Gates 3638 preamp

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.
yep; 0.05 on 100K is getting too small.  But it's a clue.....about......something. 
 
After a reboard I'm happy to report this heathen is behaving properly - at least it doesn't oscillate anymore.

The original wiring had been correct.  It looks like the grounding scheme/layout may have been the problem.

Under the old conditions I had tried bypassing the 1st stage and injecting straight to the 2nd grid with no

problems.  It's hard to say exactly what it was.

Got to do some more reading on Feedback - haven't gotten very far into that yet. 

Thanks again

 
Ok -

To the ear , 330K or 470K seems a bit better than 220K for the FB resistor. The transients sound better in addition to the increase in volume and I can still run the IS vol pot wide open without getting hard clipping.

That was simple enough. I'm still at a remedial level with FB theory so I wanted to ask a few questions that popped up during readings.

First would be - Is this voltage or current feedback or both?  One would work to lower plate resistance, one to increase it. 

What I think I'm seeing theory wise is a 6N7 whose plates see 25K (OT) // (220K + 750r) or 22.4K load, so as you increase the value of the FB resistor the effective plate load approaches 25K, thus an increase in gain. If you decrease the FB resistor, you lower the effective plate load which = less gain.

So is this the correct explanation for why current FB decreases gain for this plate to cathode scheme?  Or does current & voltage FB act as an independent variable outside the rules for parallel resistors?

Apologies if I'm over thinking this - just trying get a good grasp on how to break down single stage FB.

Second, in another post by NYDave - http://www.groupdiy.com/index.php?topic=4808.msg57828#msg57828 -  he describes using his general rule of "20db less than open loop gain" for advantageous FB using a plate to cathode arrangement.  The SA-134 plan at 200K gives about 13db measured FB.  To get 20db on the 134 would take a resistor value <<100K which pretty much squeezes the life out of this circuit.  Any thoughts on why a lower value seems more optimal on the SA-134?

Also, does anyone know of a tube circuit gain calculator online that takes single stage FB into account? 
 
> "so as you increase the value of the FB resistor the effective plate load approaches 25K, thus an increase in gain. If you decrease the FB resistor, you lower the effective plate load which = less gain." Apologies if I'm over thinking this....

You are over-thinking.

The in-circuit plate resistance is never the value on the data-sheet. (Find a triode sheet which plots Rp against current.)

In triodes, wide change of load has little effect on gain.

> just trying get a good grasp on how to break down single stage FB.

Don't get specific (yet).

The idea is: you have an amp with "infinite" (or very-high) gain.

You put a 10:1 divider from output to input, it has gain of 10 (or very nearly 10).

So by diddling that 220K-330K-470K resistor, you are changing the NFB division, and (as long as no-NFB gain is greater than NFB ratio) the with-NFB gain.

> ...how to break down single stage FB.

It is not "single stage" (if we are looking at the plan Doug posted in #13). Gates' drawing-style is more pictorial than analytical, and I'm not going to re-draw it, but I've shorn some irrelevancies:

wj9n5f.jpg


V3 is a Pentode with signal input at grid and NFB input at cathode. Its plate drives V4, a fat triode. Signal path is blue line. NFB runs from V4 plate to V3 cathode, red line.

What is gain of V3? Probably 100. What is gain of V4? Probably 10 or 13. Total (no-NFB) gain, over 1,000.

What is NFB ratio? 220K/750 293. (Well, +1 or 294.)

The forward gain is 3 or 4 times higher than the NFB ratio.

If the forward gain were more like 100,000 (chip), 300:1 excess gain, then the closed-loop gain would be very-very close to 294. No matter if the actual gain were 50,000 or 500,000, we'd still get 292-294.

We only have excess gain of 3 or 4. Closed-loop gain will be closer to 230 or so.

Change 220K to 470K. NFB ratio is 470K/750= 627. Open-loop gain hardly changes (V4 gain is all about its Rp and load, the 220K-470K change is nothing). Still about 1,000. That is greater than 627, but not by a lot, less than 2:1. Closed-loop gain will be around 400.

Changing 220K to 470K has made closed-loop gain go up almost 5dB. The original value reduced THD 3 or 4 times, the higher value reduced THD by half. The original reduced output impedance by 3 or 4, the revised by only half. The higher THD may fail some arbitrary broadcast spec. The higher output impedance may allow a level-jump if you drive two loads and one of them falls off. None of that is important when using vintage gear today.

The original network was 220K+0.05uFd. So it isn't 220K to infinitely low frequency, the ratio rises below 14Hz. Alone, that might be bad. But we have other roll-offs. It may-or-may-not be that the 220K+0.05u was selected to cancel some of the coupling and OT roll-off. In this case, that probably does not matter. even if they were trying to fudge the 32Hz and 20Hz response for broadcast tests or cascade use, such errors are unimportant for a single "flavor" amp in modern studio use.
 
Thanks PRR for the excellent info!  Much appreciated.

I'll try and digest all of this before the turkey hits the table tomorrow.

Happy Thanksgiving to all.  ;D
 
Good analysis. 

The original network was 220K+0.05uFd. So it isn't 220K to infinitely low frequency, the ratio rises below 14Hz. Alone, that might be bad. But we have other roll-offs. It may-or-may-not be that the 220K+0.05u was selected to cancel some of the coupling and OT roll-off.

Gates seems to have always used caps for tailoring of low response (sometimes high too).  I see it almost universally, where other manufacturers moved the cap more out of band. 

 
Can you post the numbers on the top of that big gray can in the middle -- or any other numbers on the can sides.  That's what failed on our unit and we have no numbers visible at all other than the characteristics I measured.  Koz
 

Latest posts

Back
Top