EL84 Amp - Baby Huey
On a couple of threads I have promised to post the schematic for my latest (and best) version of my EL84 Ultralinear Amp. This design is based upon an ECL86 Amp design posted by Yves quite some time ago. This design uses shunt feedback to reduce the output stage rp and thus better drive the limited primary inductance of cheaper output transformers. Regardless of that, the result is good enough that its worth using reasonable quality transformers and while the common Hammond 1608 (Raa = 8K) will work well, it really deserves something a bit higher quality (more expensive). It can be used without the Ultralinear connection BUT performance would be degraded.
It UNASHAMEDLY uses Solid State to assist the tubes toward their maximum performance.
I believe it is suitable as a "newbies" first project and it will certainly out-perform most of the simple "suitable for newbies" designs I have seen here and elsewhere.
I'm off the air for the next 5 days BUT feel free to discuss, criticise etc.
Having said that some component choices were what I had in my parts bins.
All the 0.22uF/400V are Wima FKP1 - I just bought a batch of them
The volume pot could be 50K log or 100K log rather than the 220K.
The MJE340s in the bias blocks don't need heatsinks if mounted against the chassis (they dissipate about 1/2 a watt) else a small heatsink would be advisable.
Set the BAL.(ance) pot for equal anode voltages at pins 1 and 6 of the ECC803S (premium 12AX7)
The 16K resistor between the 47Ks from the EL84 Anodes sets the shunt feedback level. Too much and the input stage will clip and output power will reduce.
The 12K and 470R network in the grid circuit (pin 1 of ECC803S) set the global feedback. I've used minimal global feedback.
Hope yopu like it!!!
Looks good except you have an error - your ECC803S would be woefully underheated since you are connecting it's filaments in series to 6.3V, and the series connection is intended for 12.6V.
Pins 4 and 5 should be tied together as one end and pin 9 used as the other end of a 6.3V filament on ECC 81, 82, 83.
Well spotted. Drawing error - the ECC803S heaters in the amp are actually wired as you suggest.
For those wanting to understand the shunt feedback I offer the following opinion/best guesses and a few facts as to what is happening.
Aside: Newbies don't let the following "techno babble" put you off building one of these. You don't need to understand whats going on to build a great sounding "simple" amp. I believe that this is a fantastic choice as a first tube amplifier - the only thing wrong with it as a first amp is that its going to be VERY difficult for your second and third etc. amps to better.
There will be a virtual earth at the midpoint of the 16K which cross connects the two 47Ks from the anodes of the EL84s. (point with AC voltage = 0). This also somewhat divorces the AC and DC requirements of the diffamp and steps around one of the problems usually associated with shunt feedback using the diffamp anode loads returned directly to the output tube anodes.
The effective power to the diffamp sides therefore is the DC rail plus an AC feedback signal which is provided via an effective AC source impedance of about 8K. The virtual earth point will "shift" with any AC imbalance at the EL84 anodes and tend to correct that imbalance. This "self balancing" action is in addition to the "normal" shunt feedback action using up some of the EL84s gm to lower rp. That is: it does much the same as cathode feedback would do BUT with the additional AC balancing action. The lowered rp of the output tubes better drives moderate Lp of cheaper Output Transformers AND reflects to the secondary as lower amplifier output impedance and thus giving better damping and low frequency response. Similarly High frequency response will be improved by better drive of leakage inductance and shunt capacitance by the lower rp - a win - win - win situation.
Perhaps I should have called the amp "Serendipity" rather than "Baby Huey". Cetainly the tube choices and required operating points and the circuit topology and interactions just happen to all suit each other.
The circuit demands the correct tube choices. The Input (diffamp) devices must be low current and high rp which pretty well limits us to 12AX7ish tubes. The outputs must be high gm devices which will tolerate largish Rg1 values. That limits us to Cathode biased EL84 or if using fixed bias it needs a bias servo or similar to allow the use of that large Rg1. The original ECL86 triode and pentode sections provides both these things and works very well - unfortunately (to the best of my knowledge) ECL86 is not in current production. That was one of the reasons I decided to try the 12AX7 EL84 version (plus I had a Morgan Jones "Bervois Valley" amp which could be readily modified).
BTW it was originally built with a single transistor current source on the diffamp. As with ALL single transistor current sources the AC impedance was limited by EARLY Effect. Changing the current source to a cascode gave immediately noticable improvement due to lifting the AC impedance by (probably) an order of magnitude.
As an intellectual exersize ONLY at this stage I've been thinking about how this might be applied to ECL82 (6BM8) and I think it might be made to work by using the triode sections in conjunction with JFET or BJTs to turn the input diffamp into a cascode diffamp with the JFET or BJT as the lower device in the cascodes. This would satisfy the high rp, low current (and reasonable gain) requirements. I'll leave that for someone else to play about with if interested (except to note that some modelling I did suggests that base current will cause balance problems using BJTs so if you want to give this a go I'd suggest jfets).
As a practical exersize I have built a 6SL7 plus fixed biased 6V6G experimental version but low 6V6G Rg1 requirement (imposed by the fixed bias), lower 6V6 gm, higher currents for the 6SL7 etc. have meant that it just isn't in the same league - I keep running into limiting conditions everywhere as parametrers are pushed outside of acceptable limits - that is what I meant about SERENDIPITY with the 12AX7/EL84 version, in that case everything just ended up suiting everything else. This one is destined to have the shunt feedback deleted and cathode feedback from OT secondaries substituted instead. Even then I'm probably going to have to add MOSFET source followers to buffer the diffamp outputs into the 6V6s. It LOOKS stunning, just doesn't have sound to match - YET. Even so I don't believe it will ever match the ECC803S, EL84 version.
In case I forgot - Thanks again to Yvesm for the original "damn clever" idea.
Ian, nice one! I think that this feedback arrangement is very interesting. Have you tried it with pentodes? High mu, high rp.
One minor point- with a high mu triode in the first hole and its attendant Miller capacitance, you need to keep an eye on the value of the pot. 100K or less would be better.
Yes you are correct. The 220Ks were what I had in the parts bin. I have some notes on construction that I intended to post with the schematic and in those notes was a commect that 100K or even 50K pot would be better for just the reasons you state. I haven't noticed any lack of highs with 220K BUT no sense in compromise when its not required.
Those building from scratch use 50K unless driving from a passive preamp in which case 100K or the 220K shown on the schematic might be better.
I havent tried it with pentodes but I believe Dave (the frugalphile) intended to give this a go at one time. It certainly should work.
Other comments from my notes:
1) Wire the 100uF/0.22uF HV filters for each channel directly between the Output Transformer primary Centre Tap wire with the 0V returned to the same point as the EL84 bias block 0V connections. These are fed from the common HV Supply via individual 47R resistors to give (I hope) a bit of channel separtation. Of course if you have a couple of 100mA rated transformers it would be better to build as dual supplies, one for each channel. The HV secondaries need to be 240-0-240 or there abouts. It will work with HV between about 270V on the low side to 350 on the high side although with 350V DC I would perhaps up the EL84 screen resistors from 33R to say 270R
2) If you don't have a separate 5V winding to generate the -12 to -13V supply for the diffamp CCS then use what ever low voltage winding you have. The voltage isn't critical - anywhere between about -7 and -30 would be OK. You may have to adjust the 1K6 and 1K values in the current source to keep LED current at around 2 to 5 mA.
3) The optional components the CCS are just a filter for any LED noise. I felt they made a very marginal improvement (but it might have been a case of "I expected to hear an improvement and therefore I did" which is a trap for young players to be aware of). The 22uF is a 22uF/6.3V Blackgate NX. If you are leaving them out the 470R needs to be replaced with a short circuit.
4) As a last resort if you have no other low voltage windings (which is what I ran into on my ECL86 version) you can just ditch the hetaer balance resistors (the 2 x 100R) and voltage double the 6.3 volt heater supplies. This will result in the heaters having about a -8V DC bias. I did'nt notice any detrimental effect (hum and noise) on the ECL86 version and you MAY get away with this in the 12AX7/EL84 version as well.
Q: can one used this ckt in triode mode with an ECC83 and EL84? Are there any limitaioned or apphrensions you have in this configuration?
Q2: If one wanted more power would you recommend an EL84 Parallel Push pull with this ckt. Would one have to change the ECC83 to provide more drive?
"Can it be used for EL84 in Triode Mode?"
The main reason for using Triode Mode is to get low rp (tubes internal impedance) which reflects to the output transformer secondary as reduced output impedance (better speaker damping) and a more linear response (less distortion) from the EL84. Triode Mode also presents a problem in driving the increased grid capacitance (Miller Effect) at the EL84 grids and drops the power output to about 1/2.
The combination of the Ultralinear Connection and the Shunt Feedback of the Baby Huey circuit forces the EL84 into Triode like behaviour (in that it both reduces rp and linearises the response) whilst maintaining the higher power output and keeping grid capacitance low.
So the answer is MAYBE but its not required.
This circuit gives you all the performance benefits of triode connection with none of its disadvantages. Its one of the things I like about it. So NOT RECOMMENDED.
"Can I increase Output Power by using Parallel Push Pull"
I think you probably can although I haven't tried it. You would certainly have to manipulate some of the component values and change the output transformer. I would suggest that this might be beyond an Ultra Newbie but I would'nt want to discourage anyone from learning by "diving in the deep end".
As is, it will give you the same power as parallel push pull triode strapped EL84s from its single pair of EL84s, that is 10W per channel.
If I were going to try to scale it up for higher power I would probably use EL34 instead of EL84 and proceed from there. The EL34 has the required parameters (high gm and tolerance of high Rg1 values when cathode biased) required by the circuit BUT I would be worried about the increased signal voltage drive required to be delivered by the front end and the increased grid capacitance of the EL34s compared to EL84s. To pinch an idea from SY above, EL34 with pentode diffamp may well do the trick and deliver something like 25 to 30W. Any takers???
In the unlikely event that I get enthusiastic and want to try an EL34 version - does anyone know of a dual pentode which will run happily at anode currents of 0.5 to 1mA. Else single pentodes like EF86 (6BR7) may work. After all it worked for Quad.
el84pp with different front end
To overcome the problem with a negative supply I wonder if I can use the 5670 tube at the front with the neat smm2210 8-pin chip as a ccs. See attached file. I know the 5670 / 2C51 / 6N3P is a very good sounding tube.
It makes the amp still easier to construct. Would this be a feasible solution ?
If you want a pentode as a driver you should look at the Gary Pimm amps which use the 6au6 and also local feedback.
I have taken the liberty of putting them on my site with the changes of the heater and pot. Makes it easier to post elsewhere as well.
(Updated 7 October 2009 see: http://www.diyaudio.com/forums/tubes-valves/80603-advice-el84-push-pull-amp-schematic.html#post930809) ( 4700uF caps removed from the bias networks shown)
I think you could get around the grid capacitance problem using a pentode connection. In ultralinear, your capacitances are already much higher (maybe about half those of the triode connection?), as the screen grid is not at fixed voltage any more. Running pentode will reduce this and possibly compensate for the larger tube - or more tubes in parallel.
Also, your feedback network would have to be redesigned because you need higher B+ for the EL34, in the current incarnation, it is effectively the B+ for the ECC803S. In order to bring Va down inside specs, you need higher resistors in the anode, which worsens the Miller issue, the high Rg requirement for output tubes, and may still result in too high a voltage on the ECC803S anodes in cold filament conditions. The modified version would be less elegant than it is now for sure :( but it lets you use lower Rp triodes in the front end - for instance, you could put a resistor to ground from each end of the 'common mode' resistor (16k) to ground, creating a voltage divider for the virtual B+ of the front end. The equivalent Rp seen by the triodes can be made lower this way, which means that the feedback resistors must be made lower as well, and that in turn means a lower Rp triode should be used to keep all things equal.
|All times are GMT. The time now is 08:11 AM.|
vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2014 DragonByte Technologies Ltd.
Copyright ©1999-2014 diyAudio