My parallel stereo single ended triode is now ready :)

[Jez1234]

I had one of Audio Note's cheapo ones in ("only" about £3K as a kit!) which was rated at IIRC 10W but had reached 16% THD by 8W.... It's owner thought it sounded marvellous...
I was expecting to hear the cricket results through it any second now!

[ms660]

Originally Posted by GrimJosef

The datasheet https://frank.pocnet.net/sheets/084/3/300B.pdf says on p2 that a single one at 7W has a THD of 5%, but on p5 there's a graph which covers what appears to be the same operating point (4k load, 350V on the anode, -74V grid bias) and it indicates 7W out and 2.5% THD.

The graph on page 5 reads as 5% THD to me for a 4k load.

Lawrence.

[GrimJosef]

Apologies. You're right of course. I was carelessly reading the left-hand y-axis scale rather than the right-hand one.

Cheers,

GJ

[thermionic]

Originally Posted by Jez1234

I had one of Audio Note's cheapo ones in ("only" about £3K as a kit!) which was rated at IIRC 10W but had reached 16% THD by 8W.... It's owner thought it sounded marvellous...
I was expecting to hear the cricket results through it any second now!

I had in for repair a couple of years ago, a Gamma Acoustics ’Rhythm’ amp using a pair of 211 Triodes as a stereo pair. The build quality was terrible, and considering the HT was 1200 VDC and the 211 triode Heaters each consumed over 3 amps it was a hot running amp.

The output transformers looked cheap - just like a bog standard mains transformer, and I had already written the thing off in my mind as a cottage industry, expensive folly.

I re built the power supply & re engineered the thing to be as safe as possible within the confines of the existing circuit, ready to be returned to my HiFi dealer customer. As with all repairs that landed on my bench, it gets listened to. WOW..! I was hooked! It sounded effortless and full, and I was reluctant to hand it back. Of all the many different amps I have had the pleasure of listening to, this was one of the best, IMHO. There was no NFB as such, and none from the secondary of the output transformer.

I did hastily jot a circuit down, and if anyone’s interested (Gabe)? I’ll try to find it.

Cheers, SimonT.

[GrimJosef]

I've worked on a couple of Gamma amps too - one Rhythm and one Reference. They needed the same treatment - serious tidying up of the high voltage and very high voltage wiring. The Reference also had some work done to get the hum under control as the owner had very sensitive speakers. I replaced the smoothing resistor between the reservoir and first smoothing capacitors with a choke, tuned using a parallel capacitor to enhance the reactance at 100Hz. I also did some work on the grounding.

Measurements showed that the hum and noise could be got down to 1-2mV RMS into 8ohms. The distortion at 1W out and 1kHz was 0.6-0.8% and was still less than 2% at 10W. Since it was nearly all second-harmonic it was 'sweet' to listen to.

In case anyone should ever need one I did commission Techno Transformers in Royal Wootton Bassett to replicate the amp's output transformer, to replace one which had some shorted turns. It's quite a challenge as they have to handle 10W+ and hold off all those primary-to-secondary volts while still being compact enough to fit into quite a small space under the chassis.

Cheers,

GJ

[Gabe001]

I'd very much like to see that circuit Simon.

I've got a question. The ecc82 in cascade, even with the gain block, is doing about 35db of gain (x55 voltage gain). Although there is no hum at all, there is a bit of hiss (white noise). It's very minimal - only audible up to about a foot from the speakers, depending on how you turn your head. The speakers are 89db @ 1w. I think the noise floor works out about -74 dB or thereabouts -its hard to be 100% accurate with my equipment. There's definately no oscillation on scope and the hiss on both channels is equal. The valves are tested and in good working order. I'm confident nothing is amiss.

A quick Google shows that this is probably normal for high(-ish) gain amps without feedback, and the ecc82 isn't one of the particularly renowned 'low noise' valves. All resistors are metal film and adequately rated, diodes are snubbed etc. Its equal on both sides and unaffected by volume. It disappears when the ecc82 valves are pulled.

I know I'm splitting hairs here, but I'm a perfectionist, and I don't exclude getting more sensitive speakers in future. Is this a limitation of valves used? My feedback amps are about 10db quieter, and it's noticeable. The hiss is by no means offensive, but I wonder if there is an easy fix?

My other option could be to reduce the gain block and run a tiny bit of feedback (maybe 6db) from the OPT to a voltage divider around the grid of the second half of the ecc82, which would also significantly extend the HF response, not cause any stability issues, and keep the input sensitivity and overall gain the same.

[kalee20]

It should be relatively easy to see where the hiss is coming from.

Ground your EL34 grids to chassis with shorting leads - I'll bet the hiss disappears. So you'll know it's not the output stage.

Remove, and ground the grids of the driver stage - you'll probably get the same result. So you'll know it's from the first stage.

If you ground the grids of that (with large capacitors! You don't want to upset DC conditions here!), you'll know it's grid noise (possibly from noisy resistors, maybe grid current shot noise though I doubt it). But I reckon that the hiss will stay, you're stuck with it and it's valve noise.

[Jez1234]

The one SET amp that I thought sounded good (I mentioned above that there was one..) was made by Longdog Audio FWIW.

[Diabolical Artificer]

Grid stopper can add Johnson noise - hiss, but if I remember right your not using big stoppers. It's possible to reduce the stopper value and pop in a small cap, something like 100p to 470p, see - http://www.valvewizard.co.uk/gridstopper.html

Andy.

[Gabe001]

Originally Posted by Diabolical Artificer

Grid stopper can add Johnson noise - hiss, but if I remember right your not using big stoppers. It's possible to reduce the stopper value and pop in a small cap, something like 100p to 470p, see - http://www.valvewizard.co.uk/gridstopper.html

Andy.

The grid stoppers are only 1k. The hiss goes without the ecc82.

I'll try your suggestion, but it's in the auditable spectrum.

I think the only way around this may be to get a nominal amount of feedback in, somewhere in the range of 6-10 dB,to lower the overall noise floor to -80db or thereabouts. This will also theoretically vastly extend the FR, and hopefully not cause any stability issues. Here is me optimistically hoping feedback will work for Johnson noise too!

Something like the snippet shown. Coupling cap values are very crude.

[Gabe001]

Originally Posted by Jez1234

The one SET amp that I thought sounded good (I mentioned above that there was one..) was made by Longdog Audio FWIW.

Just had a browse and they don't seem to have a valve one for sale. Was it a custom build?

If you were closer to me, once I'm 100% satisfied with this I'd be keen for you to have a listen.

[Jez1234]

Originally Posted by Gabe001

Just had a browse and they don't seem to have a valve one for sale. Was it a custom build?

If you were closer to me, once I'm 100% satisfied with this I'd be keen for you to have a listen.

I can't seem to find the specific unit I heard (which had what looked almost like DIN rail modules, several of, under the chassis and I presume were regulators of some sort. I only had a quick glance) but here's another SET from them and there are more at the bottom of their web site.

https://www.longdogaudio.com/valve-regulated-300b/

[Diabolical Artificer]

You might have an issue with locally applied FB, leading to instability and might not be needed anyway. FB is a two edged sword. If it were my amp I'd be looking for the cause of the issue rather than applying a broad fix.

I'd scope the ECC82 stages and try jumpering out those grid stoppers. I'm not a fan of grid stoppers as they reduce band width & together with the valves Miller capacitance just plonk another RC filter in the signal path.

Also, ECC82's with their low gain are well behaved and probably don't need grid stoppers, ECC81's & ECC88's might need them, but in my experience not 82's.

I'd also look at what signal source I'm feeding the amp with, another possible culprit.

Andy.

[Gabe001]

It's not a fault, it's just me wanting to hear silence with my ear to the speaker. My other single ended amp distorted more but the noise floor was -84db. I may get speakers with a sensitivity of 91db so it'll need to be quiet.

Jez thanks for the link.

[Gabe001]

Take 2

I wasn't intending to build a low distortion amplifier, but here we are, mostly by accident. I do not exclude getting sensitive speakers in future 93db or so - so I was aiming for the lowest possible noise floor. This, in addition to the the HF loss with the first version, prompted the second version of this amplifier. I completed one channel last week and the second channel today. It's been a fun journey, and I simulated dozens of circuits including a 12ax7 srpp and parallel triode.

Essentially I dropped the gain block and applied 8db of global negative feedback to the first stage - then compensated the HF response for the best square wave response. I also DC coupled the ecc82 cascade to minimise phase shift and components in the signal path. If I were starting again I would probably have used some posh audiophile valves like the e80cc, but I'm content with my long plate 12au7a.

Pdf attached

Anode/screen el34 voltage 445v
12au7 Vbb 400v
Grid and cathode of second half of 12au7a 102v/110v

HK voltage is well within safe limits. Diodes are there to prevent arcing on switch on and are out of circuit completely once the valve warms up

There don't seem to be any stability issues

Let me know if anything looks problematic. Meanwhile, some real life measurements in next post.

Not shown are 2 thermistors in the transformer primary, and I also added a 10w resistor to drop some volts as the transformer (rated 230v) was buzzing a bit when the mains voltage exceeded 245v

[Gabe001]

Real life measurements:

fR 20-20khz 0.5db
Max power approx 9.5w
Square waves 1khz and 10khz (for Jez) attached (at around 70% of full power)

Noise floor around -80dbv with signal shorted to ground (see attached), from REW with a 20db attenuator. The 250hz spike is an artefact

THD 1w 0.047% (yes , I couldn't believe it) see picture

[Gabe001]

And finally THD at 8w was 0.5%, rising rapidly afterwards to 2% at 10w.

It was a very interesting experience using LT spice to simulate various iterations of different circuits for distortion, frequency response and output impedance. The best performing was this one, followed by ecc83 in parallel, with plate to plate feedback. Interestingly, the cleanest signal at the el34 grid doesn't translate to lowest overall distortion!

[Jez1234]

Excellent stuff there! Very decent square waves 'n all

[trobbins]

The plots show a REW Measurement result with a frequency response. Typically I would undertake that form of measurement in REW using the RTA window, as that allows realtime display of harmonics and noisefloor for a single tone, as well as the ability to set up the FFT and window and averaging settings for better harmonic and noisefloor display, and to show an overlay of the Distortion panel with levels for all harmonics. Being realtime, a camera 'capture' allows the measurement to be recorded along with all harmonic data in the overlay Distortion panel, and the window/averaging data in the overlay Graph Controls panel.

With respect to distortion and the 12AU7, I measured a noticeable spread of 2nd harmonic between individual valves and their triode halves for a 10Vrms output in a Williamson driver circuit location. But the same measurement setup showed 6SN7 samples had significantly lower 2nd harmonic.

[Gabe001]

Thank you. Yes I use REW in the same way,but to upload the images here, and to retain then for future reference, I used the save function from the RTA window to capture the waveform and loaded this up, which is what you're seeing.

The noval 6sn7 equivalents (6CG7 and 6FQ7 if I'm not mistaken) would have probably been better choices THD wise but I don't feel like reworking the DC supply (which is on a PCB the underside of which is not easily accessible) given they're 0.6A filament valves, and the 12au7a is 0.3A. I didn't know they existed when I started this project