P2P Redd 47 - a few questions

[guavatone]

The zobel section should not concern you unless you have an A92.  So if you are using a cinemag, use their values or tune the components yourself.

 

[earthsled]

The Zobel circuit makes sense to me...
If using an A92, then C13 = 100pF and R1 = 100K
If using a CMMI-7C, then C13 = 20pF and R1 = 30.1K (according to the spec sheet)

Speaking of... what's the purpose of 100K resistor in parallel with the Zobel circuit in the P2P layout? Is it needed?

 

[abbey road d enfer]

Why not a good old log pot? And there is certainly a possibility to reduce gain by increasing NFB; this has not been implemented by the EMI engineers probably because there was no need for less than 34dB gain at the time, but today I find myself quite often using only 25-30dB gain on very close-mic'd loud sources.

A log pot is a fine idea. Would you suggest adding one between the input transformer and the grid of V1? What value would be best?

This would require some experimentation, since th epot would load the secondary of the IT, but I think 220k would be a good starting point. Resistor R1 would probably need to be increased to about 180k.

From what I understand, NFB (net feedback) is controlled by the 3-position gain switch. Are you suggesting that the switch could be modified to have additional steps for lower gain settings? Is there any disadvantage to this approach?

The FB network loads directly the output of the second stage; at the lowest gain setting, this load is already quite significant, so it is not possible to reduce it anymore. It would take increasing R5 significantly. I would altogether remove R5 and use a real 5k anti-log pot for VR1. The gain variation there would be larger, with a smoother taper than a Lin.
http://www.tubeampdoctor.com/de/shop_Potentiometer_Potentiometer_mit_Loetoesen/C5k_Reversed_LOG_ALPHA_Potentiometer_731

 

[earthsled]

This would require some experimentation, since th epot would load the secondary of the IT, but I think 220k would be a good starting point. Resistor R1 would probably need to be increased to about 180k.

Out of curiosity, would there a better location for the log pot? Perhaps at the grid of V2?

I would altogether remove R5 and use a real 5k anti-log pot for VR1. The gain variation there would be larger, with a smoother taper than a Lin.

Using this config, how much larger would the gain variation be? Would this mod alone be enough to achieve the desired gain reduction?

For folks in the US, Mouser has rev-log pots too: 313-1560F-5K


Taking a step back here, if one wanted to add an input attenuator - before the input transformer, what would the ideal input and output impedances be for the pad? Something like 1.2K input and 120R output?

 

[abbey road d enfer]

This would require some experimentation, since th epot would load the secondary of the IT, but I think 220k would be a good starting point. Resistor R1 would probably need to be increased to about 180k.

Out of curiosity, would there a better location for the log pot? Perhaps at the grid of V2?

The problem is that V2 is inside the open-loop, so you would reduce the open-loop gain and the NFB would try to force it. The end result would be a restricted gain control range with  a variable THD and frequency response. Not my idea of a dependable unit. I think the input pot or attenuator at the secondary is the most usable option.

I would altogether remove R5 and use a real 5k anti-log pot for VR1. The gain variation there would be larger, with a smoother taper than a Lin.

Using this config, how much larger would the gain variation be? Would this mod alone be enough to achieve the desired gain reduction?

I would have to sim the entire circuit to give an accurate answer, but I think that would give an additional 6-10dB range. The actual answer relies on the limits of stability (too much NFB induces oscillations), which is very difficult to sim accurately because it is based on ultra-accurate characterisation of the transformers and tubes.

Taking a step back here, if one wanted to add an input attenuator - before the input transformer, what would the ideal input and output impedances be for the pad? Something like 1.2K input and 120R output?

IMO 1.2k is too low for most dynamic and ribbon mics; I would choose 2k/200R.

 

[Kamel]

I've made a P2P REDD47, maybe 3 years ago, and i tryed to change feedback of NFB to increase Gain, and also change the R5 pot, the only gain i won is to return to the original schematic, there is not enough gain win changing NFB, maybe around 3-4 dB, easily available by changing the output transformer from 7:1 to 5:1, with a 1k max load everything is ok, for the R5 you win only 2 or 3 dB, as it suppose to match the gain of 2 units, with loss of distorsion... everything said come from my memory, but when i tryed to toying with it, i've seen that REDD engineers use the circuit as it maximum possibility, that explain certainly why this pre is so "magic"

my 0.02 cents

 

[abbey road d enfer]

The original question is how "to have additional steps for lower gain settings". It is already clear that increasing gain is at the detriment of BW and THD. Reducing the gain by increasing NFB is another challenge, since it involves rescaling some component and testing the limits of stability margin.

 

[guavatone]

I did the work for the NFB with 12 steps.  the lowest 2 are farther apart.  Feel free to "steel" my switch.  I didn't notice any difference.  The lower settings are for loud line levels or DI.

http://www.groupdiy.com/index.php?topic=33657.0

 

[earthsled]

Thanks for offering this! 12-steps is great!

It looks like you've reduced the value of VR1 from 5K to 500R. Is this to accommodate the stepped-gain control mods? What is the range of your gain switch? You say the lowest 2 setting are further apart, how many dB do you have for each step?

I've attached guavatone's switch schematic for discussion.

 

[abbey road d enfer]

All positions below 8 suffer of loading the output stage in a way that reduces headroom as much as it reduces gain. As a consequence, these positions do not offer higher input admittance, which is the essence of reducing gain.

 

[earthsled]

I suppose each method of reducing gain will have compromises.
What are the drawbacks of using an attenuator before the input transformer?
Attached is one example...

 

[abbey road d enfer]

I suppose each method of reducing gain will have compromises.

Using sensitive NFB is the one with less compromises, but has its limits (closed-loop gain can't be less than unity).

What are the drawbacks of using an attenuator before the input transformer?
Attached is one example...

Look at the -6, -12 and -18 positions: the input impedance is 400r, 700r and 1k5; proof once again that it's almost impossible to design an input pad of less than 20dB attenuation without sacrifying something.
On a more general POV, an input attenuator will always increase the Noise Factor. Which may or may not be an issue.

 

[earthsled]

What about this idea --

1. Modify the original gain switch (46, 40, 34dB) to have 3dB steps (46, 43, 40, 37, 34dB).
Then, add one additional setting for 31dB, giving a total of 6 steps (46, 43, 40, 37, 34, 31dB)

2. Add a switchable input pad of -18dB -- effectively reducing the gain settings to (28, 25, 22, 19, 16, 13dB).
Resulting in a continuation of the first 6 steps.

3. Preserve the fine-gain adjustment for stereo applications.
(Adjusting the values of VR1, R5, R6 if necessary)


This seems like a decent compromise allowing for 12 steps of gain control versus the original 3.

...or perhaps a 7-position gain switch w/ a 21dB input pad?

 

[letterbeacon]

Here's a high quality version of the minimum specs of the REDD 47, if it's any use to you, Earthsled...

 

[earthsled]

Nice! Where'd you find it? Looks like I got it pretty close.

Does R2 really need to be 2W?  ???

 

[letterbeacon]

I think I found it in the Drip Electronics documentation from his forum.  They're worth a read if you're planning on building the pre amp.

 

[abbey road d enfer]

What about this idea --

1. Modify the original gain switch (46, 40, 34dB) to have 3dB steps (46, 43, 40, 37, 34dB).
Then, add one additional setting for 31dB, giving a total of 6 steps (46, 43, 40, 37, 34, 31dB)

In order to do that, you need to alter the value of R5. I don't think you can do that. You have to think in terms of gain ranges, meaning you take into account the gain trim, so you end up with something like step 1: 31-37dB, step 2 34-40dB, step 3: 37-43dB, step 4: 40-46dB 

2. Add a switchable input pad of -18dB -- effectively reducing the gain settings to (28, 25, 22, 19, 16, 13dB).
Resulting in a continuation of the first 6 steps.

That is a good concept.

 

[earthsled]

The EMI documents state the range of the fine gain setting (or "pre-set gain control") is only about 1.2 to 2.6dB. So, I'm thinking this would only be useful if your were trying to calibrate the outputs of two channels so they match up. The range isn't much, but it would be enough to fill in the gaps if the main gain control could be altered to have 3dB steps (rather than 6dB).

Couldn't the value(s) of R15 be lowered slightly to reduce the "34dB" position to 31dB? Then, with the lowest value set, couldn't one calculate 3dB steps until they add up to the original highest position of 46dB?

If not, then sacrificing the fine gain control wouldn't be much of a loss.

 

[earthsled]

It just occurred to me that these modules must have had faders controlling their outputs in the desk. Why not just put a log pot on the output? Or a combination of this an the 18dB input pad?

 

[abbey road d enfer]

The EMI documents state the range of the fine gain setting (or "pre-set gain control") is only about 1.2 to 2.6dB. So, I'm thinking this would only be useful if your were trying to calibrate the outputs of two channels so they match up. The range isn't much, but it would be enough to fill in the gaps if the main gain control could be altered to have 3dB steps (rather than 6dB).

Couldn't the value(s) of R15 be lowered slightly to reduce the "34dB" position to 31dB?

Not really. The actual value is the minimum possible that does not affect headroom (in fact it does, but not exceedingly). Reducing R15 to 4.7k would reduce gain by 3dB and reduce headroom by 3dB also, which would not give any increase in input headroom.

Then, with the lowest value set, couldn't one calculate 3dB steps until they add up to the original highest position of 46dB?

If not, then sacrificing the fine gain control wouldn't be much of a loss.

IMO, the fine trim is not a necessity. I certainly would increase the range of fine control.