Yes, that's what will sort of happen if the output stage overloads first.
There is a definition issue here. In intentional AB2, grid current occurs before the output stage overloads (ie clips).
I believe I pointed out in a much earlier post that it should be arranged that the driver overloads at a signal level just a little bit above the level required to overload the output stage if you want the best overdrive character and zero post overdrive recovery issues. Because you then will not get grid current.
There is a subtle difference however, between an output stage correctly loaded (ie correct output transformer impedances) and driven to grid current by a strong driver) and an output stage with too low an output transformer primary impedance. The first case goes into clipping at or before grid current, the second case may go into grid current before clipping. In high quality audio, the second case sounds worse.
Last edited:
OK that clarifies it. Then its just a case of deciding how you apportion things.
The interesting part about the Radford is, despite its reputation, there were few proper sets of measurements done on it. Radford themselves said 35W into 8Ohms @1% THD but the woodside version, which uses the same transformer but with only one secondary will do 50W/4Ohms with 6550s. So I think we can say the output is good enough not to worry about. I'm happy with EL34 (esp the price) and chasing the last 0.5dB output power seems pointless so very tempting to set the operating point so 2V in is 35W out (passive linestage). Then its only vinyl I have to worry about.
The interesting part about the Radford is, despite its reputation, there were few proper sets of measurements done on it. Radford themselves said 35W into 8Ohms @1% THD but the woodside version, which uses the same transformer but with only one secondary will do 50W/4Ohms with 6550s. So I think we can say the output is good enough not to worry about. I'm happy with EL34 (esp the price) and chasing the last 0.5dB output power seems pointless so very tempting to set the operating point so 2V in is 35W out (passive linestage). Then its only vinyl I have to worry about.
Do the calculations. 13k x 16mA = 208V. With the necessary bias, that will be more like 212V, but close enough.
Why is it necessary to use such a high current?
In an amplifier without any silicon, the driver currents are typically in the range 1 to 3 mA. Anode loads are in the range 50 to 200 kohm.
One of the highest quality tube amplifiers ever produced was the GEC "88-50" - a refinement of the Williamson design with UL output. It has a push pull driver (12AU7) with a tail current of 12 mA. However, the 50 watt output rating required a HT voltage of 500 V. So even with a tail resistor dropping 200V, there is still 300V between the driver cathodes and the HT rail, clearly more than sufficient to provide enough clean drive to the KT88 output tubes.
But there is no high resistance tail in the 88-50. The tail is just a 680 ohm resistor - the amount needed for cathode bias. That's because the input to the drivers is balanced - which is also the case with the Bulwer-Lytton. (though with the B-L, a somewhat higher tail resistor will be advantagous for a CMRR point of view - the amp input is direct to the LTP. 10 to 13 kohm is entirely sufficient).
There is no need for a high tail current in the B-W in any case as the LTP doesn't drive the output stage directly.
Sound like a case of sloppy design requiring extra complexity to overcome the consequences sloppyness. Get it right and you don't need the CCS.
Linearity. If you had bothered to read the book before slagging it, you might have caught that part.
I wondered if you would mention that.
It's true that linearity improves as you move conditions away from cut-off. But beond a certain point it makes little difference.
Three things:-
1. It's an LTP. The dominant 2nd harmonic gets cancelled. And if the LTP doesn't remove it, the push-pull output stage will.
2. If you have to use an excessive current to get sufficient linearity, you are using the wrong tube.
3. I accessed the MJ book via Google Books. The resolution doesn't let me see what the LTP tube type is, but common candidates are:-
12AU7/ECC82: Datasheets show that linearity is about as good as it can get with Ia > 5 mA approx (LTP tail 10 mA). The Rp of this tube is such that fully driving a typical output stage requires a peak to peak swing of around 2 mA. Ref Mullard Book 2-1.
12AX7/ECC83: Requires Ia > 5mA approx for linearity about as good as it can get. The much larger Ra means a much smaller signal swing is needed, resulting in improved linearity. In practice, to get LTP driver distortion small compared to output stage distortion, Ia can be reduced to 2 to 3 mA (tail current 4 to 6 mA). Ref Mullard Book 2-1.
Once LTP/driver distortion is made small compared to the output stage distortion, there is no point in trying to bring it down futher anyway.
Some folks use a 6SN7GT or 12SN7GT. For the purpose of this discussion, it's the same as a 12AU7. (not the same in certain other aspects, not the least being octal).
There are various other triodes designed for audio that can be used. They are either directly equivalent to either the 12AU7 or 12AX7, or have characteristics somewhere between the two.
As I said, a triode's linearity improves the higher you set it's operating current. But that does NOT mean that doubling some value always halves the distortion. There is a law of diminishing returns. It's the old let's use ultra pure water in cooking thinking again.
Before you post, it might be good if you actually consulted some authoritive textbooks and understood tube circuits so that you can read Duncan with a more critical eye.
There is sanother factor in this. There is a published by a tube manufacturer in part to show how their tubes can be employed to get the best results. If you have too much impedance in the cathode circuit, you leave yourself open to hum intermod. You also leave yourself open to the tube operating conditions being upset. This is because there is always some coupling of heater AC into the cathode. Over their service life, tubes tend to develop heater-cathode leakage.
The GEC 88-50, with its' low value tail resistors, will meet specifications not just with new tubes, it will still meet them when the tubes are at there end of life point. A twin triode with some heater-cathode leakage just passing a tube test will, if plugged into the 88-50, will result in the 88-50 working entirely normally. The same triode plugged into the MJ amplifier will result in audibly bad sound - hum and distortion.
A good engineer doesn't just devise a circuit that works well with new tubes. A good engineer devises a circuit that works just the same with any tubes that pass a tube test. The tubes then give their full service life.
Last edited:
7N7. I look forward to your data showing that it can work better at low currents.
I'm not sure who Duncan is, but I've managed to get through one or two books in my day. Even designed and built an amp or two. Huh.
This is discussed at length in Valve Amplifiers. And means to prevent that issue are incorporated into the designs.
I'm not sure who Duncan is, but I've managed to get through one or two books in my day. Even designed and built an amp or two. Huh.
This is discussed at length in Valve Amplifiers. And means to prevent that issue are incorporated into the designs.
A 7N7 is a 6SN7GT with a different base and different glass shape - Loctal instead of octal.
Please read my comments more carefully. I didn't say the thing works better at low currents. I said there is no noticeable improvement above a current much lower than you seem to think is appropriate.
Some pros have chosen a 6CG7. This was originally designed for television sets and is made to tighter specs for performance and reliability. It offers much better heater-cathode insulation than the 12AU7/12AX7 types. It will give lower hum and noise, but for the purpose of this discussion requires the same operating conditions as a 12AU7.
My typo. I meant Jones. Ben Duncan is another chap who has sold a lot of books. Appologies to Mr Duncan.
By adding more complexity no doubt. Complexity no doubt unnecessary if the design was engineered properly in the first place.
Not that any such measures are shown in the B-L circuit on page 544 of MJ. The constructor seems to be left to discover the need later, if he doesn't realise the tube "failed" early (he probably won't). A common floating transformer winding is shown powering all filaments - which is itself very bad practice.
Last edited:
I explained in previous posts that there is a law of diminishing returns, and nominated tube currents such that no higher current will produce noticeable improvement.
If you look at the tube grid curves for audio triodes (please do look), it's generally pretty obvious where the grid volts/anode current transfer departs from the three-halves rule near cutoff, at the bias levels appropriate for this discussion. So long as the LTP operates in the three-halves region, harmonic cancellation is effective regardless of standing current - almost. But if the signal takes it outside the three-halves region, then the two triodes will soft clip alternately, dramatically raising odd order harmonics.
If you are still not satisfied, I suggest you look at tube curves and use the standard method of calculating the approx 2nd harmonic level. Measured data in forum such as this has to be taken on trust. You don't need to trust me - you can do the calcs and work it out for yourself.
Or, you can do the test. If you have the B-L amplifier or some other similar amp with an LTP driver and a CCS tail, connect the juction of te two ctahodes and the CCS to a pot, say 100 K log in series with a capacitor to earth. Twiddle the pot while doing a THD test or spectrum plot. Or just listen. You'll find tail impedance to earth makes little difference except at values well within what a resistor can provide. If the LTP tube is well used, you might even find a low tail impedance works better.
Let us know how the test goes.
If you look at the tube grid curves for audio triodes (please do look), it's generally pretty obvious where the grid volts/anode current transfer departs from the three-halves rule near cutoff, at the bias levels appropriate for this discussion. So long as the LTP operates in the three-halves region, harmonic cancellation is effective regardless of standing current - almost. But if the signal takes it outside the three-halves region, then the two triodes will soft clip alternately, dramatically raising odd order harmonics.
If you are still not satisfied, I suggest you look at tube curves and use the standard method of calculating the approx 2nd harmonic level. Measured data in forum such as this has to be taken on trust. You don't need to trust me - you can do the calcs and work it out for yourself.
Or, you can do the test. If you have the B-L amplifier or some other similar amp with an LTP driver and a CCS tail, connect the juction of te two ctahodes and the CCS to a pot, say 100 K log in series with a capacitor to earth. Twiddle the pot while doing a THD test or spectrum plot. Or just listen. You'll find tail impedance to earth makes little difference except at values well within what a resistor can provide. If the LTP tube is well used, you might even find a low tail impedance works better.
Let us know how the test goes.
Last edited:
Which has never stopped me from gleefully pointing out his errors when he makes them. That's one of the reasons we get along so well. I was delighted to see how many of my corrections made it into the 4th edition.
The last time he stayed with us, my wife was horrified. "I can't believe the way you guys talk to each other!"
The last time he stayed with us, my wife was horrified. "I can't believe the way you guys talk to each other!"
You are the guy who decided to dispute me. It's up to you to back up your posts. Continually referring to MJ's book doesn't cut it - my point is that MJ is not a tube amp guru, and the B-L circuit is clearly the result of someone who doesn't understand tube circuit design.
Its up to you to do look at the tube data, do the calcs, do the practical test. I assume from all your posts that's within your capability.
Look at the grid curves, honestly, and tell us what you see. At what current do they depart from the three-halves rule?
I've already done the calcs and the tests. But posting data on this forum means nothing. If you do it yourself, it will mean something to you.
Well, I''ve engineered more than a few amps and stereos etc. For commercial production - back when we had an electronics industry (everything comes from China now, same as for you guys), as well as my personal home systems. But whether or not you or I have designed amplifiers is kind of irrelevant isn't it? Neither or us has any way to tell how good/bad they were. Or whether you & I wholey engineered from the ground up, or wholey or partly lifted from someone else's circuit. What matters is whether posted explanations stack up.
Last edited:
I don't see that as a problem (if that's what you meant to imply). I don't agree with everything my friends say. And if someone else points out where a friend has an unsupportable view, it makes no difference to the friendship. With either one.
Last edited:
Since you've presented no data, nor a concrete and tested alternative to the design, there's really nothing to dispute.
That's a powerful argument.
The 7N7 used has a mu of roughly 20. So either
1) a CCS is needed (perhaps $1.50 in parts) or
2) a large resistor and a high voltage negative rail (perhaps $50 in parts) or
3) running the 7N7 at 1mA and arguing that this won't cause significant distortion or
4) arguing the the resultant imbalance from a short tail is unimportant.
Frankly, to me, the $1.50 and essentially perfect balance solution seems best.
The more parts the more there is to fail. And in getting that supposedly perfect balance you have added hum intermod susceptability, plus uncertain operating conditions.
And I already showed that even if tail currents are high, with sufficient HT for the output stage, there is sufficient HT to have a high value tail resistor without adding a negative supply.
Did I say you had to run the 7N7 at 1 mA? No, I did not, I said run a 6SN7 (same thing except for the loctal base) at around 5 mA. Although in a balanced input triode output amp with no feedback like the B-L circuit there will be enough output stage distortion to swamp out LTP distortion even with LTP currents somewhat less than 5 mA. Yes, some LTP inbalance is unimportant. In an ultralinear amp there would need to be more care in LTP design.
If a high mu triode does a better job, then use a high mu triode.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.