'Tube Gain Controlled Amp' mastering box(s) project

[alexc]

Some more noodlin' ....

-----

Looking more closely at the transconductance curves for various tubes, I'm getting a feel for what I'm after.
ie. a shallow transconductance curve where it  can spend 'most of the time'  (around 5mA), increasing to max -1V at some (high) current in 'deep limiting'  ie. 'very occasionally' and way down to -25V at the min usable current (no nothing) 'very occasionally'.

Thats a range around 0 .. 25V of CV to the grids.

Its clear that I will benefit from :

a) a higher pentode screen voltage  ie. closer to 250V than to 150V
b) a sizable screen stopper resistor to really flatten out the gm 'dive' over the decade (or two!)
   
That will allow  for a decent 'idle' current at signficant grid voltage whilst allowing the CV some 'room to move'

----

Right now its a 6BJ6 which holds up well even when discounting the appeal of it's crazy low heater current.  ;D

It has a strong max gm (4500 ish), a decent max current approaching 15mA pretty easily and can be made to flatten out pretty well to provide a broad grid range (to around -20V meaningfully) before current gets too low.

If one is able to afford a more common 300mA at 6.3V heater, then the usual candidates are usable too.
This time, it looks like a russian candidate leaps out. Off to ebay (again)

-----

NOW,

in 'my' gain-cell, there is some plate loading of the push-pull cathode followers - this is to allow a balanced feed with a little voltage gain to the sidechain.

HOWever,

I am first going to go *without* plate loads - pure CF  

That means sourcing the sidechain input signal further downstream, which in this case, is all unbalanced.

SO, either a phase splitter tube stage OR an additional (highish Z input at least) transformer.

----

I like the simpler, 'purer' version ( for want of a better term) with no plate loads.

Even choke loading might be possible, but I'm leery of unshielded anything, ever, anymore Even despite external psu!

The action of the varying current causes, of course, the plate voltage to vary due to a plate load resistor. This can work against the compression action ie. varying the gm, in a pentode mode setup.

ie. it is a useful simplification to have a constant plate voltage, at least at first until I get a feel of things

----

SO - looks like the vari-gm gain cell is going to be pure CFs  more likely with the 6k13p remote pentode or it's slightly less power rated (2.5W) us type, 6eh7. I have a few nice nos ge ones to give a workout in this here build.

As always, it can look rosy on paper but turn out poorly - fortunately I haven't experienced that, so onwards and upwards here.

------

NOW to decide - an additional pair of tubes or transformers  !  Either way, its a win.

 

[alexc]

I'm thinking operation points something like  ...

curves for plate 200V, screen 200V (24K screen stopper resistor)

typical op point , say low dB limiting  > spend quality time here

-> 5mA thru each side of the push pull cathode followers
->  requires -9.0V grid-cathode difference and corresponds to 770umhos (or uA/V)
-> Zcath approx 1300 ohms

max op point, max limiting  > dare not go beyond here under heavy GR!

->-1V grid-cathode difference results in  approaching 18mA
-> corresponds to  approaching 18000 umhos 
-> Zcath approx 60 ohms


min op point, min limiting  > not much going on!

-> 1mA each side and -16.0V grid-cathode difference
-> corresponds to 292 umhos
-> Zcath approx 3450 ohms

That's the basic idea :

- vary the Zcathode each from  3450 ohms at cutoff to 60 ohms at burnout with from 3000 ohms to 300 ohms typical

So thats a 2.7K ohm range of variation, at the cathode of each leg,  in series with a fixed resistance  of 5K5 ohms (approx 1/2).
By 'potentiometer action', result is around half the voltage in the cathode circuit to be potentiated away now!

ie. around 6dB of GR.  Which is about right.

Here's the curve attached.

I figure to go that way for now.

I'd like to push the plate and screen a bit higher; which slides it all a bit further out to the left towards increasing Vgrid-cath.
ie  strive for a typical CV range,  say  -25V to -10V or so for a 6dB GR variation.

I guess I'll study the curves some more, experiment around those values.

The 6eh7 have an upper limit of 2.5W - a little shy of the 3W of the 6k13p ruskies which limits of min/max operating conditions a bit.

However, how can I say - I just love nOs ge tubes - I've had nothing but good luck with them!

 

[alexc]

And here is an update of the vari-gm gain cell schem.

I still need to claculate the cathode circuit properly  but I'm narrowing in on a solution ....

 

[Heikki]

I'm thinking that if you take the input signal for the vari-gm gain cell sidechain further downstream the compression ratio will be very low. I might be wrong though. I probably am wrong because I don't know what your sidechain is like. 

To me it looks like this kind of limiter has to be either feedforward or like in your earlier schematic where sidechain feed was taken from vari-gm gaincell anodes. The tubes idle at low current and the signal is very small on the anodes. As it goes in to gain reduction the anode current rises and the gain increases on the anode circuit and we get a high compression ratio. With a simple sidechain amplifier I think you have to go with feedforward or with something like you had in mind earlier.

 

[alexc]

Side chain has both voltage and current gain so can be tailored to a reasonable degree for 'threshhold' and output impedance (proxy for fast 'attack' time) as well as the 'release' time.

The ratio of the gain cell should be dependent on the slope of each applicable (driven) segment of the S curve. It can be set to a reasonable degree thru choice of plate/screen voltages and of course the grid-cathode voltage.

Further tweaking can be done by varying the screen stopper resistor from a low value (<1K) to a higher value (around 24K).

I don't intend to have  a 'variable ratio' as it's quite hard to do (and test!), but if I were I'd either go  a SS sidechain.

Feedforward isn't my intention for this build - it can be done but is somewhat a different fish altogether. More difficult to do right and ultimately I don't see the point unless done like the Trimax (or the Finalizer!) - with a delay line to enable super-super fast limiting.

And I'm not using any anode load in the gain cell - pure CFs with regulated B+ to plates and a bit less to screen.

Thanks for the interest! I'm doing the sidechain now and I'll put up shortly.

ps - I particularly liked your sidechain solution in your recent 6V6 build - I am likely to do that style in another build

 

[alexc]

Looking at the cathode circuit of the vari-gm CFs, my changes will be

- add a 100R trimming pot leg for balance across the legs in place of the 33R
- use a variable regulated bias
- adjustable  metering

At low-dB GR typical operation of 5mA thru each side, and with no CV applied (grid is grounded), Vgrid-cath  should be -9V.

That means that suitably biased, the cathode string  will equilibriate to
    9V (caths) to 34V (after 5K) to 35.5V (after 300R)  to 35.75V (after 1/2 of 100R) 

At that point, there is a metering of the combined CF currents, and a final series resistance  to  cathodes return.

The cathodes return can be ground, similar to 'self  bias' or a  variable 'fixed voltage' circuit - similar to lots of amps and varimu type circuits.

I like an active (constant voltage) current sink  - something the Analag Clamp from the famed Analag/Silent Arts Poorman 670.

Imma put a SS positive adjustable current sink there, which will allow me to easily position the 'typical usage' op points on the S curve. It could be seen as a kind of a 'ratio' control - indeed if it makes much of a diff in practice to vary, I may well repurpose  it as a 'feature'

So, thats the basic idea of the cathode circuit - I'm now taking a closer look at the fixed resistances : the 5K and 300R.

I may tweak them some - the ratio of the high fixed to the low fixed, and their ratio  with the varying Zcathode  - will  determine of the GR achieved.

That in turn will determine the sidechain CV 'mapping' required, off the highZ audio signal tapped from the 'output amp'.

This gain cell is going to take a high level line input, up to +24dBu  (35Vpp)  at 20Hz with low distortion.
After the stepdown, thats  typical max.  around 9Vpp into each cathode circuit.

This CF could handle an input much larger than that!

Should result in low THD  ie. not so harmonically 'enhanced'

Output of each cathode circuit will be around 8Vpp max (no GR) to 4Vpp (6dB GR) to 2Vpp (12dB GR).

Not too much level for the gain cell output transformers - they can do 18dBu or so with no probs for 20-20K. The low Zout should have no problem driving the 15K/15K Altec there.

I'll update shortly.

 

[alexc]

One thing to explore is where to place the required fixed gain stage.

At present, I plan to do that in the driver stage of the line output amp. The 'makeup gain' control is easy to implement there, as well as some local negative feedback if desired.

I could also place that in the gain-cell as a diff amp. That would possibly give the advantage of some cmrr noise cancelation.
But it makes more demands of  the output traffo and is not easily variable.

So -  the gain goes in the output amp section along with a 'Level' control and some local nfb - driving a push-pull cathode follower. 

Same as the final part of the input amp but uprated to 6UA8 triode+pentode. I like the 'psu noise sampling' aspect of that circuit (from Silent Arts/Analag D-OAC tube opto limiter).

Also to put up shortly 

 

[alexc]

With regards to the psu - looks like I want :

It has 6.3Vac CT at 5A min    12Vac at 1.5A min  270Vac at 200mA min   

I like to have the sidechain tubes 12V dc heater (at lower current) and the signal amp tubes 6.3V CT ac heater (for lots of cathode followers and a noise adjusting pot)

That spreads the load a bit and means I don't have so much ac current flowing all over the place. As well as fewer dc filter caps and dropping resistor heat/real estate as if I went all dc.

I could go all ac for heaters but I don't like more than 4A or so on one winding. > 6A on one winding is too much for me!

-----

So, the total heater to be used  going to be 4.2A of 6.3Vac and 1.2A of 12Vdc.

That takes up 4 of my psu cable conductors  (hac, hac, hdc, hGND)

2 more for the 140mA of 300V (hvdc, hvgnd) and a last one available for a Vreg/utility voltage.

That's all on some hefty guage wire - no problem whatsoever with current capability. The socket/connector set I chose is cheap. I know it works for the 5 pin version, but the 7 pin may be a little too flimsy. Should be OK but I will see.

Thats a grand total of 12 audio tubes, 7 sidechain tubes and  8 audio transformers.

All up  8)

 

[alexc]

And, or course, in-circuit characterisation of a few sets of  6eh7s will be required to match up for best balance 

 

[alexc]

Ordered an Edcor XPWR173 for the psu - it has 400mA ! of 360V ct HV, 7A of 6.3V ct, 1.5A of 12V and 10mA of 100V.
The HV is way too high and also way too much current - but that's my choice

Not cheap ( at around 200usd delivered), what with todays shipping prices (all the way to the lower part of the globe), but well worth it.  Only the mighty Hammond 372JX is a worthy comparison, a beast of a traffo but also very quiet in terms of radiated field (and vibration wise))!

-------

Will have 7 pin psu cable -> hv, hv gnd, heater ac1, heater ac2, heater dc1, heater dc gnd, util

ps - the 7 pin connectors I have are probably a touch too small for this psu unit. They look fine to use in something of lower currents and lower HV. They are however, quite bit more heavy duty that the tube mic psu connectors I've used, for example.

pps - the 5 pin is better, with heavier pins and more clearance, so that will work in most cases, however I do need 6pins min, preferably 7.

So to be looking for the next size up. Failing to find a decent cheapy, I may have to spring the bux for something a little more industrial. Either that or use some octal socket+plug sets. I may be able to go that way.

---

The 350Vac HV is Diode rectified and smoothed to around 450V unloaded and around 420V or so loaded.  For this project, will need to drop around 70V  (7W or so)  to 350 V for B+ of around 120mA max.

Regulation will be used for the hv at the psu module,  with a pcb series pass mosfet plus discrete error amp.
Depending on what heat I can live with at the mosfet, I want to get to to around 300V B+reg around 60mA max.

Further regulation on the limiter unit, with a  string of 150+85V glow tubes (SG1P, SG16P) to give 235Vdc at 30mA max  - one string for each channel's vari-gm gain cell plate and screen voltages.

There is also the Vbias for the gain cell cathodes - a zener regulated, RC filtered setup with a variable bias voltage of around +35Vdc at some tiny current. This comes in on the 'util' pin of the 7pin connector and will be set at the psu I think.

------

So  final (starting!) tally for the 'Esquilax Master Box'  is 120mA of B+, 4.9A of 6.3Vac and 1.3A of 12Vdc, which is around 125VA all up.

That leaves around 200mA of HV spare - this is an external modular psu, so I can use it for another of the 'modules'.
I'll team up with an additional heater transformer and do the final unit in this series : a 10W+10W parallel triode pp amp.

Tube lineup at the starters gun are :

(x2 for L, R -  26 tubes all up!)

- input amp      : 6JW8, 6U8A
- gain cell          :  6EH7 x2, SG1P, SG16P
- output amp  : 6922, 6SN7
-sidechain        :  AT7, EL95x2, 12AL5, 5963

I have the main parts on the way 

 

[alexc]

Side chain is sourced from the signal output amp, at the plate of the first (voltage amp)  stage -  a 6922, which has a reasonable capability to drive 2 loads, all of hiZ.  ie. it drives the grid of the push pull CF  line amp as well as the grid of a phase inverter

It has very low plate resistance (2600ohms) and will use both stages 'paralleled' for an 'unparalleled' low imped, as well a decent current capability.  Biased sufficiently, it should have no problem here driving two hiZ loads and will provide an amplification that is 26dB or so.

Enough voltage gain to cover the stepdown of 16dB at the output traffo,  and also 10dB or so for the 'Level'  pot at it's grid.
If I want some extra gain, for nfb, I will change it to something with a higher u, even though I will give up some it's stellar Rp

The remainder of the gain makeup will likely come from the the line amp configured as a mu follower.

The GSN7 is my choice here .It can easily handle 10 to 15mA thru it and sound good nearly all the time.  I spent some good dough here for a pair of nos rca matched jobbies   

Load is a 20K trasffo primary with what I think should be a good, high inductance; combined with a  really good 10uF film cap

As a mu follower, it can do an 26dB, if it's required, otherwise just leave as a regular 'cathode-following-triode-with-active load-and-psu-noise-sampling' setup (after Analag, again)

I really wanted to use a pentode+triode here, good for 3.75W max dissipation each part - like a 6U8A but a 1/3 more W.
Couldn't find within my heater budget so I went for the old mainstay of good ol' sound, the 6SN7.

Same for using a seperate envelope triode and pentode - couldn't do better within my heater budget. So - 6SN7 it is!

-------

Indeed I've gone for 'matched' tubes or sufficient set to select for a match, each type. A mastering box should be pretty close left-to right and it starts with every link in the chain!

One of the things I liked about the Silent Arts/Analag D-OAC box is the very high degree of L-R matching I measured, with no selecting; just stock EH tubes. It was extremely close, both signal amp and control amp.I was very impressed after having just learned about matching in vari-gm gain cells and sidechains!

Of course, the vactrols are not so accommodating, so here I am taking some steps to 'do what I can' to match the opto GR across L-R.

I'll see how everything  fares when I get there!

----

So the hiZ, high ac signal amplitude,  single phase side chain feed drives the grid of an AT7 phase inverter, long tail type.

That drives a push pull pair of cathode followers with around 23mA per leg - EL95s. These are a cool smaller-to-mid size pentode, 200mA at 6.3V heater, with a max 6W plate dissipation. Good for 20mA as a CF, which is what they are here.

There will be some feedback around the 'phase-inverter'+'current-amp' stages to vary the gain of the phase inverter - this is the 'AC Threshold' control.

The side chain amp does  up to 20mA per phase, thru some cathode resistance, to provide  a variable voltage signal of medium  driving Z, to  12AL5 plates with up to 10mA per diode anode.

There is going to be a 'DC Threshold' arrangement here;  the usual 'glow tube regulator with a  variable resistive divider' to set an offset voltage will be somewhat different here as I am using the diodes reverse, for a positive going CV pulse.

I will use -Vthresh sourced from the 100V 10mA winding with a zener on the 'Vutil' psu cable pin.

The combination of  AC Threshold and DC Threshold, similar to my version of the venerable FC660, can hugely change the character of the vari-gm gain cell, so its a 'Must Have'  for a mastering limiter 

Cathodes of the diodes are combined to provide a positive going CV pulse, of varying amplitude, with a max 20mA driving current for the  for vari-gm gain cell time constant network.

I think the CV range in practice will be around +5V to +20V or so, but with the ability to get there in a hurry! and drive both grids of the 6EH7s.

Time network is fairly standard, the usual vary of R attack, C release and what not; should end up pretty fast  with a decent whack of current from a moderately lowZ driving source (CFs).

-----

What interests me (this time around) is getting a critically damped connection between the sidechain CV and the gain-cell.
It's really easy to get massive overshoots that give GR more reminiscent of a tremolo oscillator than a limiter 'grabbing and letting go'.

So - I'll be looking more at the impedance match of the signal amp and control amp networks.

Of course, thats where a lot of the magic lies - how the CV modulates the gain cell  What range of gm is 'moved' by what range of 'CV' in what range of 'time'!

And most importantly, how smooth this modulation translates into audibly discernable gain changes, on the 'grab' and the 'let go'

Such is the magical allure of toob lims  8)

-------

But wait, there's more!

an additional au7 type tube, the 5963,  per channel, configured as the vactrol led current driver - another push pull cathode follower with the cathode resistance providing a voltage signal to the 2 vactrols'  (in parallel) leds, 5-10mA or so, per leg.

Paralleling two vactrols provides some 'averaging' of the optical characteristics, as well as summing of the phases into single phase  ... um  ...  'light'   

The vactrol driver should be able to vary the intensity of the 'light' so something that way is also required. Again, I'm taking Analag's setup of driving the vactrol leds with a voltage signal - parallel-fed via cap, off the CF cathode resistance. I spent a lot of time down there on the D-OAC and came to appreciate the way it was done there.

The alternative would be to have the led in series with the cathode and cathode resistance, similar to EJ Jurich. I'm sure his scheme would work, but as I said, I've spent a long time down the other way 

That, plus his driver uses another double-diode tube to rectify a voltage signal, then pass it to a cathode follower with leds in series. Analag doesn't use specific rectification before applying signal to the vactrol.  It's kind of an ac-threshold, dc-threshold, 'peaks-only-amplification' kind of thing. If you know what I mean.

 

[alexc]

So far, I have fully independent audio channels for L, R and fully independent side chains. I am interested in doing some tricks with regards to 'power summing' of the left and right sidechains.

The EJ Jurich paper describes such a thing in his design and it relates to the summing across phases  at the side chain detector double-diode plates. He uses L channel and R channel as 'phases' in his design, so that works as a L-R 'power sum' in the sidechain.

My gain cell scheme is all balanced, per channel, so it's  going to be different, but essentially, the same idea. It's a little trial-and-error here : how to combine a portion of each channels' detector output such that the sidechain provides a consistent 'GR image' across the stereo field.  tbd you might say.

It's an exciting area, that of 'stereo field imaging' in general - one day I'd like to try an add-on to this series of tube processors - some L, R, L+R, L-R tomfoolery.    With hybrid mixing coils, If I can    8) 8)

----

There is also the matter of EJ Jurich's 'two stage' time-constant approach within his vari-gm side chain - what he describes as a form of 'adaptive setting' or 'auto' attack/release. I'll see how that goes in practice, but I like the concept.

----

Here's the vari-gm sidechain so far;        vactrol side chain to follow ....