Help me pick an amp design: P-P EL84 with 12ax7 driver

[gadget73]

A 22K resistor doesn't really tell you what the feedback will end up being. Its unfortunately a little more complicated than that. Part of it is the ratio between that resistor and the cathode resistor, some of it also involves the open loop gain of the circuit, as well as what tap the feedback comes from. If you moved it to the 32 ohm tap, you'd have more, moving it to the 8 ohm tap would be less. Ultimately the amount of feedback it will stand is mostly determined by the output transformer. Some just will not tolerate large amounts before they lose their mind.

Since this is pretty close to the original design, its a fair assumption that it will take the level of feedback it has. Also, since this was part of an integrated amp, its entirely possible they made it with low input sensitivity to match up with the output of the preamp circuit. You could try less feedback if you'd like, a pot in series with the 22k resistor would let you decrease it as you see fit. I probably wouldn't go with more feedback than it has though.

 

[kirk57]

What am I looking for if I substitute a pot for the FB resistor? The max amount of FB before oscillation sets in? The minimum amount of feedback before <fill in the blank here>? Something else?

I guess the question is: Feedback is good? Other than oscillation there is no downside?
The goal being a good sounding amp with reasonable amounts of distortion.

 

[s-petersen]

If the volume is too low you can increase the FB resistor for less feedback. If the volume is enough, I would work with the step network to reduce the ripple in the top of squarewave

 

[BinaryMike]

This post in another thread might be useful: http://audiokarma.org/forums/index.php?threads/balance-pot-how-to-connect.767348/#post-10488415

 

[gadget73]

What he said. Add a pot in series with the existing 22K resistor. I'd probably go for a 25k or 50k, but more depending on what my junkbox produced than any special reasoning.

 

[kirk57]

A 22K resistor doesn't really tell you what the feedback will end up being. Its unfortunately a little more complicated than that. Part of it is the ratio between that resistor and the cathode resistor, some of it also involves the open loop gain of the circuit, as well as what tap the feedback comes from.

For those like me who don't already know this stuff, I found this page to be a pretty clear explanation of how this works. Although the amount of feedback for guitar amps is typically less than for hi-fi amps, the principle is the same.

http://www.aikenamps.com/index.php/designing-for-global-negative-feedback

 

[gadget73]

Same basic principle, sure. A lot of it is the resistor ratio, but if you moved the tap from say 16 ohms to 4 ohms, you'd have 1/2 the voltage coming back. If you had a cathode bias amp and pulled the cathode on the output stage, that would change it too since the total gain of the circuit is less.

Honestly, I don't calculate things like this. It gives me a headache. I find it much easier to just measure stuff instead of wondering if I've gotten the calculation proper.

 

[kirk57]

So anyway back to why this amp doesn't sound so good... I hooked up both channls via a Y connector to the signal generator and the outputs appear to be slightly out of phase:

Is this something to be concerned about and if so any idea why this might be? If this is accurate it might explain the 'out of focus' sound that I'm hearing.

I suppose I should also check/compare all voltages across all tubes, just to do a sanity check.

 

[Son-of-Vere]

I want to say coupling cap variances. What freq. are we looking at? That time interval could be very short.

 

[kirk57]

I want to say coupling cap variances. What freq. are we looking at? That time interval could be very short.

1 kHz

 

[gadget73]

Would be worth hooking that direct to the input to make sure the scope isn't funny before making too many assumptions.

 

[Son-of-Vere]

From my rough math that delay is 1/30,000 of a second. Would it be possible to detect a delay that short with your ears if it actually is coming from the amp? I figured the delay was very short, that's why I thought cap variance. Moving a speaker back a few inches could create the same effect. I personally don't see this delay as an issue.

 

[gadget73]

I'd tend to agree, its not significant, but I'd still like to see the two scope probes connected to the same source just to verify the scope itself isn't where the lie comes from. I have old test gear, and I've learned the hard way that when things don't make sense its a good idea verify the test gear before assuming what its connected to is the problem. I have gone down that rabbit hole before, its maddening.

 

[Son-of-Vere]

I'll agree, verifying the equipment would be the thing to do here. This traced peaked my curiosity as to what could create delay in a tube amp. So maybe I should ask; could cap value variance create such a delay?

 

[gadget73]

you can get some phase shift from caps, but usually its from very different values being installed. To be fair though, I've never wired one of my own amps up to do this same test.

 

[BinaryMike]

So maybe I should ask; could cap value variance create such a delay?

I figure the delay difference at about 20uS, which equates to an F-3dB of 8KHz. That would be painfully obvious in a frequency response check, but the cause is unlikely to be coupling cap differences. Let us know if if the effect is really there.

Oh, and it's piqued, not peaked.

 

[kirk57]

Gadget, you're right. The input straight from the sig gen to the scope looks the same, maybe the probes?
So it's not a concern, in any case.

I had another look at my wiring, and I missed a &^%*$ solder joint (the 15k resistor from the EFB pot to ground), fixed that. Also found that the 3.5mm jack to RCA pair I was using to connect the I Phone to the amp
have been melted in several places by a soldering iron (gosh how did THAT happen? )
Got a new connector and now all is well: amp sounds fine, lush EL84 goodness!

Thanks so much to everybody for guiding me through this build. This is AK at it's best.

Before I call it done, are there any other post-op things to check? I did run test tones from 1k up to 15k and didn't see any obvious signs of stress. Did somebody mention connecting a capacitor to the speaker inputs as a test?

I'm also mostly done with making a cabinet for this guy; I'll post pics when that is done.

 

[kward]

Glad you got the beastie working. Here are the full battery of tests I usually run:

• Check HF stability by hooking a cap only load of .01 uF to .1 uF across speaker terminals and run say a 5 Khz square wave through it at 1 watt equivalent output. Output will naturally look ugly, but it should not break out into full blown oscillation.
• Check LF stability. Pulse a signal that would generate 1 watt equivalent output if it were to be a sustained signal. But pulse it quickly on then off. See how quickly the output settles to 0V on the scope after signal is removed. Try it with resistor load, cap load, and no load. It should not bounce around too much getting back down to zero. And it should not break out into low frequency oscillation under any of those load conditions.
• Check frequency response. With a resistor only load, and output set to 1 watt, with 1 Khz sine wave as reference, vary sine frequency from 1 Khz down to 20 Hz, and from 1 Khz up to say 50 KHz (if your sig gen will go that high). We'd like to see flat frequency response from about 10 Hz to 30 KHz, but your output transformers may restrict that to maybe 20 Hz to 25 Khz, which is still pretty respectable. Try this test first measuring your signal generator directly, to see if it has any frequency variations within the range you will vary frequency. This helps distinguish sig gen anomalies from amp anomalies.
• Check full power output capability. Ideally you would run both channels simultaneously for this test. Run a 1 KHz sine wave through the amp with a proper sized resistor only loads on each channel and see what voltage it peaks at. Power output then equals [(Vpeak/1.414)^2]/R. I'm guessing it will be somewhere around 12 watts, which would be 13.85V peak on the scope on the 8 ohm speaker tap with an 8 ohm resistor load.
• If you have a distortion analyzer or a spectrum analyzer you could check distortion characteristics.

 

[kirk57]

Pic of cabinet, as promised:

I'm encountering what is to me a very puzzling situation when using this amp in my main system:
When connecting directly from I-phone or DAC, everything is OK. When driving the amp from either of my Pilot preamps across the room via a 30ft pair of Mogami cables/WBT connectors, one channel is very much reduced, as is bass output. This is regardless of how I connect the inputs (I've swapped L/R) and it follows the speaker connections.

Since the behavior is different between the channels under these circumstances, it would seem there are wiring differences, but so far I can't find any. FWIW the run of Mogami cable measures 1.1nF, 0.3ohms.

 

[gadget73]

hm, wonder if its the capacitance making it oscillate. Did you by chance try loading the output with a cap across the dummy load to see if it went bonkers on the scope?

Kward's test battery would turn that up pretty handily.