Reverb Noise

[Daver]

Enzo,
I already tried a 12" guitar speaker and got a clean signal. Are you suggesting using something like a small radio speaker?
The reverb sounds fine unless the Volume or Dwell are near full. Remember this amp has a master volume. I realize that using both near wide open won't happen very often, but I'm also concerned that if a boost or heavy overdrive is used, the crappy noise will show up.
Science is fun. Just look at that freaky waveform. Looks cool, sounds bad. Thanks!
Dave

[jmaf]

Dave, this may not be the only problem, but as you and Enzo mentioned this is science and you have -14V on the grid relative to the cathode, you are running in maximum distortion zone.

The blackfaces run that anode at 430+ Volts, so take a look at the attached curves, your BIAS should be from -6 VDC to -7 VDC maximum. At -14 VDC BIAS it's 100% pure distortion zone, that BIAS is not even on the charts.

This might not explain the resistive load versus inductive load kickback thing, but it is likely part of the problem.

By more than halving the cathode resistor, you will halve the voltage drop across it, making that -14 V into somewhere near -7V, take a look at the curves for -7 V BIAS at 400+ VDC anode. You'll have increased current, so you may wish to add an 1 W anode resistor of 47K to compensate somewhat.

[tubeswell]

The blackfaces run that anode at 430+ Volts, so take a look at the attached curves, your BIAS should be from -6 VDC to -7 VDC maximum. At -14 VDC BIAS it's 100% pure distortion zone, that BIAS is not even on the charts.

Hi jmaf

I am curious about this because I am similarly fiddling with a reverb circuit and I can't find much information on biasing a 12AT7 for a reactive load, so your post was helpful thanks.

At the moment I have a cascaded 12AT7 with the 1st half biased with a 1k1 cathode resistor and a 47k anode resistor with an HT of 330 (plate voltage of 200, cathode voltage 2.7) and this cascades into the 2nd half via a 22nf coupling cap and a 1M grid load resistor. The 2nd half is driving a BF RT (15k load) with a plate voltage of 436. I tried a range of cathode resistors between 1k5 and 4k7 and got a range of voltages between 6.8 and 8.6 respectively - it is cleaner with higher cathode resistors - as you would expect. I understand that I need to get the RT working as hard as possible for maximum effect, whilst preserving a cleanish pan input signal, so I want to achieve the optimum bias point for the driver stage, but I got not idea what the tube characteristics look like, because I'm still not clear on how you draw a load line for a reactive load for a triode. (I know how to do a load line for a resistive load, and I think I know how to do one for a reactive load for a pentode - from Merlin's on-line article). Are you able to enlighten me on this? Cheers

[jmaf]

Hi jmaf

I am curious about this because I am similarly fiddling with a reverb circuit and I can't find much information on biasing a 12AT7 for a reactive load, so your post was helpful thanks.

At the moment I have a cascaded 12AT7 with the 1st half biased with a 1k1 cathode resistor and a 47k anode resistor with an HT of 330 (plate voltage of 200, cathode voltage 2.7) and this cascades into the 2nd half via a 22nf coupling cap and a 1M grid load resistor. The 2nd half is driving a BF RT (15k load) with a plate voltage of 436. I tried a range of cathode resistors between 1k5 and 4k7 and got a range of voltages between 6.8 and 8.6 respectively - it is cleaner with higher cathode resistors - as you would expect, but I got not idea what the tube characteristics look like, because I'm still not clear on how you draw a load line for a reactive load for a triode. (I know how to do a load line for a resistive load, and I think I know how to do one for a reactive load for a pentode - from Merlin's on-line article). Are you able to enlighten me on this? Cheers

tubeswell, I use an approximation to simplify things, just so I know if I'm on a linear region or not, approximately. Then I'll run the rest of the tests on the bench, using the ear-o-scope (and on fridays, a beer).

I don't know if the following is standard practice, reading other senior member's engineering foo I can only say this will seem like a Mayan method from 3000 years ago. Here's what I do: I draw 2 load lines, one for 80 Hz and one for 1400 Hz at the reflected impedance.

For my "load" I use : Rk + Ra + coilDCresistence + (6.28 * frequency * reflectedL)

Were reflectedL is derived from the usual output transformer reflected load rules. In your case, the squared turns ratio times 15000(right?). Also, you'd be looking at the far right part of the chart, very tight area to work on at that anode voltage as you can see, the linear region there is very short, your load lines may be off the chart really, just try to keep the ruler parallel to them for the frequencies you're examining.

There I just move a ruler in parallel from line 1 to line 2, remembering the higher line is for lower frequencies and the lower one for the high end. If my high end is too far off the scale, I know I'm messing that up, same for the higher line(low end).

Dunno if this is formally right though. Is that somehow near what you asked?

[tubeswell]

Were reflectedL is derived from the usual output transformer reflected load rules. In your case, the squared turns ratio times 15000(right?).

Okay thanks for that. W.r.t. the above I'm not quite following you - my understanding (limited as it is) is that the reflected load is 15k/8R which is 1875:1 impedance ratio. The turns ratio is the square root of this, so that is 43.3:1. (Pr:Sec) (But I don't understand why the turns ratio would be relevant in your equation)

[Daver]

jmaf,
I can lower the cathode resistor all the way down to 500 ohms and still only get my grid to cathode voltage down to maybe 10 volts. With this change alone, the wavefrom is deformed on the bottom half and very pointy. If I add in the 47K resistor between the plates and transformer, the bottom half cleans up nicely. The rest of the waveform looks good but the top half has the peaks cut off. I've gone up to 100K and down to 22K and I can't get rid of the top half clip. I then tried changing the cathode resistor from 500 ohms up to 4.7K, but that changes nothing.
It's encouraging to see some positive change here. If I can juggle the values to get a cleaner waveform I might be OK. Thanks for the suggestion. I'll report back.
Dave

[jmaf]

Okay thanks for that. W.r.t. the above I'm not quite following you - my understanding (limited as it is) is that the reflected load is 15k/8R which is 1875:1 impedance ratio. The turns ratio is the square root of this, so that is 43.3:1. (Pr:Sec) (But I don't understand why the turns ratio would be relevant in your equation)

(With apologies to Dave in advance for detouring this thread a bit, I hope this illustrates why I think -14V DC BIAS is causing at least part of his distortion problem.)

That, my friend, was my rushed attempt at an explanation. If you have the impedances then you don't need the turns ratio. I rushed to reply and for some reason I computed the 15k as your secondary load without thinking of what it meant(obviously not an earthly reverb tank). (In that alien case you'd multiply 15K by the square of the turns ratio to find the primary load. )

Another mistake, from my earlier explanation was to throw your cathode resistor in there in my formula, which was wrong too because it skews our anode voltage up in relation to the cathode. Now it's all quiet in here, so I can attempt to make some sense.

Summary: Basically, I find the total impedance of that load, reactance which depends on frequency + the resistive part(series resistor + dc resistance of the coil). From that, I draw the loadlines for the 2 frequencies as if it was just a load resistor.

(If you have the loadline, you'll know the plate resistance and then you can calculate the current and voltage drop with a cathode resistor on. )

Note I used reflectedL in the formula, not Z. L = the reflected inductance to calculate the impedance (reactance + resistive load) for 80 Hz and then for 1400 Hz.

Both the 15K and 8 OHMs that are given to us by datasheets are reactance as measured from a 1 KHz signal. We need the reactance for 80 and 1400 hz, our boundary frequencies, not for 1khz. Remember we chose 80 and 1400 hz for my theoretic guitar, you can choose any 2 bounds.

Again, I don't know if this is formally correct, this is how I do it based on AC to DC approximations just to find out if my tube is way off or working in or near a linear region.

So how do we find the impedance for 80 Hz and 1400 Hz to drive an 8 OHM load if you already know the primary(15K)?

We divide 15K by 1000 and multiply by 80, then by 1400.

Z @ 80Hz = 1K2 OHM
Z @ 1400Hz = 21K OHM

Now we know our reactive loads for those 2 frequencies.

To find our approx. load impedance in each frequency:
RL @ 80Hz = 47K plate resistor + 1K2 reactance = 48.2 K
RL @ 1400Hz = 47K plate resistor + 21K reactance = 68K

(Add your DC coil resistance in there, it's all part of the load and will determine our current.)

So now you can draw those 2 load lines for those 2 frequencies and know if you're in a linear region both for 80 Hz and for 1400 Hz, bottom and high end. Anything in between can be determined by moving the ruler linearly from line 1 to line 2 keeping it parallel to both.

I hope this version make any more sense.

[jmaf]

jmaf,
I can lower the cathode resistor all the way down to 500 ohms and still only get my grid to cathode voltage down to maybe 10 volts. With this change alone, the wavefrom is deformed on the bottom half and very pointy. If I add in the 47K resistor between the plates and transformer, the bottom half cleans up nicely. The rest of the waveform looks good but the top half has the peaks cut off. I've gone up to 100K and down to 22K and I can't get rid of the top half clip. I then tried changing the cathode resistor from 500 ohms up to 4.7K, but that changes nothing.
It's encouraging to see some positive change here. If I can juggle the values to get a cleaner waveform I might be OK. Thanks for the suggestion. I'll report back.
Dave

Nice to hear Dave. You might try changing the grid leak too, lower values will lower the BIAS. Good luck.

[tubeswell]

Summary: Basically, I find the total impedance of that load, reactance which depends on frequency + the resistive part(series resistor + dc resistance of the coil). From that, I draw the loadlines for the 2 frequencies as if it was just a load resistor.

(If you have the loadline, you'll know the plate resistance and then you can calculate the current and voltage drop with a cathode resistor on. )

Note I used reflectedL in the formula, not Z. L = the reflected inductance to calculate the impedance (reactance + resistive load) for 80 Hz and then for 1400 Hz.

Both the 15K and 8 OHMs that are given to us by datasheets are reactance as measured from a 1 KHz signal. We need the reactance for 80 and 1400 hz, our boundary frequencies, not for 1khz. Remember we chose 80 and 1400 hz for my theoretic guitar, you can choose any 2 bounds.

Again, I don't know if this is formally correct, this is how I do it based on AC to DC approximations just to find out if my tube is way off or working in or near a linear region.

So how do we find the impedance for 80 Hz and 1400 Hz to drive an 8 OHM load if you already know the primary(15K)?

We divide 15K by 1000 and multiply by 80, then by 1400.

Z @ 80Hz = 1K2 OHM
Z @ 1400Hz = 21K OHM

Now we know our reactive loads for those 2 frequencies.

To find our approx. load impedance in each frequency:
RL @ 80Hz = 47K plate resistor + 1K2 reactance = 48.2 K
RL @ 1400Hz = 47K plate resistor + 21K reactance = 68K

(Add your DC coil resistance in there, it's all part of the load and will determine our current.)

So now you can draw those 2 load lines for those 2 frequencies and know if you're in a linear region both for 80 Hz and for 1400 Hz, bottom and high end. Anything in between can be determined by moving the ruler linearly from line 1 to line 2 keeping it parallel to both.

I hope this version make any more sense.

Thanks jmaf

You might've missunderstood my original description of my circuit (schematic attached) - my fault probably for making it poorly worded. I think this because I can't quite figure out why you factor in a 47k plate resistor. If I am correct in my guess in this regard, and if I take that out of the equation, does that mean the two boundary load lines are 1k2 and 21k (assuming those specific 'boundary' frequencies)?

(Or am I missing the point completely, and is there some reason to factor in a 47k value in the load line function for the RT driver stage, which I as yet can't understand)?

Cheers

[dai h.]

what's the other half of the 12AT7 doing?

[Daver]

Dai,
Apologies for my poor drawing. Both sections are paralleled like a regular old Fender.
Dave

[JoeM]

Dave,

Just wondering, when you got the 'noise' above on the scope trace, did you have the reverb return plugged in too? Does it change if unplugged?

[pdf64]

I worked on a buddy's late 70s pull boost TR with this problem a few years ago. I can't find my note book from then right now, but from memory, things that fixed it were:-
1/ grid stoppers on the tube sockets - 10-22k for each pre-amp grid, move the 68ks at the input sockets on to the tube socket terminal also. Is that 47k on the 12AT7 grid on the tube socket?
2/ seperating the ground return for the 12AT7 cathode from anything else and moving it over to where the ground return for that node was soldered on to the brass plate.
3/ removing the 560pF cap across the reverb transformer primary, and increasing the reverb driver cathode resistor from the SF 470 ohm to 2k2 - not bypassed.
4/ tidying up the lead dress, which is a major problem with the pull boost circuit.
5/ I remember bypassing the reverb driver power supply node cap with a 0.1uF ceramic, but not whether that had any benefit.
I think that the cause of the problem was the inductive flyback when the reverb driver was overdriven, creating a momentary parasitic oscillation/positive feedback via ground return.
Think of the cable from reverb transformer to the 12AT7 plates as a transmitting aerial, and keep it tidy, away from signal wiring, tucked against the chassis.
Also bear in mind that the reverb driver isn't supposed to get overdriven - accutronics advise against it. With good reason, it sounds bad (but not as horrid as the thing you're getting!). Your dwell control will help avoid this, as will an unbypassed cathode on the driver.
IMO there is far greater benefit in good lead dress and use of grid stoppers, than using screened cable for grid wiring. Peter.

[Daver]

Joe,
It makes no difference if the return is plugged in or not.

Peter,
The 47K grid stopper is right at the socket being fed by sheilded cable.

The grounds for the driver stage are star grounded to the filter cap negative for that stage and then a wire run to chassis ground. The amp is quiet overall.

Changing the cathode resistor up or down from 2.2K by itself hasn't changed much. Removing the cathode bypass cap didn't make much difference.

The wires from the reverb transformer are very short and run flat against the chassis.

There isn't any distortion coming from the second stage. Yes, lowering the Dwell helps.

OK, I'm an idiot. When I added the 47K plate resistor I was watching the signal at the grid, not the output of the reverb transformer. D'oh! I've been probing too many places and none of them fun. So, no, the added plate resistor doesn't help. The waveform is less spikey, somewhat smaller but still distorted. Crap. Back to the drawing board.

Thanks!

Dave

[dai h.]

Dai,
Apologies for my poor drawing. Both sections are paralleled like a regular old Fender.
Dave

okay thanks. Was wondering if it had something to do with the other section, but apparently not.

what type of cap is the 500pF? Ceramic? Class 2 ceramic can change in value with signal. Can also be microphonic. (Maybe totally unrelated but just putting it out there just in case.)