Tube Amps for Studio Monitors

[emrr]

Cool.  There's at least one example I can think of where the same 10W PP transformer was used SE at about a 1/4 watt in professional broadcast, getting good response with unbalanced DC of 25 mA. 

I have been listening to vintage PP cathode biased 10W 6L6 with some Klipsch La Scala's recently; wow.  When the speaker does 105dB/1W/1M, 10W will get plenty loud to hurt.  The same amps sound pretty terrible with modern low efficiency speakers. 

 

[gridcurrent]

I can't think of a serious tube monitor amp that did use cathode bias on the output......
Marantz 9 (which we used at Oceanway for mids & HF) fixed bias
McIntosh 275 fixed bias
EAR 549 or 509 fixed bias
there are probably others I've never run into

Yes indeed David.
At least in the Americas, fixed bias was the norm.  And McIntosh amps were in just about every major recording studio.  NBC's order of fifty 50W-2 amplifiers was instrumental in getting Frank's company started.  40 years ago, I saw plenty of MC-30's that used to power LC-1a's, retired from studio use and destined for what your British friends call the "skip."  McIntosh amplifiers were so popular and in demand that RCA offered them through their broadcast division, both the MC-30 and 50W-2 are in the 2nd edition catalog.  And not just the big studios, even Norm Petty had Macs.  I suspect there were few if any self bias amps on this side of the pond suitable for studio monitoring, even going back to the 50's.

 

[alexc]

The sub of  a 12B4A GE nos certainly sorted out the hum and microphonics 

With the 5687 setup for gain of x17, I had 50Hz peak at approx -75dbU and 100Hz at 51dBu (!!).

After sub 12B4A setup for gain of x6, I have -85dBu and -75dBu respectively. Massively reduced.
Reduced gain is not a problem, I have plenty available which I will tailor with the feedback network.

Microphonics much reduced.

As well as the tube, the problem may have been that I didn't have the 9pin tube socket centre lug grounded and I had the unused half of the dual triode floating. I'll test that on my other channel.

So - I'm happier. The 12B4A has a high reputation with a lot of folks again with the caveat that high microphonics can also be a problem. This one is way less than the 5687.

Now for some listening 

 

[emrr]

A few professional post-war cathode biased monitor amps of USA make, designed and sold to recording studios.  You wouldn't see many of these by the mid-'60's, but some were still on the market.

Altec
127
324
332
429

Collins
6X
6V
356B

Gates
6C
SA-10
SA-66
SA-95

General Electric
4BA4
4BA12
4BA14

Langevin
110
117
128
130
138
610
5117

RCA
76 console monitor amp
82
BA-4
BA-14
BA-24
BC-2 console monitor amp
BC-3/5/6 console monitor amp

Western Electric
124
133
142

 

[bockaudio]

A few professional post-war cathode biased monitor amps of USA make, designed and sold to recording studios.  You wouldn't see many of these by the mid-'60's, but some were still on the market.

Thanks for that. I didn't really start until 1979 so my firsthand  knowledge is limited.

Yes indeed David.
At least in the Americas, fixed bias was the norm.  And McIntosh amps were in just about every major recording studio.  NBC's order of fifty 50W-2 amplifiers was instrumental in getting Frank's company started.  40 years ago, I saw plenty of MC-30's that used to power LC-1a's, retired from studio use and destined for what your British friends call the "skip."  McIntosh amplifiers were so popular and in demand that RCA offered them through their broadcast division, both the MC-30 and 50W-2 are in the 2nd edition catalog.  And not just the big studios, even Norm Petty had Macs.  I suspect there were few if any self bias amps on this side of the pond suitable for studio monitoring, even going back to the 50's.

Thanks to both of you for filling in the gaps.

 

[emrr]

No firsthand knowledge here, just a paper historian.  I'm probably the young guy in this thread.

 

[Mike Cleaver]

MC 30's were pretty standard in radio studios back in the '50s and 60s, often driving disk cutter heads and for monitor amps.
10-30-50 watt amps were plenty powerful for driving the high efficiency speakers used in those days and the amps were simple straight forward designs that were easily repaired and maintained.
RCA console monitor amps were usually 10 watts or less, I think the monitor amp in my BC2C console puts out about 8 watts from two 6V6's.
But that was plenty loud enough to drive an RCA LC-1 dome cone 15 inch speaker to shake the room.

 

[alexc]

Been listening a while to these amps with the 12B4A in the input position (instead of 5687)

Well the listening tests agree with the bench - mains hum is now very low  (approx 12mVpp at the dummy load) and the overall gain is slightly less (no adjustment of feedback done yet). I can actually hear the fizz at full volume being the dominant noise (as opposed to the hum)

But more than that, the sound is more relaxed and less ...  uptight  Not as fatiguing at higher volumes.

So the 12B4A is a keeper. It is still a little 'pingy on the tap test but much improved. Just have to do the final tweeks on the operating point (20mA and 160V on the plate, 310V B+, 17V bias, gain x6.1) and then to focus on the feedback network which is completely un-optimised so far. I want to minimise the feedback if I can.

So - a big, big improvement with the 12B4A in the input position.

 

[alexc]

So I've done some feedback measurements and  turned up some surprising  results (to me

Firstly, just to refresh, I have a standard 3 stage Mullard topology - input stage, phase inverter and finals.

Input stage is a single triode (12B4A) with gain of  x4.33 (+12.7dB)
Phase Inverter (6SN7) with gain of x4.98 (+15.5 dB)
Finals pushpull (EL34) with gain of x10.0 (+20dB)
6K6ohm to 8ohm step down traffo with gain of /15.26  (-23.67 dB)

Input to 8R dummy load overall gain is x17.1 (+24.7dB) with a 0.17dB error in meas. and calc.

OK. So I add feedback thru a resistor from 8ohm tap to the input stage cathode circuit.
The cathode has 1K with bypass cap to the feedback node then 120R to ground.

All pretty standard except :

- I have a very low gain input tube with a very large input signal margin
- I have very low overall gain as a result of the above

---------------------

What I see is this, with respect to feedback :

- open circuit (>1M feedback resistance) : no effect on output voltage
- decreasing feedback resistance down to 40K : output voltage starts increasing
- decreasing feedback resistance down to 10K : output voltage increases +2.3dB
- decreasing feedback resistance down to  5K : output voltage increases +5.5dB
- decreasing feedback resistance down to  3K : output voltage increases +8.5dB
- decreasing feedback below 3K down to 1K and to short circuit : output rises fast and goes bonkers

I checked the phases :

output is 180deg out of phase with input
feedback is in phase with output when not connected to the input stage cathode circuit

and

feedback connected (15K) cathode circuit feedback node is 90deg out of phase with input grid
increases this difference as resistance is decreased down to short circuit

========

So - my question is this:

WTF?  I thought negative feedback is supposed to decrease the output voltage

I guess my gain (Av =17) is too low to allow use of nfb?

Instead what I guess that I am seeing is feedback becoming positive as I decrease the feedback resistance to essentially short the output to the cathode circuit ?

And having a capacitance in parallel with the feedback resistance makes essentially no difference.

------------

Anyway, at my original feedback resistance guesstimate of 15K, I essentially have no feedback.

Basic performance is very good :

It sounds really good
Frequency response is flat as a ruler  with overall (-1dB)10Hz to (-3dB) 24KHz or so.
Noise and hum at the output is around 6mV and hum is independent of volume knob
No frequency instability at any output level to 30W rms
15Wrms distortion is 2nd harmonic 40dB below, 3rd+ harmonics 50dB below the 1KHz test tone
Square waves 1KHz pretty good - some rounding leading edge but very little overshoot or ringing

(still testing all this ...)

------------

What's going on?  This is my first time measuring a feedback network like this

(Before I keep it as is with no feedback and button it up for now) 

Cheers
AlexC

 

[ruffrecords]

The one aspect of your design that is nonstandard is the input stage. Could you post a schematic?

Cheers

Ian

 

[alexc]

Thanks Ian

Here is the schem - hand drawn, still challenged by schemo cad
** Corrected schem to v1.2 - PI values updated, output phase corrected for nfb **


To clarify some component values :

- Rfeedback is 15K, Cfeedback is not used

- input stage plate resistor 6K8, plate voltage 185V, bias voltage 21V
- input stage cathode circuit is 1K to feedback node bypassed with 80uF and 120R to ground

- phase inverter has around 190V on the cathodes

I have been testing with around 1Vpp on the input stage grid and 17.1Vpp on the 8ohm output tap
Feedback node started at 60mVpp or so (15K Rfb), starting at 90degrees phase diff with grid

(it can go to some 20+Vpp on the grid without distortion and some 43Vpp on the output without distortion into 8ohms)

Mighty mysterious to me - I'm certain I have the phase of the output and input 180degs and the feedback is definately taken from the 8ohm tap of the secondary.

Ricardo schooled me - the feedback should be in-phase with the input. Thanks! 

Cheers

 

[ruffrecords]

Something is not right in the phase splitter. The 6sn7 cathode volts ought to be higher than its grid by a few volts. Also given the volts drop across the plate resistors the plate current ought to be about 4mA. That's 8mA through the 12k cathode resistor which should therefore drop less than 100V.

Cheers

Ian

 

[Rob Flinn]

My ignorance. What are the Leaks, what was their power, what were they driving and what era?

Check out this link:
http://44bx.com/leak/PointOne2.html

and this:
http://en.wikipedia.org/wiki/LEAK

Cheers
Ian

I used to have a very nice pair of LEAK TL50 monoblocks that sounded great.  50W, cathode bias, using 2 KT88 (6550).  Sadly I had had to sell them when times got hard a few years back .......

 

[emrr]

Without digging up any reference materials, off the top of my head, might this be an issue related to the differences between current feedback and voltage feedback? 

Maybe try NFB from the output primary and leave the output transformer out of it. 

 

[ricardo]

...
- decreasing feedback resistance down to 40K : output voltage starts increasing
- decreasing feedback resistance down to 10K : output voltage increases +2.3dB
- decreasing feedback resistance down to  5K : output voltage increases +5.5dB
- decreasing feedback resistance down to  3K : output voltage increases +8.5dB
- decreasing feedback below 3K down to 1K and to short circuit : output rises fast and goes bonkers
...
output is 180deg out of phase with input

You have the speaker windings connected the wrong way round so you have positive instead of negative feedback.

In fact, what you describe was a common way to check if the speaker windings were the right way round.  If adding a bit of feedback increases the output voltage, it is the wrong way round.  If it decreases it, it's negative feedback.

On this type of amp, the output should be in phase with the input.

 

[alexc]

Thanks for the help - I'll look into it further today and get back to you 

Ian, with regard to the phase inverter, yes you are right. My drawn figure is wrong. I'll correct that shortly.
In addition, I forgot to adjust the PI cathode resistor since I changed the input tube.

It should bias around 6V or so up from the grid.
I also need to do some more biasing of the input tube to lower that plate voltage some more.
It should be more like 165V or so.

Ricardo, I'll check again the polarity of the secondary of the output transformer and the polarity of the feedback.
What you say makes sense, although I can't see where my polarity error is yet.

Doug - I could do that but the Mullard design is what I'm following for now and that calls for this feedback scheme.

I'm sure there's something simple going on 

Cheers
Alex

 

[alexc]

Updated the schematic and post with corrected phase inverter voltages and added phase diagrams.

In terms of phase, I have :

input                      positive phase
stage 1 grid            positive phase
stage 1 cathode    positive  phase
stage 1 plate          negative phase
output 8Ohm tap    negative phase

stages 2 and 3 are differential so each phase present

Edcor  CXPP60-8-6.6K traffo spec shows yellow wire as '8ohm' tap, white wire as 'common', which is grounded in the mullard design.

Which all seems correct to me - output is reversed wrt to input due to the stage 1 inversion of phase.

Correction
-----------

As Ricardo says, the feedback fed into the stage 1 cathode circuit's feedback node should be in phase with the grid signal.

I just reversed the white/yellow on the Edcor output traffo and now it's just Jim Dandy.

Corrected phase relationships are now :

input                      positive phase
stage 1 grid            positive phase
stage 1 cathode    positive  phase
stage 1 plate          negative phase
output 8Ohm tap    positive phase


Now can I get back onto measuring up the feedback. I updated the schemo post to v1_2 with Ricardo's and Ian's corrections.

Thanks guys  ;D ;D


Cheers

 

[alexc]

My Rfb values range from 150R for 20dB nfb to 20K for 1dB nfb.

I think that 12dB would be the max and 8dB the min I would go for in this design.


So I'll do some evaluation at 10dB nfb which is 1K feedback resistance. (no cap yet)

I guess I'll be looking at stability issues first, then improvements in linearity, noise floor and distortion in particular, which I think a little on the high side with the zero nfb I have used up till now.



..

First thing I see is the max OP before visible distortion is now 47Vpp into 8R dummy load.
Thats 34.5Wrms up from 30Wrms before.  Cool! More clean jus
Input grid signal is 8.7Vpp for max output (1KHz sine). No sign of any distortion at the input.
With a bias around 20V on that 12B4A it should swallow a whole lot more too!

Frequency response is now extended all the way beyond 40KHz where my anaylzer stops.
Something like 35KHz for -1dB.

Analyzer is using a dummy load which has maxes out at around 10Wrms or so.
I need to up my shunt resistor in the load  box to allow full power operation with an analyzer output of around 10vpp, which is the max my Motu interface likes to see.

CRO confirms this at full power frequency response to around 35KHz before any output level drop.
And all the way below 10Hz

..

OK so I can see now the effect of Cfb in parallel with Rfb. On 1KHz square wave, at half power or so, something between 3300pf .. 4400pf removes the little overshoot and squares up the leading edge somewhat. Seems like an improvement, so I'll go with that for now.

..

One thing I'm struggling with is how to 'map' the output voltage range of the amp of 0 ..48Vpp (into 8R) to the available RTA voltage range of 0..10Vpp ( into 5K ?)or so.

I am trying a few different of shunt resistance across the  8R power  resistor in the load box.
I'm just trying to work out what makes sense.

With my current settings, I seem to see a big kick up in the distortion components displayed on the RTAS beyond a certain input voltage level and I can't tell if it's the amp or the RTAS!

Anyway, all is looking good, so I'll go back to listening with this feedback network for a while.

In the meantime, I need to do some more learning on measuring power amp performance.

That and completing my neve summing bus

Cheers

 

[alexc]

After listening for a while, I've decided on more changes

The amps with 12B4A in first position are not sensitive enough with the 10dB feedback I have applied.

They sound great and super quiet, but .. yep .. not sensitive enough.

I miscalculated (again) with the first position amplification factor. I now think I need something with a mu of 30 or so.

Back to the drawing board to hunt a suitable tube (which I haven't used before - like to mix it up a bit   )

I will move to mu 30 via some intermediate steps, starting with 6S4A with a mu of 16.5
I also have some nos GE 6BZ7 at mu 36 which I have used before and quite like.
Other candidates are 6CG7 mu 20, ecc99 mu 20, 6n6p mu 22
And of course, AY7, AV7 and 6922

'like' in this context being a function of :

-> is it NOT a 12ax7 or 12au7 ? proceed.
-> cost, how nos-usa is it?, is it a groovy russian?
-> does it chew up too much juice? If so, good!

and then

-> noise+pinginess, suitability for the task

and finally

-> SOUND



----------------

Doing some more reviewing of datasheets, I've decided my next try at the input tube will be the russian 6N6P.

It's reasonable cost for nos and supposedly mil spec and combines the features that originally attracted me to the 5687 (and 12B4A) but with a higher mu of 22.

It retains the necessary properties of :

- highish grid voltage (say -8V or so needed in a studio monitor amp with quite high input signal)
- reasonably low plate voltage (say 160V needed for direct couple to phase inverter with elevated heater)
- not too high current (max 18mA or so) which seems to go with the above

and also has ok capacitance cag of < 3.5pF (lower than 5687)

I noted the 12B4A was great with no feedback. 10dB of feedback is a x3.16 gain. So my 12B4A mu of 6.5 needs to be more like 19.4. And a little extra as I felt the pre feedback case could use a little more level. So mu 22 is about right. I can reduce the nfb slightly to give a little more gain if needed. Say down to 8dB from 10dB. That would be my minimum.

So 6N6P is next my choice. Narrowly edging out the 5687 and 6S4A becuase of it's higher mu which I need. And beating the ecc99 on cost - which gives me a chance to hunt for reasonable microphonics.

I'm also going to try some cheap tube damping rings. The reason being I have these amps poorly situated on top of the speakers! Dumb I know but I don't have a whole lot of options.

I'll let you know how it goes in a couple of weeks when I get them from the Ukraine!

 

[alexc]

I sub'd in the 6N6P tubes in the first stage now - adjusted the B+ to the first stage, the cathode circuit and the feedback network to suit.

Microphonics are not there at all with these. Worlds better than the 5687s and 12B4As I tried.
The 6SN7s are the next microphonic ones now.

Now I have the sensitivity of the amp about where I want. With 9dB of feedback around 13Vpp at the grid at the onset of visible clipping at the speaker terminals into 8R load box at around 47Vpp.
That's around 16.6Vrms and 34Wrms. EL34 finals have -40.4Vdc on  each of the grids.

So the volume knob is at around 2 o'clock for a strong, comfortable listening level - probably 10Wrms or so with an input signal of around 12Vpp, which is pretty typical output of my Finalizer. More or less.

Performance is similar to what I previously described except I have no Cfb - I found 3300pF was good with the 8R dummy load box but not with the speakers - I would get oscillation. Much lower, like 100pF would be fine with respect to oscillation but made little difference to the square waves. The load box square waves were great, but with my intended speakers (high sensitivity, double 15"s+mid+tweet) square waves 'triangulated' somewhat at higher levels.

My first stage cathode circuit could be a little better - I'm still not quite sure how to claculate the value of the cathode feedback resistor, so I used ratios from other designs. Around 15% of the total cathode resistance.
In this design, it may be that 50R in a total of 350R cathode is too low. Not sure yet. Next time I fiddle I'll do some more refinements.

I'll be looking at that some more going forward, but for now it is fine. Low distortion, full frequency spectrum, big bottom end, whisper quiet hum independent of level, very low hiss, dependant on level.

These are now in service

Here's the updated diagrams ...