4P1L DHT Line Stage

[mogliaa]

Hi all,
I´ve been following the 26 DHT pre-amp thread with a lot of interest for some time. Hence, I ended up playing with LTSpice on a 4P1L line stage circuit as this valve is more than attractive due to its superb linearity. I´m sure that there are multiple and varied opinions around which DHT valve is best for a pre-amp, but the 4P1L seems a great candidate bearing in mind its cost.

As I don´t have any choke or transformer at hand, I´m looking at a DN2540 cascoded CCS as anode load. Target quiescent operating point is Ia=10mA and Va=150V/Vg=-10V. Not sure if this point is the best for this valve, I will try its sound unless anyone has any suggestions...

I was looking at fixed bias, but as there are multiple opinions around which bias setup is best, I just simulated cathode bias with 1K resistor (see diagram).

I guess that I could use Rod´s DHT filament board or a simple LM317 as a current regulator?

This line stage is intended for my 45 SET amplifier which has 47K as input impedance. Hence the 47K load resistor.

The stage gain si around 7.7 as expected. I have 220nf teflon capacitors so frequency response will be limited to this, which is not bad at all.

Based on the 26 pre-amp thread, many have tried LED bias, cathode resistors, filament bias and fixed bias with batteries. What will you recommend for this setup?

I´m throwing this out as an idea, so hopefully haven´t made too many mistakes .-)

Thanks
Ale

[Rod Coleman]

Ale, It's well worth trying this DHP in a line-stage, and I think your basic architecture is a fine starting point.

The power-supply rejection will not be so good, and consideration for a shut-regulator, eg.: SSHV will be good value-for-sound, I suspect.
Noise rejection in general will be the difference between good and GREAT sound. Building with the mains transformers in a separate enclosure is surely needed.
Multi-turn ferrite beads in the grid may be worth trying, and a Farnell 1635783 bead in the power feed (both sides) may help keep MW/HF radio out of the circuit.
Noise voltage on the filament supply will be amplified in a grounded-grid style - same gain as at the grid! Noise currents are added directly to the anode current - since they share the same internal conductor and terminals. With a LM317, you can measure the noise (voltage) with a scope, and compare with a line-level grid signal.

The whole circuit is best wired compactly, using wires no longer than strictly necessary. This minimises the EMI loop area - the "grip" that electromagnetic noise can get on your circuit.

[tomchr]

Originally Posted by Rod Coleman

The power-supply rejection will not be so good, and consideration for a shut-regulator

Curious here... What makes you say that the PSRR won't be good?

Originally Posted by Rod Coleman

Noise rejection in general will be the difference between good and GREAT sound.

Noise floor will need to be reasonably low, yes.

Originally Posted by Rod Coleman

Building with the mains transformers in a separate enclosure is surely needed.

I question that, honestly. The leakage field from even a cheapie toroid - like the ones from Antek - is minuscule. For an EI core this is different. I actually measured the output of an amplifier versus orientation of the (toroidal) transformer. I wasn't able to change any of the 60 Hz harmonics' amplitude by moving the transformer or rotating it. Granted, this was in a smaller circuit (SMD components, small inductive loop areas), but still... I was quite surprised by this result. I had expected the transformer orientation or proximity to have some impact. I was planning to use an external supply for the amp, but could place the toroid right up against the amp board without seeing any increase in 60 Hz (or harmonics) on the spectrum analyzer. Instruments like spectrum analyzers (HP 3562A, quality sound cards, etc) or wave analyzers (HP 3581A) are your friends for these tasks.

Separating the amplifier from the power supply opens up a giant can of worms (or whoop-a$$ if you're not careful). You have to really plan out the grounding scheme (both from an electrostatic shielding perspective and a safety perspective). Then add that you'll have 250 V DC running through a fair amount of cable. Of course, like any other engineering challenge, this can be solved at some expense (both money and time). I would build the amp on a prototype setup in the lab first. Get it to work on an old sheet of FR-4 first. See if a separate supply really is needed. Then decide if that's the route you want to take. It's one of those things that look great on paper, but can come back and bite you if you're not careful.

Originally Posted by Rod Coleman

With a LM317, you can measure the noise (voltage) with a scope, and compare with a line-level grid signal.

As you can with any other regulator. A multimeter with relatively wide bandwidth (HP 34401A or the older 3478A for example) are real handy for this. Scopes work too.

Originally Posted by Rod Coleman

The whole circuit is best wired compactly, using wires no longer than strictly necessary. This minimises the EMI loop area - the "grip" that electromagnetic noise can get on your circuit.

Absolutely agree. For inductive coupling, reduce the loop area. Inductance is proportional to the loop area. More inductance --> more coupling. Signal return currents should be in close proximity to signal forward current.

~Tom

[Rod Coleman]

Originally Posted by tomchr

Quote: Originally Posted by Rod Coleman
Building with the mains transformers in a separate enclosure is surely needed.

I question that, honestly. The leakage field from even a cheapie toroid - like the ones from Antek - is minuscule. For an EI core this is different. I actually measured the output of an amplifier versus orientation of the (toroidal) transformer. I wasn't able to change any of the 60 Hz harmonics' amplitude by moving the transformer or rotating it. Granted, this was in a smaller circuit (SMD components, small inductive loop areas), but still... I was quite surprised by this result. I had expected the transformer orientation or proximity to have some impact. I was planning to use an external supply for the amp, but could place the toroid right up against the amp board without seeing any increase in 60 Hz (or harmonics) on the spectrum analyzer. Instruments like spectrum analyzers (HP 3562A, quality sound cards, etc) or wave analyzers (HP 3581A) are your friends for these tasks.

Separating the amplifier from the power supply opens up a giant can of worms (or whoop-a$$ if you're not careful). You have to really plan out the grounding scheme (both from an electrostatic shielding perspective and a safety perspective). Then add that you'll have 250 V DC running through a fair amount of cable. Of course, like any other engineering challenge, this can be solved at some expense (both money and time). I would build the amp on a prototype setup in the lab first. Get it to work on an old sheet of FR-4 first. See if a separate supply really is needed. Then decide if that's the route you want to take. It's one of those things that look great on paper, but can come back and bite you if you're not careful.

Tom, please read the thread "26 Preamp" to see how this is done correctly.

You are right inasmuch as the Toroidal trafo does not radiate magnetic stray fields, but this trafo style is only a good choice for linestage duty when you have no HF/RF conducted noise on the mains supply. Not many locations can apply! This is due to the toroidal's winding style, with all the secondary windings scatter-wound directly on top of the primary. High capacitance in other words, ideal for coupling all that HF straight into your circuit. Sure you can try to filter it, but I believe that most designers would rather just eliminate the coupling and not have the noise.

In addition, the core material of a cheapo Toroidal is usually unconnectable, and floating, it can do nothing to help suppress common-mode HF mains noise being radiated from the windings, and coupled into the signal wiring, signal interconnects, etc. And since mains feeds are long tightly-coupled cables, the majority of the noise will be common-mode.

Split bobbin EI trafos, with grounded core, offer hugely better insertion loss to high frequencies, and the connecting the core to chassis is very effective in smothering emissions. That's the budget choice. For higher cost, properly constructed EIs with interwinding screens improve the noise rejection further.

The 26 line-stage builders will vouch for the lack of difficulty in making 2-enclosure amps work. The chief concern is to keep the trafo/rectifier/choke/first cap closely coupled (in the mains enclosure) - which puts an end to any recharge-pulses in the ground wiring. After that, one ground wire to the signal stages, and a cap to decouple the B+ wiring. With CCS loading, the valve stage do not impose current pulses on the supply - so sharing between the two channels is fine. And even if it was not fine, using 2 shunt regulators would completely suppress that problem.

I am not suggesting we run 10-metre cables between the mains enclosure and the valve box. 300mm will suffice, a little more will be plenty. If you have safety concerns, just use steel-wire armour cable (= outdoor duty mains cable). In the UK this is rated safe to carry our 240V mains across the outside of walls, no further protection required. It's used in large quantities, and available cheap.

Quote: Originally Posted by Rod Coleman
With a LM317, you can measure the noise (voltage) with a scope, and compare with a line-level grid signal.

As you can with any other regulator. A multimeter with relatively wide bandwidth (HP 34401A or the older 3478A for example) are real handy for this. Scopes work too.

No doubt meters exist to evaluate the noise on my CCS/Gyrator based filament supplies, but no instrument in my possession registers anything at all. The TEK 465 shows ZERO perturbation on any scale, and the DSOs don't match the resolution of the 465.

This is no surprise, since the regulator does not contain any references, zeners, bandgaps or anything else, and the noise can be narrowed down to the contributions of a couple of transistors and a sub 1-ohm resistor: down in the nV/(Hz^0.5) in other words. The transistors are run at high current, and have low Rbb, too - so the figure is low even for the circuit structure.

[tomchr]

Rod,

I still didn't get an answer to my first question about PSRR.

Originally Posted by Rod Coleman

You are right inasmuch as the Toroidal trafo does not radiate magnetic stray fields, but this trafo style is only a good choice for linestage duty when you have no HF/RF conducted noise on the mains supply.

- Line filters would be one way of combating that as well. No filter is perfect...
- Another way would be an electrostatic shield. Antek has started to offer shields on many of the their toroidal transformers. No shielding is perfect...
- EI core transformers would suffer from the same issues regarding HF/RF coupling through the transformer. Of course, a split bobbin type will reduce the coupling dramatically as well, but many transformers are not split bobbin.

But fundamentally we agree. The toroid wins on leakage fields, but loses on RF coupling between windings. A split-bobbin (note that this is often a value-add option) EI core wins on the RF coupling, but loses on the emitted leakage field.

Whether the RF injected via the mains power supply impacts the sound quality at AF is another story. It's obviously desirable to not have the EMI in the first place, but few people live in a screen room. Those of us who frequent screen rooms often don't even bother to close the door... As you also point out in your first post, RF induced on signal cables can be an issue as well. I tend to put EMI filters right where the signal cable enters the amp.

Originally Posted by Rod Coleman

Sure you can try to filter it, but I believe that most designers would rather just eliminate the coupling and not have the noise.

I agree. It's always better to eliminate the source of the issue than to attenuate it. However, a split-bobbin transformer will not eliminate the coupling. Only reduce it. Unfortunately, physics dictate that if two conductors are in proximity on one another, there is a mutual inductance between them (as well as a capacitance). Just one of them annoying physics things.

Originally Posted by Rod Coleman

In addition, the core material of a cheapo Toroidal is usually unconnectable, and floating, it can do nothing to help suppress common-mode HF mains noise being radiated from the windings, and coupled into the signal wiring, signal interconnects, etc.

Trying to learn something here... Would you elaborate on how grounding the core helps on reducing the radiated field?

Originally Posted by Rod Coleman

And since mains feeds are long tightly-coupled cables, the majority of the noise will be common-mode.

Agreed.

Originally Posted by Rod Coleman

The 26 line-stage builders will vouch for the lack of difficulty in making 2-enclosure amps work.

Do they have measured data? I guess I should plough through that thread. Sounds like an interesting one... I've only read bits and pieces.

Originally Posted by Rod Coleman

With CCS loading, the valve stage do not impose current pulses on the supply - so sharing between the two channels is fine. And even if it was not fine, using 2 shunt regulators would completely suppress that problem.

I don't see how shunt regulators would suppress the problem of current pulses any better than a voltage regulator would for exactly the reason you mention in the quote above. Neither regulator has infinite line rejection.

Originally Posted by Rod Coleman

If you have safety concerns, just use steel-wire armour cable (= outdoor duty mains cable). In the UK this is rated safe to carry our 240V mains across the outside of walls, no further protection required. It's used in large quantities, and available cheap.

Cable and connectors is one thing. Making sure the safety/chassis ground is run properly is another. If one understands the problems involved, sits down and draws it out, it's a manageable task. But it only takes one ground loop (easy to do) to screw it up.

Originally Posted by Rod Coleman

No doubt meters exist to evaluate the noise on my CCS/Gyrator based filament supplies, but no instrument in my possession registers anything at all. The TEK 465 shows ZERO perturbation on any scale, and the DSOs don't match the resolution of the 465.

Nice!

Originally Posted by Rod Coleman

This is no surprise, since the regulator does not contain any references, zeners, bandgaps or anything else, and the noise can be narrowed down to the contributions of a couple of transistors and a sub 1-ohm resistor: down in the nV/(Hz^0.5) in other words.

Well... There has to be a reference somewhere. Even if it's a Vbe or a Vgs. But I get your point. Would you point to the most recent schematic? I'm getting curious now.

Thanks,

~Tom

[tomchr]

Originally Posted by mogliaa

Hence, I ended up playing with LTSpice on a 4P1L line stage circuit as this valve is more than attractive due to its superb linearity. I´m sure that there are multiple and varied opinions around which DHT valve is best for a pre-amp, but the 4P1L seems a great candidate bearing in mind its cost.

It's a nice-looking tube. I've been mulling over building a low-power DHT headphone amp with those. Incredibly linear and good pricing. What's not to like... ?

Originally Posted by mogliaa

As I don´t have any choke or transformer at hand, I´m looking at a DN2540 cascoded CCS as anode load.

You may not need the cascode. The semiconductor CCS'es have quite high output impedance as-is and don't suffer that much from channel length modulation (MOS analog of early effect). Of course, the cascode, generally, improve the CCS'es but there may not be much difference without it.

Originally Posted by mogliaa

I guess that I could use Rod´s DHT filament board or a simple LM317 as a current regulator?

Those are options. Constant voltage is another option. I have never been able to measure or hear any difference between constant voltage and constant current, so I go with constant voltage for cost and complexity reasons. I actually use switching regulators for this on my 300B amp. You can see schematics and measured data here: 300B SwitchMode filament supply with a follow-up in post #17. You'd, obviously, have to tweak the feedback resistors to get 4.2 V rather than the 5.0 V I use for the 300B.

Originally Posted by mogliaa

The stage gain si around 7.7 as expected. I have 220nf teflon capacitors so frequency response will be limited to this, which is not bad at all.

With 1 Mohm, 220 nF should yield 0.7 Hz lower cut-off frequency. Whether 1 Mohm is enough to prevent the grid leakage of the tube from causing issues is another story. I don't have any experience with the tube, so I can't really tell. But 1 Hz is plenty low.

Originally Posted by mogliaa

Based on the 26 pre-amp thread, many have tried LED bias, cathode resistors, filament bias and fixed bias with batteries. What will you recommend for this setup?

I like LED bias. The dynamic impedance of a RED LED (HLMP-4700 for example) is on the order of 10~20 ohm so it behaves reasonably close to a 1.7 V battery (at 10 mA forward current that is). That tends to result in very low distortion (I'm measuring below 0.01 % THD+N on a 6J5 (one half of a 6SN7) running at 1 Vrms). A simple LED beats the other types of biasing schemes on performance and/or complexity 99.9 % of the time.

One thing, though... The power supply (B+) needs to be really clean. I recommend using a voltage regulator. You can also use heavy RLC filtering or other regulator types.

~Tom

[Rod Coleman]

Tom, we can put some numbers on the difference in HF insertion loss.

For some trafos loafing around on my bench, the capacitance measurements were:

split bobbin:
- 240V pri;
- 370V sec; 150VA: split bobbin. C(pri-sec): 70pF. C(pri-core): 85pF

Toroidal:
- 12V 150VA C(pri-sec): 300pF
- 54V 100VA C(pri-sec): 720pF

We might expect from this that the interwinding capacitance of a Toroidal B+ trafo might be 1.5nF or more. If that's anywhere near correct [don't see why not] - we will get 20x better insertion loss using the split bobbin EI.

The split-bobbin type I measured was an industrial type available new at a very low price:

150 VA

Builders in the UK need not hesitate with these - you can specify winding voltages for secondary AND primary as any voltage you like. The quality and service from this company is first rate.

As for connexion to the core: Consider the toroidal trafo usually has [for safety] the primary wound next to the core. Judging from the interwinding capacitances, we will get > 1nF of capacitance to use as a filter, which would reduce the noise out to much higher frequencies than ordinary capacitors can. But not if the core can't be connected to chassis.

Personally, I regard split-bobbin trafos as a minimum standard for DIY audio. Not to mention that the JMS types are lower cost than any other solution! Here in the UK, anyway.

Mileage may vary in different locations, and not everyone likes the look of industrial trafos, but then if you have more spend available, there are many high quality options, including screened secondaries, ornamental end bells, even fancy core materials.

[Matt.B.H.]

I am thinking of a building a linestage, especially one using the 4p1l as I have almost completed gathering all the parts for my all loktal power amp. Its in prototype stage at the moment but it is giving a nice 100W RMS straying slightly into AB2.

Anyway to confirm some comments by you guys. I initially used a crappy mains torroid for phase splitting duty (actually listening to again now). I stuck it on top of the Antec mains toroid and can confirm the radiated field is minimal. With an old relay coil accross the scope probe as a search coil it shows a very small amount of radiated field compared to an EI.

One problem though. any mains born noise comes straight through. I put a filter in line and this has cured a fair bit(I cant hear it but its there) The trouble these days is that more and more domestic appliances use switch mode supplys. Also another really nasty one is fridges they use a capacitor run type induction motor that causes alsorts of hash on the mains when they start up at random.

Another problem with toroids is the fact that as they have such a good core, they are wound so they are near their limit for saturation(penny pinching scroungers). Any DC offset on the mains will screw them bad and you can hear the buggers hummmmm then. If you live near an industrial site that uses a lot of straight off line inverters for motors you will know what I mean. I have a variable 10-20VDC bias on my supply. The place I lived before had the 11KV sub at the bottom of the garden and I had the luxury of a very clean supply. Not so now.

So, Rod, how did you get on with JMS? Are they any good?

I will be reading this thread with interest.

Cheers Matt.

[tomchr]

Originally Posted by Rod Coleman

We might expect from this that the interwinding capacitance of a Toroidal B+ trafo might be 1.5nF or more. If that's anywhere near correct [don't see why not] - we will get 20x better insertion loss using the split bobbin EI.

Dang. Nearly 1 nF (possibly more for higher voltages) between primary and secondary. I wonder what it reduces to with a shield between primary and secondary (shield grounded). I'll drag my Antek tranny with me to work one of these days and measure it.

Originally Posted by Rod Coleman

150 VA

Looks like a pretty good deal. Nuttin' wrong with that!

Originally Posted by Rod Coleman

As for connexion to the core: Consider the toroidal trafo usually has [for safety] the primary wound next to the core. Judging from the interwinding capacitances, we will get > 1nF of capacitance to use as a filter, which would reduce the noise out to much higher frequencies than ordinary capacitors can. But not if the core can't be connected to chassis.

Yep. And the capacitances you measured also indicated that the cap from primary to the core was higher than that from primary to secondary, so if you ground the core, less than half the signal should couple to the secondary.

I hope you're willing to share the schematic for your DHT heater regulator (or at least link to it). I saw some of the earlier revisions (I assume) in the DHT heating thread. But something tells me you have the thread and post number memorized... (Or maybe that's just a projection ).

~Tom

[mogliaa]

Tom, Rod,
Thanks for your invaluable contribution to this post! I can't believe the exchange of ideas and suggestions between the two of you :-). I have a lot to read, though.

I was looking at separate chassis as well.

It seems that the key point here is to design the cleanest HT and filament supplies as possible. Given the options thrown so far, I'm still confused as to whether a Salas Shunt HT supply or Tom's supply is the way forward? I'm keen to go for a simple design with components easy to get hold of (at least in the UK).

Rod, what about your filament regulator, do you have one suitable for this filament voltage/current?

Tom, thanks for your tips as well. I will try LED bias then...

Thanks
Ale