First valve amplifier build

[Radio Wrangler]

I'm not sure the feedback in the small signal stages counts with the overall feedback. It looks to be a bit of a balancing act (pun, sorry) in the phase splitter.

The 100pf lead capacitor in the feedback path works into a 100 Ohm shunt resistor at high frequencies and increases the feedback fraction in a way that would have given unity gain at 16MHz, had other factors not intervened.

With the 4.7k feedback resistor it gives a zero in the feedback resonse - adds 45 degrees phase lead at 338kHz. By which point the 100k loads on the phase splitter//Ra of the ECC83 is rolling off a lot into the output valves. So that 100pf massages the phase to keep it stable well above audio. Must have a really good output transformer.

Anyway, all these things have loads more feedback than the OP tried out, but they were designed with a lot more forwards gain in order to support it.

If we put the horse firmly in front of the cart. A decision is made of how much gain an amplifier is to give when in use. Then a decision is made of how much feedback is to be applied. Then the forwards gain of the amplifier path is calculated. For large-is gains and feedback amounts, the result is equal to the finished gain times the feedback amount (add them if both in dB, of course). Only then do you engineer the forwards path to give the wanted gain. THEN you start worrying about phase shifts and stability.

Feedback isn't just something bolted on to an existing, working amplifier. The amplifier has to be designed for that amount of feedback and is not going to be house trained or even useable without it.

Thanks for pointing to that Brimar circuit, GJ. I think thet are the non-topcap version of the 5B/254M, in other words an 807 in a smaller bottle. I still prefer Dr Bailey's splitter.


David

[Gabe001]

Thank you both for your comments. I decided to leave things as they are (no NFB) mostly because the status quo works fine for me, and because I'm way in over my head

From an academic perspective, I'm struggling with the concept of dB. From where I'm standing, and I believe I've got this horribly wrong, -10dB = 0.1W (according to Google). When my power output dropped from 4.3vRMS (i.e 4.6W with a 4ohm resistor) to 4.1vRMS (4.2W) due to the now defunct feedback circuit, isn't that a 0.4w drop which is more than -10dB (0.1w)

Have I got it all backwards?

What is the correct RMS voltage or wattage drop with a 4 ohm load that -10dB of feedback is expected to produce?

Gabriel

[GrimJosef]

dB are indeed complicated. The concept itself isn't too tough, once you've got used to it. But engineers have, over the years, introduced all sorts of dBs - dBu, dBV, dBm ... here's a more complete list https://en.wikipedia.org/wiki/Decibe...ference_values.

Within specialist groups practitioners tend to drop their favourite suffix and just use the word 'decibels' because "everyone knows what we really mean". This can be a challenge for those new to the field. And Google is rarely your friend with this.

Cheers,

GJ

[Gabe001]

I think that -10db works out at -0.2v using a 4 ohm resistor and -20db works out at -0.6v. The equation is a bit complicated so I could be wrong.

Thanks GJ.

Gabriel

[Herald1360]

Originally Posted by Gabe001

Thank you both for your comments. I decided to leave things as they are (no NFB) mostly because the status quo works fine for me, and because I'm way in over my head

From an academic perspective, I'm struggling with the concept of dB. From where I'm standing, and I believe I've got this horribly wrong, -10dB = 0.1W (according to Google). When my power output dropped from 4.3vRMS (i.e 4.6W with a 4ohm resistor) to 4.1vRMS (4.2W) due to the now defunct feedback circuit, isn't that a 0.4w drop which is more than -10dB (0.1w)

Have I got it all backwards?

What is the correct RMS voltage or wattage drop with a 4 ohm load that -10dB of feedback is expected to produce?

Gabriel

A Bel is the log base 10 of a power ratio. A decibel (dB) is 1/10 of a Bel.

So..... if the ratio of output power to input power is 10:1, log10(10) = 1 so you have a power gain of 1 Bel or 10 deciBels. If the ratio is 1:10, the power gain (loss) is log10(0.1) = -1 Bel or -10 decibels. Half or twice power are -3dB or 3dB.

It all gets "interesting" when the "power" aspect gets left out and you start talking about voltage gain in dB. IF the input and output impedances are the same, then since power is proportional to Vsquared, then a 10:1 voltage gain can legitimately be called a 20dB power gain. If they're not then strictly 20dB is wrong but is often just convenient to use

Using a logarithmic measure of gain or loss is handy because you just add or subtract dBs to work out overall gains instead of multiplying all the gains together. It also results in handier numbers- a power gain of 10x is a handy 10dB, a power gain of 1,000,000x is 60dB.

Hope this helps rather than adding to the confusion

[Radio Wrangler]

No.

'-10dB' is a comparison of two power levels. It is not an absolute value of power. What it means is one tenth of the power of something. Without declaring what that something is, or it being unambiguous in the context, it is meaningless.

An amplifier can have a gain of 10dB if the output power is ten times the input power.

If you stick with powers, you are safe. The danger lies in trying to put voltage ratios and voltage gains into dB. The trap lies in differences of impedances.

An amplifier with a voltage gain of 10, with equal impedances at input and output will give ten times in voltage AND ten times in current (To keep Monsieur Ohm happy) so the output power is 100 times the input power, which is 20dB dB = 10*log(Pout/Pin)

With different impedances, the power ratio could be anything. People get rather lackadaisical with voltage ratios, so take care who you believe.

For talking of power levels, an extra letter is often added. dBm is dB with respect to 1 milliwatt - a comfortable level in a telephone earpiece which became a common ref level.

G-J has listed some others.

David

[Herald1360]

Originally Posted by Gabe001

........

From an academic perspective, I'm struggling with the concept of dB. From where I'm standing, and I believe I've got this horribly wrong, -10dB = 0.1W (according to Google). When my power output dropped from 4.3vRMS (i.e 4.6W with a 4ohm resistor) to 4.1vRMS (4.2W) due to the now defunct feedback circuit, isn't that a 0.4w drop which is more than -10dB (0.1w)

-10dB = 0.1W is sort of correct, strictly -10dB is a power ratio of 0.1 or 1/10. To stir the pot a bit, you could say that -10dBW = 0.1W where dBW means dB relative to 1W.

Your power output drop from 4.6W to 4.2W was a dB change of 10log10(4.2/4.6) = -0.4dB approx.

Have I got it all backwards?

Not exactly!

What is the correct RMS voltage or wattage drop with a 4 ohm load that -10dB of feedback is expected to produce?

A voltage drop (ratio) of 0.316 and a wattage drop (ratio) of 0.1.
-10dB = antilog10(-10/10)= 0.1 for power. The voltage ratio for the same power ratio (0.1) is sqrt(0.1) since power in the same load is proportional to Vsquared.

Gabriel

Hope this helps.

(David (Radio Wrangler) could probably put it much better!)

[Gabe001]

"'-10dB' is a comparison of two power levels. It is not an absolute value of power. "

This is key. I was interpreting it as an absolute value of power which is where I went wrong.

Thank for the help and the detailed explanation. I'm sure I'll see some of you at Retrotech this week

[Gabe001]

Just an update, I've got a few bits and jobs from Retrotech and the transformer from Jerry and spent some time upgrading. A new output transformer and a dual 100uf/50uf capacitor can rated 500v were fitted. The challenge with Jerry's transformer was keeping the voltages down, so I have a couple of 5w resistors running a bit hot. The old can was only rated at 250v so it had to go. Replacing the OPT wasn't strictly necessary but the higher impedance (turns ratio 65:1) kept the 6v6s from working at their best, and the OPT price was very reasonable. The best orientation of the OPT was found using the "headphone test"

I've done a few more experiments with the circuit which ended with me moving the volume control between the ecc83 and 6c4 - it's an idea I got from the bvws amplifier article and it works much better in that position. I'll post a final circuit diagram later.

I've put a scope on the output and with a 1khz signal via my phone it clips at 6V RMS (4watt resistor) which equates to 9watts. A new set of valves should have given me 11w with the working voltages but it's loud enough as it is.

I'm really happy with the end result. It was harder than I thought it would be, and frustrating at times,but good fun and a good learning experience. I was tempted to scrap everything and build "Lola" many times.(https://www.vintage-radio.net/forum/...ad.php?t=51211).

Some pictures below. Speakers are also from Retrotech. Swmbo gave me the thumbs up. I need to do a few cosmetic bits including a wooden base, a pilot lamp and fit an integrated Bluetooth krc-86b module, but the tough part is done.

[Diabolical Artificer]

It was harder than I thought it would be, and frustrating at times,but good fun and a good learning experience

Glad you got there in the end,valve amp designing building can be as you say frustrating but it is rewarding and fun, sometimes.

Andy.

[Gabe001]

Thank you very much Andy.
To be honest I've never heard a professionally made valve amplifier, so I have no baseline for comparison, other than my radios. It would be good to know how far 'off' the gold standard mine is. I'm sure there is plenty of room for improvement.

On a slightly separate note, I tried it today with a cheap record player with a red Chinese ceramic cartridge. Whilst it's loud enough, every little speck of dust, pop and crackle and imperfection in the record is magnified. I mean, you literally hear everything, good and bad. So whilst with digital media it's great, I'm not sure I'll be using it with a record player any time soon.

[Gabe001]

Final circuit attached, and it works really well imho. The main challenge was the master volume which was causing some hum so I moved it to just before the 6c4. The master volume circuit I used (with a second capacitor) is from the fender superchamp. Initially I tried placing it instead of the 680k grid leak and whilst it doesn't hum, it doesn't go to zero.
In its position it also serves as a grid stopper before the cathodyne which protects against distortion (apparently).

The modification suggested to create a virtual centre tap using an adjustable pot is genius.

I've installed a krc-86b in it using the second 6.3v winding from Jerry's transformer.

It delivers 10-11w at maximum volume and requires a signal of about 500mv-1v to drive it to full power. I've implemented feedback in the manner described in the Lola amplifier thread but I've made it switchable using a latching button switch. This enables me to use the amp with a variety of sources and get reasonable volume.

The krc-86b delivers 900mv RMS so it can drive it to full power without feedback, and to about 8-9w with feedback. The signal from the phone is very variable, and some apps tend to run louder than others. Even music played through Spotify has different volume levels between different albums.

I built a base for it as you can see, and the pilot lamp and the krc-86b and rectifier are within it - there was no space within the amplifier itself for the BT module and at this point I had enough of metal bashing.

One day I may build one in stereo

This was a real voyage of discovery - I've lost count of how many experiments I tried.

Thanks for the help

Gabriel

[Robert Gribnau]

Nice to read that your amplifier works well.

Just something for future projects. For this kind of feedback it's better that the volume is not inside the feedback loop.

Probably a small mistake in your last schematic but on it the first stage has no grid resistor.

[Gabe001]

Thanks Robert for your feedback.

Regarding the first stage, it's not a mistake. The signal is taken (almost) directly to grid, via an isolating capacitor, and a 75k resistor to ground. There was a volume pot in that position initially but I found the amp quietest when the volume was at maximum, so I replicated this. Classically 22k is meant to be the optimum valve for 'quietness' based upon the internal valve capacitance but I didn't want any losses. Theoretically I could have used a lower resistor value + picofarad cap in parallel but I didn't want to go there.

In the Mullard 5-10 (version without controls) the signal input is taken directly to the grid too (+1M resistor to ground) so I figured that this arrangement isn't going to cause any of the valves any harm.

[Robert Gribnau]

Are you sure about the grid resistor of the first stage?

In your schematic, there is a 75K resistor to ground but it's situated before the isolating capacitor. After the capacitor there is no grid return to ground but there should be one. How it is now, there is no dc-connection between the grid and the rest of the amplifier (in this case to ground). The grid than 'floates' while it shouldn't.

I am surprised that it does seem to work. Maybe the isolating capacitor is a bit leaky, creating some sort of dc-path to ground.

I advise you to place a resistor (of something like 330K) from grid to ground. The 75K just remains where it is now.

[Gabe001]

Thanks Robert. I can do that, no problem

[Gabe001]

Resistor suggested by Robert added. Attaching some pics and oscilloscope traces

Final circuit
Square wave 1khz at approx 1watt (feedback engaged)
Square wave 400hz (feedback engaged)
Bluetooth module installed in wooden base (the chocolate block is temporary)

There is a bit of takeoff at the angles as you can see (which I couldn't improve with nf caps) but the overall sound is clear and enjoyable and I'm not going to mess around any further.

I was worried that the krc86b was going to introduce hum and digital noise but it doesn't. I used the second 6.3v winding (which also powers the pilot lamp) and a 5v to 5v isolator, which is probably overkill.

[Gabe001]

Just a quick update, for general interest. I parallelled the ecc83 for (theoretically) slightly more gain and a lower noise floor, adjusting the resistors as appropriate. As previously mentioned 6db of feedback (50% voltage cut) was applied, on a switch.

I did some tests on the amp using the guide here
https://enjon.uk/2019/09/01/a-diy-ap...io-amplifiers/

THD with 6db feedback was 0.4% when the amp was injected with a 1k sine wave, and with a 1 watt output. THD + noise was 0.44%. So not an Armstrong point one but not terrible for a first attempt

Without feedback thd is 0.8%

The noise profile is seen in the attached picture. There is a spike at 50hz with various harmonics of this which is mains hum, even though I can't hear it. The laptop was operating on battery. Some of the 2k harmonics are from the soundcard rather than the amplifier.

The frequency response however is surprisingly very good, with and without feedback.

If starting from scratch I would use a pentode in the front end for about twice the gain of the ecc83, followed by the concertina phase splitter, and implement a bit more feedback to drop the THD further. A voltage gain of about 60 is perfect for a mobile phone source and 6v6 output valves. The laptop can drive the amp harder but I almost exclusively use it via Bluetooth/phone. There's a krc-86b module in the amplifer base.

Regards
Gabriel