[Goobley]

Well, I've finally finished my CK2III and it's now time to turn my attention to my valve project that started life as a starving student. (Today it runs off dual linear power supplies 24V and 52V, with a CCS plate load 1mA per dual triode). However I get the feeling that it isn't running at it's full potential. I don't listen to music particularly loud so I don't really need the gobs of current that MOSFET output stage on the SSMH provides. I've been toying with the idea of taking this amp full valve and over the past few weeks I've been simulating different topologies (Cathode and White cathode follower output stage, Bijou style futterman etc.) The one that seems to come out on top is the Bijou, ok, I could make a point to point Bijou like the original design but here comes the hard part: I want to keep my budget to 50€ before postage.
 
This therefore necessitates trying to reuse as many components as possible: In my starving student I currently have:
 
- 2*12V 15VA toroid
- 2*22V 4VA potted pcb transformer (currently providing the 52V to the valves)
- A mix of switching power supplies from 5 -19 V (possible heater supply)
- A few decent 63V caps (1000uF, 680uF etc. (totals to about 10mF))
- A 2.2mF 100V cap
- Resistors, assorted small caps
- 4 250V 220nF wima coupling caps
- 3 12AU7s
- Small signal transistors
- Hardware : Pot, jacks, heatsinks etc. 
 
I eventually came up with the following design using the the secondaries of the toroid in series and a voltage quadrupler:
 
 

 
This is perfectly capable of outputing 110dB (25mW) into my DT880 250ohm, which is my target (considerably higher volume than I ever listen at). This certainly comes with a price though as even the simulations happy little naive world reports 0.56% THD... The preamp 12AU7 is being run at 4mA with Vgc=2V so that end is happy, but the output stage runs at about 17mA to put the 25mW out. Whilst this is quite high current for 12AUs the plate dissipation is at safe levels due to the low voltage that each tube sees.
 
If one increases B+ to 250V then simulator THD drops to 0.02% which is rather attractive. There are however a couple of major issues here:
- How do I obtain this voltage (power transformers cost a lot of money and my little toroid can't sustain this circuit with a voltage octupler between them).
- The 12AUs will end up running way too hot in the output stage as far as I can tell (I'd like to keep plate dissipation below 2.2W if possible).
 
I would therefore like to ask you folks for a bit of advice one what the best way to proceed with my budget are.
 
Ideas I have considered (to get the ball rolling):
- Buying a bigger transformer (they seem so expensive, so this will be a large chunk of the budget). Also one can't forget the cost of filtering a bigger higher voltage transformer...
- Getting some 6n6Pis for the output (they can be run hotter but they might really need more voltage so the transformer becomes an issue again)
- Giving up, since I've gone round in circles so many times
 
I'd love to hear any ideas or suggestions that you may have, no matter how odd, feel free to chuck them out here so that I can consider them.
 
I'm certainly not anti-solid-state as the phase splitter above will attest, I just want a nice tubey sound from this amp.
 
Cheers,
Chris

 

[nikongod]

If you want an amp with a stereotypical "tube" sound you do not want best performance. The "tube" sound is specific distortions. 
 
I will begin by telling you what I don't like about your design, and then give a recommendation which I don't like. 
 
To start - no mater what you do: the input capacitance of the mosfets is higher than the input capacitance of the tubes. So T6 & T7 and all the stuff around them are just costing you money and valuable B+ current. R9, R10, R20, R21, T5, and T6 can all be removed. 
 
You can also DC couple T4 to the gain tube, which saves a few more parts - R5, R6, and C2 can all go. 
 
 
Now is where I diverge into weirdness - no global feedback. Global feedback serves to reduce distortions, and as mentioned above we want them. So eliminate R3&R4
 
Next up, since we don't have global feedback we dont need the futerman action. The resistor values people use in the Bijou are incorrect to actually create the positive feedback loop that characterizes the futterman circuit anyways. Remove R7&C7.
 
Since CCS and SS voltage sources serve to reduce that distortion (very nicely) all that SS around the gain tube is SS you can replace with a plate resistor and a cathode resistor. Do not bypass the cathode resistor. You can add a cap later, but no. 
 
I am not a fan of bypassing caps. I find it creates a weird "smear" and prefer just a straight electrolytic to a bypassed cap. Remove C6
 
 
Now that I have torn your design to shreds I must say that I actually like a lot of it - it just wont sound very tubey as you drew it. Do the first 2 things I suggested (remove extra buffer mosfets and DC couple phase splitter to tube) no matter what. After that, it is really up to you. I would personally keep the SS stuff around the gain tube. I really like tubes but don't like the typical tube sound. Weird. 
 
Without the positive feedback loop mentioned above a Futterman is basically a white cathode follower with an external phase splitter. This is a big point. With the positive feedback loop (change the size of R7 to much less and C4 to much more - maybe as low as a few hundred ohms and then pick C4 based on that - experiment) you absolutely must have a global negative feedback loop for stability but the futterman is absolutely amazing. Without the true futterman action you may as well build a Morgan Jones with a more traditional white cathode follower output stage - you can apply global feedback like the Bijou, but the total cost will be a bit less due to significantly fewer parts. 
 
Full disclosure: I dont like the Bijou.

 

[Goobley]

Thanks very much for this long in depth reply
 
Your comments about the phase splitter are actually where I started from; I meant to remove the buffers again, they used to have a purpose and then I changed the design and sort of forgot they where there...
 
I also tried running the phase splitter without the coupling cap and whilst this works for a tube, I can't get a FET to act as a cathodyne while its gate is at or below 1/2 B+. I've looked at the numbers again and again and don't quite know why. Maybe someone with a bit more EE knowledge can comment here?
 
Whilst I know I want some distortion to get a tubey sound (mainly second harmonic, I believe), I don't know how much is too much. Keep it below 5% I would assume (at 110db)?
 
With regards to CCS etc, I'll get the couple of bits I don't already have and make the CCS into removable modules that I can play with and decide on my favourite sound.
 
 
Here's the big question though: topologies,
 
I tried the white cathode follower in sims today, it seems viable.
 
Next up was the Broskie cathode follower, besides the crazy Zo when used with 12AU7s, whether this is an issue with these headphones is something that'd have to be tested in reality.
 
Just for giggles I also tested the two triodes paralleled in a standard cathode follower... It wasn't pretty...
 
White cathode probably seems the best topology that I've looked at so far, this then pulls me towards a Morgan Jones/Cavalli Jones or similar, the main difference being that my B+ is 100V lower than theirs... and I don't know whether it's worthwhile attempting something at this voltage or if the result will just be too half hearted. I did however manage to find an old EI core transformer earlier rated at 20.6V*2/1.05A, this probably has the power to be put through a voltage hextupler or octupler to make a real B+ and make a MJ viable (even though it certainly wasn't designed with AU7s in mind)
 
I'd love to hear any comments and ideas that you may have,
Cheers,
Chris
 
PS: does anyone know how the actual earmax amp generated its B+? Considering it ran off a 19V AC plug pack and it didn't really look like it had room for an internal step up transformer, so I'd guess an octupler, but actual knowledge would be cool!

 

[bidoux]

Quote:

PS: does anyone know how the actual earmax amp generated its B+? Considering it ran off a 19V AC plug pack and it didn't really look like it had room for an internal step up transformer, so I'd guess an octupler, but actual knowledge would be cool!

 

From headwize. This is supposed to be the "[size=x-small]reverse-engineered version of the EarMax".[/size]
Remember that you will also need two heater supplies, but it is already the cas with your first schematic ; on the picture above it seems that they are left floating, doesn't seem like the best idea to me
 
I built  a white cathode follower a year ago, the fun was to build it as small as possible (its inside is so messy

). Actually it isn't supposed to sound good nor to perform very well, maybe you don't want to waste your time building an amp using this topology. I am sure nikongod will have a few things to say on this as well

 

[Goobley]

Thanks for that Bidoux, so they did manage to fit a step up transformer into that small box... impressive.
I think I've decided on where I'm going with this design: a fairly high voltage hybrid. The solid state buffer that I'll be using is wakibaki's LME49600 implementation configured in unity gain, it's a nice clean amp with a bit of punch, I just have no use for it on its own, so I'll see how it works out when it'w buffering tubes. I can't quite decide whether I want to go CCS or not, so I'm gonna get the parts to do both as I said before.
 
Here are the schematics I'm going for, I'll have to use both my transformers, but wont need to buy much beyond a few caps and transistors, so I should be able to keep it in budget.
 

 

 
My most important questions are on those schematics:
 - Is it worth paralleling both halves of a dual triode per channel, does it do anything other than lower Rp and double current (inconsequential here)?
 - If I stay with paralleled triodes, should I implement any mirroring schemes to make them take the same amount of current, or will they even themselves out well enough?
 - Does the operating point look okay?
 - Is this a silly amount of current for the valves? It always looks to me like hybrids like the SOHA II don't let their valves open wide enough... 
 
I have one other request that It'd be great if someone could advise on:
I understand how one draws load lines for valves, but I haven't really worked out how to choose the best value for the anode resistor. Given that I'm looking for a fairly high plate voltage (about 120V) giving a good 15V swing, I'm not sure what the best way to choose either the anode resistor is and what voltage to aim to put the cathode at. So if someone could give me some quick advice on this area of design I'd be really grateful (particularly as I'm not really operating at standard voltages...) 

 
Cheers all,
 
Chris

 

[wakibaki]

There are a couple of immediate things that leap out of the design as being possibly problematic.
 
The first is the CCS with dual LEDs. You don't need more than one voltage reference in the cascode. Secondly, you're running a current source into a current sink. I've found this problematic to achieve a stable bias with in the past, the adjustment is so critical. Typically you will see a gyrator anode load driving a constant current sink where it is desired to use active circuitry both above and below the tube or a constant current anode load with a voltage source (LED or zener) in the cathode circuit.
 
Why do you want to parallel the tubes? They're driving a very high input impedance and there's no particular reason to think that the capacitance at that input is particularly high. You don't need the output of 2 tubes to drive an LM49600 buffer.
 
I see some kind of 2-transistor regulator in the PSU. This is designed all wrong. Derive the reference voltage from the output side with a resistor divider or diode arrangement. As it stands R1 is directly feeding the base of the first transistor and the current output depends on Hfe, which is not a reliable transistor parameter. This is directly in contravention of modern solid-state design philosophy.
 
IME, even though tube spice models can be a bit cranky, they usually offer more insight into how a circuit will function in practice than a load-line based design, and sims are a more straightforward way to try a number of (e.g.) different bias levels if you like to take a kind of scattergun approach to finding an optimum operating point as I do.
 
w

 

[bidoux]

Quote:

 
I see some kind of 2-transistor regulator in the PSU. This is designed all wrong. Derive the reference voltage from the output side with a resistor divider or diode arrangement. As it stands R1 is directly feeding the base of the first transistor and the current output depends on Hfe, which is not a reliable transistor parameter. This is directly in contravention of modern solid-state design philosophy.

I don't really understand this. The way I see it, he is building a capacitance multiplier. Helps filtering and adds a little bit of delay for the high voltage to rise more smoothly, no regulation. It wouldn't be very expensive to implement a simple regulation (just as you described), but that depends on what the op is aiming for.
 
I am not skilled enough to comment on the tube's biasing method, although I think it is how the torpedo design does it, so it could work.

 

[dsavitsk]

There are a couple of immediate things that leap out of the design as being possibly problematic.

The first is the CCS with dual LEDs. You don't need more than one voltage reference in the cascode. Secondly, you're running a current source into a current sink. I've found this problematic to achieve a stable bias with in the past, the adjustment is so critical. Typically you will see a gyrator anode load driving a constant current sink where it is desired to use active circuitry both above and below the tube or a constant current anode load with a voltage source (LED or zener) in the cathode circuit.

Why do you want to parallel the tubes? They're driving a very high input impedance and there's no particular reason to think that the capacitance at that input is particularly high. You don't need the output of 2 tubes to drive an LM49600 buffer.

The second LED in the CCS may be unnecessary, but something has to be there. The LED does not cost any more than a resistor, lights up real pretty like, and works fine. This design has been used by lots of people lots of times -- it works fine. Your "typical" design is also probably fine, better than fine -- it is probably great, but it is not typical.

The bias is a shunt regulator, not a CCS. It also works fine -- better than fine, actually. It works great. The cap bypassing the diodes is probably unnecessary, but it does not hurt anything.

Parallel tubes is also unnecessary, but again it doesn't hurt anything. At 9mA, current sharing is not a worry. If the design were for an engineering department trying to shave pennies, then one side of the triode would be better. For a diy'er, parallel as many as you like. More tubes is better, right?

 

[nikongod]

It looks like this is going in other directions, but maybe this will help someday. 
 
Quote:

Whilst I know I want some distortion to get a tubey sound (mainly second harmonic, I believe), I don't know how much is too much. Keep it below 5% I would assume (at 110db)?

 
I would aim for lower distortion if possible. 
Part of the tube sound is indeed distortion, part is the right distortion, part is none of the wrong sorts. 
Crazy. 
 
 
Quote:

 
Here's the big question though: topologies,
 
I tried the white cathode follower in sims today, it seems viable.
 
Next up was the Broskie cathode follower, besides the crazy Zo when used with 12AU7s, whether this is an issue with these headphones is something that'd have to be tested in reality.
 
Just for giggles I also tested the two triodes paralleled in a standard cathode follower... It wasn't pretty...
 
White cathode probably seems the best topology that I've looked at so far, this then pulls me towards a Morgan Jones/Cavalli Jones or similar, the main difference being that my B+ is 100V lower than theirs... and I don't know whether it's worthwhile attempting something at this voltage or if the result will just be too half hearted. I did however manage to find an old EI core transformer earlier rated at 20.6V*2/1.05A, this probably has the power to be put through a voltage hextupler or octupler to make a real B+ and make a MJ viable (even though it certainly wasn't designed with AU7s in mind)
 

 
I also dont like simple cathode followers. From what I have seen in simulations, measurements, and my experience listening to them they "fall apart" too quickly when used outside of ideal conditions. 
 
White Cathode Follower is just more tolerant of the real world. I find this ironic because people usually build simple cathode followers with HUGE tubes (6as7/6080/similar) and they still suck with low impedance loads.
 
To see if you can "deal with" the output impedance of the Broskie follower or the MJ clone you can just wire up a big fat resistor in series with your current amp. Zo has a pretty significant effect on the sound of some headphones - I find Beyers particularly sensitive to changes here. I prefer Beyers driven from an amp with a bit of output impedance, but this is totally subjective and its easy to try for yourself. 
 
 
 
Quote:

My most important questions are on those schematics:
 - Is it worth paralleling both halves of a dual triode per channel, does it do anything other than lower Rp and double current (inconsequential here)?
 - If I stay with paralleled triodes, should I implement any mirroring schemes to make them take the same amount of current, or will they even themselves out well enough?
 - Does the operating point look okay?
 - Is this a silly amount of current for the valves? It always looks to me like hybrids like the SOHA II don't let their valves open wide enough... 
 
I have one other request that It'd be great if someone could advise on:
I understand how one draws load lines for valves, but I haven't really worked out how to choose the best value for the anode resistor. Given that I'm looking for a fairly high plate voltage (about 120V) giving a good 15V swing, I'm not sure what the best way to choose either the anode resistor is and what voltage to aim to put the cathode at. So if someone could give me some quick advice on this area of design I'd be really grateful (particularly as I'm not really operating at standard voltages...) 

 

 
 
To agree with Wakibaki:
Dont parallel the sections like you did. 
If money was no object you have increased miller capacitance which sucks. 
Since money is an object you are wasting a tube. 
If you do have the B+ current and cash for another tube handy I would look into Long tail pairs. They have some compelling advantages over a simple common cathode gain stage - even if you only need a single ended output. 
 
If you do stay with paralleled sections you don't need to do anything to get them running in an acceptable fashion. With large power tubes you do, but that is not the case here. 
 
The current is fine. You may want to tweak it later, but you will have that under control. I agree that many hybrids run the tubes with way too little current. 
I have no idea if the bias setup you have is correct for your Vgk. 
 
Selecting the cathode/anode resistor is easy. 
CATHODE resistor:
Select an operating point and make a mark there on the plate(anode) curves graph. If you use paralleled tubes use 1/2 the total current and the actual plate voltage. 
Now, on the graph there are the sweeping diagonal lines which represent Vgk (voltage from grid to cathode) interpolate to find what your Vgk is. 
Calculate resistance using the current and voltage. In the case of paralleled tubes use the total current. 
 
Anode resistor is easier, but with the voltages you have it may just not happen. 
((B+)-(plate voltage))/current=resistance. 
 
Respect units. 

 

[Goobley]

Thanks a lot for the really valuable responses. I'm not really sure whether I want a hybrid or a fully tube amp. I'm happy to use either but would like a fairly tubey sound as I already have a clean sounding amp. I wasn't sure about paralleling sections, but yes thinking about miller capacitance, it wouldn't bring any advantages. Right to resolve this question for once and for all (to hybrid or not to hybrid) and avoid wasting anyone's useful time spent posting on my hesitant thread, do you believe that with the 3 12AU7s that I already own, I can build a satisfactory fully valve amp? If yes then I'll turn my attention to doing it properly, if not then I'll fiddle with hybrid ideas on my workbench. Thanks for the great information that you've already posted, Chris

 

[bidoux]

I am pretty sur you can pull off a gain stage + white cathode follower  tube amp using only 3 12AU7. But if you feel too limited by these tubes, you only have to buy other types (lots of very cheap options on ebay) and play around with these. I don't know how tight your budget is, but unless you throw in boutique parts, the pricey part is not the amp itself, rather the case and power supply.

 

[Goobley]

Right, I've been thinking about this design, I don't really just want to do a valve voltage stage coupled to a SS current stage. There a bits of the EHHA philosophy that I really like, directly coupled valve and solid state. So I came up with this (based around the fact that the bottom triode in a WCF is a regulator). I know that the two stages can be coupled directly, but if they are there's very little voltage across the output triodes as the input stage's plate sits at 95V. (Values are far from final, and I don't know how I'm going to power this yet).
 
 

 
This amp is fully class A, quiescent current in the output stage is set 25mA, simulation shows this to be very promising, any comments?
I'll be hoping to build wherever this project goes during July (after my exams) and have been really pleased to have all of you input so far, and anymore input you may have is really welcome 

 
Chris

 

[Goobley]

Sorry to bump this, but I was wondering if anyone had any comments on the hybrid output stage that they'd like to share, while it looks good in simulation a bit of real world knowledge really wouldn't go amiss.
Cheers!

 

[Goobley]

Well I actually knocked something up this weekend using an active load common cathode input stage at 2mA per triode. From my 22V transformer I doubled one winding positively and the other negatively before filtering them to obtain +/-50V giving 100V for the tubes. However I used my 63 V caps in the PS (in a RCRCRCRC filtering scheme (3.2mF per rail)), and they're seeing up to 59V across them. Is this safe, or do I need to get some 100V caps ASAP?
(They're decent Panasonic FC and Nichicon PWMs)
I have to say that I'll probably be inclined to stop here as it sounds bloody nice.
Cheers!

 

[nikongod]

Quote:

However I used my 63 V caps in the PS (in a RCRCRCRC filtering scheme (3.2mF per rail)), and they're seeing up to 59V across them. Is this safe, or do I need to get some 100V caps ASAP?

 
EEEEh, its on the edge. 
If the voltage **never** goes above where it is now (and they dont get too hot) they should last nicely. 
If the voltage climbs - perhaps due to a change in mains voltage they may fail. 
 
Mains voltage fluctuates +/-10% where I live... I usually design with the next voltage rating higher than I need for B+ caps.