I'm working on the steve bench phono pre circuit and am trying to figure out sourcing for the riaa network caps. The values to go for according to salas are 6500pf (1800||4700) and 11400 pf (1000||2200||8200). What I'm wondering is, how tight do the tolerances need to be for these. I found them on mouser.com in silver mica 1% (except the 8200pf which is non-stocked - 2% is in stock). However, it would be over 50 bucks just for those caps, so I'm wondering if I can go wider on the tolerance and/or find a better deal elsewhere.
Ben
Ben
As someone else mentioned get a LC/LCR meter and hand pick your own. I feel its as if not more important for the caps and resistors in each channel to match each other so you don't notice a difference in tonal balance between channels. ie; if one cap is at one end of its 10% tolerance in one channel and the same cap in the other channel is at the other end of its 10% tolerance you will likely hear it..
Dave
Dave
There's two schools of thought on RIAA accuracy. One is, "make each channel be as tight to RIAA as possible." The latter is, "absolutel conformance is nice but channel-to-channel matching is most important." My own views tilt toward the latter. I want each side matched to better than 0.1dB.
You can use an LCR meter, but a better way (considering tube and stray tolerances) is an inverse network (same network used in both channels' measurements) fed from a very flat generator or flat input and accurate AC voltmeter measurement of the output. Undersize the caps a bit, then trim to spec with smaller caps.
You can use an LCR meter, but a better way (considering tube and stray tolerances) is an inverse network (same network used in both channels' measurements) fed from a very flat generator or flat input and accurate AC voltmeter measurement of the output. Undersize the caps a bit, then trim to spec with smaller caps.
Hi,
A trick for that is have a larger value DC blocking capacitor (0.1uF to 0.47uF @ 630V is easier to find than RIAA valued ones) in series with your accurate cap. 50V is fine for the accurate cap because there's no way any AC value in an RIAA stage will (or should) exceed the RMS rating.
Cheers!
boxerboxer said:
A trick for that is have a larger value DC blocking capacitor (0.1uF to 0.47uF @ 630V is easier to find than RIAA valued ones) in series with your accurate cap. 50V is fine for the accurate cap because there's no way any AC value in an RIAA stage will (or should) exceed the RMS rating.
Cheers!
boxerboxer said:
In the UK, Farnell do polystyrene and silvered mica in 1%. The fact that valves/tubes have tolerances of +/-20% is not necessarily an issue. It will all depend on the design.
I would agree with SY regarding accuracy. A function generator has an inherently flat response. Or you could search the second-hand electronic instrument shops for an old component bridge.
Ouch.
Perhaps I should
, I really do not normally differ from EC8010! 
But when I take a 'holographic look' (such a nice, cultured term to use ...), I come up against worse ups-downs in loudspeaker response a.o., with steeper rise times - and these differ from model to model. My point is, it is not going to be the steepness of the characteristic that is going to suffer with tolerances of a few%, just the position of the curve. There is still basically one R.C, thus 6dB/octave in the limit - not that steep. (Then what about equalisers? One gets 12 - 18dB/octave there.) I have not simulated - it might broaden my horizons to do so! - between different 'curves' resulting from a few % deviation in C. (... there, EC8010! Now you have done it! I am now curious and will go to bed even later after having added some 'Spice' to my life. All your fault. ).
In fact, I am known to be naughty. I mostly use RIAA in a feedback topology, and I find it easier to come right with some sub-sonic attenuation, if I allow a slight rise from about 80 Hz, peaking to +2 - 3dB above correct RIAA at 30 Hz odd. What with everything else usually starting to drop in that region, I had nothing but praise for 'maintained solid bass' resulting from such (not revealed, of course!).
Edit: spelling
Perhaps I should
, I really do not normally differ from EC8010! But when I take a 'holographic look' (such a nice, cultured term to use ...), I come up against worse ups-downs in loudspeaker response a.o., with steeper rise times - and these differ from model to model. My point is, it is not going to be the steepness of the characteristic that is going to suffer with tolerances of a few%, just the position of the curve. There is still basically one R.C, thus 6dB/octave in the limit - not that steep. (Then what about equalisers? One gets 12 - 18dB/octave there.) I have not simulated - it might broaden my horizons to do so! - between different 'curves' resulting from a few % deviation in C. (... there, EC8010! Now you have done it! I am now curious and will go to bed even later after having added some 'Spice' to my life. All your fault.
).In fact, I am known to be naughty. I mostly use RIAA in a feedback topology, and I find it easier to come right with some sub-sonic attenuation, if I allow a slight rise from about 80 Hz, peaking to +2 - 3dB above correct RIAA at 30 Hz odd. What with everything else usually starting to drop in that region, I had nothing but praise for 'maintained solid bass' resulting from such (not revealed, of course!).
Edit: spelling
Johan Potgieter said:
That's exactly right, and that's why I fret about tolerances. Mispositioning one curve relative to another causes a shelf error, and although loudspeakers typically have peaks and troughs of 2dB or more, it's important that their general trend is correct. Worst of all, typical RIAA errors tend to cause shelf changes across the ear's most sensitive region - but I expect you've seen that now that you've played with SPICE.
Mouser sells Wima polypropylene film (FKP2) caps in 2 and 5% tolerances at very reasonable prices. They are cheap enough that you can buy a bunch of them and hand match them.
Very important that you match the caps so they are as close as possible channel to channel as well.
B&K used to sell a capacitor meter for about $80, and a lot of modern DVMs have the ability to measure capacitance. If you don't own one of these, make the investment.
Very important that you match the caps so they are as close as possible channel to channel as well.
B&K used to sell a capacitor meter for about $80, and a lot of modern DVMs have the ability to measure capacitance. If you don't own one of these, make the investment.
Geek said:
Yes, the polystyrene caps are rarely available in voltages above 100VDC -- you can find some 160VDC, 500VDC but these are comparatively expensive. Mouser sells PS Caps, so does Nebraska Surplus.
WRT an inverse RIAA network -- this a scope and sweep generator are all you need, but you can do it by hand. You might find that your hand-held DMM is not very accurate over a few kHz, however.
I wrote a routine for an HP3577 network analyzer does the analysis on the fly (remembering that you have to change the input impedance of the analyzer!).
You can drive yourself batty designing a tube RIAA network, and I am exempla par gratia. Gotta remember not to overdrive the darn thing!
IMHO, a properly designed circuit, ss or tube, will not change its response much with age or as tube characteristics change. You want the RIAA curve to be accurate and the best way is buying a number of less expensive polypropylene caps and measuring them. I like polystyrene in ss circuits, but the difference between the two is minimal. Silver mica sounds good to me, but prices are high and it has a high dissipation factor. Go with the film. I try to select my RIAA parts to 1%, but match the channels to half that or better. It's pretty trivial to make a single range bridge for the values you're talking about, if you have one known value cap and a good DVM to measure resistance. But, a used 1650 or a new cap meter is a lot easier.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.