Goodbye 4-65a SE, at least for now
After enjoying the 4-65a SE amplifier for many months, I couldn’t resist myself from upgrading the output stage to the 814s. I just needed changing sockets and filament raw supply transformers to fit the requirements of this lovely transmitting valve. Needless to say, my recent tests on 814s were very encouraging. The 814 seemed to perform much better than the 4-65a in delivering 10W of class A2 sound at half the distortion levels. This to me, was only worth trying.
HT supply levels are closely similar to my previous 4-65 design so biasing could be easily adjusted given the DC coupling capability of this amplifier. At least I can take advantage of an additional benefit from this monster amplifier!
Last week I prepared the UX5 sockets with the screen and suppressor grid resistor and protection diodes as a single module. It didn’t take me too long to fit them in the amplifier.
Adjusting the Rod Coleman filament regulators wasn’t a difficult task either as you would expect. 3.5A regulators could easily deliver the 3.25A needed by these filaments. After 30min of burning them in, I did minor adjustments to set the current back to the right levels: 3.25A @ 10V.
I found my stash of 814/VT-154 to be extremely consistent. I looked for a useful operating point to compare valves in triode mode. I used the following as a reference using my uTracer:
- Va=250V
- Vgk=-5V
- Ia=66mA
- Gm=4.5mA/V
- Ra=1.8KΩ
- μ=8
Not representative of the normal operating point but good enough to me to compare valves. Ra will normally vary from 1.4KΩ to 2KΩ across the loadline but mainly is around 1.6KΩ.
My 814s run at about 100mA/540V and needs about 120Vpp to deliver 10W @ THD=2.3%. Clearly they could run at 600V or higher to deliver greater power in A1 but I’d need a different power supply. So that is for a later upgrade.
THD looks very promising when operating in class A2 providing the driver is capable of handling the low impedance of the grid and delivering the grid current required. Whilst running on class A1, this stage will only deliver 3.3W@1.4%. It’s obvious we are not squeezing the most out of this valve in A1 so a higher HT supply would be needed for more power. Interesting to see that if we bias instead higher up at around 650V/92mA, at this bias point the stage will deliver about 6.7W@2.7% in class A1 or 10W@3.3%/15W@4%. The effect of curve crunching at higher voltage is more evident here.
The higher mu of the 814 in triode mode is less demanding in terms of drive. To drive the 814 at full tilt I need 120Vpp, about 40Vpp less than what was needed by the 4-65a. This is not a problem for the 46 driver as it was design to deliver 200Vpp at less than 0.2% THD.
My fear was clearly whether the 814 triode strapped could run at +540V or not given the screen voltage limitations. Some people have tested other valves in this mode with success inspired by Morgan Jones in his “Crystal Palace” design (Page 461 of “Valve Amplifiers” 3rd Edition). However the proof is in the pudding. It was all yet to be seen when the 600V HT was fired up.
814 SE Amplifier Circuit
The amplifier circuit is exactly the same as the one I used for the 4-65a. The driver is an onerous implementation of the 46 (triode-strapped) in filament bias with a FET gyrator load in mu-follower mode to deliver sufficient current and low output impedance to handle A2 needs. The great concept implemented by Michael Koster in this design is that the grid current has an independent loop without disturbing the driver and output stage bias as the grid current only goes through V5, M2 and the output valve. The MOSFET provides a high PSRR so a simple supply can be implemented for V5. The price paid for DC coupling and flexibility is the number of power supplies. On this 100% DHT implementation a total of 7 supplies are needed! A simpler design could instead include a nice pentode driver like 6e5P, D3a, 6J52P, C3g or even 4P1L. I’m a big fan of DHT and wanted to experiment with this amp. The LL7903 is a superb step up transformer which in conjunction with the 46 driver provides sufficient gain to the driver stage. It has been already pointed out that C3 can be as low as 100nF with R4 being 4.7M. I just haven’t had the time to change it 🙂
To prevent the 46 to be driven at lower current due to the lower bias requirement of the 814 compared to the 4-65a, I lowered the V4 SSHV2 output voltage to 265V. The 46 is now running at about 205V / 36mA.
After creating a new Spice model for the 814 to improve A2 grid current simulation, I run in LTSpice a THD analysis versus output power:
We can see in the simulation that the THD at 10W is just about the same to the graphical analysis from the curves. However, H2 is lower and H3 is higher. Some even cancellation of harmonics is taking place as expected but also the H3 contribution of the 46 driver and the output stage are clearly there.
I normally listen to this amplifier at about 3-5W. The fingerprint is clearly H2 but would be interesting to listen the H3 contribution in this amplifier.
Initial tests
I had a bad experience with this valve in the past. I blew my curve tracer when did first tests 18 months ago. Paul LeClerq encouraged me to try the 814 in an amplifier, so I had to do this! When finished all basic individual tests, I placed a pair of US ARMY Signal Corps VT-154 made by General Electric in pristine NOS condition. After reseting the HT levels as needed, I fired up the 600V supply with my VARIAC. Everything worked well but after some playing time one valve started to drop gradually the anode current. I suspected biasing needed adjustment but quickly realised that the valve wasn’t performing correctly. Before anything unwanted happened, I switch the amp off and replaced the valve to check whether it was the valve or not. It was. I played the amplifier for about 2 hours last night.
After some initial burn in time, and as Morgan wrote in his book, the fact that I poured myself a glass of 10 year old Aberlour, didn’t influence my perception that the sound gradually improved after the initial hour. The solid bass was still there and greater detail in the sound. A less sweeter sound given lower H2 component was noticeable but without lossing the warmth. I just like this valve, but need further listening to seriously review this modification.
Glad to see that the 814 performed sturdily at +540V in triode. The case has been proved that yet another transmitting tetrode can deliver outstanding triode performance and sound in an amplifier.
Stay tuned…
Ale
Inspiring work like always Ale. Thanks for keeping up with posting your results. I can’t find the time to do it.
Have a question about the crow bar protection seen on your simulation. What is the influence through the sensing resistor on the sound ? I am thinking to use one for my Lundahl OTs. Good idea to protect on investment.
Best,
Radu
Thanks Radu for your comments.
Regarding the crowbar question, I can’t notice any difference. Hard to say as haven’t tried running the amp without. It’s a very small resistor and cap. The SCR won’t impact unless it is triggered which will only be under an abnormal situation. I wouldn’t implement it unless you are working on a fixed bias and with very high currents and voltages.
Hi Ale, if you get that distortion spectrum from the real amp as well then the 814 would be quite perfect for a 25W class A PP where you will get rid of that 2nd harmonic easily at any level just because of the PP features. I would expect something like 1% THD at full power….
I am not convinced about a step-up transformer at the input. You need a really good preamp to cope with the high capacitive load.
Yes, 814 in PP would be very interesting albeit the heating requirements. I would try to measure the actual harmonic distribution when I get the chance.
I only implemented the step up since I have the 25 preamp in a configuration that provides output impedance of 500 ohms and is good to drive this load. So far, I haven’t noticed any sound degradation. Treble is very good so can’t hear any slew rate issues. I wanted to avoid a third triode stage and was obsessed with the 46 in this place. In a more sensible design, I’d gone for a pentode driver without a doubt. Thanks for your comments which always help me a lot!
Hi 45,
You would be surprised to see that in PP mode can deliver about 29W @ 0.05% according to the simulations. The anode load is 8K, biased at 700V, Ia=85mA, Vg=-55.5V and input signal is 120Vpp. It could get up to 50W in A2 with THD at 0.14% if driven with 160Vpp.
I suppose that driver fingerprint will be more evident on top of the harmonic contribution of the output stage here.
Nice push-pull project though. I can imagine that the bass would be outstanding in this configuration if mine in single-ended is brilliant!
If it weren’t for the filament requirements, I would be tempted to do it. I had enough with this project. The amplifier is dead quiet. I could never imagined I was going to be able to achieve such level of low noise given the filament requirements of this amplifier. But with the level of filtering and using the Rod Coleman regulators, it’s impressive to hear the results.
For a push-pull stage a single Coleman regulator could be used with the filaments in series and centre filament (cathode) grounded if using fixed-bias. This will make the LT supply to be as high as 28V @ 3.25A.
I wish I could have all the time to test these ideas!
Cheers,
Ale
Ale,
in general I prefer an integrated amplifier and find that often people make preamp + mono power amps without a real need. This is one of those cases where monoblocks are a must. At 50W output it is the same as the 845. However with 700V supply and those filament requirements to get 50 watts from the 845 you would need to enter A2 quite heavily. If you then consider the price, including Chinese 845’s…..
Cheers,
Paolo
Hi Paolo,
In my experience, I only managed to make the amplifiers dead quiet when supply chassis is separated. With regards to integrated amplifiers, I rather have a bit of extra flexibility, only because I like playing and changing the setup constantly. Of course, as a definitive setup, I’d rather go for and integrated design.
I have some 813s and GM-70s in my valve stock. I read very good reviews, but after this experience with transmitting valves, I think I would stay away from them for a while. I think is time for a bit of push pull now. I will likely pick up again my previous ideas of 6C4C or 4P1L in PPP. Have you built any of these? What is your experience?
Thanks, Ale
Well, the American 813 is one valve I like a lot but frankly is too big and better used for higher power, if one needs it. The GM70, although it is ljust linear as the best DHT’s is a waste of power. It is too demanding in terms of filament supply. I have sent you an email….
Hi,
What is the maximum dc current plate 814 in triode?
Piotr
Hi Piotr,
Have a look at the datasheet. I run it at 60W Pa maximum. Anode current will depend on your chosen operating point. I’m using it at 100mA. However higher anode voltage and lower current (e.g. 90mA and 650V) works quite nicely with this loadline.
Ale
Hi Ale,
Thank you, but the datasheet is 60ma max dc current plate.
Piotr
Piotr,
DC current is stated in the datasheet for ICAS and CCS for class B and C operations with current ranging from 60-150mA. Given class A is not listed, it’s hard to say what is the maximum DC current allowed so I always stayed within the maximum dissipation limits for anode and screen. In fact, there are no triode-connected specs at all and I found out in practice that you can exceed the maximum screen voltage given it is connected to anode for a triode-like behaviour.
Ale
Hi Ale,
Thank you. 🙂
Piotr
I am planning to build a push-pull- no compromise-mid-life-crysis amp…Should have aboit 20-30W: Candidates are tubes I have already which are: 300B, 801A, 814A, eml520v2…How would you characterize the 814A against the usual suspects like 300b/eml ?
Ok, I got a very nice stock of 814a…and want to try a push-pull class a…so you think 8k raa is enough ? or better 12k ?
Hi Frank. Embarking on a project like this is a big decision in my view. The valves should not be the excuse. The iron is what it will make a difference.
The 814 in PP will perform greatly at just 8K. Distortion will be lower than 0.05% for 30W! You will only need 40Vrms to drive it to full output in class A. Much easier to handle than the equivalent 300B thanks to the Mu of the 814 in triode mode.
If you want to go for 11k, you will get less output and slightly better distortion which is already really low. What are your speakers? What about the driver?
I have big line sources which consist per channel of 12 Fountek Fp-2 Ribbons and 18 Eton Kevlar chassis 4″. In theory about 95db, but in practise this means 12 input transformer (ribbons) to drive and 18 coils…I tried a 20W SE power amp, but could not cut it.
So, PP…
On the driver-site…well…I would love to use my 801A to drive the 814, but not sure if this will be enough gain though…my dac-line stage delivers 8Vpk-pk
So, the ideal setup would be:
-814 as oitput tube
-6k6-PP-OPT with 4 ohm-tap connected to my 6ohm speaker (its handy that Monolith has such piece in amorphous)
-Interstage 1:1-PP from Monolith which I have on the shelf, which in theory should give me the ability to drive even A2
-801A running with 30-40ma into the IT, LTP with CCS in the tail.
Only if really necessary: DC-Coupled input stage (anode chokes, LTP like Kevin Carter does)
What do you think ? Nice and simple, fully differential.
I think you will need a gain of 15-20 minimum. I don’t see the 10Y to be ideally coupled with an IT given the high ra. I’d rather use a gyrator, but that’s me. If you want to use 10Y you may want to consider a 1:2 or 1:4 step up transformer for large signal at the input of the 10Y.
There is a nice EML-20 valve which is the only DHT that can do a gain of 20. I’d expect it to be very linear, however I haven’t tested or seen measurements to confirm that.
If you DAC has differential outputs then you you can explore an LTP. I’d do it with gyrators and CCS tail. Much better performance and sound Than the anode chokes.
I’d also look into other great performing drivers such as 4P1L, 6J49P-DR, 12HL7, 6e5p/6e6p. Have a look at the tests from here:
http://www.bartola.co.uk/valves/2016/11/27/6э5п-shunt-cascode-driver/
I’d ask you to send me an email if you would like to proceed with this discussion as there are about 150 people subscribed to this blog and each post will be broadcasted to all. not sure if everyone wants to follow up on this particular conversation.
Many thanks
Ale
I just have re-read your article and you made some comments to be cautious about how high the actual anode voltage may be in triode mode given the g2-screen limit of the data sheet of 400V I guess…?
Given your pratical experience with this tube, do you think that 700V like suggested above for my upcoming push-pull-project will work ? How big should the resistor be ?
Best Regards
I haven’t tested it beyond 600V. It could arc easily. If you want to go further you can explore anode to grid feedback when ran in pentode mode.
Ale
What is the recommended output Transformer impedance for a Single 814 in A1 CLASS , with 600V, 70mA current ?