Well, in a properly configured SRPP, both triodes DO run under the exact same conditions
...
At 10V output, I had no problem getting 0.05% from 12AT7, and 6dB better than that from the best versions. That quibble aside, I am in hearty agreement with everything else you wrote.
Only in SPICEland. On our planet putting two triodes in one envelope does not guarantee identicality, but (at best) similarity.Keit said:
On your planet, maybe.
Evidence? Measurements, that is, not simulations.
I think you will find that the maths says otherwise. Have you ever worked through the theory of the SRPP? I have - although I am now re-checking my algebra in case I made a mistake.
A 12AT7 is never going to be as linear as a 12AU7 or 12AX7, because it is a television/FM radio frequency changer, where non-linearity is a requirement, to get an IF output from modulating the input from teh aerial with the local oscillator.
Take look at the anode curves for the 12AT7 and 12AX7 in the Mullard Technical Handbook. You don't need to read off values like I did for the 12AX7, the non-linearity of the 12AT7 is visually obvious.
Finding an example that is 6dB better than the average is quite unlikley unless you are testing a lot of faulty tubes with poor emission. The non-linearity deliberately built in to the 12AT7 comes from the design of the grid structure, which is made to tight limits to comply with the datasheet.
If you have one or more 12AT7's that are more linear that audio triodes, it may because you have mislabelled tubes. Either your 12AT7's are really 12AX7's, or your 12AX7's are not 12AX7's. Such things turn up from time to time.
Your claim has no credibility anyway as you did not state operating conditions. Anode current? Bias voltage?
Last edited:
Hi,
Do they?
Cheers,
Well, in a properly configured SRPP, both triodes DO run under the exact same conditions
Do they?
Cheers,
Yes I saw your comment on syclotron. I must admit I was skeptical. Indeed, I suspect you may have made a measurment error. Either that or you have some highly unusual 12AT7s on your hands...
Yes, I have found all my 12AT7s to be very consistent. I have seen 6dB THD variation in 6SN7s though.KEIT said:
The follwing is for some 12AT7s with MOSFET CCS @ 3.2mA, bias 2V, 300V HT, 10k source resistance. Measured at 1kHz with an AP1.
An externally hosted image should be here but it was not working when we last tested it.
Last edited:
Yes, I have. More than once. Have you?
Tell you what: I've prepared a worked through example. When you are ready, you post your worked though example. If you find that you were wrong, we'll leave it at that. If you still think you are right, I'l post my already prepared example ASAP (it's evening time here as I post this - you might have to wait until morning when I get up.
For the icing on the cake, you can if you like assume a 5% error in tube gm and resistor values, and calculate the degree of distortion let through as a result.
Note that a 12AT7WA (military version) probably should not be included in a comparison.
Military versions are the same basic type but may differ in materials, design of micas, and design of grids. Performance can be thus expected to be slightly different, though in general should still meet datasheet specs.
Sometimes the US military objected to certain alloys. Sometimes they just wanted a failure rate lower than the civilian version was achieving.
I posted details on the measurements here a few years ago- I'll put them up on syclotron one of these days... In any case, the measurements were consistent across several varieties of 12AT7. The only outliers were the JJ, which look like they are constructed quite differently than the usual and had about 6dB lower distortion than Mullard, GE, RCA, and 6201.
Measurements were validated by comparing spectra acquired via a soundcard plus interface to an HP 3581A analyzer. And I've repeated them several times over the years while checking other phenomena (I find that tube to be a useful test bed). So no, I'm quite convinced that the measurements were accurate.
BTW, you might try to use a lower current, maybe 2mA. This result surprised me.
edit: How do you buffer the tube from the input impedance of the AP? Or does it have a high input impedance?
If they genuinely do give such low distortion, it would be interesting to curce test and verify that they still meet Mullard/GE/RCA datasheet specifications. Especially in regards to grid curves which, if different, mean that these "12AT7's" could not function properly in their intended function in VHF receivers.
Another MOSFET source follower, DC coupled to the anode. (Yes, I checked the distortion of the source follower, which was below 0.005% at 10Vrms)
I checked one sample at both 1.7mA and 9.1mA (both -1.5V bias) and got very little change in the distortion, which I suppose is indeed a little surprising.
Last edited:
The formula I arrived at is Zopt = (mu x Rk - ra)/2. The same formula is given by someone else in this thread (see post 12).Keit said:
Typical values of Rk will be in the region of 0.5/gm to 1.5/gm, so typical values for Zopt range from ra/4 to -ra/4. Yes, make Rk too small (i.e. run the valve at too high a current) and Zopt could be zero or negative.
As an example, let us take one of the 12AT7 conditions from the RCA 1954 datasheet: Va 100V, Ia 3.7mA, Rk 270R, Ra 15k, mu 60. I calculate Zopt as 600R.
We also need to know the SRPP output impedance. For this I have
Zout = (r'a1 + Rk) ra2 / (r'a1 + r'a2)
where r'a1 is the effective anode impedance of the lower valve and r'a2 is that for the upper valve. There are two cases to consider, depending on whether the lower valve has a bypassed cathode. Unbypassed we have ra = 15k. Bypassed gives us r'a = 15k + 61 x 270 = 31470. So
Zout (Rk1 bypassed) = 4929
Zout (Rk1 unbypassed) = 7564
Both are considerably higher than Zopt.
Modern audio valves should not be assumed to have the same characteristics as their earlier namesakes, but merely similar bias arrangements and gain. I forget where I saw it, but a few years ago one manufacturer was boasting that his valve had lower distortion than others - I took this to mean that it was not an equivalent but merely a useful substitute for most audio circuits.Keit said:
The new 'ECC83' which briefly appeared from Blackburn a few years ago not only had different heater current but also different interelectrode capacitances.
If you are interested in receivers (as I am) then best to stick to NOS valves.
That formula is completely wrong. Care to post the derivation?
I'll give you a second chance before I put up my worked example.
That would indeed kill the gain and make it useless. Do you actually believe it? Have another think with good ol' pencil and paper.
Looks like a typo to me. It's almost right.
Hint: If the thing's balanced, you'd expect a certain symetricallity in the Zout formula. Otherwise the Zout will not be constant over the signal swing, and being in series with the load, cause distortion.
Having read a great number of your posts, I certainly respect your knowlege. So I won't post my version yet - you have another chance.
May I suggest you choose a 12AX7? The principles are of course valid either way, but for the 12AX7 there is datasheet information the gives common cathode distortion we can use as a benchmark.
Last edited:
Best to stick to NOS for all manner of things if you ask me. Certainly best to avoid Chinese stuff.
I am involved in restoration of vintage radio station, recording studio, and television station equipment, and test gear. We use Chinese tubes if we are fault finding and can't quickly get NOS. Doesn't always work well.
Yes, that is the formula for the 'optimum' load impedance that gives balanced conduction in the two triodes. At least if they are assumed linear.
Hmm... assuming ra1 = ra2 = ra, and Rk1 = Rk2 = Rk, and mu1 = mu2 = mu, I get:
Zout = [ ra(ra+Rk) + raRk(mu+1) ]/ 2[ra+Rk(mu+1)].
This simplifies to approximately ra/2.
...may be the most condescending thing you have said so far, yeesh...Keit said:
Last edited:
That's not inconsistent. If memory serves, I got similar numbers at -1.7V (roughly) bias from a red LED. Switching to green (roughly -2V) dropped the distortion significantly.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.