Testing parallel DHTs
After listening to a great incarnation of the 4P1L PSE in filament bias output stage from Andy Evans, I decided to have a look at the impact of unmatched pairs of triodes from a distortion point of view. Main reason was that when listening to Andy’s amplifier I noticed a bit of an uncomfortable treble with some strings. Perhaps the increase of odd harmonics, but wanted at least to see what was all about.
4P1L are very easy to match. you can easily get a pair with equal mu. Just randomly I picked from my collection a pair of valves with a difference of 0.5 in mu.:
THD is about 0.03% mainly driven by H2. It happened that one 4P1L from the pair had 0.02% where the other had nearly 0.04% distortion. The difference between H3 and H2 is about 8dB.
Then looked at a more closely matched pair (0.03 mu difference). The distortion wasn’t surprisingly different:
Again, nearly 0.03% and difference between H2 and H3 is down to 7.5dB.
Looking at the individual performance of the 4P1L, now biased at 30mA and similar anode voltage, we can see that despite having a lower THD, the difference between harmonics is just 5dB. This is the THD of the other 4P1L from the pair:
Well, how rthis compares to a 2a3/6C4C? The latter valves are two triodes physically connected in parallel inside the same envelope. So, no matching can be done:
The previous was a low distortion 6C4C I have. Distortion is higher than 4P1L PSE, but not that much. H3 – H2 difference is about 12dB.
My early thoughts:
- 4P1L are very easy to match
- 4P1L PSE performs really well. Distortion of the pair is lower than a 6C4C performing at same level.
- H3 component is higher in PSE and this could be the reason why is more noticeable when listening to strings – as I proved in practice.
Ale, not correct to compare the optimum conditions with lamps 4P1L passport regime 6S4S.
Svetlana
Svetlana, not sure if I understood correctly your comment. Only reason to compare both in this context is PSE and one 6C4C as two triodes wired in parallel. Operating point was chosen based on a similar scenario, same power handling.
4P1L cost is significantly lower and also really easy to operate in filament bias, whereas the 6C4C isn’t with the bias voltage required to deliver same power.
Ale
6C4C (or 2A3 types in general) is not made of two triodes in parallel. To be precise there is one plate, one filament and two grids in parallel.
Cheers,
45
Hi,
My understanding was that from its geometry the anode and the two grids behave as two triodes in parallel. Please correct me if wrong here. If you have pictures or info, do share!
Thanks
Ale
The H shape for the plate can be misleading because it is one plate effectively. The filament is one, it is a an upside-down W shape. To be honest the older 2A3 with a single flat plate, like the 45, wasn’t any better in terms of linearity. It was more fragile, though. I have built myself a 2A3 PSE (schematics similar to the Audio Note Neiro with some important differences) and didn’t notice any substantial difference in terms of THD and sound character in comparison to the SE. It worked actually better with its double Pout.
I have bought some 4P1L’s myself because this valve really looks good but I think the 45 is the tube to be compared to.
Cheers,
45
P.S.
If you think the 6C4C filaments make a difference in comparsion to the old 2A3 flat plate (which has the real W filament), you could also see a difference in the 4P1L when going from 4.2V to 2.1V. With DC supply I guess will hardly see (listen to) a difference, with AC 2.1V will be generally better.
In my opinion all the important differences are due to the grid(s). One one side you have the old design with a grid where Gm has to 5.2 ma/V (in thrut I have only meaqsured 4 mA/V at 250 V/60 mA) on the other side you have two gris in parallel which contribute for half of that value. In pracitce you really get at least 4.7 mA/V. Having two grids means the spacing is less critical for a given manufacturing accuracy. So the differences you have from two separate grids can well be balanced out by the fact the having one grid requires more accuracy and precision. In fact the older design was not better. If one measues the THD as given in the manual he will get a typical 5% from the old design (mainly H2 and very littele H3), i.e. just as stated. Instead one will get an average 3% from a “double-plate” (H3 a bit higher but overall THD still dominated by H2). With a “double-plate” and a proper driver one will get 5% at around 5W output! The old single plate is still able to give about 5W but THD will go up to 8-10%….!!! This is my personal experience. I just sold them because the 45 is better.
Cheers,
45
Good insight, thanks.
Question then, if two grids in parallel share same filament and same anode it’s effectively two triodes in parallel. Now, the construction and shape is improving the matching and accuracy of those two grids, but yet, those two triodes are not 100% matched, correct?
Regarding 45, I completely agree. It’s my current SE amp. I wish it could provide a tad more of output power, that’s all. I will not ask anything more out of this perfect valve. Have you tried 45 in PP? or 45 in PSE?
Thanks
Ale
Yes, they are never 100% matched but the point is we are comparing two triodes in parallel each with 2.6 mA/V mutual conductance with one which has double. The higher the mutual conductance the more critical is the accuracy you need in the construction. Best linearity is always achieved by selecting the valves and a fraction of additional THD won’t make a difference. Talking about the 6C4C I think it is a superior design respect to most American variants. It has proper filament suspensions, not just hanging them on the mica discs, which give better geometry and make them safer. In fact you can apply up to 360V plate voltage.
With the 45 I have an open PP project. The SE can be nice but the power is low and I prefer (a lot) more economical solutions for this. I am just using one….
I don’t have too much time for it now but I will surely finish this.
Have a look at AWV datasheet and you will see that, with a proper driver, you can get 15-17W in Class AB2! http://frank.pocnet.net/sheets/084/4/45.pdf
This might seem heresy but if you look carefully in some of the proposed solutions idle conditions and the OPT’s are identical to the class A amplifiers. The difference is in the driver. I am using a 5K OPT instead of 3.9K (at 275V/36 mA). I like to think I am building a 6W Class A2 amplifier with THD =1% but it can double the power, if needed, without clipping (THD around 3% at 12W)!
I have just bought some 4P1L’s to use them as transformer-coupled drivers, in place of the present 6N30P-DR’s.
Cheers,
45
Very interesting. I’m currently building a 4-65a SE in class A2. I thought about using a similar driver (e.g. 4P1L filament bias and gyrator load in mu-follower) to driver a 45 in A2. Will the 45 be ok for this? or for grid current operation AB2 is the only route to go?
Thanks
Ale
To be honest I don’t know Ale. The solutions I have always used for Class A2 (and AB2) are transformer-coupled small power valves (> 3W) or direct coupled cathode followers with resistive load. Both are enough to drive 45 and 2A3. The former is good enough even bigger triodes like 300B and 211/845. The only thing you need to match is the right driver tube. It depends on how deep into positive grid you want to go and how much the output tube grid demands as they differ from one to another..
Another solution which is an improvement in comparison to the cathode follower is the Mosfet source follower which should also work fine with a 211!
Cheers,
45
The MOSFET gyrator in mu follower is unbeatable as a driver for A2. My question was specific to the 45, working in A2. The valve wasn’t design for this so not sure how much current and grid power can handle?
Thanks for the input!!
Ale
The 45 can surely work in class A2 (i.e. negative bias and driven into positive grid beyond Vg =0 which is not a saturation point for this device, it demands power though). Up to a certain point is Class A1 (usually until the grid is still marginally negative respect to the cathode) and then into Class A2 or straight into AB2, depending on the plate current with signal respect to the quiescent current. In other words you provide a peak signal which is larger than the bias voltage. Given maximum or near maximum operative conditions, as typical with power stages, then the way to tune it into A2 or AB2 is the anode load. The higher the anode load the more it would only be Class A2.
This has nothing to do with positive grid bias which is by definition “2” class of course.
Cheers,
45
Thanks! So for A2 a load of 5k will push it further into A2 providing bias is adjusted and driving signal is provided. Seen details on input power and peak grid currents but Is there any maximum that needs to be taken into account? I usually bias 45 in 34mA and 300V or around that point. Where will you start looking at moving the quiescent point to for A2, a more positive Vgk and lower anode current with a higher anode impendance?
P.S. The power required to the driver and DC grid currents are reported in the AB2 examples in the datasheet link above. It is of the order of 0.5-1W.
A2 will start when Vgk is still slightly neative, generally. If you don’t want AB2 then leave it at 300V/34 ma and use 7-8K OPT but Pout will be lower.
However I would not be worried and use 5-6K max. Consider another very important thing about definitions. As you can see from the datasheet, RCA used to define Class A1 up to the point where the output current equals the quiescent current. Texbook definition is: peak output current equals quiescent current, basically less Pout. This is because the 45 is so linear that you are not going to see any x-over distortion when biased with sufficient idle current, say 20 mA or more and other things well implemented….this is not a transistor or a pentode!! 🙂
Cheers
Hey 45,
Started a new entry to discuss this in more detail:
http://www.bartola.co.uk/valves/2013/01/26/driving-the-45-dht-in-a2/
I have just seen the 6J5GT input stage with CCS load for your 45 SET on Diyaudio. It looks good but I don’t think you will get that low distortion not only at 100 V peak-to-peak output but also at 88 V paek-to-peak when you drive the 45 grid biased at -44V. RC coupling doesn’t work very well in reality, as I told grid current can start, and usually this is true for most DHT’s, when the grid is still sligthly negative respect to the cathode.
If you want to avoid this and actually also drive the 45 for additional 10-15V into positive grid you can use a cathode follower.
I would add another 6J5 in cathode follower so that the voltage amplifier will see high impedance in any condition from the input of the cathode follower . The CF then is DC-coupled to the 45. If you provide +/-170V to the CF using 15K cathode resistor you can adjust the bias so that its cathode is at about -44V you need for the 45. Adjusting for -44V on the cathode of the CF means that the CF itself will run at about 214V/ 8.4 mA.
For adjusting the bias of the CF you can create a resistive path on the negative rail which only takes about 1 mA. Having -170V you can make a series of resistors for a total of about 170K. From the datasheet of the 6J5/6SN7 your bias for 214V/8.4 mA would be around -6.5/-7V. This means that in your case it will -44 -7 = -52V. For example, going from -170V to ground, you can have 47000K + 10K trimmer (multi-turn is mandatory for fine adjustment) +110K (2×220 in parallel). This way you can trim the bias for the CF from -47.8 V to -58V. So your -52V will be more o less in the middle. If you start with the trimmer set so that you have -58V on the cathode then the 45 will be safe as well.
This is basically the Audio Note solution for biasing DHT’s. It has been used by Kondo in all his best (Japanese) amplifiers. There is a difference though. Kondo’s cathode followers usually run at lower currents because he was not really interested in full power drive but just avoid distortion caused by the onset of grid current. The A2 swing capability was limited to few volts only.
Of course you can do the same in a PP!
Cheers
Yes, that original design with CCS and 6J5 is not optimal. I’m currently using a gyrator in mu-follower which provides a great performance with low output impedance. Only disadvantage compared to cathode follower is that current capability is limited to 6j5 biased at 7mA.
Use that gyrator then but things don’t change if you want power drive. There is no way you can go into A2 with RC coupling and get best results. Low output impedance is only half of the story. You need direct coupling and possibly no resistors in the cathode path of the output valve. The other option is transformer coupling.
Capacitor coupling gives rise to another type of distortion at onset of the grid current. This is called blocking distortion and, in few words, means that the grid current will charge and discharge the coupling cap and thus the signal it can pass is limited and distorted. Having a cathode resistor will cause modulation of the bias point because even the most linear device is not perfect and under modulation the anode current will not be constant causing thus a chance in the voltage drop across that resistor.
Cheers
Indeed, the MOSFET Source follower is an option as well. Have a look at my last two posts where I explore the A2 operating point and the DC coupled circuit. Keen to get your feedback
Thanks
Ale
Approximate relationship of Ra Ia when power dissipation at the anode 6S4S to 15W, with a minimum of third harmonic:
Ia = 65mA, Uak = 225V, Ra = 2,4 k, Pout = 2,2 W, Rout = 2,44 Ohm, Y2 =-29dB, Y3 =-77dB
Ia = 60mA, Uak = 240V, Ra = 3k, Pout = 2,5 W, Rout = 2,2 Ohm, Y2 =-29dB, Y3 =-78dB
Ia = 55mA, Uak = 260V, Ra = 3,6 k, Pout = 2,8 W, Rout = 2 Ohm, Y2 =-29dB, Y3 =-78dB
Ia = 50mA, Uak = 285V, Ra = 4,8 k, Pout = 3,1 W, Rout = 1,8 Ohm, Y2 =-29dB, Y3 =-94dB
Ia = 45mA, Uak = 320V, Ra = 6,2 k, Pout = 3,5 W, Rout = 1,65 Ohm, Y2 =-29dB, Y3 =-92dB
Ia = 40mA, Uak = 360V, Ra = 8k, Pout = 3,9 W, Rout = 1,5 Ohm, Y2 =-29dB, Y3 =-98dB
Right. 4P1L PSE have two 1k6 ri in parallel, so Ri equivalent should be 800 ohms. I think Ri is the same as 2A3. If load is 2*ri at least then pa=(Vak^2)/(16*ri). If same Vak is applied clearly we get same power? Will be about 4.88W for the 4P1L PSE and 6C4C…
Ale
Two lamps never better than one as two resistors and everything else. Parallel path is not good.
Why is not good? Soundwise? We did a listening test recently, not same amps. One was 4P1L PSE with filament bias and the other 300B SE with cathode bypass cap (unfortunately. The PSE sounded really nice to me…
You are right! You might also try with a PP. 🙂
This is a very good test! and very valuable information for parallel 4P1L 🙂
I plan to make a PSE on the 4P1L also. What Driver stage (name valve) do you plan (or you advise) to use. Thanks.
Why not 4P1L as driver as well, you can go DHT 100% that way 🙂
If not you can look at 6e5P-triode mode, great one.
I’m very tempted in going for the DC coupled version of Torsten’s Monkey:
Ale
I’m coming back to my 2012 PSE design described here and building another one. This time instead of one supply to the PSE outputs I’ll try two supplies – one to each 4P1L. That way I can play around with matching tubes and see if it makes any significant difference. Not quite sure of the topology – in the middle of putting it together. I’ll use the same LL1660 interstage and filament bias right through, like the original design. This time I have an O’Netics OPT. Interested to hear that. I also have a better DAC now, battery driven, so the source is cleaner. Andy
Hi Andy, not sure if I understood what you’re planning to do. Are you considering one filament supply per 4P1L? Why not spending the effort in matching pairs on mu or alternatively mixing filament bias with a tad of fix bias using a FET follower? A bit more complex though. Options with TX coupling are a bit more restricted with the PSE configuration
Unfortunately I’m putting away the 814 soon as Sofia will start moving around and the whole system will be a hazard. Hope to catch up with you soon, it’s been crazy these months 🙁
cheers
Ale