RCA BA21 build

[dmp]

I picked up some original transformers for a RCA BA21 preamp. I'll post some pictures along the way.
This is a 40 dB fixed gain preamp with 20dB of negative feedback, fed from a tertiary winding on the output transformer, using two 12AY7 tubes.
Input measures 1:10 (600:60k)
Output measures 5.33:1 (secondary ~17k:600) and 10:1 (tertiary)
The output inductance measured 66 H at 100Hz on Instek LCR-821
The input measured "OVER" which I haven't seen before.

I'll investigate running this with input/output pads and variations of the feedback / no feedback.

 

[emrr]

Those are the more common UTC outputs, there's an earlier set that look like a different manufacturer, no branding. The one set I recorded anything on were:

Extech meter at 120Hz

input
UTC version
600 ohm 32H 27kΩ
150 ohm 10H 7k5Ω
50k 2445-2537h

Silver RCA version
150:50k 6.47H 6k4 to 1792H

UTC output
1030R CT pri, 82H 13K5Ω
46R sec, 5H 8k8Ω
83R tert

In several I modified, I couldn't measure anything useful from R14/C4 presence into a 10K load, and a 600 ohm load isn't required here. I was able to make a +/-10dB from stock work out fine, though higher gain quickly reveals tube and PSU deficits, if present.

Have fun!

 

[dmp]

I simulated the R14/C4 feedback path and I think it is to slightly boost the low frequencies. This simulation is showing the frequency sweep feeding back to the cathode of V1, so it will be subtracted (i.e. subtracting more high frequencies boosts lows). It will only boost low frequencies by about 0.8 dB though. If I'm misunderstanding this, please let me know

Once it is operating, I can try out C1,C2,C3 going slightly larger, and assess any need for C4/R14. And what higher values of R13 can be used.

I was thinking of using a 600 ohm attenuator on the output of T2. Bad idea?

 

[dmp]

What do your measurements mean?
For instance here:

1030R CT pri, 82H 13K5Ω

I guess you measured 82H & 13K5Ohms on the Extech meter?
What is 1030R?
And are you measuring the primary with pin 2 & 7 and all other pins open, including the CT?

When I measure L and R on my Instek I get a higher R than I get with a multimeter. Not sure why.
The Instek measured 66H and 6kohm. I forget what the multimeter said but could check it again.

 

[emrr]

What do your measurements mean?
For instance here:


I guess you measured 82H & 13K5Ohms on the Extech meter?
What is 1030R?
And are you measuring the primary with pin 2 & 7 and all other pins open, including the CT?

When I measure L and R on my Instek I get a higher R than I get with a multimeter. Not sure why.
The Instek measured 66H and 6kohm. I forget what the multimeter said but could check it again.

1030 DC resistance with a center tap present.
13K5Ω at 120Hz. LCR meters are reading the loading at the frequency, unlike a DC meter.
No loading present anywhere.
Parallel setting for hi Z, series for lo Z, as instructions specify. Useful to see what the other setting says, but I don't record it.

I measure a slight low boost even without R14/C4 present. Mine are pretty flat between 16Hz and 16K, there's a resonant transformer dip of a dB around 25K then it comes up to flat again around 32K before rolling off.

 

[dmp]

Better measurements (at 120 Hz with Instek LCR-821)

T1 (input)
PRI: 79.8 DCR, 33H, 29 kohm
SEC: 5230 DCR, 2883 H, 2746 kohm

T2 (output)
CT PRI: 1032 DCR, 90H, 266kohm
SEC: 47 DCR, 2.22H, 12.8kohm
TER: 78.5 DCR

 

[dmp]

I sketched out a turret board layout using the RCA manual, following the original layout. Followed exactly, except for moving C5 onto the board (C5 was a metal can cap in the original). Nice way to see a well designed circuit & ground routing. Standing on the shoulders of giants.
I print it out, tape it to the 1/8" board, and drill the holes with the drill press. Press the turrets on the drill press.
Materials cost about $25

 

[dmp]

Planned this to go in a 1 unit rack case, so mounted tube sockets and the transformers on a 1.5" piece of aluminum angle.
Drilled with a step bit on the drill press. The square holes for the transformers were a pain - drilled out the corners and then filed.

 

[dmp]

Wired up following the RCA layout. Next step is install caps and test with bench power supply.
The BA21 pictures I've seen look like the on board connections are underneath, with maybe a second insulating board
I did the connections above with teflon tubing over the solid wire.
Hookup wire is also teflon insulation, 22AWG

 

[heatwalk]

Nice build! Interesting seeing this layout with no resistors sharing turrets since it isnt a prefab of two parallel rows of nodes. Clean right angle components help the noise in this case? More accustomed to fender eyelets and marshall turrets with shared connections and diagonals

 

[dmp]

I followed the RCA layout - probably could have condensed it and saved some turrets. But they knew what they were doing.
Putting a whole row just for hooking up felt wasteful, but it sure made the connections simple to do and it looks clean.
Biggest thing here for low noise is the ground layout and metal film resistors, but the tube noise will dominate.

 

[dmp]

Powered it up with a bench power supply at 250v. Working well. I started with C4 out of circuit and using 0.1uF for C1, C2, and C3.
The meter points in the design measure 0.898V for V1A, 0.818V for V1B, and 0.919V for V2. These taps in the cathode circuits should measure 1V for good tubes, according to the manual. Might be low because the B+ in the manual is 285v.

Measuring 36dB of gain (low compared to the 40 dB spec in the manual). The measured noise with the input termination 150ohm is about 3.5 mV. That is -47dBu, giving a EIN of -84dB, which is in line with the -82dB spec in the manual. Sounds good with a mic.

Frequency response initially had a 2dB bump at the high end. Added a 600 ohm termination to the OPT that showed no change. Put C4=0.022uF and the bump decreased. I don't have a 0.015uF on hand. Will play with the feedback more this weekend.

 

[dmp]

I looked at the high frequency bump on the scope to see how it behaved past 20kHz. Peaks at 36kHz with +8 dB, then falls off.
Tried a few other C4 values, 0.015uF up to 0.037uF, not much effect. I don't see how that Fdbk effects the high frequency anyway.

Next looked at the frequency sweeps of the input (600:60k) and output Tx on their own. The input rolls off the high end while the output is flat. So I think the high frequency boost is happening in the circuit somewhere, but I don't see where. Maybe the V2, T2 coupling. It doesn't sound bad the way it is. Thoughts? I haven't tried swapping the polarity of the T1 / T2 hookups either, another possibility, with a capacitance to ground.

Also, I haven't tried setting the input to 150:60k, which is the way the BA21 was shipped. Might try this as a last test.
Also curious what the fixed gain of this should be? I'm measuring 36dB with the input as 600:60k. The manual sounds to me like this should be 40dB with the input 150:60k. Adding another 6dB from the 150ohm transformer hookup would be 42 dB. (edited)

 

[dmp]

Tried reducing the feedback by putting a 250k pot between the tertiary winding and R13 || R14C4. This worked well, measured 56dB of gain with the pot at max. And it gave a nice level control. Sounded good throughout. Frequency sweep stayed similar with a little (0.5dB) drop at the low end. Putting the pot in place of R13 didn't work as well. Max gain 46dB.

 

[emrr]

We’re definitely seeing some differences. I’ll try to pull up my plots again. Memory says i replaced the entire feedback RC and used something like 2k-100k range for -10 to +10.

Response worth checking with all of one or other or both transformer windings reversed.

 

[dmp]

Trying 150ohm hookup. Seeing 42dB gain. I think this is corresponds to the spec in the manual.
With 150ohm the high end rolls off. I put the 150/600 on a switch and measured these both one after the other.
Next to try swapping the transformer side to ground. Surprising the pinout of the transformer would need a modification, as RCA was producing these according to the layout. But I also am seeing the phase flipped with it built to the schematic, which would be weird.

 

[dmp]

Flipping the secondary of the input transformer did not do much. Slight difference in the response for both 150 and 600.
Don't think flipping the output transformer would have any effect since it is push pull.

 

[emrr]

My plots are somewhere in the middle of your two.

 

[dmp]

Thanks Doug, I think it is some interaction with the circuit and transformers. I'm done with experiments on frequency response.

Experimenting with the feedback loop and doing some calculations.
First, noticed the pot has noise when turning since there is DC in the feedback loop, since it is connected to the cathode directly and the tertiary winding goes to ground. Increasing the feedback would start to change the bias of V1a. So a blocking cap for DC would help if making the feedback adjustable. Making it large (similar to V76) will put the knee low enough that the feedback can act throughout the audio range.

Basic gain analysis of the BA21 for the input wired 150ohm and the output 600ohm:
Input Tx +26dB
V1a (gain stage) +24dB
V2 (cathodyne PI) -0.5dB
V2 (push-pull) +26dB
Output Tx -14.5dB
Feedback -19dB
Total +42dB

(estimated with help from Merlin Blencowe's book on tube preamps)

The feedback is returned from the tertiary winding (10:1) through the R13||R14, R1+R2 voltage divider. For the schematic values, the voltage divider is ~0.3, multiplied by the tertiary 0.1, gives a feedback beta = 0.03. This is applied to the amplifier gain (gain after T1, before T2), which is ~48 dB (or an open loop gain of 10^(48/20) = 267). The dB of feedback is then 20*log(1+0.03*267) = -19dB for the schematic values.

Putting this formula in a spreadsheet and I can make a stepped gain switch to vary the feedback. A minimum R13=1.5k gives an overall gain of 32.5 dB, while a max R13=20M gives an overall gain of 60dB. A blocking cap of 47uF or greater looks ok. For the minimum R13, the knee is at 2Hz. I'll be giving up the frequency compensation of C4+R14.
Going to test these values next with C4=47uF and the switched R in place of R14 (and omit R13 )

 

[emrr]

I went switched with some size of series cap to eliminate clicks. Always faster for me to measure gain steps on the bench with a decade box. I’m sure my cap is larger than 47 and the lowest gain steps begin to have some subsonic boost, I find a note that says +2 @ 6Hz with 1k35 NFB R. 4K there’s no bump.

Do you just want my recipe? Maybe less fun.

I felt going more than about 10dB less on NFB got significantly noisy with AC filaments. The orig layout doesn't seem up to it.

I also noted C4/R14 are reverse of drawn according to layout.