Well, I have the choke tuned to 120Hz by way of a capacitor across the leads... so I *guess* that functions like the .68u cap you suggested since in my arrangement it goes from diode bridge-to-ground across the choke, and I have bleeder resistors across the capacitors too, which should help. I think I'll bypass each diode in the bridge with a .1u cap just to be safe too.
As far as the buzz, I probed the setup, and there is very very little ripple from the PSU... less than a volt. I think my problem was a stupid ground loop with the input since with the input terminal grounded, the amp is dead silent, but with anything plugged in... or with a 50K resistor soldered across the input jack from hot to ground, it buzzez, with the buzz level being controlled by the volume pot after the input stage. The layout is fine, and the star ground scheme is the same one that I always use, so I automatically thought that the new choke-input setup was the buzz culprit. I re-flowed all the grounds and components, and tried different ground points for the input stage... even raising the signal ground with the old .22uf/22r/parallel reverse diode trick, but it didn't help. I have a good 15" between my transformers, so I don't think induction is the problem, and that would not be modulated by the volume controll anyway. The AC from the wall is filtered too.
As far as the buzz, I probed the setup, and there is very very little ripple from the PSU... less than a volt. I think my problem was a stupid ground loop with the input since with the input terminal grounded, the amp is dead silent, but with anything plugged in... or with a 50K resistor soldered across the input jack from hot to ground, it buzzez, with the buzz level being controlled by the volume pot after the input stage. The layout is fine, and the star ground scheme is the same one that I always use, so I automatically thought that the new choke-input setup was the buzz culprit. I re-flowed all the grounds and components, and tried different ground points for the input stage... even raising the signal ground with the old .22uf/22r/parallel reverse diode trick, but it didn't help. I have a good 15" between my transformers, so I don't think induction is the problem, and that would not be modulated by the volume controll anyway. The AC from the wall is filtered too.
this is what a high voltage 1n4007 diode circuit really looks like -- 210mH is the leakage inductance of an 800V Thordarson trafo I happen to have lying around -- so the leakage inductance, diode junction capacitance and all the stuff are modeled here.
when you put 100nF across the diode you bring the resonant frequency of the circuit down -- add enough bypass capacitance and you can hear it!
when you put 100nF across the diode you bring the resonant frequency of the circuit down -- add enough bypass capacitance and you can hear it!
Attachments
Aletheian,
Is there any way you could give us the schematic of your input stage (before the volume control). Quite apart from power supply considerations, the buzz you mention and the fact that the input resistance controls it sound for all the world like (feedback) instability involving Miller capacitance - but that would depend on the circuit. Then if I may go very wide; I presume your heater supply to the first stage is properly earthed (or bypassed to common)?
Also, I accept that a small cap before an LC filter will tend to snub choke switch-off peaks, but it is always better to address the source of an interference without going through other circuit parts, especially in the case of high voltage peaks. In that respect I personally would prefer to mount an RC directly across the choke.
Is there any way you could give us the schematic of your input stage (before the volume control). Quite apart from power supply considerations, the buzz you mention and the fact that the input resistance controls it sound for all the world like (feedback) instability involving Miller capacitance - but that would depend on the circuit. Then if I may go very wide; I presume your heater supply to the first stage is properly earthed (or bypassed to common)?
Also, I accept that a small cap before an LC filter will tend to snub choke switch-off peaks, but it is always better to address the source of an interference without going through other circuit parts, especially in the case of high voltage peaks. In that respect I personally would prefer to mount an RC directly across the choke.
Johan Potgieter said:
It's really nothing special... just a standard grounded cathode gain stage using a 12ax7. I'd draw a schem, but it is really right out of the RDHB. I had originally paralelled the halves, but have since cascaded them with the voume controll in the middle, then to a 12at7 LTP and into the KT88's. I had no NFB originally, but recently applied about 10dB or so, but I really don't like the effect, so I'm removing it, which had no effect on buzz.
Filaments for now are half-wave rectified, and filtered through 20,000uF, and then a .47 ohm, 5w resistor to knock the voltage down to about 6.7v DC. I tried them AC with two reference resistors forming a virtual center tap and tied that junction to a voltage divider from B+, decoupled with a 10u cap... but the buzz was still there.
So what is this Miller feedback instability thing you are talking about? Maybe I know it by another term.
PS I placed bleeder cap directly across the choke. I have measured no voltage spikes.
Alethian,
Since the "buzz" is controlable by the volume control which is after the first stage that suggests that the buzz is being introduced/generated in the first stage.
Make sure that the input socket ground side is isolated from chassis. The ground side of the input socket should connect to the bottom (0V side) of the input stage grid leak resistor. Use shielded cable for the run from the input socket to the input stage. You can also try wiring a 10nF ceramic from the input socket ground side to the chassis locally to make sure that its at AC ground at RF frequencies.
Also make sure that you have a series grid stopper resistor in the grid of the first stage of at least 4K7 with its body as close to the grid pin as you can physically manage. This will address any Miller capacitance caused instability issues.
Cheers,
Ian
Since the "buzz" is controlable by the volume control which is after the first stage that suggests that the buzz is being introduced/generated in the first stage.
Make sure that the input socket ground side is isolated from chassis. The ground side of the input socket should connect to the bottom (0V side) of the input stage grid leak resistor. Use shielded cable for the run from the input socket to the input stage. You can also try wiring a 10nF ceramic from the input socket ground side to the chassis locally to make sure that its at AC ground at RF frequencies.
Also make sure that you have a series grid stopper resistor in the grid of the first stage of at least 4K7 with its body as close to the grid pin as you can physically manage. This will address any Miller capacitance caused instability issues.
Cheers,
Ian
I am so glad that you got past the sizzling - always rewarding to be able to clear up a fault.
Instability as a result of Miller capacitance is probably as you know it. I was referring in very general terms to those main amplifiers with an ECC83 as input stage, following a 1M or such volume control. Even with low source impedance you still have 250K in serie with G1 at pot. mid-setting. The total input capacitance (Miller plus rest) could be of the order of 100+ pF, which with the above serie resistance starts to produce phase shift at 6 - 7 KHz. But with the pot at minimum or maximum this point could be orders higher. With this situation in a heavy global feedback loop (applied to the input cathode) instability could occur at mid-pot setting if the amp was originally optimised with direct low impedance (signal generator) input. This infrequently could also occur in feedback pre-amp stages. Supersonic oscillations thus generated could manifest as an audible buzz (intermodulation effects from dithering supersonic frequency).
As Gingertube suggested it is better to start with a finite resistance in the grid circuit, but then with high-mu triodes as inputs you quickly start to erode your h.f. response.
As said, not a general effect but I have encountered it, and since we did not seem to find a solution for you.........
Instability as a result of Miller capacitance is probably as you know it. I was referring in very general terms to those main amplifiers with an ECC83 as input stage, following a 1M or such volume control. Even with low source impedance you still have 250K in serie with G1 at pot. mid-setting. The total input capacitance (Miller plus rest) could be of the order of 100+ pF, which with the above serie resistance starts to produce phase shift at 6 - 7 KHz. But with the pot at minimum or maximum this point could be orders higher. With this situation in a heavy global feedback loop (applied to the input cathode) instability could occur at mid-pot setting if the amp was originally optimised with direct low impedance (signal generator) input. This infrequently could also occur in feedback pre-amp stages. Supersonic oscillations thus generated could manifest as an audible buzz (intermodulation effects from dithering supersonic frequency).
As Gingertube suggested it is better to start with a finite resistance in the grid circuit, but then with high-mu triodes as inputs you quickly start to erode your h.f. response.
As said, not a general effect but I have encountered it, and since we did not seem to find a solution for you.........
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.