dialing in your NFB tone

[tubenit]

A recent post by PRR and DaGeezer inspired me to try "dialing"  in some negative feedback on the Carolina SongBird Plus.  So here is how I went about it ................

I decided to look up SE amps with NFB ............

This Garnet had a 150R/10k
http://www.el34world.com/charts/Schematics/files/garnet/garnet_g15tr_gnome.pdf

This Champ had a  1.5k/22k
http://www.el34world.com/charts/Schematics/files/fender/champ_5f1_schem.pdf

This Champ has a 47R/2.7k
http://www.el34world.com/charts/Schematics/files/fender/champ_aa764_schem.pdf

This Gibson has a 2.2k/47k
http://http://el34world.com/charts/Schematics/files/gibson/Gibson_GA-1RT-1.pdf

So I figured ............... not much of a standard here, so I'll decide by just listening.

I got my resistor substitution box and tried 47ohm thru 330ohm for the resistor to ground

AND used a 25k potentiometer to dial in everything from 2.2k to 22k for the resistor to the + speaker connection.

Well, alot of different combinations sounded good, BUT I found there really was a "sweet spot" where the harmonics sounded more musical.  It was 75 ohm & 6.8k.   I tried dozens of combinations and kept A/B'ing ones that I liked. And I compared the NFB to no NFB.

The point of all of this is ............ PRR and DaGeezer suggesting trying something  by ear and deciding for yourself. (They were referring to what speaker ohm to hook up)

Right on!    You may be surprized @ what your ears "recommend"

 stated by DaGeezer

Anyhow, I went with what sounds best to me & feel like it was signficantly better than the "standard" values I tried. The result was a little better definition of notes when chording, IMO. It lost a little blooming ability with the controls the same, but I could dial the blooming back just as well by increasing mids.

With respect, Tubenit

[John]

As always, thank you for sharing your work! And out of curiosity, when building a SE amp, is using NFB necessary, just a Very Good Idea (to keep down hum, etc.) or simply builders choice?

[jojokeo]

tubenit,
I've always undestood that the important thing was the ratio here?!

Your is a ratio of 75 & 6.8k = 90.7 (btw, you show 73r on the schem but write 75r?)
Garnet = 150/10k = 66.7
Early Champ 1.5k/22k = 14.7
Late Champ 47/2.7k = 57.5
Gibson 2.2k/47k = 21.4

So you like "a bit less" than the normal values which I think is cool. Too much can make an amp constipated, lifeless, and dull.

[Tone Junkie]

Cool good thinking I love when i get to learn somthing from you guys another thing to use my resister sub box for, might have to use this as an excuse to build another one with a 250k pot in it.
Thanks Bill

[tubenit]

I played the amp with the NFB and liked it quite a bit.  Then I plugged in the digital delay (which isn't a very warm effect) & had a chance to play it for about 45 min and decided I like it better without the NFB with the digital delay. So I'll probably reverse this mod back to what it was.  Just an FYI.

With respect, Tubenit

[jjasilli]

I've always undestood that the important thing was the ratio here?!  Voltage divider ratio is the tail end of the matter.  The prime importance is the amount of NFB signal voltage injected back into the system (and the frequency bandwidth of the NFB signal if filtered by a cap in the NFB loop).  That depends on the voltage we start with.

Power in watts = V²/R.  E.G.:

1.  20W amp >   8Ω;  20 = V² / 8;  V² = 160; √160 = 12.6V
2.  20W amp > 16Ω;  20 = V² / 16; V² = 320; √320 = 17.9V

Note that as the impedance changes by a factor of 2, the voltage changes by a factor of about 1.414.  Here we have a 5.3V difference to contend with in designing our NFB voltage divider.

The NFB voltage we start with is a critical factor.  Objectively, the NFB voltage divider needs to be some ratio that delivers a final NFB voltage within useful operating parameters.  Subjectively, within that range, we can fine tune a result that we like.

[jojokeo]

Thanks for the detailed info. but what kind of voltage do you expect to measure at the 8 ohm output jack to where you can expect to derive this amount of voltage, especially since it's always fluctuating and it's actually more current than voltage here since that's the OT's job to convert this to the speaker?
Or is this just the mathmatical process to go through and everything will simply be scaled down in a proportional manner to where it's applicable? Is the source of this info available to read about it further?

[jjasilli]

Yes, if there is no signal then there is no output voltage at the OT secondary, and no NFB voltage.  The Power formula is giving the voltage when the 20W amp is actually making 20W. If the vol is turned down so a 20W amp is making only 5W, then the output voltage is commensurately smaller (per the Power formula), and so is the NFB voltage.  But we don't figure NFB at -0- or low signal conditions.  NFB must do its job when when the amp is driven hard, so that's how NFB is derived.  Then as the amp vol is turn down, the NFB voltage automatically lowers in step with the output voltage to the speaker. 

(Aside:  Note that most mass manufacturers tend to tap NFB off the highest Ohm speaker tap to source their NFB voltage.)

[jojokeo]

Okay, next question is how do I calculate the resistors to use to get the 12.6v I that are needed? How do I know what voltage to use to make the divider so that I can get the 12.6v?

[catnine]

I don't know why this does not compute for me. I have a later SF champ . on one end of the 2.7k ohm NFB resister is in seriers with a 47 ohm to ground , the other end of the 2.7k NFB resister goes to the OT secondary to ground. I wired in a 1k linear pot in series with the 2.7k NFB resister .

 Since both ends of the 2.7k NFB resistor go to ground one end on the OT secondary and the other end through the 47 ohm to ground. So if I add 100k to the 2.7 k in series and this in parallel. I definitely hear a difference . Instead of 2.7k ohm parallel with 47 ohm I have then 102.7k ohm in parallel with 47 ohm correct. Original I have 46.2 ohm NFB  add the 100k to the 2.7k and I have what?

[HotBluePlates]

Tubenit: do you have a signal generator?

I was gonna post a dissection of your feedback circuit, but I've run into a couple of issues with assumptions, and need real numbers to sort it out.

In a nutshell, I was interested in:
1. How much bias there is on your output tube.
2. Injecting a test signal to get the peak voltage on the 6V6 grid to equal the bias voltage.
3. Measuring the signal at the 12AY7 plate, given typical settings of the volume and tone controls for the onset of distortion.
4. Or, just measuring the signal across the 73 ohm resistor, with the feedback loop open and closed (with and without feedback).

I was looking into how we could describe the effect of feedback on your circuit, and realized I didn't have a good way to estimate the amount of loss introduced by the volume and tone control between the 12AY7 and the 6V6. I figured measured numbers would be the best solution, hence the question about the signal generator.

[tubenit]

HBP,

Always nice to hear from you. I don't have a "signal generator". And I haven't measured bias on the output tube.

As of now, I probably will just put it back to no NFB since I liked that tone with the digital delay.

With respect, Tubenit

[Shrapnel]

HBP,

Always nice to hear from you. I don't have a "signal generator". And I haven't measured bias on the output tube.

As of now, I probably will just put it back to no NFB since I liked that tone with the digital delay.

With respect, Tubenit

You *could* always make it switchable. Switch it out when you use the delay, and switch it in when you don't. Best of both worlds.

[jjasilli]

Okay, next question is how do I calculate the resistors to use to get the 12.6v I that are needed? How do I know what voltage to use to make the divider so that I can get the 12.6v?

NO, that's the source voltage; not the NFB voltage after division.  12.6 volts is what the OT secondary is feeding to an 8Ω speaker, under the following conditions:  the speaker and NFB are connected to the 8Ω tap; volume is set so the amp is making 20W clean.  If the speaker is connected to the 8Ω, but the NFB is connected to the 16Ω tap then there 17.9V to start with. 

The final NFB voltage, fed back to the tube, after division is an open question.  Plagiarize from known amps to calculate source voltage, then final NFB voltage.  E.G.: a Champ is 5W > 3.3Ω.  Use the Power formula to calculate OT secondary voltage.  Look at the schematic to see the values of the NFB resistors. Use the voltage divider formula to calculate the final NFB voltage.  Use this procedure for known amps similar to the amp that you want to build or modify.

[jojokeo]

Okay, I think the ol' ear test is best but for grins lets see if I have this right.
a 5w Champ w/ 4 ohm speaker and tap:
5=V^2/4; V^2=20; V=4.5 volts
Using late model Champ w/ 47ohm/2k7 divider:
Vout=Vin*R2/(R1+R2)
Vout=4.5v*47/(47+2700)
Vout=.077V fed back into the Champ circuit.

[jjasilli]

YES! So for this SE amp at full, undistorted vol. we have .077V fed back into the tone recovery stage.  Now, what is the signal input & output -- with and without NFB -- of the tone recovery stage.  Because we're dealing with impedance, it depends on the frequency of the signal.  So pick some frequencies"  1000Hz common for hi-fi -- 400Hz useful for guitar amps.  Then measure.  "Love is a bitch"  Rolling Stones.  So is amp design! 

[HotBluePlates]

HBP,

Always nice to hear from you. I don't have a "signal generator". And I haven't measured bias on the output tube.

When I make it back down that way, I'll give you one of the (many) Hewlett Packard 200CD's I have laying around. It's way big for an audio signal generator, since it's all-tube instead of solid-state. It also has way more range of frequency than needed for audio testing. But it will help you verify some things in your amps, if you're ever so inclined.

[PRR]

> amount of voltage, especially since it's always fluctuating

In theory, he should use the Voltage Gain of the forward path. This is (nearly) consistent, whereas the instantaneous voltage does vary.

However voltage gain around power tube and OT is tedious to compute.

In many practical cases, you can just throw "0.5" as the gain from power tube grid to typical speaker load.

If you want to get tedious: yes, start with the expected output voltage. Then look at the expected voltage at power tube grid. Since we normally take NFB to the stage before, estimate that stage's gain.

This assumes direct path from driver to power tube. If you put volume or tone networks between, the problem becomes too large for the back of an envelope. Also there is intense interaction between tone settings and NFB. Such schemes must be tweaked by ear.

5F1 Champ, 6W out, 4 ohms, about 5Vrms or 7V peak. Cathode bias is 18V and normally the peak signal is equal to bias. So 18V in, 7V out, gain is 0.39.

NFB in 5F1 is injected to 12AX7 V1b. Normal values except no cathode cap. Assume gain is 25.

25*0.39= gain of 10.

22K and 1.5K NFB values is a loss of 16.

So this amp has "hardly any" NFB on nominal load.

However the "4 ohm" load is a speaker which will bump-up to maybe 25 ohms at bass resonance. 6V6 gain will also bump-up nearly as much. So at bass resonance the 6V6+OT gain is more like 2, driver gain still 25, overall gain is 50. NFB is still 16:1. The NFB _IS_ effective at bass resonance.

That's probably how many good guitar amps are. NFB is very slight at nominal load, but (because of rising gain) more effective at bass resonance. This keeps the cone from getting over-excited at bass resonance, less slap.

So it is VERY much about the speaker. Which is yet another reason to rough-in ballpark values, get the amp working, try different speakers for possible use, THEN re-consider the NFB values.