A Precision Inverse RIAA Filter For Testing Phono Stages

[spaceistheplace]

Looking at available options for Inverse RIAA Filters, Jim Hagerman's iRIAA2 looks like it fits the bill.

Why an inverse RIAA? Well, one can cut the turntable, tonearm, cartridge and other variables out of the equation, drop a standard line input to phono levels and as a result get more useful performance data.

However, the parts supplied with the iRIAA2 give us only +/- .5db accuracy, which would not be very useful for testing a precision RIAA network. For the additional "burn in" uses Jim describes, I'm sure this would be fine.

From the manufacturer:

The iRIAA Filter is a two-channel passive inverse RIAA response filter for use in testing phono preamplifiers. Unlike traditional networks, the iRIAA Filter includes a correctly placed upper 3.18us (50kHz) corner in its transfer function. Output level is switchable between -40dB and -60dB. Frequency response is accurate to within +/-0.5dB from 10Hz to 100kHz.

His white paper on the topic: http://www.hagtech.com/pdf/riaa.pdf

The schematic:




So, we have an excellent little filter here that suffers from a lack of precision parts selection. For example, for testing the 834p that many in the forum have built this would not give us very reliable results. This would be equally useful for the CNC designs here on LH as well as most others designed for both MM & MC cartridges.

Using this PCB and the BOM below you should be able to get better than +/- .1db, which would be equal to or better than what we are shooting for with the 834 and most other precision phono stages.

All parts are available from Mouser. However, there's a shortage of inexpensive 6n2 1% micas. Why micas? I was unable to find 1% polypropylene capacitors in the required values and voltages. Luckily, there are two locations on the pcb for using 2 caps to get this value. With a little looking around deals can be found to lower the parts cost. I stuck with what was available from one common retailer.


The BOM:



iRIAA2 Purchase link:
https://www.haglabs.com/collections/accessories/products/iriaa2-inverse-riaa-filter

Assembly should be straightforward so I will not rehash here.

Now, time for some testing!

If you have experience testing phono stages using Audacity or other software, feel free to put your experiences here so others may benefit.

Many do not have access to scopes and signal generators. My hope is to reduce the complexity / barriers in testing ones diy Phono stage using more commonly available items.

From what I have found on the web there is almost no information on straightforward step by step test methods using common tools (computer, audio interface). As I make progress in this area I hope to post my findings and a how-to for others.

[spaceistheplace]

An example of how one might use an Inverse RIAA filter is shown here, beginning on page 6: http://hifisonix.com/wordpress/wp-content/uploads/2016/12/Accurate-Inverse-RIAA.pdf

[Loggie]

Highly interesting !

My Douk 834 is still under construction but I can use my Muffsy pre-amp and my signal generator and scope to try out the test mentioned in the article.
Its an old Tektronix 2215 but I'll try and make some pictures.

-Arno

Update: Erm.....ofcourse I would need to get an inverse RIAA filter first 

[Jessica_K]

Have been thinking of this myself. Accuracy is the biggest problem. Single components in the filter will not give the accuracy you need to measure the phono amp. Using multiple parts of as high accuracy as you can get will improve it but you still have the second issue of both source and measurement. The source will not be able to maintain a constant amplitude over the complete frequency range and it would require a very expensive RMS meter to measure accurately over the frequency range too

All is not lost however. If you use the network just as an approximation of the signal so that over the range the signal represents a RIAA envelope then you can spot measure a series of frequencies. Measuring the input to the amp (output of the network) and the output from the amp and plot the deviation from the calculated curve. This should give a good measurement of the RIAA

You would need 2 meters as the loading of the meter on the network will effect it and also do the test twice swapping the meters round and taking the average

Robert

[wer]

Would like to read a comment by Robin or anybody else who was involved in record production. I can't believe that a recording engineer would have given a hoot whether the RIAA pre-emphasis was followed precisely, or was off by half a dB here or there, as long as the end result was ok.

[Jessica_K]

I would hope recordings were as accurate as possible to the RIAA standard, but I have commented before how dire some recording are so never know. As a designer however can only work with the spec and get as good as one can.

Robert

[fetteler]

Isn't the RIAA equalisation part of the cutting lathe electronics? Just switch it in and off you go - perfect every time, assuming that is that the lathe manufacturer took care to get it right which I'm sure they did.

Steve.

[spaceistheplace]

Have been thinking of this myself. Accuracy is the biggest problem. Single components in the filter will not give the accuracy you need to measure the phono amp. Using multiple parts of as high accuracy as you can get will improve it but you still have the second issue of both source and measurement. The source will not be able to maintain a constant amplitude over the complete frequency range and it would require a very expensive RMS meter to measure accurately over the frequency range too

All is not lost however. If you use the network just as an approximation of the signal so that over the range the signal represents a RIAA envelope then you can spot measure a series of frequencies. Measuring the input to the amp (output of the network) and the output from the amp and plot the deviation from the calculated curve. This should give a good measurement of the RIAA

You would need 2 meters as the loading of the meter on the network will effect it and also do the test twice swapping the meters round and taking the average

Robert

Robert my understanding was that using a quality sound card and an inverse RIAA I'd at least be able to get some general "beginner" measurements using a sine wave sweep out from the soundcard, through the inverse RIAA, into the Phono preamp, into Aux in preamp and then from tape output in my pre back into the soundcard.

I believe that is what you were referring to in your last paragraph? I'm having trouble understanding what you were intending to express.

Is .1db not effective enough for inverse riaa measurements?

All of my parts in the inverse are matched to 0.1% or better on a recently calibrated Fluke 87V which I believe can also do RMS measurements? I believe that will give a much better response than the included kit parts which were 1% resistors and 5% films giving 0.5db error.

I'd be using either the Rightmark or audacity.

I was hoping to get a simple configuration together to run a series of tests at least to introduce myself.

The technical issues involved are a bit daunting from what I gather.

Is there some method that I could use what is already at my disposal (using a laptop / audio interface to generate a signal and then plot it?) or am I out of my league and require some more precision bench device?

Audacity also has an inverse riaa function it can apply digitally.

Soundcard specs: http://www.creative.com/emu/products/product.aspx?pid=20347

Preamp specs: https://www.dbsystemsaudio.com/DB-1B2a.jpg

[fetteler]

.... I believe that will give a much better response than the included kit parts which were 1% resistors and 5% films giving 0.5db error.

I'll bet the issue here is not the tolerance of the components themselves, rather it is the compromise in the RIAA characteristic that results from using standard values. Even if all your Rs and Cs were exactly the value shown on the circuit diagram the resulting EQ curve would still be in error. To get it really close I reckon (without doing the circuit modeling and getting the calculator hot) it's necessary to combine components to get non standard values.
Hope this makes sense

Steve.

PS Checking the Fluke 87 spec shows it's accuracy is 2% ish above 1kHz. This is about 0.17dB

[Jessica_K]

Parts list is good but tolerance errors are acumulative when in series so a 0.1% resistor in series with a 0.1% resistor could give an error of .2%. Where as in parallel the error is the sq root of 2 times the error so the two in series would give an error of 0.07. So you can do quite good with the original design and expect to probably with selection get to within 0.5 db all in inc temp drift

As for source and measurement these are all valid and certainly can do a great job of measuring to within a db. This well worth the effort. Also feeding back the IRAA directly into the PC you can create a curve that can then be used to calibrate the loop. This would get you very close to the real response with only thermal drift and system instability being the error factors

Steve that is true too the values are never exact and you have to take into account loading, any signal source will have an effect on the circuit and any load on the output. So it's always one big compromise. Hence calibration is better than absolute measuring
 
Robert

[spaceistheplace]

Robert,

My reasoning is to better understand the low frequency response deviations that can occur from the 2M / 680k resistor based upon which tube / resistor combination I'm using.

I imagine the method you detailed would at least give some general clues in this regard for more closely matching riaa under 100hz.

From your simulations with parts list matching you suggested a .1db error above 100hz so I don't think this would be worth analysis with the equipment I have at my disposal.

[spaceistheplace]

Could this help minimize some of the issues involved?

http://www.pmillett.com/ATEST.htm

[fetteler]

Steve that is true too the values are never exact and you have to take into account loading, any signal source will have an effect on the circuit and any load on the output. So it's always one big compromise. Hence calibration is better than absolute measuring

Of course Robert I was just pointing out that without considering any other factors the circuit values given are not perfect

Cheers,
Steve.

[fetteler]

Parts list is good but tolerance errors are acumulative when in series so a 0.1% resistor in series with a 0.1% resistor could give an error of .2%.

I started thinking about this, you had me fooled at first Robert but now I'm not at all sure you are right, sorry...

The overall tolerance of a string of 0.1% resistors is still 0.1% regardless of how many individual resistors are in the string.

The reasoning goes like this:
A 100K 1% resistor in with its worst case increase in value (1%) will be 101K. If this is connected in series with another 100K resistor which has the same error in the same direction (i.e. it too is 101K) then the total resistance will be 202K. This is still 1% high. So errors in series do not add up and (assuming the same tolerance for each component) the total error can never be more than the error of the individual components. Of course there is usually some statistical spread in the individual errors so a string of series connected components is very likely to have a smaller error than any of the individual component's tolerance ratings. 

Thinking further, for the worst case scenario of two or more components all at the same limit of their tolerance in the same direction and connected in parallel the resultant combined resistance will be in error by that tolerance. So, for 1% resistors, all high by 1% their resultant parallel resistance will also be high by 1%. Again with a spread of errors within the tolerance the combined error is likely to be lessened.

The square root thing comes in when you are doing a statistical analysis to predict variations for large numbers of the same circuit configuration. There will always be 'unlucky' combinations which have the maximum error though.

Cheers,
Steve.

[Jessica_K]

Thanks Steve think I had a bit of a senior moment there of course you are right.