• WARNING: Tube/Valve amplifiers use potentially LETHAL HIGH VOLTAGES.
    Building, troubleshooting and testing of these amplifiers should only be
    performed by someone who is thoroughly familiar with
    the safety precautions around high voltages.

6SN7 at low plate currents

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
In the input stage of an amp I have designed a 6SN7 is used with both triodes paralleled in a common cathode gain stage. This has an unbypassed 270R cathode resistor and a constand current plate load, and then feeds a direct-coupled LTP phase splitter (also with a constant current source in the tail).

In very rough terms I have a gain of 20 in the common cathode stage and 10 in the LTP. Therefore to achieve the 120V p/p needed to drive the output stage I need 12V p/p into the LTP.

Having looked at the plate curves etc I started off running the common cathode stage at 10mA for a plate voltage of 100V. This simulated well and measured OK, and this plate current or higher is the 'conventional wisdom'.

With the rest of the amp built I then measured the THD open loop, using a commercial meter (NTi Audio XL2).

This is where it gets interesting ...

At 50% ouput from the LTP (about 24VRMS) I discovered distinctly lower THD at much lower plate currents in the first stage - I adjusted the constant current for lowest THD and ended up with 3.3mA plate current for a plate voltage of 42V on the 6SN7 triode gain stage plate!!

The low plate current also simulates well - what is going in here? My assumption is that the lower plate current in the first stage means that the second LTP stage is further from clipping, therefore at medium to high outputs the composite driver delivers lower distortion.

I had read that at low plate currents the sample variability of the tubes would become higher. I substituted a NOS RCA for the tung Sol I had done the testing on and got the same result (with an even lower plate voltage of 37V!!

I did not write down the readings and the amp is currently in pieces waiting final assembly, but the THD measurements were Z-Weighted at 1kHz, and the difference was something like -40dB THD at 10mA plate current to -50dB at 3.3mA.

The complete amp with GNFB in place measures very well to extremely high outputs, but that is another subject and another thread.

Is conventional wisdom just wrong?
 
I would like to just comment your test procedure.
If you are designing an amplifier with GNFB and 120 Vp-p as required drive level.
Then you should ensure that your voltage amplifier and LTP combination is able to supply up to some 140 Vp-p with low THD.
It is not essential how low the THD is at 50 % voltage levels if the clipping begins at 90 % level.

It is not any surpise to me that very good linearity is achieved at low anode current when the load impedance is high like in your case.
 
The 120V P/P is the minimum for full drive of the output stage, I do understand that there should be a lot of headroom in the driver. In fact at 3.3mA current in the first stage the driver overall clips at about 223V P/P so no problems there.

I was testing at 50% of output simply to get some idea of typical THD, and to optimise the driver operating conditions.

I've put this on the board because the linearity at low plate voltages is amazing, in the context of the 'accepted wisdom' for currents in 6SN7 gain stages. I had not really thought about it until over the last weekend I built the two 'production' driver PCBs from the cicuit diagrams and looked at the typical voltages on them.

A CCS giving as an infinite plate resistor seems to work wonders.
 
Its bad practice to design around harmonic cancellation because the cancellation will degrade as the triodes age differently. It is also fairly unpredictable over the whole frequency range and under real signals. Far better to operate the triodes in their most linear region and minimise the distortion this way. If that means running higher currents - then run higher currents.

Shoog
 
This simulated well and measured OK, and this plate current or higher is the 'conventional wisdom'... Is conventional wisdom just wrong?

Yes, it is wrong. Take a look at this loadline (attached). This is for the 6J5, a singleton triode very much like one section of the 6SN7. There is no reason why this type can't be used as a "current trickler" if that's what your application requires.
 

Attachments

  • 6J5-Loadline-100K.png
    6J5-Loadline-100K.png
    45.1 KB · Views: 511
^^

Quite trustworthy, I'd say. I designed against an RCA spec sheet, and used Westinghouse JAN 6J5s. The DC voltages and currents were spot on, as was the AC gain. If there is any problem with this, it's the very large rp and RP that compromise wideband performance. The -3.0dbv frequency was 45KHz measured. It's trivial to get a higher cutoff frequency with any transistor circuit, even if you assemble it on a solderless prototyping board without paying attention to lead dress (lots of Cstray) So I wouldn't recommend it if you're after wideband performance. However, for the intended application, it was NBD, as the OPTs cutoff at 30KHz (open loop) with a 35KHz cutoff under NFB (local and global).
 
OK, so I will run up the driver board this weekend, measure the THD directly out of the driver board at various currents in the first stage. I was measuring the THD at the amplifier output... but I certainly was not designing for distortion cancellation, I'm not that clever!

On a lighter note the output KT88s glow blue on the glass for B+ above 500V. My science head has this down as harmless phosphorescence caused by impurities in the glass. Sadly KT120s did not glow when I tried them under the same conditions .. I am very disappointed.. I loved the glow ..
 
Also to reply to Miles Power, I have no problem with bandwidth. the whole driver is measured-3dB at 150kHz with 3.3mA in the first stage. Increased currents will increase the first stage bandwidth but I am rolling it off anyway prior to the LTP to stabilise the GNFB loop and avoid slew rate problems as the rest of the driver has to push big volts (push pull KT120s with 520V B+).

I have looked at plate curves and never really understood the common statements about linearity at higher plate currents, more than anything I ended up thinking that an accurate simulation might show considerable differences in relatively small amounts but that unless huge amounts of distortion were at issue caused by very bad operating points then there was nothing to see for reasonably linear tubes.
 
@shoog Now that we have low cost microcontrolers that can control parameters in realtime couldn't we start working with distortion cancellation again? By having a microcontroler control parameters such as bias we could keep aging tubes in the distortion cancelation envelope for a much longer time.

I would not be surprised to find out that some of the old tube audio rules of thumb don't hold up. Passive components and PSU designs have improved quite a bit as well as have the testing tools available. I have wondered for a while what would come about from someone using NwAvGuys (odac and o2) techniques for designing a tube amp.
 
I seem to recall that some of the RCA (and Mullard) recommended circuits often used very low anode currents, far lower than 'everybody knows' will work. The manufacturer ought to know how to get the best out of their products.

However, these low current designs will need a low output load (i.e. high resistance, low capacitance). High current designs will work better for the type of 'designer' who just throws together circuits he has found (or heard about from a man in a pub) without thinking about interfacing between stages.
 
I have just retested everything. I have not got time to post loads of numbers but the conclusions I put in my first post are valid, there's nothing funny going on, most of the distortion is in the LTP even running at 40V RMS out. Look at the thread 'Parallel Push Pull KT88' for the driver circuit, in this circuit the 6SN7 performs beautifully at low plate currents and voltages.

My advice to people, don't take advice like '8mA is a good current to run tube X at', Understand the application and what you are trying to achieve.

I was trying to achieve a DC coupled driver, hence I was helped by a low plate voltage in the first stage for maximum headroom in the LTP. Even with the low first stage currents the driver has bags of bandwidth and has to be slugged for a stable GNFB loop, and I am using very good output transformers.
 
I take the comments that have been made as useful input, but surely in the vast universe of valves there are better candidates to run at low current. This offers the option to actually choose an optimized loadline with flexibility. It also means the design can be more efficient.

Xpersephone - what made you select the 6SN7 in the first place ?

jZaptopa, I am certain that this could be made to work, but at what cost in extra complexity. Surely its better to make a better design in the first place where there is no need for distortion cancellation. Again the huge range of valves offers us plenty of options in this regard, and SPICE can allow us to make fairly accurate predictions before we finalize our designs.

Shoog
 
Last edited:
Shoog, I wanted to use current production tubes, and I also rather liked the idea of an all-octal build. I had decided on a simple topology for the amp overall, as it is a high-powered design with regulated power supplies and was already headed for some complexity. 'Doing the budget' for gain required, on the basis of input sensitivity, amount of GNFB likely to be operated and the output tube targets suggested a medium mu triode, and the 6SN7 is a good example.

I'm glad it works rather well! What do you think would work better, I would be genuinely grateful for suggestions?
 
Small signal octal leaves you with very few options really. I have made such esthetic choices in the past myself (an octal dual pentode) but the performance suffered as a consequence and I probably would do it differently again.

I would consider something like a triode strapped EF86, extremely linear over a wide range of operating conditions.
How about the 6SL7;

http://www.wooaudio.com/docs/tube_data/6SL7.pdf

Shoog
 
Last edited:
6SL7 was tried, too much gain. I am running 16dB GNFB for about 1.6V RMS driving the amp to clipping with the all 6SN7 design.

As it is the amp gives about 0.2% THD at 100W out (clipping at 145W) with GNFB in place, and about 3% at 100W without it, so it is respectably linear and I don't want more gain, not even wishing to start a thread about the evils of GNFB!
 
6SL7 was tried, too much gain. I am running 16dB GNFB for about 1.6V RMS driving the amp to clipping with the all 6SN7 design.

As it is the amp gives about 0.2% THD at 100W out (clipping at 145W) with GNFB in place, and about 3% at 100W without it, so it is respectably linear and I don't want more gain, not even wishing to start a thread about the evils of GNFB!

You are getting very good results.
3.3 mA is not so low if the valve doesn't have to swing high voltages into a favorable load. The 12SX7GT was a 12SN7GT selected for low voltage and low current applications in equipment with dry cell battery supply. It is not surprising that you have some 6SN7's that work well.
Some harmonic cancellation is beneficial too, if this applies and is not abused. The arguments about aging do not apply that much here as you might need to burn 2-3 sets of output valves to see some variation in the driver performance....
 
Last edited:
jZaptopa, I am certain that this could be made to work, but at what cost in extra complexity. Surely its better to make a better design in the first place where there is no need for distortion cancellation. Again the huge range of valves offers us plenty of options in this regard, and SPICE can allow us to make fairly accurate predictions before we finalize our designs.

Shoog

Sorry if this is getting to off topic.

The added complexity of an arduino and a few components would be minimal and the circuitry would actually not be part of the audio circuit. The complexity of programming the thing is what keeps me from trying. In theory you just need a way to measure the bias voltage and change the bias voltage. What I don't know is if bias is the only factor for distortion cancellation designs or if aging tubes present other hurdles.

I may have to try to build a module to do this and see if I can find someone to program it for me one day.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.