparisons between case (2) and (3) yet. But the bass is obviously weaker in (3). The mid and highs are at least as good. It’s not very useable as a full range amp.

These are just raw data. I’m not sure what to make of these yet. Perhaps someone will educate me?



On the bias currents, I had to break the connections to the cathodes to make the measurements after the amp warmed up.

On the ECC803S cathode voltages, after I balanced the voltage between pin 3 and 8 to 0, I tried measuring the voltage between the plates (pin 1 and 6). One of the channels has a difference of ~3V. The plate resistors are probably not matching. I tried balancing the plate voltage instead. I didn’t hear a difference. I guess a difference of 3V is rather insignificant. Should I balance the resisters? Should I even worry about it?


Clement
 
arnoldc,
Dashed this off as I'm rushing out the door to pack for buisiness trip. I will be off the air for the next 8 days.

Each bias block on the EL84s passes 38 to 39 mA set by the 16 Ohm resistors. They should be balanced pretty well (influenced by the match in the 16R resistors and the match in the bottom transistors - I used the hfe tester on my multimeter to select 2 transistors with similar and high hfe and 1% resistors for the 16 Ohms). I would expect the bias currents to be the same within say 2 to 3% - that is within 1mA.

The diffamp CCS passes 1.0 mA for 500uA per side

Sorry I'm not familiar with 2C51 tube so can't comment.

ctaudio,
Thanks for posting the overload recovery stuff - have printed it out and will read on the 5 hour plane trip.

Cheers,
Ian
 
ErikdeBest said:
Can I put two matched resistors in serie with the plate of each EL84 and match the bias block measuring the voltage across them?

Another uncommon but simple method is to tie an insulated milliameter (eg handy DVM) ACROSS each plate winding of the OPT.

This works because the internal resistance of the milliameter is much lower than the one of the winding.

And also because the slight plate voltage difference is "don't care" with penthodes or tetrodes.

Yves.
 
Ex-Moderator
Joined 2003
ctaudio: Thanks for posting your interesting images. The problem is that although by changing the value of the capacitance you clearly change the overload recovered time, by not bypassing the CCS you also change the AC conditions of the output stage before overload. That's not a criticism, just an observation that it makes it harder to interpret your results.

To me, your images #1 and #3 are nice illustrations of why I like to avoid capacitors interfering with the bias of output valves.

I have seen a similar waveform to your image #2 (the "square" wave with crossover ringing), and that was on an amplifier with conventional RC cathode bias. I'm assuming that the ringing is something to do with the output transformer (mainly because I can't think of any other obvious culprit).

ErikdeBest: There are two reasons why you shouldn't measure current at the anode. First, it has 300V superimposed on it, so it's downright dangerous. Second, you're dealing with a pentode, and it's cathode current (which is the sum of anode and g2 currents) that is important.
 
EC8010 said:

. . .ErikdeBest: There are two reasons why you shouldn't measure current at the anode. First, it has 300V superimposed on it, so it's downright dangerous. Second, you're dealing with a pentode, and it's cathode current (which is the sum of anode and g2 currents) that is important.

I can't disagree !

However, it could be a way to know the screen current ;)

Yves.
 
I chose the DIY route to learn the effects of the different circuit elements so I can tune the sound to my taste. When I hear a clear sonic difference by removing the 4700uF bypass, I want to understand why.

The Morgan Jones book talks about overload recovery. In his EL84 design, a bypass of 2200uF is theoretically needed to have cut off down to 1Hz. He compromised on a 470uF because of overload recovery reasons. I wanted to see and hear what he was talking about. If there is an “optimal” value for bypass (for my taste anyway), I’d like to find it. If the best bypass is no bypass, I want to convince myself of that before I start making circuit mods.

I was expecting to see a much longer recovery time with the 4700uF, but I didn’t. It’s longer only by 15% or so. But the sonic difference is very clear. It tells me that something else must be going on. But what?

The experiments with the 1uF are really for my own curiosity. If removing the 4700uF makes such a difference, I wanted to see what would happen if I take that further. (100uF may have been a better choice, but I didn’t have any lying around.)

The crossover distortion could be due to the output transformer. I don’t know. One thing the 1uF bypass experiments tell me is that the bypass is somehow related – perhaps indirectly due to the shifted AC operating point. Again, I don’t know. Much more to learn; and that’s what’s fun for me.

Someone on this forum pointed out an interesting article I am still digesting:

http://www.18watt.com/storage/18-watter_buzz_info_311.pdf

I’d appreciate any comments and pointers.


On measuring the bias currents, breaking the circuit is indeed inconvenient. I have jumper wires I can desolder and resolder quickly. I thought about inserting a 1ohm resister into the path. At 38mA, the voltage drop across the 1ohm is only 0.038V. That’s pushing the accuracy limit of my DMM. I’d like to have another digit of resolution. I’m not sure what a 10ohm in the path would do to the sound – especially when I’m trying to study the effect of bypass. Perhaps an experiment for another day. I’d welcome any suggestions.
 
Bias point

In post #134, I talked about reducing the “wasted” power on the voltage dropping resistors by raising the plate voltage and dropping the bias current. That was a bad idea and was the source for the crossover distortion. As I decreased the bias current, I shifted the amp towards class B and the non-linearity near cut off results in the distortion.

I experimented with the bias point a bit in the last couple of days. This time, I took it the other direction. I reduced the plate voltage to the low 200V and increased the bias to 42mA range. At the higher bias currents, I can drive the amp deep into overload with the output looking almost like a square wave and still have crossover distortions that are less severe than before. Sonic-wise, I noticed a very significant increase in bass. Overload recovery time was also reduced. But as the amp tightens up, it also looses “air”.

I now have the amp at a plate voltage of 234V and a cathode current of 38mA. The voltage drop across the 33ohm screen resistor is 0.119V. This gives a screen current of 3.6mA, leaving a plate current of ~34mA and a dissipation of ~8W or ~67% of the max dissipation of 12W.

Gigertube, I don’t know how you selected the bias point of 38mA cathode current and 270-350V plate voltage. I’ll continue to experiment with the bias point. What I have now gives me the best balance between bass and air so far, and is a significantly improvement from before.
 
G'day Guys,
Back on deck - after my trip to sunny FNQ (Far North Queensland).

Ctaudio - the EL84 is a 12 Watt maximum Anode dissipation rated device - So measure anode to cathode voltage (no signal). Then take 12 watts divide by that measured voltage to come up with the maximum allowable idle current.

Note that idle current is anode PLUS screen current so this calculated maximum idle current will be a little on the conservative side BUT therefore good for tube longevity.

I tried pushing up that local feedback set resistor (the cross coupled one) to 47K. Amp sound became very "correct" - it sounded just like a good solid state amp, did'nt care for it much and so am back to an 18K resistor. Much more "air" and "live" sound. It seems to me that just as too much global feedback kills an amp so does too much local feedback.

I'm about to order some 2200uF/35V Blackgates to try as the cathode bypasses. Will be a while - for us poor folks down under the best price/service in obtaining things like Blackgate Caps is to order them from PartsConnexion in CANADA. Strange BUT true.

Cheers,
Ian