ARC SP-6A Tube Preamp

[Atmasphere]

Older resistors of high values (1M or higher) are always suspect in any renovation.

 

[chiily]

Older resistors of high values (1M or higher) are always suspect in any renovation.

Indeed, a lesson learnt

I don't think I've come across a resistor that has become open cct before. So burnt ones yes, but just "naturally" I've not seen before.

This old SP-8 just keeps on pulling surprises.

 

[Atmasphere]

Indeed, a lesson learnt

I don't think I've come across a resistor that has become open cct before. So burnt ones yes, but just "naturally" I've not seen before.

This old SP-8 just keeps on pulling surprises.

Whenever I do a renovation I look for higher value resistors. Its rare I find one that is actually the value that its supposed to be. I first noticed this when working on solid state gear 50 years ago. Even new, cheap carbon film +1M resistors would just open up. I never replace carbon film or carbon composition parts of these values with similar parts. I always use metal film for that.

 

[chiily]

So here we are....

In the PSU I replaced R63 as it was O/C, added a 47uF cap for C17 (I have some low ESR 22uF caps that I might swap in), and, using a 20k, multiturn pot attached to the leg of R51 determined that R51 could be replaced with a 348K (should be 342K). What I doesn't understand is how come a 332k was factory fitted?


In the MM stage replaced both R9 and R50 on both channels.


So, now B+1 is 403V and B+2 is 393V...yeah...ripple is around 3mV.

Previously, when switching over to MM input and turning the volume up I could hear the rustling, crackling, about to burp/splat noises. Now I can't hear anything, other than a bit of hum and hiss, nothing out of the ordinary.

The anode voltages in V4 and V5 still oscillate over about five voltages whenever I first measure them, they do settle after about 15 secs.

Just a couple of albums in, but no return on any erroneous noises, just music...could this be the end? Can I finally put the front back on with the handles?

I did do a little cosmetic work, I know only a tiny bit of the black comes through the front panel, but I know it is all black underneath.



Going to run a couple of soak tests...

Thank you everyone who had pointed and poked, suggested where to look... especially @Atmasphere

 

[Atmasphere]

Good work!
I think you have it sorted. If you still encounter the popping sound, it might be from a source (AC line disturbance) that is overcoming the regulation of the power supply. With it working properly, those events are less likely to do what they were doing- if that is what was going on at all.

 

[DonH50]

Good work!

The 332k original value depended upon everything else in the circuit, so changing values of various components (as well as tube bias, gains and such) over time, requiring a slightly different value now is not unexpected and would not concern me.

IME it is normal for voltages to vary as the tubes (and everything else, like zeners and resistors) warm up so again would not concern me. The rails of my SP3a1 took a good five to ten minutes to stabilize, and my old D79 was not much better.

 

[chiily]

Thank you both for your kind words, I wouldn't have been stuck without your invaluable help. It has taken a little time with plenty of false leads, however I have learnt so much as I've gone along.

An usual pairing in my office cum workshop, SP-8, Naim NAP 160 and a pair of DIY speakers. Sounds fabulous with a bit of Monk and not a splat or unusual noise from the SP-8 all day yesterday! If the SP-8 behaves itself for a week, I'll take it upstairs and plug in to my main system of a pair Radford MA15s and Quad 57s. I do also have a Radford TT100 and Klipsch Forte IVs, lots to try



And for metal film resistors as grid stoppers, not a problem at all!

Here's to another 40 years service from the SP-8.

 

[Atmasphere]

Excellent!! Happy to hear things are sorted out. That took a lot of perseverance on your part.

 

[chiily]

As a follow up. I removed the two 47uF capacitors across the B+1 and B+2 rails and I'm sure the sound quality improved a little. I may experiment with a smaller values.

I ran a couple sweeps with REW measuring BW and distortion. BW was flat from 20Hz to 40kHz, with THD, with noise, not more than 0.07% across the BW.

I'm really enjoying the pre-amp

 

[Atmasphere]

As a follow up. I removed the two 47uF capacitors across the B+1 and B+2 rails and I'm sure the sound quality improved a little. I may experiment with a smaller values.

I ran a couple sweeps with REW measuring BW and distortion. BW was flat from 20Hz to 40kHz, with THD, with noise, not more than 0.07% across the BW.

I'm really enjoying the pre-amp

I suspect you won't want more than about 22uf. Glad its working out!

 

[DonH50]

Why would the sound quality change with decoupling caps? Ralph @Atmasphere ? I can think of a few things, but barring a problem with the rails or regulators I have a hard time seeing an audible change, except perhaps "looser" rails affecting transient behavior.

For broadband decoupling, more smaller caps are usually better, but resonances among caps can be an issue...

 

[Atmasphere]

Why would the sound quality change with decoupling caps? Ralph @Atmasphere ? I can think of a few things, but barring a problem with the rails or regulators I have a hard time seeing an audible change, except perhaps "looser" rails affecting transient behavior.

For broadband decoupling, more smaller caps are usually better, but resonances among caps can be an issue...

Re.: your question, The output impedance and bandwidth of the regulator isn't properly matched to the caps, which are supposed to bypass the regulator at the frequency at which it is rolling off. I doubt that test was ever done when this circuit was designed.

If the caps are too large, and the output impedance of the regulator thus too high (my theory is that's happening here), the regulator can't refill them properly so you can get low frequency drift as the regulator hunts.

Another problem is if the cap is too large and its just not that good of a part, it might be less effective at high frequencies. You really want the regulator to handle as much of the noise generated in the supply rails by the audio circuit as possible. So unless the regulator is truly bad, the filter caps at its output are likely to be smaller rather than larger.

My surmise is thus the power supply has lower noise with the parts removed than with them installed. But its likely it can be even lower noise with the right values installed instead.

 

[DonH50]

Re.: your question, The output impedance and bandwidth of the regulator isn't properly matched to the caps, which are supposed to bypass the regulator at the frequency at which it is rolling off. I doubt that test was ever done when this circuit was designed.

If the caps are too large, and the output impedance of the regulator thus too high (my theory is that's happening here), the regulator can't refill them properly so you can get low frequency drift as the regulator hunts.

Another problem is if the cap is too large and its just not that good of a part, it might be less effective at high frequencies. You really want the regulator to handle as much of the noise generated in the supply rails by the audio circuit as possible. So unless the regulator is truly bad, the filter caps at its output are likely to be smaller rather than larger.

My surmise is thus the power supply has lower noise with the parts removed than with them installed. But its likely it can be even lower noise with the right values installed instead.

Got it. Regulator and noise bandwidth, you'd think the design engineer would have analyzed and measured that carefully, but my own ARC experience argues otherwise. Have to watch loop bandwidth and compensation with the caps gone to ensure it doesn't oscillate, but he'd likely know that by now.

Thanks! - Don

 

[chiily]

The output impedance and bandwidth of the regulator isn't properly matched to the caps, which are supposed to bypass the regulator at the frequency at which it is rolling off. I doubt that test was ever done when this circuit was designed.

I'm curious. How would one go about measuring the impedance of a regulator like that in the SP-8?

My surmise is thus the power supply has lower noise with the parts removed than with them installed. But its likely it can be even lower noise with the right values installed instead.

And that's interesting because, I think as we spotted before that originally the B+2, for the line stage 402Vdc, had a 200uF capacitor across its rails. This cap was then removed in later versions of the amp, for sonic benefits, as the schematic note says.

 

[Atmasphere]

I'm curious. How would one go about measuring the impedance of a regulator like that in the SP-8?

And that's interesting because, I think as we spotted before that originally the B+2, for the line stage 402Vdc, had a 200uF capacitor across its rails. This cap was then removed in later versions of the amp, for sonic benefits, as the schematic note says.

To measure the output impedance you'll have to inject a frequency onto the supply rail using a capacitor of such a value that its essentially 0 Ohms at that frequency- such as a 0.47uf at 1KHz. The source of the AC signal, driven by a sine wave generator, might have to be fairly low impedance (so the generator might or might not do on its own) and you'd put a variable resistor (probably no more than 25K) wired as a rheostat in series with the lot. Then you measure the 1KHz tone across the control vs that on the supply rail and adjust the control until they are equal. At that point the control (as long as the source impedance is low) should be the same value as the supply rail.

Then you would increase the frequency until the signal Voltage increases on the supply rail by 3dB. This is the roll off point of the regulator. Then you would have to calculate the value of the bypass cap such that its impedance is low at that frequency but increasing at lower frequencies.

Its tricky because if you're not careful you could damage the source and/or the regulator by a misstep. This approach is not going to be perfectly accurate, but as they sometimes say in the business its 'good enough for rock and roll' or 'good enough for government work'. My uneducated guess is this value will be between 2uf and 22uf. You could just install a 10uf cap and call it good...

 

[DonH50]

I used a network analyzer (sometimes in conjunction with a high-power amp, but with a good four-port VNA you can use a scheme that obviates the need for the amp) but that's not a practical solution for most hobbyists (or me now I've retired and don't have access to a nice lab of toys).

Personally I would not try to measure the impedance over frequency; it's tricky as Ralph implies and too much risk of things going wrong (from accidentally shorting something to causing the regulator to oscillate and destroy itself during the probing and signal injection portion). It's also hard to get the true loop response with output decoupling capacitors in circuit, and breaking the regulator's feedback loop is often non-trivial and even riskier to measure (though is most accurate). If you've measured the rail with a 'scope or spectrum analyzer (as I believe you have), and it's quiet, then I wouldn't worry about it. Or throw a 10 uF cap on it as Ralph @Atmasphere said and call it good.

If you just want the DC impedance, unload the rail (or pull a few tubes to lighten the load), measure the voltage and current, then repeat with load, measure again, and turn the crank to calculate the output impedance from the difference in voltage drop under load. For rails with a series resistor in the circuit, you can measure the resistor (with the amp off!) and then calculate current draw by the voltage across the resistor (I = V/R).

 

[Atmasphere]

@DonH50 makes a good suggestion here- likely easier than what I was thinking!

 

[chiily]

Thank you for the detailed explanations. I think I might leave the measuring of the AC impedance of the regulator to a professional. I would not like to break the SP-8.

Ah, I have previously used a 24k 10w resistor across the rails to act as a load without any valves installed, so that's about 16-17mA on each rail. I have then inserted all the valves, to steadily increase the load on each rail, the regulated voltages didn't drop at all. Not an impedance test, but indicates that the regulator is regulating the voltage with different loads.

I could decrease the resistor valve to a point that the regulator stalls, dropping the voltage. I would have to be very careful...maybe I should relish the music and stop being so curious about how it works

I will try different cap sizes, I have 10uF, 22uF and the 47uF 450V at hand to gauge any perceivable change in SQ.

I did manage to source three replacement toggle switches, so I swapped out the mute switch as a precaution, based on the fact that it was probably the most used.

I just need a cheap, broken SP-10 to fix now

And I moved the ARC upstairs...excuse the mess...

 

[Atmasphere]

I could decrease the resistor valve to a point that the regulator stalls, dropping the voltage. I would have to be very careful...

Yes indeed! That's something I would be hesitant to try, given the fragile nature of ARC regulators from that era.

But it will need some kind of bypass. That is why I suggested 10uf. If you hear no changed in sound quality, that's a good thing. If it gets better, that's good too.