A lo-cost HT-supply true transformer-coupled tube headphone amplifier - Page 7

Yes, I've got a couple of irons in the fire as far as development is concerned.

These things seem to be multiplying...

I'm still interested in this design, so I'm experimenting with what can be achieved at the lowest cost (hence the reverb transformers) and with the greatest simplicity, but also seeing what the best performance obtainable is without pushing up the price too much.  

I've got quite a lot of components for this on hand including some Supertex LR8 450V variable linear regulators. You can use them to build CCSs too. They probably don't work quite as well as a cascoded LM334, I've got some DN2540 MOSFETS to try that with though. I also got hold of some true LDO regulators (MIC2941) for the heater supply.

I've done a lot of sims with different tubes including dissimilar twin triodes. I've also done the smaller circuit board which has a low-ripple HT supply, and the CCSs and voltage regulators on it. This will fit in the small chassis and really simplifies the construction, and it's more versatile than the larger PCB I drew up earlier, you can wire it to identical or dissimilar tubes. You can take this board and build it with minor changes to the resistors and the zeners and run the tubes at currents from 1mA to 15mA. It's got a CRCRC filter with smaller components than the choke filter and each tube has an independent regulator or plate CCS. It's a kind of modern shell with the tubes embedded in it, that lets you set the operating conditions more-or-less without reference to a conventional (resistor) bias calculation, and it extracts the best the tubes are capable of in a common-cathode, cathode bias configuration. The only consideration really is the quality (impedance) of the CCSs, and that can be modified.

I'm angling toward the 6SN7s ATM, but I haven't bought any yet. It's a different socket too, and a different diameter. It has a high Ra, ~7k. Triode load for best power is 2*Ra, which would be 14k. The 6N6Ps have an Ra <=2k.

It's also possible to use 12AX7s for the gain stage, I have one somewhere, it has a LOT more gain than the 6N6Ps or the 6SN7s, then you (I) usually end up adding some GNFB (in the sim), which is necessary to avoid too much swing on the grid of the second tube. Even though the 12AX7s are not known for linearity, the THD still works out pretty good with a bit of NFB, which is to the good in terms of output impedance.

The good thing about the 6SN7s and 12AX7s is that they are available new, which I think is a big consideration. The other good thing is the current saving. The amp I've got 'working', although it's behaviour remains a bit problematic, wants 50mA for best performance, where the 6SN7 + 6SN7 or 12AX7 + 6SN7 only wants 14mA. Which makes a big difference to the amount of ripple attenuation I can get for a reasonable resistance drop using RC instead of a choke. With 360 ohms I get -58.8 dB with a voltage drop of 11V, the LC filter delivers -61.9dB and a voltage drop of 0.44V. 3dB difference and we lose 10V. We don't want to be giving away any B+ that we can avoid.

The present amp sounds reasonably quiet. I'm doing my best to isolate the tubes from the power supply, the PSRR should be better than the existing arrangement, with it's resistor anode load being replaced by a CSS. The B+ at the transformer is now regulated. The PSU as built is long on capacitance, but traditional in concept. It way outperforms the RC filtration, but there is the ripple rejection offered by the regulators yet to take into account, plus the fact that the tubes each enjoy a greater degree of isolation, constant current and stable voltage supplies. Anyway the net impact of all that is that I think the amplifier will have somewhat better hum performance than it did in it's previous incarnation, and it is pretty good already.

The changes make the whole thing easier to build, regardless of tube selection.

One reservation I've got is that Morgan Jones suggests >=8mA for best linearity with the 6SN7, but I really like 5mA for best performance in the sim. I will be looking at that again. I can get 0.06% THD for a milliwatt into 16 ohms with the 12AX7 + 6SN7 with a gain of 4. That is better than many solid state amps, and it's all second harmonic, although obviously it's not hard to better in a transistor amp with a bit of effort..

Let's think about what we're doing here. We're gradually improving the power output at a notional transformer primary impedance of 10k and notional output impedances of 4, 8, 16, and 32 ohms, with a load of 16-600 ohms distributed across the transformer taps at the operators discretion, with the main axis of the the performance optimization being output into 16 ohms as simulated, with the simulated output into 32 (and sometimes 38) and 600 ohms being taken into consideration. The circuit delivers a minimum gain of 4 into 16 ohms with the volume fully advanced. Minimum input impedance is 25k.

These are not dissimilar to the performance criteria chosen for the O2, although only in terms of what voltage (9v pk-pk) is required and what impedances catered to. It's what is generally recognized as the majority of the range of loads a headphone amp is likely to be exposed to nowadays. Obviously output impedance is 'notional'.

We're trying to keep the components to off-the-shelf, and minimize the cost and effort. We've got a power transformer, this limits our other choices to a degree, although it's in other ways a good match for our requirements.

Tube amps are more touchy-feely than solid state. You never know really what's going on in there in terms of the tube, primary and secondary impedances, so there's no point in getting overly worked up that things aren't working out exactly.

Although efficiency (power output) falls as a triode anode load is increased, this does not mean that output disappears. One of the consequences is a reduction in distortion. I'm using transformers rated for 45mA, the 6SN7s will run 5mA, so DCR should be good. (I don't know the current rating of the reverb tx's, they say 6W).

After this you have to up the power output by looking for a bigger output tube to underrun, and/or change OPT. There are larger dual triodes out there, but we're at the limit of our heater supply, and we want really to stick in the 10k primary range of transformers.

You can approach this from a different point of view. That's to build an amp that will deliver your required voltage (or a bit more) into a (say) 8 ohm load, and slug it with a resistive shunt of (say) 16 ohms. So when anybody plugs a pair of 16 ohm (or greater) phones in, the amp is happy to deliver 9vrms, or whatever it is you want and the load is a minimum of (16 || whatever you plug in) ohms. Which is 10W.

No. We have to look for some finesse, instead of just beating the problem to death. 

w

Not really. The first prototype was built in under 2 weeks IIRC. What I can tell you is that since I stopped treating it like a job I had to finish to a price, I've had some fun.

I don't want this thread to spiral off into another disagreement about costs.

Whatever it's cost me, it wasn't so much I couldn't afford to put it all in the public domain.

I'm not about arguing with people who think good SQ should be expensive, I'm about trying to make it available to those who don't.

I just filled a BOM for the major parts. Tubes, sockets, regulators, caps, chassis, transformers, jack sockets, fuse, IEC socket, pot, switch, knob. U$214, so we're looking at U$250 all in. Add on a bit to make a full kit with a punched chassis.

J Epstein, thanks for the encouragement. Some interesting stuff on your site.

I managed to get the heater supply board for the second build done. It should be working in the next couple of days.

Unfortunately some new symptoms have meant that the oncologist sent me for a CAT scan, the surgeon sent me for a PET scan and I'm keeping my fingers crossed that it'll be no worse than a second round of surgery. It could still turn out to be nothing significant. Lung function test tomorrow. I don't really feel ill, but it has made it a bit hard to maintain my focus on audio design. Hopefully it's only temporary, I have got plans for at least one more prototype with dissimilar tubes, probably 12AX7 and 12AU7 with feedback. I've got 4 * 4P1Ls coming from Russia too, I don't know if these will go in a headamp or maybe a low-power push-pull to drive some full-range speakers after Morgan Jones' design on diyaudio. I've long had a fancy for building a hybrid push-pull with a SS phase splitter.

w