"Radical" twist on an all DHT phonostage

[Binaural]

Hello all,

I've been browsing the net for the past year trying to find unique tube circuits to learn all I can from them. I should state before I go any further that this project is primarily for learning and to try some things that perhaps aren't viewed as orthodox or common in today's electronic mindset.

I stumbled upon Dmitri Nizhegorodov's 3a5 all DHT phonostage and this really piqued my interest. I then found a couple of threads (one on DIYAudio) of people who have actually built the unit and praised it's great sound.

Since I now know it is possible to build a very good all DHT phonostage, I'd like to try some new things with it. I have read much about CCS's and how they work, and it sounds like something that would work very well with this circuit. It should give the triodes an ideal (well, close) load and allow for greater mu from each stage, increasing the gain and should help to lower the noise floor a bit.

There were 2 other ideas I had to try with this, and it may mean that a whole new design might need to be had, but like I said, this is going to be a learning experience and hell, we might just stumble upon something special!

The first idea was to balance the circuit. Since I will be using this with an MC cartridge and transformer, the outputs of the transformer can feed a balanced circuit. Would this mean simply constructing 2 of this circuit per channel for a total of 4 for stereo, or are there other simpler ways to achieve this?

Finally, and this might help with the above idea, but being from the pro audio side of things, I have come to adore transformers and dont' view them with the same "hatred" that some in the pro audio community do. But what about instead of using capacitor coupling between the stages, use transformer coupling? Also what about differential parafeed? In this case the transformers would be smaller and (possibly) cheaper.

What are your ideas? I fully intend to build and experiment as we go. I just don't have the knowledge required to design something like this, and haven't found anything close to what I have described here to try and cut and paste something together. I have just always felt that glass and iron were soul mates and work wonders when properly paired together, and with modern advances such as CCS's this should make for a (hopefully) very good phonostage. I look forward to your ideas and comments.

P.S. Not to waive a finger at anyone, but I'd like to keep this thread as positive as possible, I know people disagree but I'd like to stay away from comments such as "this isn't possible" or "this is a stupid idea." You never know what's possible until you try.

Thanks all,

Joe

[kevinkr]

If anything you would derive the greatest benefit from running balanced from the cartridge to transformer, and unbalanced in the active stages.. A differential input stage is about 3dB noisier than the unbalanced equivalent, two unbalanced stages (pp) running on an input transformer might be quieter or noisier depending the implementation. The additional complexity probably will have an impact on the transparency.. (usually my experience)

Otherwise it does sound like an interesting project. To derive the most benefit from the CCS you might want to consider driving the EQ (passive hopefully!) with a mu follower type of configuration (Ixys 10M45 cascode or a suitable high transconductance pentode like the D3A, C3G, C3M, 7788, 6688) - this will allow you to minimize the contribution of johnson noise by the resistors in the equalizer network by selecting much lower values than would otherwise be possible.

You could also consider fixed bias on the input tube fed through the secondary winding of the input transformer. (No bias components in the signal path - although IME Leds are pretty good for this duty.)

Filament supplies must be astoundingly quiet because the filament in this case is in the signal path. A diy ronan style filament supply is not overkill. Ripple on the filament supply needs to be in the low uV region in order to achieve a good snr. Possibly batteries could be used, note that they are not necessarily noise free either, but at least they will not hum or buzz..

I hope this is at least a little helpful.

[Binaural]

Note: forgot to add a link to the page about the circuit in question:

Phono Preamplifier with 3a5 Directly Heated Triode

[Binaural]

Kevin,

Thanks for the quick reply.

I understand what you mean about the added complexity. My philosophy is usually to make everything as simple as possible, but if there is a real benefit in the added complexity which translates to positives in my ears, I usually have no objection then. I have always heard that there is a benefit to the balancing of the circuits especially in the noise department, could you tell me how you have come to your conclusion?

Yes, the eq will be passive, and since the original design uses 3 stages, it will be split between stages 1 and 2 and then between 2 and 3 (the final might not be this way, so who knows!) I do see the advantages of splitting this up, and since the 3a5 only has a mu of 15, i don't see this being converted into a 2 stage amp. I am familiar with the 10M45, I have used it in the building of my Tubelab SE and that sounds great! I think I'll stick with solid state CCS's unless someone can convince me otherwise.

As far as the filaments are concerned, I will be using batteries. eBay has 1.2V 10A/Hr NiMH batteries that will be perfect for this purpose. I'm not going to worry about building a charger, as I'll just keep a spare set in a separate charger and have a meter monitor the voltage to let me know when they will need to be changed.

I just don't think in this situation anything will be able to outperform the batteries for the filaments. Of course, this is unless I am convinced otherwise.

Thanks again Kevin, I look forward to any additional ideas you have.

Joe

[Binaural]

]I have a couple of questions about this power supply here:



The one I will be using is identical but instead of using the EZ81, I'll be replacing it with 2 6CJ3 damper diodes because of their lower noise. Other than that, I'm sure a little tweaking will be in order, but I haven't been able to find those large value capacitors in high voltages. I'll need at least 500V right? Also how could I string multiple caps together to get these values if I cannot find them or they are too expensive?

Joe

Also, I should probably post the schematic of the circuit in question to make it easier to reference.

[leadbelly]

You don't need 500V caps; 240 * sq root 2 = about 340V. You could go lower for some of the caps further down the chain.

[kevinkr]

Originally Posted by Binaural

]I have a couple of questions about this power supply here:



The one I will be using is identical but instead of using the EZ81, I'll be replacing it with 2 6CJ3 damper diodes because of their lower noise. Other than that, I'm sure a little tweaking will be in order, but I haven't been able to find those large value capacitors in high voltages. I'll need at least 500V right? Also how could I string multiple caps together to get these values if I cannot find them or they are too expensive?

Joe

Also, I should probably post the schematic of the circuit in question to make it easier to reference.

Given the load currents required I think those first two caps are ludicrously large, something on the order of couple of hundred uF maximum would be much more reasonable. I am not sure any tube rectifier could survive the inrush currents generated by the large values shown.

I am also not sure about the rating of that input choke, a minimum current is required for proper regulation and I bet this pre-amp circuit draws nowhere near enough current for the choke listed. (Which should really be designed as an input choke, for the specific load current, and not any old off the shelf type.)

I don't understand your comment about the 6CJ3 having "lower noise" than an EZ81, can you explain how you came to that conclusion?

[Binaural]

Sorry, I should have originally stated that the posted power supply design is not of my making and was copied from another forum. The person who designed this has built it and is currently using it as shown.

His selections were from his junk box so that explains some of the values posted. I have purchased parts with similar values for mine, I'll see how it turns out.

The 4k7 50W resistor is there to keep a constant current through the choke to maintain regulation.

TV damper diodes aren't commonly used in HI FI but from what I have read, their switching noise is the lowest of all tube rectifiers. I know in this case these things are way overkill, but they have been selected for the reasons stated above.

Here's a link to one of the articles that discusses the damper diode as used here. The Amity, Raven, and Aurora It's about 1/2 way down the page where he starts talking about the B+ supply.

Joe

[Binaural]

Well, I've ordered all the parts for the pre, and have received most all of them. I'm waiting for the 10ah D cells, all the tube sockets, and the power transformer and chokes.

For the chassis, I ordered the biggest conductive plastic case from Hammond and plan on mounting the tubes inside with padding to shield against microphincs and the case should shield against RFI. Should look pretty interesting when finished.

[Sheldon]

Originally Posted by kevinkr

Given the load currents required I think those first two caps are ludicrously large, something on the order of couple of hundred uF maximum would be much more reasonable. I am not sure any tube rectifier could survive the inrush currents generated by the large values shown.

The supply models OK. If you make sure the series resistance is at least 200 Ohms, it won't exceed the rectifier ratings. I agree, those second two caps are at least 10 time bigger than needed. It will be quiet, but it would be a lot easier to do by just adding an RC section or two.

Sheldon

edit: BTW, if you model the supply with 220u, instead of 4400, and 2200 in those positions, you get about 15nV of ripple for B3+, and less for the B2, B1. That would be plenty quiet. So even with those values you don't need any extra sections.