I've used ferrite beads in some designs with tubes that were prone to oscillation. I also use them in my curve tracer.
They are easy to slide over the lead of the grid-stop/screen-stop resistor so they are right up against the tube socket pin.
You might give them a try as well.
They are easy to slide over the lead of the grid-stop/screen-stop resistor so they are right up against the tube socket pin.
You might give them a try as well.
Hello Scott,
When you test these valves you are now posting the place current. Please also post the plate voltage and the grid bias voltage. A first stage of a phono preamplifier has a low input voltage allowing a low grid bias voltage. A low bias voltage and low plate voltage will result in the higest possible gm and improved noise performance. Lower input voltage needs a low noise valve to function with a high as possible S/N Raito. This first stage likely will dominate the noise performance of the entire amplifier.
While you are testing these valves also consider 6ak5's and 6cl6's.
Thanks DT
When you test these valves you are now posting the place current. Please also post the plate voltage and the grid bias voltage. A first stage of a phono preamplifier has a low input voltage allowing a low grid bias voltage. A low bias voltage and low plate voltage will result in the higest possible gm and improved noise performance. Lower input voltage needs a low noise valve to function with a high as possible S/N Raito. This first stage likely will dominate the noise performance of the entire amplifier.
While you are testing these valves also consider 6ak5's and 6cl6's.
Thanks DT
This is unecessary since Scott has posted the test circuits. You can work out the voltages from that using Ohm's law without making more work for Scott. In any case, the voltages are largely irrelevent once you know the current, since the gm is more-or-less independent of voltage.
Let’s examine this assumption. First, gm=mu/rp. Now let’s look at a set plate curves for an example valve say a triode connected D3a. With a little looking and examining we see that as we look left to right in the direction of increasing plate voltage mu decreases gradually and rp increases not so gradually. Looking in the direction of increasing plate voltage we see that of the three parameters; mu, rp and gm we see that gm changes at a greater rate with a change in voltage than either mu or rp.
What do you mean “plate voltages are irrelevant”?
DT
I went through this in post #80. Unless you're dealing with a variable-gm tube (which is unlikely on a hi-fi forum), gm barely changes as you increase plate voltage while keeping current constant. That is what I mean when I say the plate voltage is irrelevant. If you know the current, then you also know the gm (within 10% say), which is all you need to know for noise calculations. OK, if you know the plate voltage as well then you can pin down gm to the nth decimal place if it makes you feel good, but it doesn't give you any more accurate noise information.
Last edited:
For an ideal triode (constant mu, 3/2 law) gm depends on current alone, so it should not be surprising that this is approximately true for a real triode.
It has been already mentioned by Frank but I'm wondering if there is any serious (non-empirical) study or collected technical data's about tubes used in starved (underheated) mode and the impact on the noise specifications in the audio band. This trick was used by NEUMANN long ago in their venerable U47 microphone which used a severely under-heated VF14 in order to reduce noise. I also noticed that FISHER and H.H SCOTT (a.o) deliberately used under-heated tubes in their preamp circuits (usually 12AX7's heated @ 10V instead of 12.6V), maybe for similar reasons ? (lowering noise).
Yes, in my paper I mention briefly how I have found noise figure to improve with a starved heater if the anode current is less than about 1mA. Above 1mA the noise figure degrades if you starve the heater.
Pioneer SM-Q300B tuner-amp used 5V AC heaters on a couple of 12AX7 pre-amp tubes that were wired for 6.3V--I think they were the phono stage.....
Lowering heater ACV can lower hum contribution - perhaps that was significant enough to make it a design change for Pioneer.
An update on my D3a measurements. My test circuit shows no indication of RF instability. I examined the output with a 100MHz analog scope, and I probed all the internal nodes with a low-capacitance, 100MHz 10X probe. There is no RF output, even when touching or tapping the tube. So the microphonics appears unrelated to RF stability.
I did notice that the tip of the tube was touching the metal enclosure I use for shielding, since the D3a is taller than other 9-pin tubes I’ve tested, and that was exacerbating the microphonics. I cut a hole to eliminate this problem.
That said, the four D3a samples were still the most microphonic tubes I’ve seen in my study. My noise test setup has so much gain that I can clap my hands while standing across the room and see the microphonic behavior on the oscilloscope. ALL the tube types and brands I’ve studied show some microphonic output, but the four D3a samples had the highest by far, under exactly the same conditions as the other tubes. Maybe I just got 4 bad D3a??
As I mentioned last time, I also tested a bunch of NOS 6GK5 / 6FQ5A tubes I have on hand. These are high-transconductance frame-grid VHF triodes, sort of a mass market cousin to the D3a. Unfortunately, the low-noise magic of the D3a does not extend to the 6GK5 – I tested more than 10 and the best of them had input-referred noise around 1.5µV rms in the audio range (they’re not particularly microphonic, however……). If there is any value to these measurements, it is just to reiterate the conclusion that low RF noise does not necessarily translate into low audio noise. I did not add the 6GK5 to my list, since I suspect only I would be silly enough to consider the 6GK5 for audio.
Next in my measurement queue are more samples of current production E88CC and 6SL7. I’m also continuing my study of hum induced by AC on the heater – but that requires care to do correctly, and I am reworking my test fixtures.
Cheers,
Scott
I did notice that the tip of the tube was touching the metal enclosure I use for shielding, since the D3a is taller than other 9-pin tubes I’ve tested, and that was exacerbating the microphonics. I cut a hole to eliminate this problem.
That said, the four D3a samples were still the most microphonic tubes I’ve seen in my study. My noise test setup has so much gain that I can clap my hands while standing across the room and see the microphonic behavior on the oscilloscope. ALL the tube types and brands I’ve studied show some microphonic output, but the four D3a samples had the highest by far, under exactly the same conditions as the other tubes. Maybe I just got 4 bad D3a??
As I mentioned last time, I also tested a bunch of NOS 6GK5 / 6FQ5A tubes I have on hand. These are high-transconductance frame-grid VHF triodes, sort of a mass market cousin to the D3a. Unfortunately, the low-noise magic of the D3a does not extend to the 6GK5 – I tested more than 10 and the best of them had input-referred noise around 1.5µV rms in the audio range (they’re not particularly microphonic, however……). If there is any value to these measurements, it is just to reiterate the conclusion that low RF noise does not necessarily translate into low audio noise. I did not add the 6GK5 to my list, since I suspect only I would be silly enough to consider the 6GK5 for audio.
Next in my measurement queue are more samples of current production E88CC and 6SL7. I’m also continuing my study of hum induced by AC on the heater – but that requires care to do correctly, and I am reworking my test fixtures.
Cheers,
Scott
Scott - you may be able to characterise the microphonic output using a spectrum analyser. It's likely that there are a few dominant resonance frequencies that are being excited.
Given that you are plumbing the depths of noise floors, it would be a good cross check that there were no such extraneous signals surreptitiously being added - which could be easily missed by a meter, compared to a spectrum view.
Given that you are plumbing the depths of noise floors, it would be a good cross check that there were no such extraneous signals surreptitiously being added - which could be easily missed by a meter, compared to a spectrum view.
That's an interesting idea! I'll see if I can capture spectra of the D3a microphonic modes, and compare to the other tubes. Of course, we need to keep in mind, I may just have 4 bad tubes..........
Do you have a PVDF strip? That could be useful in two different ways, as excitation and as the reference vibration pickup to normalize the spectrum of the tube output.
A shock (ie. tap) test, with the amplifier stage output captured by a spectrum analyser with some form of waterfall display mode would be a simple technique - perhaps REW V5 can do that.
The PVDF strip looks like a great sensor for differential vibration testing. I recently bought two MMA7361L for just such testing - their response is limited to about 400Hz, but should be good for checking damping between say a desk, and an amp chassis using isolating feet. The PVDF may be useful for sensing the differential vibration between an amp chassis and the valve glass wall when using a damper socket of some type.
The PVDF strip looks like a great sensor for differential vibration testing. I recently bought two MMA7361L for just such testing - their response is limited to about 400Hz, but should be good for checking damping between say a desk, and an amp chassis using isolating feet. The PVDF may be useful for sensing the differential vibration between an amp chassis and the valve glass wall when using a damper socket of some type.
I have tested a lot of 12AX& variants for noise and microphonics in my tube mic pres. I have a Lindos MS10 audio test set. In noise measuring mode you can listen to the noise through its internal speaker and quite clearly hear the characteristics of microphony when you tap the tube. You can also hear pings as the tube heats up or cools down, as the electrodes expand and contract.
I have mainly tested significant number of current production types of 12AX7. The most consistently quiet type with minimum microphonics is the Sovtek 12AX7 WB. Both the 12AX7LPS and the 12AX7EH are quiet but much more microphonic.
Cheers
Ian
I have mainly tested significant number of current production types of 12AX7. The most consistently quiet type with minimum microphonics is the Sovtek 12AX7 WB. Both the 12AX7LPS and the 12AX7EH are quiet but much more microphonic.
Cheers
Ian
I updated the measurement summary on my "downloads" page to include a frequency domain spectrum of the microphonic response of a D3a. It was obtained by tapping gently on the aluminum box of my test circuit (with the cover removed), then waiting 0.3 – 0.5 seconds and capturing the waveform at the output of the noise measurement amplifier. There are distinct vibrational modes at 3.2 kHz, 4.2 kHz, and 4.8 kHz, along with a broad spectrum from 12 – 17 kHz. The modes at 3.2 – 4.8 kHz damp out within a few seconds, but the spectrum from 12 – 17 kHz can persist for 30 seconds or more. All four D3a tubes showed nearly identical behavior.
In contrast, identical tests on the JJ E88CC and Tung Sol Reissue 6SL7GT showed at least a factor of 10 lower microphonic response, which damped out so quickly that it was difficult (impossibly, really) to capture by this method.
Keep in mind that these results on the D3a are based on only four tubes obtained from a single source. And other forum contributors have reported good results with the D3a. I am looking for more samples to test.
Scott
Downloads – Tavish Design
In contrast, identical tests on the JJ E88CC and Tung Sol Reissue 6SL7GT showed at least a factor of 10 lower microphonic response, which damped out so quickly that it was difficult (impossibly, really) to capture by this method.
Keep in mind that these results on the D3a are based on only four tubes obtained from a single source. And other forum contributors have reported good results with the D3a. I am looking for more samples to test.
Scott
Downloads – Tavish Design
That is interesting stuff Scott. Perhaps that higher frequency persistent signal band is related to local anode-grid capacitance feedback.
Hello Dr. Dave,
Agreed, depending on the selected operating point the plate curves can be straight parallel and approximately evenly spaced.
Take a look at the D3A plate curves that we are talking about. At an operating point of 115 volts on the Anode and 40ma the grid voltage is -0.5 volts, the plate curves in this neighborhood are close to following Child’s Law.
Now look at an operating point of 165 volts plate, 7ma and -2.0 volts grid. In this neighborhood the plate curves are all stacking up on bended knee, hardly “Ideal Triode”. Parameters u, rp and gm are changing rapidly at this operating point.
An added thought. An “Ideal Triode” has straight evenly spaced plate curves.
Being straight and evenly spaced gm would be constant and not vary with current.
DT
Last edited: