Design parameters for feedback

[ayan]

Hi Teo,

Assuming that the taps are rated correctly, I can't see how power scaling would change anything. The 50 watt amp has less gain than the 100 and therefore the db drop would scale as well... no??

{snip}

The actual feedback ratio is what counts and where it is tapped from. Not amp total output.

On the other hand, I am just a player and I could be really wrong. I just don't think so.

But maybe I am.... Sheesh this is confusing..... LOL

I have to confess that I hadn't given this any thought until I heard from you, caninus auricularis. This is what I think: the negative feedback loop takes a sample of the output signal as seen at the OT secondary, scales it down and feeds it back to the phase inverter. I need to make a couple of assumptions here, which I believe are reasonable:

1. For all practical purporses, the signal generated at the phase inverter remains unaltered -- if no feedback is applied to it -- regarldess of the power being dissipated by the power section.

2. Assume that the OT has perfectly sized taps such that going from using one pair of tubes to two pairs of tubes can be compensated for by using another tap of the OT, such that the power transfer to the speaker will be optimum for a given load rating (call it 4 ohms, since that's where Dumble taps the NFB loop from).

Now, if you call V1 the voltage seen at the 4 ohm tap when two tubes (50W) are being used, the power dissipated is (V1)^2/R. And we said that R is the load impedance.

If we now turn on two more tubes and change OT taps, the new voltage appearing at the OT secondary is V2 (100W case). The power now dissipated by the load is (V2)^2/R, and R is still the same load impedance.

It's pretty easy to see that V1 and V2 are not the same since:

(V1)^2/R = (1/2) (V2) ^2/R

It follows that V2 = 1.4 V1...

The above, coupled with the assumption that the PI is not loaded by the power section (#1 above), means that if I need to feed back the same "amplitude" voltage in the 100W case as we did in the 50W case, the scaling factor will have to be 1.4 times smaller, since the signal that appears at the secondary of the OT is 1.4 times bigger than in the previous case. Conversly, the scaling factor for a 50W amp needs to be 1.4 times bigger than in the 100W case.

Since the NFB loop is a voltage divider with the FB resistor being much larger than the shunt resistor (390 ohms in Dumbles), then it follows that the feedback resistor for a 100W amp should be 1.4 times higher than that of a 50W amp. We know that value is 4.7K for the 100W amps, so for the 50W amps it should be 4.7/1.4 = 3.4K.

I believe this result correlates nicely with Teo's simulation findings.

Cheers,

Gil

[dogears]

Interesting..... I shall experiment with lowering the feedback resistor size in my 30 and 50. Dropping from 4.7k to 3.9k or less has a very noticeable different tone and feel. We shall see if I like it!

Thanks Gil and Teo!

Btw, I have Gary measuring the actual voltage at the 4 ohm tap with a signal inserted into the power amp in jack. For both a Glaswerks 50 and 100. This should yield the exact conclusive answer.

[ayan]

Btw, I have Gary measuring the actual voltage at the 4 ohm tap with a signal inserted into the power amp in jack. For both a Glaswerks 50 and 100. This should yield the exact conclusive answer.

For what it's worth, for any measurements to have some kind of meaning, both amps need to be identical up through the PI, and set up identically in terms of signal level present at the PI. And, their power sections need to be perfectly scaled, i.e., the 100W amp should be able to put out exactly twice the power the 50W puts out. But then, if this is true, you already know the answer and need not measure a thing, right?

The findings may or may not prove be conclusive, they seldom are in the real world. And, of course, there is no such thing as an exact answer... there is in physics problems in college, but not in actual day-to-day engineering -- that's why engineers came up with "factors of safety," better known as fudge factors.

* * *
Some historical data:

1. 100W Dumbles use a 4.7K feedback resistor, 390 shunt resistor, so the feedback voltage transfer function is 390/4700 = 8.3%.

2. 50W Dumbles used a 8.2K feedback resistor and 1K shunt resistor, so the feedback voltage transfer function is 1/8.2 = 12.2%....

With a little more math, their ratio is: 12.2/8.3 = 1.47, so there is that same old magic number again. What do you think of that?

Gil

[dogears]

WHoa! I never saw the feedback resistor as different in a 50 watt Dumble! I guess there it is! Neat how every schematic is wrong and most if not all "clones" are too.... If going for accuracy....

So, would you use the 3.4k or use the 8.2k with a 1k shunt?

Or just tap the feedback from the 8 ohm tap (easy solution)

Kind of explains why my 30 is gainy when I crank the master.....

Btw, I have Gary measuring the actual voltage at the 4 ohm tap with a signal inserted into the power amp in jack. For both a Glaswerks 50 and 100. This should yield the exact conclusive answer.

For what it's worth, for any measurements to have some kind of meaning, both amps need to be identical up through the PI, and set up identically in terms of signal level present at the PI. And, their power sections need to be perfectly scaled, i.e., the 100W amp should be able to put out exactly twice the power the 50W puts out. But then, if this is true, you already know the answer and need not measure a thing, right?

The findings may or may not prove be conclusive, they seldom are in the real world. And, of course, there is no such thing as an exact answer... there is in physics problems in college, but not in actual day-to-day engineering -- that's why engineers came up with "factors of safety," better known as fudge factors.

* * *
Some historical data:

1. 100W Dumbles use a 4.7K feedback resistor, 390 shunt resistor, so the feedback voltage transfer function is 390/4700 = 8.3%.

2. 50W Dumbles used a 8.2K feedback resistor and 1K shunt resistor, so the feedback voltage transfer function is 1/8.2 = 12.2%....

With a little more math, their ratio is: 12.2/8.3 = 1.47, so there is that same old magic number again. What do you think of that?

Gil

[ayan]

WHoa! I never saw the feedback resistor as different in a 50 watt Dumble! I guess there it is! Neat how every schematic is wrong and most if not all "clones" are too.... If going for accuracy....

So, would you use the 3.4k or use the 8.2k with a 1k shunt?

Or just tap the feedback from the 8 ohm tap (easy solution)

Kind of explains why my 30 is gainy when I crank the master.....

Scott, that was just one amp, I have no info about any other 50W (and the one I mentioned was an old one). You may want to wait to see what Gary has to say based on his test, or just go with whatever sounds right to you.

Gil

[llemtt]

...I can't see how power scaling would change anything...

My O'Connor style power scaling works by reducing supply voltage to the power amp (only power tubes or power tubes + PI) while properly tracking the bias voltage. Unfortunately this is far from perfect and while it scales power it also scales power amp gain, this is more noticeable at very low power settings like 1-2W. I think this is because at low plate voltage, less than 100v, tube mu is no longer fixed but it starts to decrease.

I am also quite new to the feedback design concepts so thanks a lot Gil, your reasoning makes things a lot clearer.

Best sources on this subject seems to me operational amplifier design and solid state hi-fi amp design, but sometimes it's really hard to read and translate concepts into the guitar tube power amp world

Cheers
Teo

[Structo]

Hehehe,
I was digging through some old threads to refresh some of my notes when this one popped up.
Scott, I was wondering if you ever came to a conclusion about the NFB on the 100w vs 50w amps?
With Gary doing the measurements?

I was going to do some experimenting with it in a day or two.

I know when I played with it before, I left the same tail resistor (390R) when I went back and forth between the 4K7 and 8K2 FB resistors.

This time I will try the 1K shunt resistor in addition with the 8K2 and see what it brings.

[David Root]

I picked up on this just now.

You know, some of the early 100W ODS amps used 8K2/1K, #13 for example, same as #4 which was a 50W. #40 was 50W and it used 820R/100R, like a Twin.

Ratios are all the same, and all three used 4 ohm OTs, so at least at that time HAD thought output power made no difference?

[thyx]

So if the voltages are all pretty close, why go from a 4.7k/390 to a 8.2k/1k? Perhaps voltage isn't the only issue?

[Structo]

I also have a schematic called Music Man RD 50 dated 12/2006 that has a 8K2/ 1K with a 1uF and 2K pot.

Not sure who designed the circuit as it isn't signed but it is a high plate 6L6 50 watt amp.

The D'Lite as designed with 6V6 called for a 6K2/ 390R 1uF 2K pot.

Then of course you have the Bluesmaster that shows 100K/ 4K75 .1uF 25K pot on ver 167 I have. But I believe that is a Marshall thing.

All of these taken from the 4 ohm tap.

As Bob said earlier it would be easy enough to sub in a pot to see how it the NFB affects the tone and or distortion.

[LPSGME]

I just tried the 8.2/1K on my 50 watt and it sounded slightly more compressed (and better) compared to the 4.7k/380 that was in there.

In the end I opted for a slightly lower values, which were 7.2/680.

That actually divides out close to the same ratio as 4.7k/380 - but to me sounded slightly better.

RE: the other kind of feedback
I can't get my amp to feedback which it did at one point - but then I couldn't leave well enough alone and changed things. I'm pretty sure I had lower plate resistors in V2 at the time (180k/150k) and not the 220k/180k that's in there now. Does anyone know if going lower again would help.

Shifting the plate voltages doesn't seem to help, as I have 3 switches that allow me to widely vary them; and the PI has a 5-6v difference. But no feedback.

Would perhaps using small value cathode bypass caps in V2 (like 1.5 and 5) help? - as I also increased them.

[Structo]

Did you perhaps mean 220K and 150K on the plate load resistors?

I currently have both V1 and V2 with the high plates and cathodes.

All 4.7uF

At one point I had a 1uF on CL2, I think I may revisit that.
Seemed to tame the bass a bit.

I also want to try 100K on V1 and leave V2 high, to see how it changes the clean and OD.

I have the LNFB on a switch which is usually off.
Also have a switch on the mid cap for .01 or .06uF.
But it stays in .01uF 99% of the time.
Mid boost is also nice for low to medium volume but I find it a bit much
when approaching gig level.
Something I didn't realize before was that on the HRM amp, you pretty much want PAB on when in OD. Otherwise you have two different tone stacks in circuit. I suppose that could be made into a good thing but most likely having a relay that switches PAB on with OD would be the ticket on the HRM.

What I find with these little tweak switches and stuff is that it allows me to dial in a decent tone at low volume levels so I can play and not bother the wife when she is home.
Kind of like a Loudness switch you used to see on the old stereos that compensated for low volume levels.

I have the bright cap on clean only relay setup and I like that a lot.
Thanks to Scott and Tony for that one.

So I spent about 8 hours yesterday reading the Dumble section from about 2006 forward.
This is in preparation for my 100 watt ODS build. Getting all my ducks in a row so to speak. I'm really seeing that the ODS 101 series is probably the best schematics for the amp I want to build.
That is a lot of material. It is interesting to see how the group pulled together and combined learned knowledge about these amps which allow mere commoners like me to build an amp that sounds pretty close to the real deal.

Can't wait to get started on my 100w ODS.
I think I will go with the Non HRM 100w.

[ChrisM]

So I spent about 8 hours yesterday reading the Dumble section from about 2006 forward.
This is in preparation for my 100 watt ODS build. Getting all my ducks in a row so to speak. I'm really seeing that the ODS 101 series is probably the best schematics for the amp I want to build.
That is a lot of material. It is interesting to see how the group pulled together and combined learned knowledge about these amps which allow mere commoners like me to build an amp that sounds pretty close to the real deal.

Can't wait to get started on my 100w ODS.
I think I will go with the Non HRM 100w.

This is why I love building these amps. There is so much work and research to do before the build. Come on, anyone can look at an old Fender layout and build from it!

[LPSGME]

Did you perhaps mean 220K and 150K on the plate load resistors?

I currently have both V1 and V2 with the high plates and cathodes.

All 4.7uF

At one point I had a 1uF on CL2, I think I may revisit that.
Seemed to tame the bass a bit.

I also want to try 100K on V1 and leave V2 high, to see how it changes the clean and OD.

I have the LNFB on a switch which is usually off..................

My amp started out as an HRM with V1 @ 220/180 and V2 @ 180/120.

Then I tried 100/100 on V1 and increased the V1 to V2 coupling cap to the non HRM larger .01 and increased the cathode bypass caps for some low end growl.

Still, I kept wanting to see if I could get more of that quick bloom and ramped attack envelope that Scott's and Brandon's amps seem to get -and I understood Brandon to say the demoed amps had high plate (resistors). I also read someone suggesting V2 @ 220/180. So I've had the resistors all over the place, but have to think they were lower when I was able to get feed back and more bloom.

I had the LNFB on a switch but recently removed it. But it was connected when the amp was able to feedback. I still have the V2a snubber on a switch though.

I guess I need to go back to the lower plates and smaller by pass caps and see what happens.

[Structo]

Yeah, one can certainly end up chasing their tail on the tweaked values.
Since the OD is dependent on what the clean stages are doing it can become a vicious circle.
But rewarding as hell when you get it right.