First post here, and my first build of a tube amplifier. Hoping someone might steer me in the right direction. I built an amp from scratch based on the Dynaco ST-35 schematic, and it does work. But when I hook a scope up to the speaker output terminals (8 ohm dummy load attached) and feed in a test signal, I see some garbage on the positive peaks of the sign wave. This is the same for both channels. I discovered that when I started probing the circuit with my DMM to check voltages, the signal would clear up whenever I probed the areas of the circuit circled in red on the attached schematic. These areas are only on the EL84 tube without the feedback circuit. Could anyone educate me on what I'm seeing here and what I might do to correct it? Thanks.
Attachments
This is what some call partial or conditional oscillation. It makes high notes sound awful and not very good for tweeter. A reason, why one should not hook up an expensive set of speakers initially. When you probe around the data path, the loading and capacitance in the DMM can take it out of the conditional oscillation condition. A number of things could cause this. Start with post #2.
Last edited:
Thanks for the reply rayma. I am using Edcor transformers. The output transformers are CXPP25-8K/23%. I thought these were a pretty close match to the Dynaco. I started out with stock part values in the signal portion of the circuit, but have modified these slightly because I used a different power supply design. Attached are images of the amplifier and a schematic that shows the current part values. My B+ is 352V, B++ is 327V, plate voltage on the 4 EL84 is from 341V to 344V, EL84 cathode voltage is 14.4V. On the two 12DW7: pins 1 are 194V and 199V, pins 2(&6) are 130V and 125V, pins 8 are 0.9V and 0.94V. I'll have to check on what the voltages on pins 3 are and post later.
Attachments
Since your amp is significantly different from the Dyna, you'll have to determine empirically
how to stabilize it. A series RC network across the first plate resistor can help, though you
may have to remove the positive feedback 150k resistor, and also the nfb 27pF capacitor, first.
Try something like 20k in series with 1nF to start with.
Last edited:
Before adding an additional compensation network, I'd try the following:
Change the value of the cathode resistor on pin 3 of the 12DW7 from 27K to the Dyna specified value of 33K. Use at least a 1 Watt resistor. This will improve the AC balance, and additionally it has an effect on the amount of positive feedback, which could contribute to your issue.
I'd add 1Kohm "grid stopper" resistors to the control grids of each output tube. This is accomplished by placing the resistor between the junction of each 470K grid resistor and .1mfd/630V coupling capacitor, and pin 2 (control grid) of each output tube, in place of the present direct connection.
If this doesn't help or resolve the issue, then I'd look at the network that Rayma mentioned.
Change the value of the cathode resistor on pin 3 of the 12DW7 from 27K to the Dyna specified value of 33K. Use at least a 1 Watt resistor. This will improve the AC balance, and additionally it has an effect on the amount of positive feedback, which could contribute to your issue.
I'd add 1Kohm "grid stopper" resistors to the control grids of each output tube. This is accomplished by placing the resistor between the junction of each 470K grid resistor and .1mfd/630V coupling capacitor, and pin 2 (control grid) of each output tube, in place of the present direct connection.
If this doesn't help or resolve the issue, then I'd look at the network that Rayma mentioned.
Grid stoppers and Zobel networks are the first things for sure. Dynacos were pared to the bone and even a few very inexpensive parts were removed if possible (possible in their layout and their transformers). The common cathode resistor was another cost saving measure that really shouldn't be continued for a DIY modern design. These are wonderful amplifiers, even nowdays, but weren't delivered from the mountaintop on stone tablets.
All good fortune,
Chris
All good fortune,
Chris
Once you get the OPT's issues sorted out this site has great info to improve that amp. See Dave's Lab for his EFB mod.
Dave's Lab
Dave's Lab
Sorry to join late.
As this is NOT a dynaco, Not even using dyna iron you are on our own track. Not necessarily bad or wrong but your own. Said that, to some suggestions:
- remove the positive feedback between the triodes. It's very dynaco specific and will create problems if not perfect. Then make both end of the concertina 33k for symmetry. If you need more sensitivity, increase the feedback resistor.
- add screen and grid stoppers on the el84 ( 1k grid, 100ohm screen). Mount these
direct on the tubesocket.
-using a scope and low level square input, adjust the feedback cap for best square on a resistive loaded output. -Double check the amp with some capacitive load (1uF in parallel with 8ohm) Adjust feedback components until good.
As this is NOT a dynaco, Not even using dyna iron you are on our own track. Not necessarily bad or wrong but your own. Said that, to some suggestions:
- remove the positive feedback between the triodes. It's very dynaco specific and will create problems if not perfect. Then make both end of the concertina 33k for symmetry. If you need more sensitivity, increase the feedback resistor.
- add screen and grid stoppers on the el84 ( 1k grid, 100ohm screen). Mount these
direct on the tubesocket.
-using a scope and low level square input, adjust the feedback cap for best square on a resistive loaded output. -Double check the amp with some capacitive load (1uF in parallel with 8ohm) Adjust feedback components until good.
lfs,
I am also sorry to join late.
I see your schematic of post # 4 has 375V on the screens with a cathode self bias of 13 or 14V. 375 - 14V = 361V screen to cathode.
You may not get very much life out of your EL84 tubes that way.
The EL84 maximum screen voltage specification is 300V (screen to cathode) at quiescent (no signal) conditions. The EL84 maximum plate voltage specification is 300V (screen to cathode) at quiescent (no signal) conditions.
The EL84 maximum screen voltage and plate voltage specification(s) are 550V (plate to cathode and screen to cathode), but Only when the tube is completely turned off (no plate current and no screen current) which only occurs when there is a very high negative voltage on control grid 1. This kind of voltage (550V) on the screen would only occur in Ultra Linear mode and Triode mode when the tube is cut off. In Pentode operation, the screen can not be more than 300V.
I am also sorry to join late.
I see your schematic of post # 4 has 375V on the screens with a cathode self bias of 13 or 14V. 375 - 14V = 361V screen to cathode.
You may not get very much life out of your EL84 tubes that way.
The EL84 maximum screen voltage specification is 300V (screen to cathode) at quiescent (no signal) conditions. The EL84 maximum plate voltage specification is 300V (screen to cathode) at quiescent (no signal) conditions.
The EL84 maximum screen voltage and plate voltage specification(s) are 550V (plate to cathode and screen to cathode), but Only when the tube is completely turned off (no plate current and no screen current) which only occurs when there is a very high negative voltage on control grid 1. This kind of voltage (550V) on the screen would only occur in Ultra Linear mode and Triode mode when the tube is cut off. In Pentode operation, the screen can not be more than 300V.
Last edited:
Adding an EFB mod, and no other change is not going to increase the life of the EL84 tubes.
With 375V on the screen, and 370V on the plate, in order to get enough bias voltage to cut the plate to cathode and screen to cathode voltages to 300V, the bias voltage would have to be 75V, which will put the tube into cutoff (no current, no output power).
The B+ voltage needs to be reduced first. The solid state rectifiers would have to be replaced with a tube rectifier, and there still would not be 75V B+ drop. So keep the solid state rectifiers. Then . . .
To get reduced B+ would be to use a choke input supply (no capacitor before the input choke). The formula for critical inductance is 350/Load mA. For example: A mono-block with 60mA B+ current would require a 5.8 Henry input choke. A stereo amp with 120mA B+ current would require a 2.9 Henry input choke. Do not forget to include the current of the output stage, the input stage, driver stage, And the Bleeder resistor (a bleeder for 'safety first').
With 330V 0 330V B+ secondary, choke input gives 0.9 x 330V = 297V There will be some loss in the choke DCR, and any other series resistor in the B+. So you can use a small capacitor just before the choke, perhaps a 0.5uF, 1uF or 1.5uf to get the B+ to 300V at the Screen of the Ultra Linear Tap. You do not want the first cap and choke to resonate at 2X power line (50 = 100Hz or 60 = 120Hz) For a 3 Henry choke on a stereo block, you have to use a 1 uF or larger (0.5uf and 3 henry resonates at ~ 120Hz, 2 times 60 Hz power) For a 6 Henry choke you can use the 0.5uF cap. Because of the smaller input capacitor before the choke, you will probably need 1 more RC filter after the choke to get the B+ ripple low enough, especially for the input tube.
Good luck getting the B+ set for long tube life.
With 375V on the screen, and 370V on the plate, in order to get enough bias voltage to cut the plate to cathode and screen to cathode voltages to 300V, the bias voltage would have to be 75V, which will put the tube into cutoff (no current, no output power).
The B+ voltage needs to be reduced first. The solid state rectifiers would have to be replaced with a tube rectifier, and there still would not be 75V B+ drop. So keep the solid state rectifiers. Then . . .
To get reduced B+ would be to use a choke input supply (no capacitor before the input choke). The formula for critical inductance is 350/Load mA. For example: A mono-block with 60mA B+ current would require a 5.8 Henry input choke. A stereo amp with 120mA B+ current would require a 2.9 Henry input choke. Do not forget to include the current of the output stage, the input stage, driver stage, And the Bleeder resistor (a bleeder for 'safety first').
With 330V 0 330V B+ secondary, choke input gives 0.9 x 330V = 297V There will be some loss in the choke DCR, and any other series resistor in the B+. So you can use a small capacitor just before the choke, perhaps a 0.5uF, 1uF or 1.5uf to get the B+ to 300V at the Screen of the Ultra Linear Tap. You do not want the first cap and choke to resonate at 2X power line (50 = 100Hz or 60 = 120Hz) For a 3 Henry choke on a stereo block, you have to use a 1 uF or larger (0.5uf and 3 henry resonates at ~ 120Hz, 2 times 60 Hz power) For a 6 Henry choke you can use the 0.5uF cap. Because of the smaller input capacitor before the choke, you will probably need 1 more RC filter after the choke to get the B+ ripple low enough, especially for the input tube.
Good luck getting the B+ set for long tube life.
Last edited:
There isn't a tube out there that can not be run somewhat higher than it's maximum ratings. The good thing is that doing that may not destroy it, but may cause early failure or early performance reduction over time.
A lot of solid state devices will have very short lives when you exceed maximum voltage, maximum current, maximum power, Safe Operating Area (SOA), maximum temperature; or two or more of these in combination. They are less forgiving than vacuum tubes.
I just prefer to run my tubes at less than their maximum ratings. Some tubes produced today are more subject to failure when driven past their specs . . . But the best of "yesterdays" tubes would have lived.
For whatever reason, lfs's EL84 plates were at 341 to 344V, and bias at 14.4V. That is quite different than 370V.
Post #4 shows a common-4-cathode self bias resistor of 47 + 47 = 94 Ohms. At 14.4V, that is 153mA for 4 tubes (38mA/tube). 341V - 14V = 327V 327V x 38mA = 12.5W, at least the plate dissipation is only a little above spec. at 370V - 14V = 356V, and 356V x 38mA = 13.5W plate dissipation.
And, there are no individual self bias resistors for each tube, so currents can vary quite a bit from tube to tube. Push Pull Output transformers do not like very much DC imbalance. There is no provision to test the imbalance current in this circuit. Use 4 individual self bias resistors and 4 individual bypass caps. That way, testing current balance is easy, and most tubes will balance quite well this way without even using matched tubes.
This amp is built from scratch to 'duplicate' a classic amp, but planning ahead would allow for improvements (but some improvements like 4 individual bias resistors and 4 bypass caps will take up more real estate).
There are 3 feedback loops in the amplifier:
Input tube cathode to driver cathode loop. Ultra Linear Tap to input tube cathode loop. Output transformer secondary to input tube cathode loop.
All of these 3 loops, plus the output transformer characteristics, determine the overall stability of this amp, both with a resistor load, and particularly with a loudspeaker load.
A lot of solid state devices will have very short lives when you exceed maximum voltage, maximum current, maximum power, Safe Operating Area (SOA), maximum temperature; or two or more of these in combination. They are less forgiving than vacuum tubes.
I just prefer to run my tubes at less than their maximum ratings. Some tubes produced today are more subject to failure when driven past their specs . . . But the best of "yesterdays" tubes would have lived.
For whatever reason, lfs's EL84 plates were at 341 to 344V, and bias at 14.4V. That is quite different than 370V.
Post #4 shows a common-4-cathode self bias resistor of 47 + 47 = 94 Ohms. At 14.4V, that is 153mA for 4 tubes (38mA/tube). 341V - 14V = 327V 327V x 38mA = 12.5W, at least the plate dissipation is only a little above spec. at 370V - 14V = 356V, and 356V x 38mA = 13.5W plate dissipation.
And, there are no individual self bias resistors for each tube, so currents can vary quite a bit from tube to tube. Push Pull Output transformers do not like very much DC imbalance. There is no provision to test the imbalance current in this circuit. Use 4 individual self bias resistors and 4 individual bypass caps. That way, testing current balance is easy, and most tubes will balance quite well this way without even using matched tubes.
This amp is built from scratch to 'duplicate' a classic amp, but planning ahead would allow for improvements (but some improvements like 4 individual bias resistors and 4 bypass caps will take up more real estate).
There are 3 feedback loops in the amplifier:
Input tube cathode to driver cathode loop. Ultra Linear Tap to input tube cathode loop. Output transformer secondary to input tube cathode loop.
All of these 3 loops, plus the output transformer characteristics, determine the overall stability of this amp, both with a resistor load, and particularly with a loudspeaker load.
It is no problem at all driving an EL84 at 375V. Don't let the suggested operating conditions in documentation fool you to believe that this is
all that is allowed. Look at this documentation, page 9 : https://frank.pocnet.net/sheets/030/e/EL84.pdf here we have anode curves up to 600V, Vg2 = 300V Dynaco ST35 runs the EL84 at 375V, so does Dynaco SCA35. None of
these are known to destroy EL84, none of these uses the tubes outside spec.
Lack of bias adjustment in suggested amp is a moot argument, use a matched quad and no adjustments are needed.
> here we have anode curves up to 600V
Curves are traced with short-term tests; often with 100/120Hz rectified AC. The test voltage will be over 300V only about 1/3rd of the time. The DC value of the test waveform is under 300V.
Curves for power tubes "should" be traced to almost twice the safe value for steady DC, because in-use they try to swing to twice the B+.
That said.... I would not worry about any new-made, or most vintage, EL84, working closer to 400V than 300V. There are too many well-loved designs which work them at high voltage. Aside from Dynas, there is a flock of guitar amps which ignore the rules. Anybody selling a EL84 today has to be designing for high voltages.
Curves are traced with short-term tests; often with 100/120Hz rectified AC. The test voltage will be over 300V only about 1/3rd of the time. The DC value of the test waveform is under 300V.
Curves for power tubes "should" be traced to almost twice the safe value for steady DC, because in-use they try to swing to twice the B+.
That said.... I would not worry about any new-made, or most vintage, EL84, working closer to 400V than 300V. There are too many well-loved designs which work them at high voltage. Aside from Dynas, there is a flock of guitar amps which ignore the rules. Anybody selling a EL84 today has to be designing for high voltages.
And as I mentioned The EFB mod has been proven to help with output tube life.
You can even set it up with a pot to adjust a pair/per channel or use 4 pots.
JJ are about the only new production that can last in the Dynaco ST35/SCA35 and other amps that run the tubes hard. Most Russian and US NOS (new old stock) also last.
You can even set it up with a pot to adjust a pair/per channel or use 4 pots.
JJ are about the only new production that can last in the Dynaco ST35/SCA35 and other amps that run the tubes hard. Most Russian and US NOS (new old stock) also last.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.