January 28, 2012 - 09:23 pm|
|I think your modification instructions are first-class. Really well-done.|
I've modified an ST-70 and a 2080, albeit a bit differently. I followed all of your recommendations on the 2080 (except regarding the line / tone stage which I removed entirely); I also added a remote control pot and mute circuit, driven by a 6SN7 cathode follower. The 2080 works fine, and your changes are a marked improvement over stock.
I did most of the same mods to the ST-70, but left off the 6SN7 balancing pot and resistors.
I built the ST-70 for a friend, who made a really nice enclosure and face plate. He used it for awhile and brought it back with a motorboating problem. I didn't check out the amp again before I started working on it.
I also noted an abnormal of hum with the volume pot at almost minimum position. The hum does not increase with the volume control setting. The phono stage was really noisy and full of hum. The hum pots seemed to make no difference at all.
I noticed that I had left C37 & C38 without increasing them to 1 MFD, so I replaced them with film type caps.
I had some of the unused OPT leads a bit to close to some of the inputs, so I bundled them together out of the way.
I made some additional mods:
1. I added a 62K 2W resistor to drop the voltage to the plates of the phono section to the SAMS specs.
2. I measured about 24 VDC riding on the filaments, which I thought was too low. I removed the DC from the filaments and removed the hum pots. I ground one lead of each of the 6.3 VAC supplies.
3. I installed a 4 terminal strip and added a voltage doubler consisting of (2) 1N5408 rectifiers bypassed by 0.01 caps. I installed (2) 3300 MFD @ 25 VDC caps for filtering, and rewired the two phono 12AX7 tubes for 12 VDC. It made a huge improvement.
4. I added a 10 ohm, 2W resistor to drop the filament voltage from 14.4 to 12.0 VDC.
5. I lifted all of the line level input grounds with 10 ohm resistors.
6. I added a CL-80 thermistor to drop the B+ a tad (roughly 10 volts).
I've still got the hum issue at low volume. I had installed new electrolytics, and used additional capacitance in the PS. Hmm . . . the amp hummed and puttered. I think I fixed the "puttered," but both could be caused or exacerbated by a bad filter cap. So I'll check that next.
I also noted that if I removed V5, the hum went to "normal" on the left channel, but still remained on the right, unabated. That makes me wonder if I grounded the wrong wire of the 6.3VAC secondary on the tap intended for the R-channel. It also made me wonder if some of the noise is from the AC filament on V5.
The next steps will be to add V5 to the DC filament supply, and try swapping the grounded lead on the right channel's AC filament supply.
Any other suggestions or troubleshooting ideas would be welcome!
BTW, thank you Steve, for hosting this excellent site, and for contributing a great deal to the hobby of tubecraft!
December 15, 2011 - 08:41 am|
Thank you for your post. Actually, the compensation around the driver stage (V7) is balanced, as is. Also, what you describe as a "double whammy" of rolloff is precisely what is intended. The RC feedback from V7-2 to V7-1 develops a virtual ground at high frequency, at the plate of V5-1. This suppresses the secondary pole at that plate, as well as performing pole-splitting at the V7-2 plate. That results in less excess phase shift at high frequency.
The reason that the rolloff effect is balanced is that the V7-2 AC plate current is coupled through the cathodes, ending up in the V7-5 plate. The small reduction due to the R65 loading effect is corrected by the increased value of R63. So the AC plate currents match. The only remaining unbalance would be the the fact that the RC feedback network itself loads only the V7-2 plate. Hence, the RC network on V7-5 is added to balance that. (It doesn't significantly affect the primary rolloff.) As a result, the mod reduced distortion at 20kHz from 0.85% to 0.33%.
Regarding your suggestion that degeneration resistance be added at the cathode of V5A, I doubt that this will help, since it reduces loop gain in about the same proportion that it reduces distortion in the stage. Adding the degeneration would increase overall distortion, unless V5A were the primary distortion contributor of all the stages in the loop, which is unlikely.
Finally, the idea of reducing power supply hum from the V5A stage by coupling into V7-4 is interesting. However, the power supply at point-V (feeding V5A) is far better filtered than the voltages feeding the other power amp stages, so they are the limiting factor in power supply rejection, not V5A. So the effort to reduce V5A hum would be fruitless.
I do appreciate your stimulating comments and the opportunity to clarify the operation of the compensation networks.
December 14, 2011 - 10:32 pm|
|I noticed that the RC networks from plate to grid of both drivers won't act in a balanced way since one side is coupling to a high impedance input, the other to a grid at AC ground. Besides unbalancing the load on driver plates at higher frequencies, there's also the effect of the driver input impedance (feed back components magnified by the Miller effect) rolling off the output response of the V5A. I doubt that a double-whammy was the intent. I think the intended pole-zero in the response for stability could be achieved with a single network placed across R33 or between the driver plates but not both (values to be recalculated). Keeping the drive stage fully balanced will reduce distortion at high frequencies. (Of course the value of the series resistor in the RC compensating network must not be so low as to starve the stage for current / available high-frequency voltage swing. That was the cause of many early solid state amps and op-amps developing slew rate limiting.)|
Having a portion of the V5A cathode resistor above the feedback point unbypassed would provide local degenerative feedback to reduce distortion due to V5's transconductance varying over the operating current range. That matters because the global feedback helps with everything else in the power amp except for the V5 input characteristic.
Another idea - Since any a.c. hum on supply V feeding R33 is divided by R33 and the plate resistance of V5 and is added to the signal, it might be reduced by introducing the same hum (as a common mode signal) to the grid of V7B (pin 4). Essentially make a capacitive divider using C37 and a newly added capacitor between V and C37. I read that space is tight, since electrolytics are usually too poor on tolerances and have poor long-term stability, the capacitive divider formed by C37 and the newly added capacitor would probably be better done adjacent to the filter for V using non-electrolytics, with a wire from there to V7B pin 4. Low leakage caps are better anyway since leakage would upset the grid bias. The ratio of the size of the caps could be calculated, but the dial in a value with a decade box could work as well. Beware that connecting/changing any caps to the part of the circuit when live carries both a shock risk, and a certainly of large brief output spikes. Have a dummy resistor on the amp output when testing to avoid arcing and speaker/ear damage.
April 08, 2011 - 07:49 pm|
Metal film resistors will be fine for those resistors. R41 provides negative feedback around V3A. So R33a and R41 determine the gain of the inverting amplifier formed by that. One problem with the original design was that negative feedback around the stage was shunted to ground when the level control was set low. That left the stage gain wide open, causing extra noise (and distortion) at low level settings. Let me how it goes!
I enjoy listening to an HF20 system too, by the way.
April 08, 2011 - 06:55 pm|
|Still working on it as I have time; I'm changing all the electrolytic caps and doing the loudness control mod at the same time, as well as working on one of my HF20s and my HF81. Just a quick question: Are R33A, R34A, R41, and R42 grid stoppers or what is their function? Reason I ask is that I've heard the suggestion that metal films should not be used for that purpose because of their inductance, so I was wondering if I need to use carbon comp at those locations? |
March 20, 2011 - 09:50 pm|
Thank you for your thoughtful comments. Addressing them individually:
1. You are quite right that you could replace R27,28 instead of paralleling. The only reason for paralleling would be to give finer resolution in setting it, given the available resistor values. Since we are only reducing the existing 68K by a factor of about 1.2X, a parallel resistor has an effect "geared-down" by about 5:1. The ideal net value varies with the tube, so the suggested alternative fixed value is pretty rough anyway. Your choice of replacing them with 56.1K should be fine.
2. It is true that good metal film resistors produce less excess noise than do carbon film ones, when there is appreciable DC voltage drop. I'm not sure which resistors in the first stage you are referring-to but R7,8 would probably benefit from being metal film. Since R27,28 don't have much DC drop, there would be less to be gained there. Metal film certainly can't hurt, though.
3. Adding the terminal strips seems like a good idea. The "flying lead" works well, given that it is properly insulated. Certainly the terminal strip looks classier. Thanks for the suggestion.
4. Good points.
5. I guess you actually mean C37,38, which are 1uF 450V (can be 250V). They were changed when the value was increased from 0.25uF. (Required by low frequency stability considerations.) A film cap can certainly work well there too but will be much larger. Things are pretty cramped there if you add the AC balance pots. As far as reliability, good quality 105C electrolytic caps should have a decent service life. My unit still has its original power supply filter caps; going strong after nearly 50-years!
6. C35,36 is NPO ceramic because NPO implies the excellent linearity (low distortion) that we want there. Mica is also an excellent choice but is more expensive. Hence, I don't tend to stock as many of them.
7. Yes, the value of the feedback resistor (R85,86) and the value of the compensation cap (C35,36) would need to be changed, if a different transformer tap is used for feedback. Working out the optimum compensation was the most difficult part of this project, involving many hours of measurement, experiment and simulation. Those components work in concert with the RC networks added around V6,7, to achieve the desired stability and performance results. Unfortunately, one can't always predict what compensation will be needed on a different output tap, based on behavior measured at another tap, so I would recommend leaving the feedback on the 8ohm tap, if at all possible. That also results in lower distortion on that tap. If you must move it to another tap, here are the theoretical values for R85,86 and C35,36. 16ohm tap: 6.08K, 85pF. 4ohm tap: 3.04K, 170pF. High frequency stability into a capacitive load should be checked if you try one of those.
Regarding the question about 120pF versus the original 125pF: The fact that the new value came out close to the original is pure coincidence. With the new values of the other feedback components and the networks added to V6,7, the compensation strategy is very different from what it was before.
I hope that your rebuild turns out well and will be happy to help if you have further questions. Would love to know how it goes.
All the best, Steve