Reducing the Pre Amp gain- cleans the 321 sound

[AR_sound]

Hi lads,

Since I use high sensitivity loudspeakers (92dB),

I have always seen the volume level in my NAC 32.5 (72 boards) to be a problem - too high, and it was not comfortable to use the volume pot. At very limit range.

A while ago I replaced the volume pot (ALPS) with a stepped attenuator (12 positions, Grayhill Switch with HOLCO resistors and CADDOCK shunt Resistors).

(Pic of the Grayhill switch attached below)

Ok, this volume control worked very fine, it cleans the sound comparing to the conductive plastic ALPS, AND more ... now one could control the volume range as he wish by selecting the resistors range, and By changing only two resistors (the 20K ohm shunt) you could adjust the volume to be lower or higher.

So what is the problem with this?,

It is seems to be not the best arrangement when one is using very sensitive L.S. , In my case the resistor to ground was at the range of few tens to few hundreds ohms, and the shunt resistor was 25K ohm!,
This arrangement provide gain of down to -58 db at the lower vol. position,

At this low level, the signal is more prone to noise and I thought it will be useful to low the gain of the 321 board in order the make the vol. work a bit higher level. there are two benefits when doing this:

1- The higher signal to the 321 is less prone to noise (power supply noise, and induced noise since the tracks are not shielded wire).

2- The 321 will have lower distortion at lower gain.

Until now nothing is new, as most of us know it is easy to control the 321 gain by reducing the 12k feedback resistor,

So, what is new in all about?

You have to do it properly-

After I reduced the 12k feedback to 4.7K the sound was much cleaner .. But I noticed I slight roll off of the low frequencies,

When I looked at the 321 scheme the problem have been found:

321 scheme

In the feedback there is 1k ohm res. and the 47uF silmic cap , and.. in parallel there is 330 k ohm resistor,

The 1k and the 47uF cap. controls the feedback at audio frequencies,
while the 330k res. controls the DC feedback to the first stage.

So in order to maintain the same DC feedback to the first stage we need to reduce the 330k res. at the same level we reduce the 12k feedback resistor.

After I reduced the 330k to 130k (with 4.7k feedback) the bass came back to the tune and the sound was much cleaner and much better in every way,
The improvement is huge in my system,


Avi.

 

[Mus]

You would have acheived the same result (attenuation) by adding a 27K series and the 15K shunt resistor inside your 32.5 thusly:

This thread covers the finer details, I tried different values in my 32.5 and 27K/15K worked well.

Regards,

Mus

 

[AR_sound]

Hi Mus,

It is not the same result, since my post deals with a new way of attenuation by changing the feedback, without changing the sound of the 321 stage.

Lowering the 321 gain is the issue here and not simply the attenuation that can be done with some ways, like the way you have mentioned.

Regards,

Avi.

 

[Mus]

Avi,

I can see that your way is different, but I can't see any benefit, is the end result not the same?

Regards,

Mus

 

[AR_sound]

Hi Mus,

The benefit is that it is sounds much better!

By reducing the 321 gain the NAC sounds much better - more refined and the speed of the music increases tremendously, more details, tighter bass, clearer highs, at short, better in every way.

I have done a lot of changes in my pre (including 4 super-regs, splitting the grounds, ECT..)
But IMO this simple one brought the biggest benefit to the tune (The speed and the low resolution detail is amazing),

If you can decrease the gain just a bit (with insufficient l.s.) The improvement will be small, but if one have sufficient loudspeakers, he can gain a lot by lowering the gain of The nac to a grater level.

IMHO, the nac was designed by NAIM at the times when the dominant source was LINN lp-12 with low output moving coil. Nowadays when CD's are much in a use, there is no point at developing pre amp with a big gain like the nac range have.

You can try and see yourself; you only have to change two resistors, as mentioned above in my first post.

Cheers,

AVI.

 

[Mr Tibbs]

Hello AVI,

Very interesting.

Have you ever tried using a pair of buffer (non filtered) boards in place of the 321s?

Mr Tibbs

 

[teddy_pardo]

Avi,

By changing the resistors, in addition to changing the gain, you also change the feedback frequency curve (the R/C acts as a filter). It explains why changing just the feedback res changed the sound a lot, and changing the other reg 'fixed' the problem. I would assume thus, that the 'improvement' you got is rather a 'change' due to the frequency responce changes, but if you like it then it's perfect.

 

[AR_sound]

Hi Mr Tibbs,

Have you ever tried using a pair of buffer (non filtered) boards in place of the 321s?

No, I have not, since I don't want a passive pre + buffer,

I prefers an active pre, but with less gain.

(From your question, have you tried this ??).

Teddy,

IMHO only changing the 1k resistor or changing the 47uF cap will change the RC feedback filter response, as you change the 12k and 330k at the same rate.
The 12k defines the rate of the feedback, while the 1k+47uF defines the frequency curve, since the output of the transistor 'see' only the 1k+47uF in the output. Well, this will work as much as you will not dramatically reduce the 12k resistor say for 1k ohm.

Maybe there is very minor freq. response change due to the increased feedback, since the 47uF is more dominant now, But who care for this 'minor change' if the result in the tune is much better

Any way, in order to be sure if I am right concerning the freq. response of this stage, some measurement with spectrum analyzer will have to be done,
In such case, some minor correction with the value of the feedback cap can be done.

Don't forget that as we changing this cap for silmic for example, we also make some changing of the feedback response. But if the result sounds good who cares?

Regards,

AVI.

 

[AR_sound]

I would assume thus, that the 'improvement' you got is rather a 'change' due to the frequency responce changes, but if you like it then it's perfect.

The improvement is due to lowering the noise and distortion of the 321 stage,

active stage with a small gain sounds better then a stage with bigger gain!,
(As passive pre has the lowest noise figger for example).

Anyway, I trust My ears to tell the difference between 'response change' and other change in the tune

Cheers,

AVI.

 

[Mr Tibbs]

No, I have not, since I don't want a passive pre + buffer,

Why not?

Besides, surely a preamp that has an output buffer is no longer a passive preamp anyway? What is the advantage of using a buffer before the pot, when any decent source can easily drive a 20K (or even 10K) pot?

If you feel the need to keep the input buffer, then simply use another buffer after the pot, in place of the 321 stage.

The reason I suggest this is that I think the design of the (unfiltered) buffer stage is simpler and may offer some sonic advantage over the 321 stage. I haven't tried it though, so I'm just interested to know if anyone has and what results they got.

Mr Tibbs

 

[yairf]

IMO this is remarkable stuff and an excellent thinking: with today's high-output sources, no one needs a massive amount of gain in the line-stage (hence the popularity of passive line stages with active buffer). so instead of having to leave with a volume control that you can use to only fifth of the way, one can reduce the line stage gain and get a more usable volume control.

This line of thought can be taken one step further: instead of altering the volume with an attenuator (the traditional way), maybe the gain resistors can be selected and as a result, have an adjustable gain.

Can anybody think why this would NOT be an interesing idea?
Yair

 

[AR_sound]

I think the design of the (unfiltered) buffer stage is simpler and may offer some sonic advantage over the 321 stage.

Mr Tibbs

You made a Good point, But before placing the buffer in front of the NAP we have to be sure that it's (the 324) finel stage has similar impedance like the 321.

I may try this for some day (In order to do this I must revise all the resistors at the stapped attenuator- not a small work at all..

Regards,

Avi.

 

[teddy_pardo]

Yair,

I tried this some 30 years ago, the problem I had then was that if the moving contact of the pot is not perfect you get a momentary very high gain when you turn it (no feedback at all), also the behaviour is somehow different (linear/logarithmic)

(BTW, I'm getting my new C-core transformers on Monday, my invitation to come and listen is still open)

Teddy

 

[AR_sound]

Yair,

The problem I had then was that if the moving contact of the pot is not perfect you get a momentary very high gain when you turn it (no feedback at all), also the behaviour is somehow different (linear/logarithmic)

Teddy

Teddy,

You must use a Make before Break switch!
If you will not the result will be catastrofic to the loadspeakers

My four years trusty GRAYHILL switch have not saw any 'switching noises' till now, like some ELMA users reported.

Regards,

AVI.

 

[AR_sound]

Yair,

Also the behaviour is somehow different (linear/logarithmic)

One good thing about stapped attenuator is the fact you can control his logarithmic behavior as you wish by selecting the proper res. values!

cheers,

Avi.

 

[Mr Tibbs]

This line of thought can be taken one step further: instead of altering the volume with an attenuator (the traditional way), maybe the gain resistors can be selected and as a result, have an adjustable gain.

Yair, changing the feedback will change how the amp sounds, and too much will cause instability in any given design - not a good thing when all you want is a change in volume.

You made a Good point, But before placing the buffer in front of the NAP we have to be sure that it's (the 324) finel stage has similar impedance like the 321.

Avi, the input impedance of both will be so small as to make no difference.

The buffer stages don't use loop feedback, instead they use something called a series feedback pair - so no 47u cap needed. Anyway, I'm sure this would sound different to your modded gain stage. Will it sound better? Only you can tell us

Mr Tibbs

 

[Mr Tibbs]

Avi,

I'm not sure, are you using 729 TA buffers or the older non-TA type?

Mr Tibbs

 

[AR_sound]

Update:

Yesterday I have made some measurements in order to check if my revised 321 works ok.

I calculated the gain, and checked bandwidth using tone generator and scope.

The results are:

Gain of the pre (NA321 with 4.75k ohm feedback resistor) is +7.6db , totally flat from 15 Hz to 20 kHz, with a slight roll off begins at 14 kHz and roll off of - 0.3 db at 20 kHz.

The 729 TA filter is the cause of the slight roll off near 20 kHz,
when I bypassed the 729 the pre was totally flat to 20 kHz.

Changing the 130k res. back to the original value (330k)- gave the same results!

When I Tried lower feedback values, I found the first stage to become unstable and oscillate at values lower then 3.5k Ohm, while reducing the 330k to 30k Ohm resistor helped the stability only down to 3.3k Ohm.


Conclusions: It is possible to reduce the 321 gain to some level with good results, but increasing the feedback too much will make the stage not stable and Oscillate. Values down to about 5k Ohm (X 6 gain) will be safe.

IMO, The 330k ohm is there for keeping the dc feedback for the first stage, while lowering this value dosen't help much to the stability of the stage with increased feedback.

are you using 729 TA buffers or the older non-TA type?

Mr Tibbs,
-I use 729 TA. (324 after the 729 are out of the question for me since I will not be able to use stepped att. with shunt resistor without changing the impedance of the switch).

Avi.

 

[Mr Tibbs]

-I use 729 TA. (324 after the 729 are out of the question for me since I will not be able to use stepped att. with shunt resistor without changing the impedance of the switch).

Ah, I initially assumed (wrongly) you were using the old non-TA buffers - sorry for the confusing advice I gave earlier. I like the 729s and think they're much better than the old boards (or no boards), when working into a Naim power amp.

Conclusions: It is possible to reduce the 321 gain to some level with good results, but increasing the feedback too much will make the stage not stable and Oscillate. Values down to about 5k Ohm (X 6 gain) will be safe.

That's the main thing - so long as you know the stage will remain stable.

Cheers

Mr Tibbs

 

[Linnik]

I have done a lot of changes in my pre (including 4 super-regs, splitting the grounds, ECT..) But IMO this simple one brought the biggest benefit to the tune

AVI.

This is interesting idea, the FB mod, Avi! Lowering the gain.
Certainly high gain have been a problematic thing and my Linn Aktiv box even lifts the gain a bit higher.

First case EVER I see somebody telling he is improving the TUNE!!

Interesting. This is interesting IMO. Must start looking for the suitable resistors...

BTW there have been quite big range in the true uF values of the FB caps I have been tried through. Obvious is they have been changing the curve, too. I have known this. Guess I will now dig up the original Maroons from my stock and have them measured..

Thanks for the thought-fuel and a bit more understanding, Avi!

Oz