My poor man version of the Raven Preamplifier.

[piano3]

but to actually put some numbers on it; the anode is loaded by a step down tranformer which has a load of perhaps several hundred kilohms so the impedance seen at the anode is in the megohm range and the mu of the valve is less than 20. About the current mirror I cannot comment as I have not used this circuit and I would normally use very high impedance bjt cascode sinks in such a circuit.

 

[Zen Mod]

I'm too lazy for actually putting some numbers there (except mentioning fact that putting several hundred of kiloohms as load is right path for filling wallets of cable guys)

it's easy to calculate what's input impedance of cathode node , so , according to result , calculate minimum value for bridging caps ......

probably Shoog will chime in ....... but I can't see any necessity for that , because value of that cap I'm seeing as of secondary importance

 

[piano3]

ok, my interest is that I am trying various ways of getting round large value coupling caps when interfacing valve mic preamps with SS equiptment. Shoog's idea is similar to one of the things I am trying.

 

[Shoog]

I think that Pianos analysis is correct in this instance - 10uf is probably more than enough when the high anode load is the dominant factor. I added the 1000uf caps to see if they made a difference - which with the stiffer current mirror I was reluctant to say that they did. This was why in the initial discussion when Lars suggested I try at least 470uf to reduce the resonant hump induced by the cathode resistors - I pointed out that this would not be necessary when the current mirrors were in place.

The caveat is that listening tests showed this over analytical presentation and the only way to get rid of it was to either bypass to ground (which I didn't want to do) or try a low impedance load (100R) at the cap junction - for which a junction was necessary. However by our previous analysis it is probably true to say that the caps could be reduced down to 20uf back to back, allowing the use of film caps and creating the needed node.

Listening tests dictate the ultimate design decisions, and the various stages of building this preamp have produced radical shifts in the overall presentation with each change. I think this version is coming to a completed implementation which achieves the goal of hyper detail, absolutely dark quiet background and smooth presentation. I have to add that the high output impedance doesn't seem to have the expected effect on drive capability - it is delivering 300mA to the output load - this gives it massive authority in delivering tight detailed bass. The step down ration is also sufficient to overcome the inherently high capacitances within the transformer.

I was listening to a Kind of Blue a day or so ago and was staggered by the fact that I could almost see the drums been brushed so lightly that they easily get lost in the background with my other preamps.

Shoog

 

[piano3]

Shoog,

It is great that with the help of semiconductors these inexpensive magnetic parts can perform so well! (and outperform easily available push-pull x-formers) I have a couple of 125 watt ballast chokes(DCR 12 ohms) and I am thinking that they might do for passive LC filtered heater supplies for series heater chains.
I am trying to understand the details of the current mirror; given that your HT supply is unregulated and allowing for the usual spread of valve parameters how did you choose 164R as the programming resistor? Did you just experiment? And what about the 100R collector load resistor?
I am very keen to understand this as I am tired of building negative supplies when I might not need to and which are just more to cram in the cases.

 

[Shoog]

A Current mirror as implemented has a few special characturistics - but it is not a CCS as usually implemented. The current mirror enforces current balance on its two loads - but it doesn't set that current within the external load. The DC current is set by the bias which is applied to the valve just as a normal valve is biased. In this case the bias point is defined by the voltage drop across the transistor (0.7V) and then the voltage drop across the resistor. The resistors are chosen such that each transistor drops the same voltage and hence has the same dissipation (for temperature stability). In this case the total cathode bias on each side is -4V which sets the 5687 current at about 15mA. In doing this the impedance of the current mirror is set high which has the effect of blocking the AC signal from been passed by the valve. Also the two loads can only pass the same current (which is effectively the same as having high impedance). So the current mirrors have to be bypassed either to earth in order to allow the signal to pass, or to the other cathode of the LTP to allow the signal to pass between the two LTP valves.
I believe the principle of 100R cathode node bleeder is intended to allow any error signal generated by the LTP to escape. Without it any error is integrated back into the signal via the valve cathode acting as an input (as intended in a the operation of a LTP). The value of 100R is just one suggested by Allen Wright in his PP1C power amp. In my previous power amps the issue of stridency has not arisen and so I have always stuck with a 1Meg bleeder (this is essential to their stable long term operation of these designs).

I hope that long and converluted answer helps clarify. Essentially the Current mirror ensures DC current balance in the transformer but doesn't set the current. Its advantage over a CCS is its lower drop out voltage. The caps are essential to allow the signal to pass.

Shoog

 

[piano3]

Thank you Shoog, it's not convoluted at all but very clear,particularly as I had not even heard of a Wilson mirror a few days ago and will be very pleased to add this to my repertoire, so to speak. The details of the cap junction bypass will require some reading to understand the theory in detail but I am happy simply to take advice for now.
One detail though; the collector of the right hand transistor is clamped at 2 diodes drops above ground but the voltage of the upper left hand transistor's collector is only set by it's load resistor which at present will be about 2.5 volts. Is this your intention or is it a misprint in your latest schematic?

 

[Shoog]

I am in sort of the same situation - I only discovered this current mirror out of necessity and am only just getting my head around its internal workings. You are correct in saying that R2 & R3 should be equal, but its not that critical since you could remove R3 altogether and it would still work.

Having just measured it - the single transistor is actually loaded with a 62R resistor followed by a 100R resistor on the valve pin and measures 2.32V at the top of the 62R resistor. The twin transistor leg is missing the 62R resistor and measures 2.3V also - so it is compensating internally.

I was thinking about that myself only moments ago and realised my original intention (to keep dissipation in all transistors equal) isn't possible in the Wilson Current mirror. I will probably equalise them to see if it make any difference.

Shoog

 

[multi-volti]

Corrected to include volume control

Am I remembering correctly the round element below the grid represents a cold cathode, like a VR tube (let's not argue about grids)?

 

[multi-volti]

I was reading that air has a quarter of the dialectric figure than epoxy - and this is where I think the difference lies . Interwinding capacitance is always going to be high when using toroidals and increasing it by (potentially) a factor of 4x is going to kill the circuit. I found that response was tailing off above 10K when I attempted to use the encapsulated one on the output.

Response is acceptable down to very low frequencies. It may not have much drive into the 5687 grids - but its an easy load and seems adequate. Configured as a CTC increases overall performance by an order of magnitude.

Shoog

Reading material for insomniacs:

I was reading that air has a quarter of the dialectric figure than epoxy - and this is where I think the difference lies . Interwinding capacitance is always going to be high when using toroidals and increasing it by (potentially) a factor of 4x is going to kill the circuit. I found that response was tailing off above 10K when I attempted to use the encapsulated one on the

Shoog

Many kinds of epoxies...

Here is an article discussing the dielectric properties of about six categories of epoxies that have dielectric constant ranging from about 2.9 to about 4.7. As you probably know, air dielectric constant is close to 1.

http://www.adhesives.org/docs/pdfs/electrically_insulative_epoxies.pdf?sfvrsn=0

I'm pretty sure potting a transformer doesn't simply multiply the interwinding capacitance by the ratio of dielectric constants.

Magnet wire insulation varies in dielectric constant, depending on the insulation selected for temperature 'class'. Most are 3.7-3.9, except for Formvar which is 7.4 (with the lowest temperature rating among the common insulations. It's the dielectric that is most intimately in contact with the winding conductor. See page 13 of the following. https://issuu.com/mwswire/docs/mws_wire_techbook.

Toroids likely use some kind of poly-nylon, according to MWS Tech Book. The toroid shuttle winding process may be harder on the insulation film than on an EI or C-core type winding. I used Polyurethane-Nylon 130 in the 90's, but it appears to be no longer made for some reason. It was solderable. (750 F dip in a solder pot removed the insulation then tinned the wire ends. Removing the non-solderable insulation films takes a whole different dedication to process...chemical or mechanical abrasion, typically...avoiding damaging the insulation mechanically, and neutralizing any harsh chemicals used to dissolve, to avoid corrosion later.

Potting won't change the geometry, particularly the spacing between wires. If it is vacuum impregnated (many winders simply dip and drain...Heyboer gave me a partial tour of their plant south of Grand Haven Michigan about 20 years ago when I picked up some prototype electronic fluorescent ballast wound bobbins I had them make for me (actually for my employer back then ). I don't know what they do now, but they used that method in the mid-90's, and the conveyor that moved dip-impregnated windings (on bobbins or 'paper tubes' through a baking oven used a VW transmission...I don't remember the motor (electric, not internal combustion!). Cool and functional in a Frankenstein way).

Where was I? Oh, the dielectric of the magnet wire is probably dominant as it can't get replaced or double-coated by a potting dielectric over 100% of the area between two adjacent turns if they are touching and immobile during potting. The space the potting dielectric fills is complex compared to an unpotted winding which has air in the non-contact space between turns and layers. Kind of a complex mirror symmetrical opposite (well, 'different' is probably more accurate) shape than the adjacent conductor geometry of two adjacent conductor insulation layers. Empirical measurements are probably the only way to gauge the difference in capacitance between potted and unpotted...too many winding and interleaving geometries.

Remember also that inter-layer insulation is usually (should be) used, not always every layer for lower voltages, but for power transformers, more often than not, each layer. Paper or polymer film typically have lower dielectric constant than magnet wire insulation.

Some people don't like potting OT's based on the increase in capacitances, but I wonder how reliable, long term, unpotted windings are...they can expand and contract, vibrate, etc. and rub.

Lack of potting and voids in potting may be areas where corona can occur...it occurs due to electric field. The existence of insulation may not protect against it. That is likely one process by which some transformers develop insulation system failures after decades of use (if lucky).

I also vaguely recall from an electromagnetic fields class (not my strongest subject) that sandwiched layers of differing dielectrics cause the electric field to divide unevenly, in proportion with the dielectric constant (simple case of flat conductors)...it vaguely reminded me of the care in thinking about different value capacitors in series...

Still awake?

If that's not ugly enough, there was a series of tech briefs from a capacitor manufacturer decades ago, I think ElectroCube, that discussed various interesting factors about capacitance. One discussed air space between conductors that form variously modeled capacitances. Do you remember the dielectric constant of water? Pure water, = 80, salt water about 78. Water that makes up relative humidity is pretty likely closer to pure condensate than otherwise. Of course, the density of humid air reduces the factor, but in some cases, variation in relative humidity can cause undesired variation in 'parasitic' capacitances that occur in transformers.

My guess is that there are also many winding and interleaving geometry variables in toroidal windings as well. Each manufacturer winds for their own purposes, and our desired bandwidth for 'off-label' repurposing is an unfortunate victim of circumstance. Our frequency response mileage (kilometerage?) for toroidal power transformers as OT's may vary :O(