RCA OP-6 Analysis

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.


Well-known member
Nov 8, 2005
When a mic pre-amp gets as famous as the RCA OP-6, I thought it might be interesting to see what we can find out about it, what makes it special.  The techniques I used to get the data may be new to some of you and maybe useful on your own projects.

There is mostly hype on the net and a youtube demo, but what hard facts there are have been supplied by Doug at EMRR.

This is what we know:-

OPT is 48k:500,      IPT is  50/250:50k,    Choke DCR is 4k,      Choke inductance is 400Henries.

All we have to help us is RDH4 , Excel and the schematic.

First of all I made some assumptions about the three stages from the resistor values given.  V1 and V2 are very low current and the output stage V3 is high current.  With any pentode, the ratio between the plate current and G2 current changes depending on the overall current, so we have to make a guess here using the handy chart from RDH4 page 515.

From this chart I decided to use a ratio of 4 for V3 and 4.3 for V1, ( V2 would be similar).  Starting with V1 first, I started a spreadsheet with a range of B+ values ranging from 260 to 200 (for the table I deleted the least relevant).

Next I used the info from RDH4 which shows that the G2 loadline can be drawn on the triode characteristic when you change the scale according to the ratio:-

Here you can see that I divided the current scale by 4.3 to give those odd values and at the bottom there is the 1.2M g2 resistor load line.

I put the values for 200V into the spread sheet and calculated the voltage drops and hence the current.  You then multiply the g2 current by the ratio to get the plate current.  With the total current you have the cathode current and hence the cathode voltage.

From the table you can see that the measured G1 voltage from the chart and the calculated voltage from the table coincide at somewhere between 0.75V and 0.8V, this seems like sailing too close to the wind to me and I would not have designed it that way myself.  I checked these figures on an actual circuit and they are correct at ~0.7mA.

Actual voltages for V1 are:-

B+208V,  Vp=76V,  Vg2=25V  Vk=0.74,    current ratio=3.93

I did the same thing with V3 but as I had very little idea of what the B+ was, I had to use the maximum dissipation figures to give me an upper limit.  I used the Radiotron version of the data from Frank's site.  Here it says a 6J7 has 0.75W plate max and 0.1W g2max.

These figures are all obtained with simple ohms law from voltage drops, nothing fancy.  The red figures exceed the spec and the blue figures are the estimate.  After this I made the circuit with a 4k load (no choke handy) and the actual figures are:-

B+ 222V,  Vp=208V, Vg2=100V,  Vk= 2.07.  Wp max =0.728W, Wg2max= 0.0938W, in the end the current ratio was 3.73 which accounts for the error between 110V and 100Vg2.

I will be posting more on the gain of these tubes later on, but I must say that it was a surprise to me that the designer went to such extremes.  I think that he/they maximised the gain of all the stages so that they could use lots of feedback to kill the noise but still have sufficient gain.  Whether it is actually 90/95dB we shall see.

I hope this is not too nerdy for most of you, but electronics is that way inclined!


Nice examples here! 

I could have given a bit more data to work with.  The manual states B+ of 270V at 6.1mA for battery operation with six 45V batteries.  Total rectified voltage across C-11 should be approximately 255V with AC power.  These are nominal values which shouldn't vary more than 5%. 

I misspoke from memory slightly on the IPT, 30/250:50K. 

Max input -24dB*
Max output +19dB*
Overall gain 90dB
Obviously max input to max output is operation at 43dB gain, so tossing 47dB on the gain control. 

An addendum for units preceding MI-11202-A version changes:
R-4 to 1500
R-7 to 2700
R-8 to 2M2
R-10 to 680K

Note C-13 is a 300V part.

V1 P:106V G2:37V K:1.2V (was 80/35/0.93)
V2 P:68V G2:26V K:1.15V (was 82/25/0.93)
V3 P:240V G2:110V K:2.23V

DaveP said:
I hope this is not too nerdy for most of you, but electronics is that way inclined!

Nah; this is good. Sure beats another 'which cap' thread.
> pentode, the ratio between the plate current and G2 current changes depending on the overall current

The /ratio/ stays semi-constant over a wide range of current and voltage. Close-enough for design, or reverse-design. (Not counting aligned-grid tubes like 6L6 and a few TV types.) 4:1 is a fair bet for many small pentodes. (Larger G2 current uses significant power, lower G2 current would not save much.)

> somewhere between 0.75V and 0.8V, this seems like sailing too close to the wind to me and I would not have designed it that way myself.

I do not understand your close-wind thought. 0.7V is often a fine value for an input device's bias.

And note that, here, 0.7V is not the maximum input, because there is NFB in front of the grid. Does it work? Ballpark a hi-gain pentode should give Gv=50 to 100. The NFB ratio is around 25:1. There's not a lot of feedback, but the signal at T1 pin 7 can be about twice the grid bias. 1.4V there is probably 100mV off the microphone. You NEVER find 100mV coming off an RCA ribbon! (Not before mike-lickers and Marshall 100Ws.) If you did get stuck close-miccing a DC3, every tech had pads.

Another way to get "near 0.7V" bias. Ass-ume the plate sits at half of B+. That's not max-gain, but is near max-output, for most loadings. The Raw DC off 230VAC will be a bit over 300V, there's drop-resistors but not large. Take 300V at C10, 150V across R5. 150V/220K is 0.68mA. In R4 1K that's about 0.7V.

Check for impossible: 0.68mA in R5, 4:1 Ip:Ig2 ratio, is 0.17mA in R3 1.2Meg. 0.17mA*1,200K is 204V drop. Vg2 is about 96V. Very fine. (Saying Ip:Ig2 is 3.73 does not change this result.)

If it were less than zero, or even less than 20V (Vgk times Mu), we'd have doubts. 100V is a fine bias for a small-signal pentode's screen. (Is in fact the datasheet suggestion.) For large-signal we need higher, for max-gain we want as low as possible. Here we have uncontrolled small-to-medium signals and want a fair balance of gain and output.

> 6J7 has 0.75W plate max

You are right, that is what the sheet says.

Similar but later-documented pentodes are 2W-3W. 6J7 in triode is rated 1.8W. Something funny here. May be over-conservative rating. The guys in Camden could call Harrison and get more aggressive numbers, all in-house.

Taking 300V supply and 48K load, for max output (and one pentode is not big output by line-level standards) we expect tube V/I to be near 48K. 300V/48K is 6mA. This is 1.8 Watts, violating pentode-spec but right-on triode spec, and probably "allowed" if you knew the tube-guys.

If we ballpark V3's Vg2 as "100V" (book likes that number), we have 200V across R-26 130K, 1.5mA Ig2, so 7.5mA for the whole tube. This causes 3.6V drop in R-13 470r. 3.6V times Mu of 20 suggests 72V on G2. Lower than assumed. Iterating, 72V on G2 says 1.75mA Ig2, 8.7mA total, 4V bias, estimated 82V on G2. Fair agreement. If Ip:Ig2 is only 3.7, then this result should be lower, maybe 8mA.

If we believe EMRR's addendum of "6.1mA" (AT 270V!), and allow ~1mA for V1 V2, we have 5mA in V3. 5.5mA at 300V. A bit cool for best output, a bit hot for book-spec, but not absurdly so either way.

So 5mA to 8mA, and pentodes are tricky enough that you have to go to the graphs and even bench-rig it.

> lots of feedback to kill the noise

Negative feedback does not reduce hiss.

Here R1 contributes non-negligible hiss; acceptable(?) because a pentode is quite hissy and R1's contribution just-barely raises the hiss level.

> gain.  Whether it is actually 90/95dB we shall see.

Without finding another matchbook, I'm sure this thing does not have 95dB voltage gain (from the 200-250r input taps to the 600r output). That would be absurd. (The 30r input adds 9dB voltage gain re: the 250r input.)
Very hesitant to contradict PRR, but if a bias of 0.7V was OK, then why did they revise the cathode bias upwards with the mods EMRR posted?

I was always told that grid current could arise anywhere from 0V to 0.5V so 0.7V seems too close for comfort.  In the end they chose to reduce the gain of V1 and V2 and compensate by reducing the feedback from V3,  they must have had a reason I guess.  Maybe for outside broadcast the power supply voltages they encountered were too variable?  Batteries discharged too quickly?

More on the gain later
Just to clarify the issue of noise and feedback.

PRR is right to say that negative feedback will not reduce noise in the input, but it does reduce noise within a feedback loop.

Extract from tutorial:-  http://www.learnabout-electronics.org/Amplifiers/amplifiers33.php

The Role of Negative Feedback
Noise at frequencies above and below the required bandwidth of the amplifier can be reduced by the use of high and low pass filters, but negative feedback can play a part in improving the signal to noise ratio within the bandwidth of an amplifier. The feedback signal from the amplifier output contains both an anti phase portion of output signal and an anti phase sample of any noise generated in the amplifier. When this anti phase noise is added to the input signal, it subtracts from the noise generated within the closed loop, reducing it by a factor of 1+Aβ compared to what it would be without NFB.

These are the results of tests done on V1 in the early configuration before EMRR gave us the addendum.
With no load the gain was exactly 100, with a 3.3M load it was 97.3, with 2m load it was 92.5 and with 1M it was 86.25.
To find out the gm and rp of this set-up,  I used Steve Bench’s formula, which is:-
Gain = Equivalent Plate Resistance/ Cathode Resistance.

The EPR  = 220k//rp//RL  and the CR = 1/gm  The cathode resistance is 1/gm when the cathode resistor is by  passed.
There are two unknowns here so I used simultaneous equations to solve the problem.
1/(1/220+1/rp+1/1000)/1/gm =86.25  and  1/(1/220+1/rp) /1/gm =100

The 1/gm’s cancel so there is only rp to solve for:- so the simultaneous equations become:-

1.15942/(1/220+1/rp+1/1000)  =  1/(1/220+1/rp)

This works out at rp = 579k after a page of maths!

Putting the rp back into the equations gives you the gm which is 1.595.

I shall be repeating the exercise for the other tubes. and thereby uncovering the complete gain structure of the amp.

Notes on battery life say:

270 hours for the B
16 hours with ten A, 34 hours with 15 A.  This starts with full charge being 7.5VDC. 

I measured gain from V2 grid to output transformer 600 ohm terminals (before output pad) at 38.5dB.  No idea how 'correct' the condition of this example was at the time. 
PRR was right (as usual) about the maximum power handling of the 6J7 8)

I had been using the GE data sheet figures, but when I checked with the Radiotron data sheet it was spec'ed  at 1.4W plate and 0.35W G2 for use as a RF power tube.  For some reason Frank's site lists the Radiotron 6J7 data as "undefined".

This checks out with EMRR's voltages so the B+ is settled as 270V, whether by batteries or power supply.  This means that the voltages of V1 and V2 will also rise to his figures, so I will have to re-test the tubes at the higher voltage.

yes, but feedback also decreases the gain, so the signal/noise ratio remains constant

Yes, that's why we often refer back to the E.I.N. figure.
Thanks for pointing that out
Remember two points of operation.

Total rectified voltage across C-11 should be approximately 255V with AC power.
followed by 6K8 to reach the same point 270V is found with fresh batteries.  The vast majority would be run today with AC power. 
Ok, I will check that out because 270V B+ gave me 253 plate and 121V G2 and 2.51 on the cathode.
Now with B+ 254V I get Vp=238 (with 4k DCR), Vg2 = 114V, Vk = 2.3V,

So we are in the right ballpark now

I've finished testing V1 at the new higher voltage and with the later version 1.5k cathode resistor.

B+ = 239V,    Vp = 93V,  Vg2 = 34V,  Vk = 1.24V.

Gain is 130 with no load and 112.5 with a 1M load, this works out at an rp of 530k and a gm of 0.8357.

With the 1M load and the 1.2M FB resistor the gain goes down to 19.75 which is 15.1dB of negative feedback.

That's all for tonight.

Is the above gain measured or calculated using the Bench formula?

Gain was measured on an oscilloscope.

I just used the Bench formula to calculate the rp/ri and gm.

The Bench formula gives the same result as the normal one using mu, but it's handy for pentodes where mu is rarely given in the spec.

Thanks, for verification purpose, could you please calculate the gm and rp at the operating conditions provided in the datasheet: Vp=Vg2=100V, Vk=-3V? I would do it myself, but I don't have a 6J7 on hand... :-[
OK, jazbo8, here are your results:-

I guessed at somewhere near 250V for B+. This was not an easy test to set up so I had to use pots for the screen and plate resistors and I measured them afterwards as 70k for Rp and 249k for Rg2.

Vp was 100V as was Vg2, with a 1.2k cathode resistor, Vk was 3.15V

Gain with no load was 72 and with a 1M load it was 68.

This gave rp as 368.4k and a gm of 1.224 (book says 1.185) I have a good tube I guess, others may vary.  mu would then be 450.

I can't devote any more time to this as it's off topic, but I hope the results are helpful.

Back on topic, I don't have a suitable inductance coil to measure the gain of V3 , so I will have to work it out from the data we have.

The DCR is 4k so on the plate chart, it's going to be a near vertical line.  The total current is Vk/Rk which is 2.23V/470 which is 4.745mA.  The screen current is (255V-110V )/130k=1.115mA.  So plate current is 4.745-1.115=3.63mA.

We can draw a horizontal line across the chart at 3.63mA and a 4k load-line from the B+ of 255V, the crossing point is the tube's working point.  The chart is not totally accurate as its for 100V screen current not 110V, but it will be near enough for our purposes.

The reactive inductance of a 400H coil is a massive 2.5M at 1kHz and the rp of the tube is also very high compared to the OPT primary impedance of 48k, so I just drew a loadline of 48k across the working point as shown on the chart below.

From this chart we can drop down verticals to the x axis at 0V and 4V on the loadline, these are the max/min before clipping.  From these points we can see that the choke can swing from 60 to 380V or a total swing of 320Vp-p.  The input for that swing is 4Vp-p so the gain is 320/4=80.  This does not have to be super accurate as it will be going into a feedback calculation with the gain of V2.

Now to test V2,