[Goobley]

Hi all,
 
While the original Starving Student was an incredibly high value for money amp, it is fair to say that since disappearance of 19J6 and the standard tube becoming the 12AU7 it has lost some of it's bang for buck, with most 12AU7s costing well over $10 for a decent tube. Since building my Starving Student with CCSs on the tubes and a pair of linear regulated PSs (~48V for the plates and 24V for the MOSFETs (powering the tube heaters of course)) I've been trying to think of ways to shrink the amp and increase its performance while still retaining a lot of the characteristics of the amp such as the single MOSFET per channel providing current gain but while removing the blocking cap on the output which significantly modifies the amps dynamics.
 
Obviously these sorts of changes require the addition of a fair number of components, and there is the obvious fact that if the output capacitor is removed then the tube heaters can no longer act as current sinks to bias the FETs into class A. If the tube heaters are no longer needed for this purpose then there is very little reason why one couldn't only use 1 valve for the entire amp with one triode per channel. Given the cost of these valves, the extra $10-$15 gained can go a long way in transistors, regulators etc. to get this amp running without a blocking capacitor on the output and a CCS on the tube.
 
Below is my current schematic, B+ is provisionally 60V to be provided through a a couple of PI filters and the bipolar supply for the output will be either +-12V or +-15V:
 
 

 
The load on the MOSFET is similar to that which is used in the SOHA II (as is the servo, although they're just very standard).
 
I don't yet know if it's best to take a page from the SOHA IIs book and pull the cathode negative while attaching the CCS to the tail of the valve. This is the area I'd like to for help in, as I don't really have any experience of these sorts of modifications. 
 
This design was the reason why I requested a various complementary mosfet buffer designs, but of course class A does seem like a nice and pure output stage in keeping with the original design.
 
Any comments on the design would be welcome so that I can avoid any really foolish mistakes,
Cheers,
Chris

 

[dsavitsk]

It is a fine idea, but there are some things I'd do differently. First, I think that if you are going to bother with a HT supply, it might as well have enough voltage to matter. 200V or more is probably a good idea here. Additionally, there are lots of ways to bias the driver tube, but a resistor bypassed by a cap is probably the worst of them. Similarly, CCSes are a good idea, but those CCSes are not so good -- look at the 10M45 or DN2540 at a minimum. Finally, I think you will want a relay on the output so you don't have DC during startup. Oh, and a voltage regulator is cheaper and better than a couple of pi filters any day. And even that is unnecessary if the CCS is sufficiently stiff -- which this one is not.

 

[Goobley]

Okie Dokie, I'll look into those suggestion, thanks very much. This is exactly the reason why I posted here!

 

[Goobley]

Thinking about this now, were you suggesting not using pi filters for the B+ or the output power supply?
Obviously I'd regulate the LV side with standard silicon regulators (although this wasn't that clear in the original post), but are you suggesting a HV (atleast semi discrete) regulator for B+? 
 
With regards to the CCS, from what I can tell, you're suggesting cascoded depletion mosfets with a resistor on the source such as you used in the L'espressivo. Is this the case? Is this a nice high quality, cost effective type of CCS then?
Cheers!

 

[dsavitsk]

I took your original post as suggesting that you were going to use a CLCLC filter. A choke is not a bad thing at all, but a regulator after it can be a huge improvement -- much more than a second choke. So yes, I am suggesting a b+ reg -- it need not be fancy to be effective. The one on this pageworks pretty well.

On the CCS, yes, that would be my initial suggestion though there are certainly numerous other options. For instance, I like using a BJT to set the current as Vbe is more predictable than Vgs on a FET, but cascoding it with a FET (D or E does not really matter) for the impedance boost.

 

[Goobley]

Over the past couple of days I have been fiddling with the Torpedo CCS scheme, are you really running the Grid at -2V? Does one not just lose a lot of the input signal at that  voltage? I've been trying to get my plates to about 130V with a b+ of 150+V along with 12mA through the tube (from the datasheet).
I have had no joy using a torpedo style biasing scheme with a transistor at the bottom of the tube due to not being able to set the current through the reference string without changing the plate voltage.
 
I have however had success with cathode load and a resistor over the plate or purely a plate load and bypassed resistor/capacitor on the cathode. For that I used the CCSs outlined in this document http://bildarkiv.hififorum.nu/Ryssen/Pic/R%C3%B6r/diyAudio-CCS-beta2.pdf are they up to snuff?
 
I'd rather stick to BJTs for the CCS if I can because I still don't feel I understand D-MOSFETS well enough.
 
Cheers,
Chris
 

 

[KimLaroux]

Nice to see you are moving forward with your project. I'll be following your progress. I have a feeling this thread will be filled with relevant information.
 
I actually got the parts to upgrade my starving student to something similar to the SOHA II. I can easily create a 60 V supply for the tubes and a 30 V supply for the MOSFETs, without any voltage multiplier. 60 V is simple as I can just use the SOHA II tube stage. I kinda of put the project on ice wondering if I should just go ahead and multiply the HV to 200V or so. But I'm not confident I can modify a tube circuit to work at the voltage I want it to. Maybe this thread will clarify a few things for me. Or I should just go back to reading every single tube design pages I can find...
 
 
 

 

[Goobley]

SInce I'm going the whole hog with this design (next iteration to be released soon). I wondered whether you guys think it would be sacrilege to add a class T (TA2020 to be specific) mini power amp and use this as a valve preamp. The decent class Ts are supposed to be fairly neutral and with the valve sound 'preamp' i think the combo may be fairly decent.
 
I want class A for my headphones and I also want to be able to use the same amp for whatever small speakers I'll have. 12W into 8ohm is pretty good considering the modest requirement of the class Ts. And there's just no way I could get the quiescent current in my buffer high enough for a class a speaker amp (even 3W RMS per channel).
 
So i was just wondering what your thoughts were on the subject (and also if anyone knows if a ta2020 can be configured for unity gain)
Cheers!

 

[nikongod]

As a preamp why not do the following:
Build it like a headphone amp. Dont even worry about the preamp until the end. 
 
After that,
If you are extremely confident in your output biasing, just re-route the output to the back with a couple RCA's Maybe add a switch so you can go back and forth from speakers to headphones. 
If you are less confident - Cap couple the headphone outputs to RCA's on the back with very nice film caps. 
If you are somewhere in the middle and hate ground loops - transformer couple it. 
 
I'm biased towards transformers, even with a very good DC biasing setup, because I hate ground loops. 

 

[Goobley]

Right well here is my fairly final provisional design, looking for advice on what to improve before getting bits to try this beast.
 
The main amp (1 channel):
 

 
B+ supply:
 
 

 
The multitude of LV supplies:
 

 
Here are a few basic simulation results, taken with a 1kHz input signal and with the amp outputting 1mW into 250Ohm (due to me using DT880s).
 
 

 

 

 
All in all it looks rather promising, I'll also be using the modified epsilon 2 from the SOHA 2 or CTH, but felt no need to redraw it (it seems to take forever!).
 
The tube operating point is currently intended to be about 9mA at 190V with the cathode about 6V positive giving just under 2W plate dissipation.
 
I've been thinking of replacing the OPA134 servo with a TL081 as I see no reason why it wouldn't be good enough and the opamps are much cheaper (it also simulates fine). The SOHA II did however choose the OPA134 for the servo, so there must be a decent reason for it. Could someone enlighten me? Oscillation?
 
I've attached my rough BOM which shows that cost is around €160 pre post including at HiFi 2000 case and a pair of toroids (2*15V 30VA and a 100mA 200V, 2A 6.3V). This also includes a pair of high quality Audyn Cap Plus interstage coupling caps
 
 

Let the criticism commence! (Hopefully)
 
 
 

 

[wakibaki]

What's the servo authority, i.e. what's the range of offsets for which it can correct? I don't think it's a good idea to tie the output to the negative rail in a DC application, but I'm prepared to be convinced. What's your thinking here?.
 
w 

 

[Goobley]

'Twas Mr Cavalli's design originally, tying the opamp to the negative voltage line forces it to always draw current and then force it into class A. Since the buffer is inverting everything on the output, at higher amplitudes it's going to be switching on and off a lot. At least thats what the sims said I was personally expecting offset to settle down and stay fixed... Anyway most good modern opamps have almost no crossover distortion so I wouldnt expect any difference really. It's just been proven to work nicely in the SOHA II.

I've simulated the buffer with up to +/-12V offset and run it normally up to its clipping point (about 27Vpp) and it's coped fine, I'm fairly sure that this servo range is limited only by the opamps power rails and how close to them it will swing.

 

[KimLaroux]

Quote:

[...]
 
The tube operating point is currently intended to be about 9mA at 190V with the cathode about 6V positive giving just under 2W plate dissipation.
 
[...]

 
What made you choose this operating point? I've looked around and there does not seems to be any evidence dictating the best operating point for those tubes, so I'm curious what made you choose this point over another. Also, I've never seen those tubes operating at such a high current. It seems scary, to be honest. How's 9 mA better than 2 or 4 mA?

 

[Goobley]

Quote:

 
What made you choose this operating point? I've looked around and there does not seems to be any evidence dictating the best operating point for those tubes, so I'm curious what made you choose this point over another. Also, I've never seen those tubes operating at such a high current. It seems scary, to be honest. How's 9 mA better than 2 or 4 mA?

 


12AU7s have a fairly low Rp and therefore require a fair amount of current to function as intended, if I were to use 12AX7s, I wouldn't need anything like as much current 2-3mA would suffise. However if you look at this datsheet http://pdf1.alldatasheet.fr/datasheet-pdf/view/225928/ETC2/12AU7.html  you'll see that normal operating conditions for a B+ of 250V is 10.5mA. I find that the graphs on this datasheet are very good. http://www.wooaudio.com/docs/tube_data/12AU7.pdf
 
The operating point was taken to keep the tube a fair way away from max plate disspation (2.7W), whilst still making it nice and linear (I've read in several places that AU7s want 6-8mA through them before they open up and become linear. In other hybrids there is of course no way that you'd see so much current but that is due to the B+ much lower (look at the graphs). Plate current is also, from what I've read a criticism sometimes levelled at the SOHA II (unless it's used with 6DJ8s) as these tubes require more to open up fully.If you look at the torpedo, (which obviously needs more power than this amp then you get (1.8(led)-0.7(Vbe))/68(resistor) which gives a current of 16mA, although I'm not sure of their B+, but I know its around/a bit over 200V. As far as I'm aware for low µ tubes this it's not uncommon to be so high.
 
I'm fairly sure the above is all correct but a lot of the information has been gathered from shady forums such as this one...

 
Thanks for your input
 
Cheers!

 

[wakibaki]

Hey, Goobley...
 
It doesn't say on your profile where you live. You should stick it in, because it makes it easier to make recommendations about components if the responder has some idea of where you can source things e.g Europe or USA.
 
Have you seen these:- http://www.tube-town.net/ttstore/product_info.php/info/p3772_Supertex-LR8N3-Linear-Regulator-450-V.html
 
...450V regulator/current source, only 10mA, but dissipates 0.75W, just like LM317 to use.
 
Or look here:- http://www.diytube.com/phpBB2/viewtopic.php?t=5615
 
..for details of source/sinks built with Ixys (10M45s) or other Supertex (DN2540) parts.
 
Anyway, looks like you are the only one working on a design ATM, looks OK. Just push ahead with it, whatever doubts you have, you have to put aside because that's how you make progress in the real world. Sometimes there are errors in a paper design, you just have to accept that as part of the process, because it is a process that sometimes involves going back and debugging what you have done.
 
w
 
I just ordered a couple of reverb transformers, 3.5W, they cost me <U$20 the pair, going to see what I can get out of them for headphone use, with a chokeless PSU with these Supertex LR8's, one per plate supply.