Testing a Preamp's RIAA frequency response

Since you can't really trust a test record (I can identify the flaws in mine's recording, starting with tape head alignment), I thought I would do some frequency response testing on the rest of my hardware.

First up, I create a five minute white noise 96/24 file, and play it through my sound card while recording it with a loopback cable - let's see how good the frequency response of my sound card is (this would be output + line-in frequency response, so we can't really know which is contributing more).

After recording, I did a maximum size FFT analysis (for low frequency accuracy) of the whole recording in Audition 3:



The response of my sound card is +0/-0.5dB from 20Hz to 40+KHz. More amazingly, the levels between channels are within .01 dB when I get RMS stats. If I recorded for over an hour and analysed, it would get rid of even more of the random "fuzz" in the frequency response graph.

Now, I do the risky thing, plug the output of my sound card directly into the phono preamp (no capacitance or load compensation, etc, just pure voltage), and slowly turn the volume up... nothing goes poof and I still have a -60dB noise floor. What does this analysis look like?



What do we see? Why it's the RIAA curve, the equalization that is applied to LPs when cutting, which is removed by your preamp on playback. I can see that the channel balance is off in the bass frequencies, confirming my previous needle-drop observations, and giving me an answer to why - it's not the cartridge, it's the preamp!

For further analysis, we need to "fix" this RIAA curve. Am I going to equalize the sound or something? No - I am going to export the data points from the FFT analysis and see how the measured frequency response compares to actual calculations for what the filter curve should be at specific frequencies.

This site gives us the formula for calculating the RIAA frequency response compensation for a particular frequency. it looks like this:



where f is the frequency, and the t's are the three different time constants 75, 318 and 3180 microseconds used for the combination of filters.

Fortunately that site also has an Excel spreadsheet so I didn't have to type all that into Excel myself. Some cut and pasting, and we can calculate the frequency response for each of the frequencies of the FFT analysis that was exported:



The FFT raw data is columns A-C. In columns D and E, I normalized the response by adding 37dB. The calculated RIAA response is column F. The difference between the calculated curve and the analysis is columns H and I. Now we just have to chart the differences:



Voila! That's what my preamp is doing to the sound! It looks like the preamp filter components (with a small contribution from the sound card response) don't quite create the right curve, but now that I know this frequency response, with some equalization I can compensate and balance the channels to turn this back into a flat line.

With this knowledge, you have further topics to investigate - eliminate the preamp's contributions, and you can judge your cartridge by itself or your test records.

 

What specific test record are you using? If it's a name-brand -- CBS Labs had some great ones in the 1960s and 1970s, and even the newer ones from Hi-Fi News & Record Review are good -- it should be fine.

Every time I've played back the test signals through my Krell KSP-7B preamp, it's flat within about .1dB all the way to about 18kHz or so (the limits of the signals). I was floored by how accurate it was.

BTW, there are software programs that will just take the raw output of a turntable and apply a "virtual" RIAA curve to it to apply the correct playback EQ after the fact.

 

I have a cheapo, so it's more a matter of having a quality one I would suppose. If the same lathe and equipment cuts your test record as is cutting your music, that would be the best way to analyze how well their music is getting to you.

I may go direct line in and check if the quality is better (although many purist threads here talk about correct loading for the cartridge and such). I would have to see how rumble and crackle compares to a -108dB noise floor at such low input levels and if I still have 16+ bits of sound.

 

I've been able to get cartridges that flat as well with my Levinson preamp. I used to use a HP-400 Meter a pencil and some graph paper to record my results.

I have the following test records:

CBS STR 100 - Professional Test Record
CBS STR 112 - Square Wave Tracking & IM
CBS STR 120 - Wide Range Pickup Test
CBS STR 130 - RIAA Frequency Responce
CBS STR 170 - 318-MicroSecond Frequency Response
Shure TTR 102 - Phono Cartridge Test Record
Shure TTR 103 - Phono Cartridge Trackability

 

I have those three, as well as a couple of Stereo Review test LPs from the 1970s and a couple of the HFNRR records.

The pink noise passages are useful for checking phase and tracking, and the response sweeps are pretty conclusive. As I think I mentioned one other time, there have been moments where I was sure an LP was down about -5dB or so, but the test record showed me it was as flat as can be. Just a decision by somebody in mastering.

The CBS Labs record is widely available for like $5 on eBay. They must have sold hundreds of thousands of them back in the day.

Man, I've never heard a record that could even get close to a S/N ratio of even 60dB at best. And quite often, I could still hear the tape hiss cut in and out on (very clean) vinyl, so the analog tape for the master is at a higher level than my system. This is why I kind of question people who insist on 192kHz/24-bits for recording vinyl.

The other scary thing is separation. Some of those test signals make it pretty clear that you're very lucky to get even 25dB of separation between vinyl channels. Whole lotta leakage goin' on.

 

Just so you know, I created a new testing methodology and silently updated my charts in the first post.

Instead of using white noise, which has random bumps in the frequency response even with a long sample time, I created a test tone. So that the tone would cover all frequencies in a short amount of time and also be correctly analyzed by FFT, I created a logarithmic sine wave sweep from .48Hz to 48KHz, and then applied a logarithmic volume envelope to weight this correctly, from -51dB at 1Hz to -1dB at 48KHz. (these levels somewhat mirrors the RIAA preemphasis too, so it keeps levels reasonable when analyzing a preamp). Analyzing the tone gives me a frequency response graph that looks like this:



That would be +/- .0005 dB for you and me. There's just some negligible aliasing artifacts around 2k.

Here is the test tone:Analysis-sweep-1Hz-48000KHz-96-16bit.flac

0-50 seconds is a 300hz+3000hz+30000hz tone at -1dB (with -0.3 dB modulation every 5 seconds). This allows you to adjust recording/playback level, and detect clipping or distortion by looking for 600hz or 6000hz signs of second-order distortion.

50-60 seconds is full band white noise at -20dB, this plays to make sure that frequency response goes above 24KHz (I had a problem where Windows 7 would keep dropping the output sample rate to 48KHz if anything else had played audio)

The remaining six minutes is the tone sweep. It starts with a very low level so you likely won't hear it for a while. A full FFT analysis of this is the ultra-flat frequency response you see above. Play this through your equipment and analyze...

--

I am thinking my thirty-year-old Parasound preamp might need some re-capping to fix the filters (but I might just make it worse...). Fortunately the manual includes a schematic, here's it's phono stage with lots of capacitors!:

 

Back in the predigital day, we used a simple passive Inverse RIAA stage, which had the side effect of causing enough loss that we could test with a conventional oscillator (actually the oscillator half of a distortion analyzer) and with hand chosen parts in that stage and carefully calibrated measuring equipment honestly spec a properly operating phono stage to be within .2db or better across the 20-20K spectrum. I don't have the circuit we used back then, but Jim Hagerman's circuit here is easy to build and should work fine

http://www.hagtech.com/pdf/riaa.pdf

 

Problem identified - Problem solved. I replaced my 1983 Parasound with a 1998 mid-hi Kenwood receiver for doing needle drops (that I forgot had a phono in). Here is it's response compared to the Parasound in gray:



As a bonus that will never be needed, the phono input can take a lot more level before it starts distorting.

 

oops, didn't mean to double-post...

 

Bravo for your good work here. I completely agree that a test record would be of no use in determining the frequency response of the preamp. Actually, since there is no way to objectively know what is cut into the vinyl, a test record is of very limited use at all.

At any rate, there is more to a phono preamp than just a flat frequency response. Are you sure the Kenwood receiver is superior sounding to the Parasound preamp (other than RIAA correctness)?

Again, good job!

 

The sound seems just fine, it's not like you could make vinyl worse, with it's crappy channel separation, poor sound stage, sibilance, and crackle, (and backcueing groove damage from DJing...); given qualified mastering, 1949 tech can't beat quiet 44/16 (see if I can make somebody here mad). I wanted to like the Parasound more, since there's nothing in there but analog transistors, resistors, and capacitors.

With a cartridge, there is a more dominant 60Hz (and all orders of harmonics) in the noise above the hiss of transistors than I had previously noted in the other preamp, likely the turntable being next to and the preamp being in the same chassis as 500 watts of power amplification. With a 12" 45rpm disc with peaks at -3dB, this noise is -85dB RMS unweighted though, at least an order lower than the -75dB weighted turntable rumble spec.

 

It's the equalization just OPPOSITE to the equalization applied to LPs when cutting:
http://www.smartdevicesinc.com/riaa.html
(FIGURE 1),
and the RIAA playback curve that your preamp precisely displayed, applied to the signal sent by a magnetic cartridge from an LPs when playing, makes the resultant FR flat. Except for the roll-off of the FR below 30Hz (appr. 12 dB from 35 Hz to 10 Hz and so on) which is to apply to the playback FR (“subsonic filters” etc.) to avoid subsonic parasites, heavy inter-modulation, rumble etc.

I didn’t understand what you meant by this. “Fix” what and equalize what, if your preamp displays precisely the RIAA playback curve?

 

This is why real test records are not cheap.

Preamps' RIAA accuracy is generally measured by feeding in test signals from electronics; these days a computer is used.