Yes, so there is no wide band noise measurement.
On mine the maximum bandwidth is 300Hz so there's going to be a lot of measurements to cover 20KHz. It is not really a practical way of measuring wideband noise. I use a Lindos Minisonic MS10 audio test set which measures noise directly to the current IEC specification. I have to say I have never seen an ECC88 CF come anywhere near -100dBV.
Cheers
Ian
Yes, a dedicated test set is a nice thing to have if you can afford it. But it's not terribly difficult to use other instruments to get the numbers if you're reasonably resourceful.
There's no reason you shouldn't get a -100dBV noise; the ENR of a decent ECC88 is 250R or so at 10mA plate current (the ones I use are Amperex, made for Tek scope plug ins). Maybe if you post the circuits you're using and the waveforms you're getting, we can troubleshoot them.
There's no reason you shouldn't get a -100dBV noise; the ENR of a decent ECC88 is 250R or so at 10mA plate current (the ones I use are Amperex, made for Tek scope plug ins). Maybe if you post the circuits you're using and the waveforms you're getting, we can troubleshoot them.
That's just the problem, the ENR of an ECC88 is nowhere near 250R at 10mA. The theoretical ENR is that figure but that formula applies only at RF. At audio frequencies is it of little use on its own. In practice at audio frequencies the ENR is much higher.
As for circuits, no mystery there, a regular self biased CF set for the same plate voltage and current as your design will illustrate the point.
Cheers
Ian
As for circuits, no mystery there, a regular self biased CF set for the same plate voltage and current as your design will illustrate the point.
Cheers
Ian
Perhaps if you could post actual measurements and circuits, we could discuss this more intelligently.
And I need to apologize for one oversight on my part (working from memory is not a great thing to do)- my measurements are referenced to 2.5V, not 1V. That shifts things about 8dB, but doesn't significantly change my point.
Most of my measurements are on circuits with gain. I last used a CF in a low level circuit a long time ago and was disappointed with the noise performance. I will try and dig out the figures but from memory there was a gain stage with a measured output noise of 40uV (20KBW). I added a CF to buffer the output and the noise jumped to about 80uV which essentially means most of that 80uV co
So looking at your last graph if we say that averages about -145dB across the audio band then the rms noise is 43 + 8 dB higher or about -94dBV or about -92dBu which is a tad more believable since I have been able to measure -88dBu myself.
Cheers
Ian
Yes, details count. Input shorted, 10k input impedance sound card as load (Echo Audiofire 2), driven through a 1m coax cable. Make sure to use the input transformer (low DCR secondary) or a short to a quiet ground, CCS biasing, a quiet heater supply, and a DC output, all of which have impact on the noise. Of course, your choice of ECC88 is critical as well (as I mentioned, I used a version specifically made for Tek scope input stages). It's easy to introduce extraneous noise, and the whole point of this project is squeezing everything out of cathode follower you can by paying attention to the details.
Even if the ENR is 1k (and a transformer can have that much self-noise if the DCR isn't nice and low), the unweighted 20-20k noise is well below a microvolt.
Even if the ENR is 1k (and a transformer can have that much self-noise if the DCR isn't nice and low), the unweighted 20-20k noise is well below a microvolt.
I shall not be using a CCS. I plan to use a 275V regulated dc supply, dc heaters, 18K cathode load with 150R cathode bias, 200K grid resistor and input shorted via 1uF capacitor. Even if the ENR looks like 18K that's still -112dBV in 20K and I would be surprised if is gets that low. Test measuring will be with my Lindos test set. I will let you know the results.
Cheers
Ian
Cheers
Ian
Hello Guys,
This topic; noise is interesting.
The way I have studied it, ENR is a rule of thumb used in the RF world, and flicker or 1/f noise increases with each decreasing octave out weighing thermal noise below 5K Hz.
Sy the plot in your post 233 it looks like the worst offender is the 60Hz and harmonics from the power supply.
All for Fun
DT
This topic; noise is interesting.
The way I have studied it, ENR is a rule of thumb used in the RF world, and flicker or 1/f noise increases with each decreasing octave out weighing thermal noise below 5K Hz.
Sy the plot in your post 233 it looks like the worst offender is the 60Hz and harmonics from the power supply.
All for Fun
DT
Last edited:
GROUNDING
Hey, I just built this schematic without the input transformers for now and I'm a bit confused about the input grounding, hv grounding and how it all connects to the rail splitter ground. I have the negative rail going through the 470k to the output ground as per shematic (no dethumper yet). Do input and output grounds just get connected to this point?...or to the splitter ground....? Splitter ground and HV ground connected???
As well I have separate filament power (I'd love to try a 6n6p here...) and I'm using a tda 2030 to split the 24v. Would it be worth it to add the Lm317, 337's or perhaps a pair of tl431's at a later date. Tda 2030 seems to work ok?
Ready to apply power and assess but really don't want to smell smoke.
Tom
Hey, I just built this schematic without the input transformers for now and I'm a bit confused about the input grounding, hv grounding and how it all connects to the rail splitter ground. I have the negative rail going through the 470k to the output ground as per shematic (no dethumper yet). Do input and output grounds just get connected to this point?...or to the splitter ground....? Splitter ground and HV ground connected???
As well I have separate filament power (I'd love to try a 6n6p here...) and I'm using a tda 2030 to split the 24v. Would it be worth it to add the Lm317, 337's or perhaps a pair of tl431's at a later date. Tda 2030 seems to work ok?
Ready to apply power and assess but really don't want to smell smoke.
Tom
Well, I'm getting ready to build a whole mess of these buffers, I'm building a hafler matrix/ quadraphonic converter with all signal circuitry being tube, so on top of a few common cathode gain stages I will be building eight of these buffer circuits total.
In order to save sockets and board space for such small and trivial circuits, I was going to use either dual opamps (burr brown 2132) or quad (lm324) since I already have a dozen or so of each. Would the lm324 be adequate, or should I go with the opa2132?
Can I increase the servo .68uF cap to 1uF, or reduce it to .47uF to use parts already on hand?
Also, I have seven 6n1p, and five 6cg7, other than gain and current capability, shouldn't both work equally well in this application? I will be building a maida style regulator (from the RLD article) to supply everything, but my surround processor needs a b+ of 250 volts, so I would likely be using that voltage for everything.
I have only one 6dj8 that is reserved for the signal processing portion of the project, so I'd rather use the tubes I already have. A pair of the 6cg7 are NOS RCA clear top, which seems almost wasteful to use for this application, while the 6n1p are NOS voshkod all from the same lot.
I won't be driving more than a few feet of cable, and it will be into tube amps with ~50~100k inputs most likely.
In order to save sockets and board space for such small and trivial circuits, I was going to use either dual opamps (burr brown 2132) or quad (lm324) since I already have a dozen or so of each. Would the lm324 be adequate, or should I go with the opa2132?
Can I increase the servo .68uF cap to 1uF, or reduce it to .47uF to use parts already on hand?
Also, I have seven 6n1p, and five 6cg7, other than gain and current capability, shouldn't both work equally well in this application? I will be building a maida style regulator (from the RLD article) to supply everything, but my surround processor needs a b+ of 250 volts, so I would likely be using that voltage for everything.
I have only one 6dj8 that is reserved for the signal processing portion of the project, so I'd rather use the tubes I already have. A pair of the 6cg7 are NOS RCA clear top, which seems almost wasteful to use for this application, while the 6n1p are NOS voshkod all from the same lot.
I won't be driving more than a few feet of cable, and it will be into tube amps with ~50~100k inputs most likely.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.