120w valve amplifier.

[Diabolical Artificer]

Started to build a 100w valve amplifier last year but put it away because although I managed to get a good clean 90w or so out of it, try as I might I couldn't get more. Why bother with 100w valve amp? Because there's a little kid inside me that likes big loud amplifier's is one reason the other one is because I managed to get a Hammond 1650TA 120w OPT off ebay cheap. I wanted to use 807's not because they are noted for their sonic
capabilities, but because I have a bit of thing about 807's. So from the start the wrong way to go about designing an amplifier, but what the heck I'm stubburn and it's good craic.

So back to this time last year I could have settled for 90w ish but the main reason for trying to get more watts is because the OP stage uses 6 807's, an expensive OPT not mention an expensive mains tfmr etc I know 100w isn't or 120w isn't going to make it that much louder than 90w a few dB at most. I used a pretty bog standard topology, EG 6SN7 common cathode gain stage front end, 6SN7 LTP PS and a 6SN7 cathode follower. I learnt a lot from building this it being the first valve amp I built.

Got it back out about 2 month's ago and try as I might, I couldn't get over 90w. In fact I was getting a lot less. I tried using a cascode LTP, an LTP using two 6SN7's their anodes and cathodes strapped and an LTP using two EL84 triode strapped. All this resulted in piles of datasheet's, loadlines and sheet's of calculations. In the end I was about to give up again when I thought about using a 12BH7 as the two cathode followers. This is a beefy triode I'd seen John Chambers use to drive his big multiple valve amps. After more experimentation I got a good 130w, albeit briefly; don't want to kill the OPT.

Using 6SN7's with the 12BH7 wouldn't look right so I used a ECC83 as the front end and a ECC83 as the LTP PS with a CCS instead of tail R.
Did a quick test today and the frequency response is pretty flat from 100hz to 30khz with a dummy load. Now comes the tricky part of tweaking things, applying NFB and making sure it's stable. Also need to build a chassis and build a power supply - I'll wind my own mains tfmr. I'll use a separate PSU to power two monoblocks, that way only one tfmr and one choke less caps etc. Obviously will need to build in some protection too.

So lot's more to do but am getting there. Here's a schematic. It has bits missing and a few mistakes but you should get the gist.

Andy.

[Murphyv310]

Hi.
In the circuit above I'd strongly suspect you are running Class "A" Push-Pull. The quiescent current will be high. I'd be inclined to either go for Fixed bias or higher HT and increase the values of the cathode resistors to get the amp into class AB or even "B".
The 10K cathode follower resistors could be raised to say 15k which could give a little more drive.
An interesting project but make sure you feed the leccy meter with plenty of shillings and turn off the heating in the house.
Good luck.

[Diabolical Artificer]

It's running in fixed bias Trevor, there's -35v next to the 807 grid resistor. 807's are set to 50mA Ia though on my load line they should be at -40/45v. I have a pot I use to set bias, -35v at the moment is about right, but I'll probably set it lower. The 12BH7 runs at -10v bias, 15mA Ia. To get it to run higher I'll need to get some bigger cathode R's as the 10k 2w R's in there at present are getting hot.

The amp has developed a problem, which is a sod, all day Sunday it ran well. Switched on Monday to do the frequency test and I heard a clicking or low frequency buzzing. Checked DC voltages and found one side or one triode of the 12BH7 was low. I suspected a diode were switching. They're there to keep the grid and cathode within about a volt at switch on. Disconnected the diodes, found one triode of the 12BH7 was low gain, bummer, they were NOS valves from the US. So new valve in DC conditions are back to spec, IE biased at -10v center biased, but now getting distorted OP.

When the driver's are disconnected from the OP stage all is well, connected the OP waveform distorts. Sorry camera died so no pic but the sine goes back up prematurly on the negative going part of the cycle. The sine looks like it has a small bite taken out of it, or picture two sines on your scope, both the same amplitude and frequency. Now move the bottom trace up till the peak of the bottom sine is superimposed on the negative going part of the sine above by about 1/5 th, maybe a 1/4 at high amplitude. That's the sine I'm getting, though maybe move the bottom trace 10 degrees or so forward. I thought this might be 2nd harmonic distortion and am not sure whats causing it. I wonder if the 807's are drawing grid current, maybe going into AB2? Or is there some feedback?

Will do some testing today to see if I can find out.

Andy.

[Murphyv310]

Sorry Andy I see your bias point now, I'm viewing it on a phone so it's not so easy to see. Pity it's gone on the bum but no doubt an easy fix.

[Diabolical Artificer]

Had a play with the amp. First I tried taking out the 807's = drive signal was fine. Next put two 807's back in = sine at drivers, IE 12BH7 cathodes distorted as reported before. Changed
470n caps for 100n, aha! = sine distortion has now changed. The "bite" out of the neg going half of the sine now of bigger radius, IE slighter and and higher.

Now I'm on shaky ground, perhaps you can help Trevor. The grid reference R on the OP stage forms a high pass filter, right? So we want all signal's over 50hz say to pass. I have trouble interpreting calculator results in micro farads, but I make 1/ (2.28 x 50 x 150000) = 5.8 -8 or 5n . Nearest value is 4700p or maybe 6n8.

Or have I got this wrong, do I need a a low pass filter here to restrict OP of amp to 20k or maybe 17k? I've seen some transistor amps specs given as f response up to 100khz or something similar, why? Seems reasonable to restrict OP to 17khz or 20khz as this will stop HF instability, yes?

Thanks for taking an interest Trevor, Andy.

[Doodlebug]

Why bother with 100w valve amp? Because there's a little kid inside me that likes big loud amplifier's

I'm glad that I don't live next door to the noise !

[Mark Hennessy]

100 watts per channel isn't really all that much if you're using high quality loudspeakers that are inefficient. I have many (!) pairs of speakers that vary between 83 and 87 dB/W at 1m. And 100 watts is only 20dBW, so maximum levels from each at 1m would be 103 and 107dBSPL. Obviously, I don't sit that close to them.

But main factor is the programme material. Ignoring modern hyper-compressed stuff, well-recorded music has a PMR (peak to mean ratio) of at least 20dB. That means that if you adjust the volume so the amplifier is just on the verge of clipping on the peaks, the average level is only 1W. And when we as humans assess loudness, we're more concerned about the average level rather than the peak level.

Loudspeaker designers are very happy about this scenario. Given that more than 95% of the electrical energy is converted to heat, it's a good job that tiny voice coils aren't asked to dissipate >95 watts of heat when driven from a 100W amplifier.

Flip this argument on its head - how come you can fill a room with a 3 watt Mullard amp? Partly because you're hopefully using more efficient speakers, but mostly because when you turn it up, the thing is in clipping. When playing pure tone, we are quite sensitive to clipping, but when playing music, we're surprisingly tolerant to it. For it to sound bad, clipping actually has to be gross.

More efficient speakers might make better use of watts, but nothing is free, and they usually have "interesting" measurements. Luckily, power is cheap.

The sad irony for hi-fi designers is that most people judge loudness by the level of distortion, and if you demonstrate a powerful amplifier, they don't think it's all that loud because it's clean and undistorted. People are used to distortion, and quite like it. Of course, people especially like valve distortion, which is where the fallacy about "valve watts" being better than "solid-state watts" comes from.

All this explains why hi-fi amplifiers are rarely fitted with clipping LEDs, which of course are the norm in professional applications.

[ppppenguin]

Why is frequency response limited in a power amplifier? No simple answer but here are a few points.

If there is a negative feedback loop then additional low pass or high pass filters within the loop can cause nyquist instability. This is where the phase shift at some frequency approaches 180 degrees while the gain is still greater than unity. Thus turning negative feedback into positive. This is one reason why overall negative feedback in valve amps is limited: there's a huge great high pass filter called the output transformer. Transistor amps don't usually have much by way of high pass filtering, just normal AC coupling. If they are DC coupled then there's none at all. You normally don't want to pass DC through a power amp so there's usually at least 1 coupling cap.

What about the HF end? In transistor amps it's all too easy to have far too much HF bandwidth. This makes the amp vulnerable to RF interference so a low pass filter at the input is a good idea. Valve amps have output transformers that limit the HF response. In a few amplifiers such as the Quad 405 it's essential to limit the input bandwidth as the output stage inherently can't cope with signals that are both high amplitude and high frequency. It won't be damaged but the distortion will get really horrible.

Mark's comment about distortion is worth re-iterating. Distorted sound is subjectively louder than clean sound.

[Murphyv310]

Hi.
Looking at the phase splitter Andy I see you are feeding the cathodes from a negative supply, if those cathodes are decoupled (cant see your PSU arrangement) then the lower phase amp is probably the problem, it's cathode could well be shunting high frequencies but as I cant see the rest of the circuit I could be way off track. The -78v would need to be a high impedance feed and not decoupled at the cathodes.
Being as I am I prefer auto bias, seen too many amps go up in smoke if a valve goes into grid current with output transformers taking the brunt of the current. Bit like thermal runaway in transistor amps. Yes I do realise the down side of auto bias but it's just a me thing.

[Diabolical Artificer]

Doodlebug - I should of said there's a kid inside that likes the thought of big amps, I'm responsible on the whole though my neighbors where treated to The Russian's are coming by Val Bennet at louder than normal volume briefly on Sunday. I had to point the speaker out the door as the windows vibrate alarmingly if not.

My main speakers are 86dBA SPL @ 1m 1w pink noise which work well with a transistor amplifier which is 70w. At about half volume the bass moves your diaphram 8 feet away, it's only kicking out about 1w, I measured it and was very surprised. What I don't get Mark in your post above is how the volume should be set. To get full power ish your amplifier would need to be at 3/4 volume say, so that would work if your listening to a record with a large dynamic range, a bit of classical say where you have quiet passages then a crescendo at the end. For rock music which is full on from start to finish though it would be uncomfortable. A better set up would be with a separate pre-amp and power amp. Preamp set low, power amp high. Hope you get the gist.

Going back to the amplifier, I direct coupled the PS to the cathode follower's to keep the phase difference between IP and OP as low as poss.

I made a few changes yesterday and things are a lot better. I tried a 4700p cap and the result was terrible. Did some more calculations, did some looking online at at some schematic's I have and changed the two 150k grid R's for 47k and used 220n caps. That was a lot better. What made the big difference though was changing the way I got the negative bias to the two sides of the OP stage. Now things are a lot better.

The cathodes of the PS are fed by a transistor CCS which is powered by a separate PSU from the main HT PSU. It has a separate tfmr 50v, 78v DC once rectified. It's positive goes to ground giving - 78v DC, this feeds the negative bias for the OP stage too. Not sure if this will be classed as de-coupled. I'll draw a schematic; so far I havn't really looked at the PSU, it could do with improvement. I think the CCS could be improved too to make it higher impedance.

I'll rig a protection circuit up Trevor that will cut power to the amp in case of no bias or excessive OP stage current. To get max power fixed bias is the only way to go, though I did think of doing both; a bit of self bias and fixed, will see.

Andy.