Fairchild 660 question.

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

DaveP

Well-known member
Joined
Nov 8, 2005
Messages
3,019
Location
France
I am about to start making a Fairchild 660 for a studio.

There are a couple of issues that I'd like to clarify first.

The first is the way the bias is applied to the control amp output valves on the original 660



It has two sets of grid resistors, the other must be 330k not 300 ohms????, the resistor numbers are out of sequence so it looks like they were added later to cure some instability, they feedback anything over 170KHz, any thoughts?

By the time they came to design the 670, they kept the feedback caps and 47k's but went for a simpler standard bias feed through 150k's instead of 330k's


The original 6V6's only have a 100k limit for fixed bias operation, whereas the 6973's limit is 500k
My inclination is to go with the 670 modification.

The second question concerns the B+ voltage to the GR tubes, its 240V for the 670, but only 200V for the 660.  There are also slight differences to the bias arrangements to the cathode balance..


With the 660, the CR goes into the CT of the IPT but the -19V  6V6 bias goes into the cathode balance pots.


With the 670 the bias of only -6.2V goes into the balance pots, and the 2x 100k grid resistors are deleted.

I'd be interested if anyone can unpick the logic of these changes
Best
DaveP

 
DaveP said:
It has two sets of grid resistors, the other must be 330k not 300 ohms????,
Certainly; with 300 ohms there would be no signal.
the resistor numbers are out of sequence so it looks like they were added later to cure some instability, they feedback anything over 170KHz, any thoughts?
That's if you neglect the impedance of the node here, but it's about right.
By the time they came to design the 670, they kept the feedback caps and 47k's but went for a simpler standard bias feed through 150k's instead of 330k's
That's typical of the evolution of products over time. The first units have more trimming range, and as the designers get more feedback from the field, they figure out they can cut some corners.
My inclination is to go with the 670 modification.
That's also my opinion, since the 670 benefitted from the learning of the 660.
The second question concerns the B+ voltage to the GR tubes, its 240V for the 670, but only 200V for the 660.  There are also slight differences to the bias arrangements to the cathode balance..

With the 660, the CR goes into the CT of the IPT but the -19V  6V6 bias goes into the cathode balance pots.
These are minor differences; I wouldn't think it changes much to the performance (well the increased plate voltage may have been a tentative to extract a little more performance from the duet of 6386, compared to the quartet in the 660).
 

With the 670 the bias of only -6.2V goes into the balance pots, and the 2x 100k grid resistors are deleted.
  And the trim cap (C114) has been added. It probably proved more adequate at controlling the HF bump than the modest load of the 100k's.
All we can do is conjecture, unless someone unearths the ECO log...
 
> bias is applied to the control amp output valves

You pruned a lot of context off that part.

As shown there is evidently NO bias on the power tubes. 300r grid to cathode... WTF??

You would expect 300r in the cathode return (not grid-cathode). Or a negative voltage from lower space.
 
Rein Narma did the 660, George Alexandrovich  did the 670, Rien is gone but George is still around and would probably be able to answer a few of these questions via email,

there are also 4 triodes on the 670 on each side, but they draw 2 to simplify,

that means if you use 6BA6 your signal tube count goes from 8 to 16 yeah!

getting cold out,

 
Thank for your input so far.

I was hesitant to put the entire schematics up in case the detail was not clear enough, but here goes.


Clear copy.

EMI Report copies with voltages.


Other half


Best 670 with voltages


CJ,  This will be a mono unit so only 4 x 6BA6 per side, 8 in all, but 15 tubes in the total compressor!!

Best
DaveP
 
The other point that I noticed which may affect the sound and gain, is that the signal amp cathodes are not by-passed in the 660 (except by 2x 680pF caps which do nothing).

Whereas in the 670 they are by-passed by 4/8uF caps which pass down to 14/28Hz -3dB point.  Never could work out if they were 2x 4uF. :-\

There is AC cancellation with the joint 1800 resistors in the 660, but the working point of the tubes has been changed between the two amps;  Cathode resistors 1800 ohms for the 660 and 680 ohms for the 670.  I guess they were looking for more gain?

DaveP
 
I wonder if there's a difference in distortion when in deep gain reduction with the different cathode arrangements.
 
Off the top of my head, where does the fixed gain cathode setting interact with the side chain bias in each?  Does that affect distortion point, or ratio, or anything else?  On one level, since many comps use a variable resistance in the cathode to trim the GR meter, we can say there is no 'fixed' cathode resistance across the board, only a requirement for a given current for metering, which will change depending on the set of tubes.  Not the case here, but similarities apply. 
 
DaveP said:
Whereas in the 670 they are by-passed by 4/8uF caps which pass down to 14/28Hz -3dB point.  Never could work out if they were 2x 4uF. :-\
My Rosetta stone says it's 2x4uF.  ;)
So it's 8uF decoupling 1410 ohms (680+680+100//100)
The -3dB point calculation should take into account the cathode impedance of the 6386's, which makes it actually somewhat higher than 14Hz. But also remember that it's not a simple HPF with a response that falls asymptotically to -infinity, so the attenuation at the characteristic frequency would be less than 3dB. However, the resulting low-frequency cut may introduce a significant effect in cleaning subsonics.
There is AC cancellation with the joint 1800 resistors in the 660, but the working point of the tubes has been changed between the two amps;  Cathode resistors 1800 ohms for the 660 and 680 ohms for the 670.  I guess they were looking for more gain?
It's actually a tad more complicated, since the two 500r balance pots in the 660 dominate over the 1.8k's. As a result, there is slightly less gain in the 670 to start with, justifying the bypass caps. It's not only the working point that has changed, but also the topology. In fact there are so many things that are different that it's almost impossible to make a proper evaluation of gain. You may note that on the 670, the 600-ohm T-pad is loaded by 360 ohms, where on the 660 it is under-loaded by about 2.5k (the reflected impedance of the 150k through the 1:9 xfmr is about 1.25k); this would account for a few dB difference.
 
see that there are 3 unknown currents in the 670 cathode equation,

to complicate matters, we do not know the sign, (+ or - ) of one of the currents, (i-2)

to get the sign you could breadboard the circuit if you have a dual supply

you can bisect the circuit to simplify it down to something we believe to represent this from a DC point of view>






 
Thanks CJ,

I am working on this as we speak.  I've made a breadboard of the signal amp and got it working ok, 660 style.

The negative supply for the 660 is simpler as you know, and it uses the regulated supply as the meter zero :eek:.

The 670 uses a negative supply and pot to shift the entire timing circuit up or down to accomplish the same purpose.

Whichever negative supply you choose, 660 or 670, it must be able to source over half the current to the signal amp, as being more negative than the ground pot, more current will go in its direction from the cathodes.  I have decided to copy the 660 circuit entire (that's what the customer wants) but it's interesting trying to get my head around what's exactly going on as you know only too well.  I hope to post some results here when my tests are complete.

Best
DaveP
 
> we do not know the sign, (+ or - ) of one of the currents, (i-2)

Sure we do, or can.

Un-superposition. Take the right loop off.

14V to 680+50. Counting on thumbs, call that 10:1 or 20:1 divider. The junction is 1.4V to 0.7V. Call it +1V?

We already see that this is "lower" than the 6.2V battery, so the 6.2V batt will be "draining". It will pull the joint node more negatively.

680||50 is 46.57, call it 50? We now reconnect the 50+6.2V, but call it 6V?

We have a 1:1 divider running from +1V to -6V. That is 7V total. 3.5V each resistor. The joint node is now 3.5V down from +1V, and 3.5V up from -6V. Smells like about -2.5V.

We rounded some numbers. Less than 10%, so it is all good for tube-work, but let's be obsessive.

After re-drawing it +_up and -_down, and combining resistors to a series string, we can calculate the current and use the TI-30Xa and a 4-digit scratchpad (matchbook) to get more better numbers. Working with I*R twice, we would like to see the two answers come out very similar. (I screwed this up bad, but in a do-over I got lucky.)

670k-anal.gif


For an encore I called my idiot assistant. SPICE gives a number that agrees within the expected rounding-error of my 4-place scratchpad. (SPICE does everything to 8+ places.) As my Simpson VOM don't know 2.49V from 2.51V, I say it agrees exactly.
 
PRR,

Thanks for that, but I think that both CJ and I have worked that out some  years ago around Feb 2009 from the file properties I've archived, I think CJ forgot that.

What we are not sure about is the mechanism taking place here.  The fact that the cathode current has two paths to take and why that is beneficial.

DaveP

 
Happy Thanksgiving USA.
Here is something to look at after you've finished your turkey,  :D



With these voltages from the EMI schematic, it gives 10.5mA / side.  Tubes are 8x balanced 6BA6's triode -wired

Control voltage had to be -14.8V to get these figures.

5.5mA goes to the ground and 15.5mA goes to the negative supply.

With an ac input of 4.28 g-g the output is 1.81 without a cap across the cathodes and 1.84 with one.

Output voltage taken before the pad.

So compression is 7.47dB with no cap and 7.33 with a cap......only 0.14dB difference.

Now this is set-up I'll prepare a table of various control voltages.

DaveP
 
i like the brute force method, did not know you could do it that way,

seems more solid than pulling out RDH4 and messing up Kirchoff's equations,

they might have been trying to get that balance pot to work better by running some current thru it,

forgot the equations for this problem, trying top figure them again without cheating with the old thread,

forgot a lot of things, looking over an old cheat sheet for series, don't even know how to work one, but we do know how to bake a pie and let it cool in the window like Aunt b, Happy Thanksgiving!

 
well we just spent 3 hours messing up equations that would put us on Mars instead of the moon had we not the simplicity of the problem to see the errors of our ways, should have went to Denny's and scratched it on a dinner napkin while flirting with the waitress, but no, we had to OCD it in our lonely room,

but it did come out all right, as long as you do not turn the bal pot and apply a complex signal, and we already ate a whole apple pie for the price of a single slice  ;D

some dead Russian would be proud, and it seems to agree with Lord PRR,

 
This is what you did way back on the napkin at Denny's with no waitress distraction.



DaveP ;)

 
Back
Top