Hoping I'm not opening a Pandora's box but here goes 
I'm still in the process of designing my Rega Elicit phono clone and been wondering about this. In one of my books by a reputable author, he describes how to achieve perfect network component values for the resistors and capacitors by paralleling each component with a triming one, with the requirement that the ratio of the main and trim values be greater than the ratio of their tolerances. Of course the value measurements have to be made with a precision bridge. This is particularly helpful if one has come up with "awkward" values outside of the standard series, which is not my case.
I don't have a bridge and honestly I think it might be overkill to do this. Reason is, one of the components of the network is the output impedance of the gain stage which I believe will have worse tolerance than the 0.1% resistors and 1% capacitors the author suggests unless output transistors and their associated circuit components are also selected to be within these tolerances. And I seriously doubt manufacturers go to these lengths.
But out of curiosity I simulated my design with 1% variation factors for the passive components of both passive and active eq filters; see frequency response output plot below for 27 seven different combinations. Yes the frequency response is less flat but the maximum deviation from the theorical values is on the order of 0.3dB peak-peak for the extremes. Can this difference really be heard?
Just wondering if I need to go to the trouble of adding space for extra trim components on my PCB. I might do it for resistors as this is not that complicated but for the polystyrene caps I'm not sure as there are large variations in footprints of the parts I can get my hands on off ebay.
Thanks in advance for any insights.
-Joris
I'm still in the process of designing my Rega Elicit phono clone and been wondering about this. In one of my books by a reputable author, he describes how to achieve perfect network component values for the resistors and capacitors by paralleling each component with a triming one, with the requirement that the ratio of the main and trim values be greater than the ratio of their tolerances. Of course the value measurements have to be made with a precision bridge. This is particularly helpful if one has come up with "awkward" values outside of the standard series, which is not my case.
I don't have a bridge and honestly I think it might be overkill to do this. Reason is, one of the components of the network is the output impedance of the gain stage which I believe will have worse tolerance than the 0.1% resistors and 1% capacitors the author suggests unless output transistors and their associated circuit components are also selected to be within these tolerances. And I seriously doubt manufacturers go to these lengths.
But out of curiosity I simulated my design with 1% variation factors for the passive components of both passive and active eq filters; see frequency response output plot below for 27 seven different combinations. Yes the frequency response is less flat but the maximum deviation from the theorical values is on the order of 0.3dB peak-peak for the extremes. Can this difference really be heard?
Just wondering if I need to go to the trouble of adding space for extra trim components on my PCB. I might do it for resistors as this is not that complicated but for the polystyrene caps I'm not sure as there are large variations in footprints of the parts I can get my hands on off ebay.
Thanks in advance for any insights.
-Joris
Last edited:
As far as I know, some people can hear 0.2 dB differences in the midrange under double-blind conditions. I'm not one of them.
As far as I'm concerned, the point is really to make the RIAA correction considerably more accurate than the cartridge, so you don't get much degradation compared to the cartridge alone. I usually assume that untrimmed 1 % tolerance components are good enough to achieve that.
As far as I'm concerned, the point is really to make the RIAA correction considerably more accurate than the cartridge, so you don't get much degradation compared to the cartridge alone. I usually assume that untrimmed 1 % tolerance components are good enough to achieve that.
There's a free online calculator tool which helps you find series pairs and/or parallel pairs of components, which together produce whatever value you desire. Even better, you can tell it to constrain itself to only use the EIA standard E24 or E48 series of discrete values, to synthesize your desired result. (link)
I checked prices at my favorite distributor (mouser.com) and found that E48 series resistors with 0.1% tolerance (not 1%), were priced around USD 0.70. I can afford eight or ten of those on a stereo RIAA board, so that's what I buy. Then put them in series/parallel to get the desired value as calculated on the software tool above.
Because I also happen to own a 4.5 digit capacitance meter, I simply cherry-pick capacitor pairs to get the exact parallel value required.
I checked prices at my favorite distributor (mouser.com) and found that E48 series resistors with 0.1% tolerance (not 1%), were priced around USD 0.70. I can afford eight or ten of those on a stereo RIAA board, so that's what I buy. Then put them in series/parallel to get the desired value as calculated on the software tool above.
Because I also happen to own a 4.5 digit capacitance meter, I simply cherry-pick capacitor pairs to get the exact parallel value required.
As a side comment, how accurate can the RIAA be in a DIY tube phono stage? The RIAA EQ components are typically in parallel with the first tube, and how different can the plate resistance of a tube be when only the tube number is specified in the design? Of course tubes can vary a great deal, be made decades apart with different tooling with different manufacturing quality. I don't hear anybody complaining.
2% capacitors were the best I could find at the time. I think the resulting +/- 0.2dB is already quite good. A MM cartridge can have much larger variations if the loading is off.
IMO, the most important thing is not to have systematic error. Your pre-amp implements the 75uS pole passively to avoid the typical "+1" error in active EQ.
Ed
IMO, the most important thing is not to have systematic error. Your pre-amp implements the 75uS pole passively to avoid the typical "+1" error in active EQ.
Ed
A chart with a smaller vertical axis would be more revealing (just restrict freq to 20-20k and center vertical axis -0.5dB/+0.5dB around response at 1kHz). I hardly see on stereophile measurements better than 0.2dB and I doubt anybody would ever be able to discern a 0.5dB accurate RIAA curve from a 0.05dB and you don't even know if the recording equalization was that accurate.
Accuracy can be improved splitting and buffering the RIAA eq network but still you can call yourself lucky if you go below 0.1dB
As accurate as a solid-state phono amplifier if you allow trimming or unusual circuits. For example, with passive equalization, you could reduce the effect of the internal resistance by using pentodes (disadvantage partition noise) or cascode circuits (possibly more hum with AC heater supply). You can also use active equalization, if you are prepared to use more than the bare minimum number of valves.
By the way, you have to be careful with AC coupling when using passive equalization, see https://www.diyaudio.com/community/...a-calculator-formula-help.401437/post-7439558
You also have to be careful with DC servos, no matter whether the equalization is active or passive, see https://www.diyaudio.com/community/...ing-riaa-correction-poles.406227/post-7525018
Thank you Ed for your reply!
Can you elaborate on this please? Is that related to the divisor of the general feedback equation A = A0 / (1 + bA0) ?
Yes, but my point was specifically about DIY, where I don't much or many of your "if" qualifications actually happening.
For the non-inverting feedback topology: Vout = Vin x ( 1 + Z1 / R2 )
So the minimum possible (ideal) voltage gain is unity, when Z1 = 0.
Attachments
Very simple discrete RIAA amplifiers with the bare minimum number of active devices are typically not very accurate, DIY or not. With passive equalization they have the problem you pointed out and with active equalization they have too little loop gain.
For what it's worth, my DIY valve phono preamplifier follows the RIAA curve within +0 dB/-0.3 dB from 20 Hz to 20 kHz (measured, with respect to 1 kHz). But it is a somewhat atypical design that uses two pentodes (one of which is triode connected) and two triodes per channel.
Tell you what - I have one of those cheap LCR meters, I don't trust it much below 50 pF but it has a nulling pot and my RIAA eq caps are larg-ish 33n2 1% so I guess I could get matched pairs fairly easily and then add trim resistors only to adjust the filter pole for the actual cap value. Would change the filter impedance a bit, not sure if it's that serious. Having both resistor and capacitor trims is only useful if you need to get non-standard values IMO. I will have to make sure to maintain the 1:9 resistor ratio for the active eq though. All resistors are 1% metal film.
Last edited:
With modern feedback design amplifiers and buffers, a tube phono stage can be as accurate as one made with op amps I believe.
Minor deviations from accurate frequency response might not even be objectionable or noticeable (except in an A-B test with a more accurate phono amp), although with careful choice of components I believe you could make a very accurate RIAA curve preamplifier.
Way back I had a "hi-fi" that had multiple phono EQ curves (at least 4 including for 78s) with a rotary switch. It also had real bass and treble controls, all from the late 1950s. It was nice but I'm certain we could build a much better example today.