Who's measured the noise on their RIAA preamp?

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I made a quick-n-dirty RIAA phono pre-amp with 3 op-amps per channel. I measured the input signal as about +-15mV, and I wanted the output at about +-1.5V (to be followed by a passive attenuator; update on actual output levels coming soon).

When the input is disconnected, the output noise voltage is about* +-10mV, and jumps up to +-30mV when the recordplayer is connected. This appears to be mostly random ultrasonic noise, and includes a 50Hz ground ripple from the 12V adapter in the wall plug. (*About 95% of the noise that's visible on a scope easily fits between +10mV and -10mV).

Improvements I can make to my design (schematic coming soon!) :
-Better schematic with fewer design errors.
-Next time not put the whole thing inside a metal box with exterior dimensions of 58mm x 64mm x 35mm. :D It was just asking for parasitic component values and parasitic feedback etc.
-Better power supply, but on the other hand I think those single-ended "battery eliminators" are quite convenient.

So, how does your phono preamp perform?

CM
 
Noise - wideband.

For a 1.5 Volt output your (output ) noise must be 0.5mV to get -70dB S/N ratio.
You will have to short the input of the preamp with say a 100 ohm resistor to get some meaningful noise figure.
As for hum:
1. In an opamp circuit you should not even see it !
2. Your shielding may be poor.
3. Your power supply must be trash - the opamp can take a bit of crap in any case.
4.Your Phono cartridge to pre-amp wiring may have to be sorted out.
5. You will need a mains earth .
6. Check for ground loops.
7. There should not be any LF noise with half decent opamps at these levels.
Put up a circuit and pictures of the unit. Then more people can give you practical suggestions.

Better not progress to tubes till you have all this sorted out.
Cheers.
 
Wall warts are crap when it comes to 50Hz (or 60Hz if that comes out of the plugs). A small lab supply should not be expensive and far better. Or build one yourself. As the riaa curve amplifies at 50Hz quite a lot of dB's more then at 1kHz, whatever your cartridge and wiring picks up gets amplified a lot. So re-arranging your cables can help.

I don't think what you get is from parasitics of your box. But it could be bad grounding inside your box.
 
As I said, the circuit was "quick-n-dirty", and I know that there are several things wrong it, so I'm not worried at all that the noise is +-10mV. It is +-10mV whether the input is open or shorted; most of the noise is picked up by the turntable, which may be crap but it's a great antenna ;) .

Unfortunately the schematic for, er, version 0.9 is still just a hand-drawn scribble, but when I post it I will also post a much improved version so people can actually see what I'm rambling on about. A lot of the noise is internally generated because of too much gain and some missing components. The mains hum within the +-10mV is due to the single-ended cheap adaptor, and I simply used a few resistors and capacitors to create a bias voltage at half-way rather than use ground as the reference. The bias voltage is a different reference to the one I used for my active crossover - which also happens to use a single-ended adaptor.

Maybe I got off to a bad start, I'm just interested in how much noise people get from their phono preamp, so that helpful comparisons can be made between different designs. It's all very well just admiring the idea of discrete JFETs on the input and using ultra-linear MOSFETs to avoid using negative feedback, but how well do all these different designs actually work?

CM
 
Small update: you made me curious, so I did a "measurement" of my riaa preamp. This is a Barry Porter balanced input design with a couple of LM394H as input transistors, rest are NE5532A/NE5534A with an LF411 as DC servo.

How did I did the measurement: made a plug with a 100 Ohm resistor and connected that to the input of the LEFT channel, the right channel was left open. The output of the preamp was connected to a RME ADI-8 AE AD convertor, and so it was piped into the pc. Recorded 1 minute of the output into CEP at 32bit float/44.1kHz. Then took 30 seconds out of the middle and had a statistic made of the taken samples. After that did a frequency scan of the same selection with 16k point fft and blackman window. This gives a very low noise floor (to get at the statistics reading, you would have to integrate over the frequency scan), so I used it here to show the form of the noise floor.

Results:

...................................Left..................Right
Min Sample Value:..........-1.15...........-12.3
Max Sample Value:..........1.16.............12.49
Peak Amplitude:.............-89.05 dB .....-68.38 dB
Possibly Clipped:.............0...................0
DC Offset:.....................0...................0
Minimum RMS Power:......-104.38 dB...-87.65 dB
Maximum RMS Power:.....-91.84 dB.....-69.31 dB
Average RMS Power:......-99.8 dB.......-79.38 dB
Total RMS Power:..........-99.59 dB......-78.88 dB

Using RMS Window of 10 ms
 

Attachments

  • riaa scan.jpg
    riaa scan.jpg
    23.8 KB · Views: 465
And now with the cartridge and the ground wire to the TT removed. Average RMS is now only -65dB anymore and that are all 50Hz harmonics! The small peak at 15kHz and a bit is probably the pc monitor.

For the record, the cartridge is a Stanton 680 HiFi, TT a JBSystems High Q 30D (a DJ type that I use for recording at 78rpm). I could use about 6dB more gain in this setup as it was made for another cartridge with a bit more output.
 

Attachments

  • riaa no gnd wire.jpg
    riaa no gnd wire.jpg
    23.7 KB · Views: 418
Tadaa!

Well, I haven't yet worked out a new schematic that has been improved enough... So for now I'll just attach the schematic of the existing circuit and list some of the things that I know could be improved with it.

It runs on +15V that's supplied to it through a Schottky on another board. Only 1 channel is shown, and the bias is common to both channels.....
 

Attachments

  • phono.gif
    phono.gif
    10.6 KB · Views: 423
The opamps are actually NE5532AN btw.

At low frequencies the signal is systematically amplified by a total of 60dB (who cares that it's inverted?). Nothing too major there, except that the DC bias needs to be regulated somehow, so that the subwoofer doesn't generate 3 - 5Hz seismic disturbances from miniscule voltage changes.

At high frequencies the signal is inverted and attenuated slightly, then attenuated some more, and then amplified about 100 times. How silly is that? Preferably it should not be attenuated at all, but simply amplified to a lesser extent, by incorporating the 2 RIAA filter stages with the amplifying stages.

I still need to consider whether an inverting topology will have a lower inherent noise level or not. The 2 stages of the RIAA filter should be combined with the amplifying stages so that noise isn't generated from redundant amplification, and filtering caps should be used so that ultrasonic frequencies are not amplified, or at least not excessively. I'd be reluctant to use just 2 amplifying stages instead of 3, because the much higher gain per stage could produce higher distortion, and more readily pick up noise.

CM
 
NFB of input stage for phono stage.

The input stage of a phono pre amp has to have voltage series feedback ( non inverting input ). Shunt feedback could be used in the following stages if done carefully.
At the input the amp will see a minimum of 47K ohms source resistance in the shunt feedback case ( like your circuit).
In series feedback the 47K will be parallel with the dc coil resistance of a few hundred ohms and the coil inductance. This will result in less source resistance and result in a lower noise floor at the input.
Since the output of the cartridge is already low , the source resistance's noise voltage becomes significant. In your case I think the S/N ratio will worsen by possibly about 20db or more as compared to a non inverting input stage.
Cheers.
 
The input stage of a phono pre amp has to have voltage series feedback

WHY?

In series feedback the 47K will be parallel with the dc coil resistance of a few hundred ohms and the coil inductance. This will result in less source resistance and result in a lower noise floor at the input.

Just like in the case of an inverting circuit. Don't forget the summing node is a virtual ground. There is indeed a noise penalty with a the chosen topology, but it can be done.

- put both inputs of the unused opamp to ref
- get rid of the last gain stage, nothing worse than 2 gain stages when it can be done with a single one (the second amplifies all the noise of the first and adds it own noise).
- use lower resistors for your last gain stage. A 5532 has no problem driving 600 Ohms.
- use a regulator to get your 15V if you don't have one already (or drop to 12V) and plenty of capacitors, something like 8x 4700uF split as 2 groups of four with a resistor of 100 Ohm in the positive will do.
 
Havoc,

Just out of interest, what is the tool that you're using to test your stage's noise?

At the moment, I'm particularly interested in collecting such pieces of software to help me out. I'm considering buying a copy of Fluke View for my Fluke scope! Yay! :)

If anyone has tried the full version of Fluke View, I'd be interested in your opinions of it. I only have the demo at the moment, and no interface cable... :mad:
 
Havoc.............
quote:
The input stage of a phono pre amp has to have voltage series feedback


WHY?.........................


With a NE55xx opamp this would be the only way to get least noise when loaded with a cartridge that needs a 47K load and to keep output impedance down..

While the inverting circuit input is a "virtual" ground , the signals do not pass to ground ! It's not a signal ground ..... its just zero signal volts for simplicity.
The 47k and the cartridge will come in series with the input ! Any text book will clarify that . Check it out.

Cheers.
 
Eeka: I wrote it in the first post with the results. For compactness: load -> preamp -> RME ADI-8 AE -> RME Hammerfall -> Cool Edit Pro. Then I use the tools inside CEP. It is not a real measuring instrument, but it gives a good idea where you are. To be complete, I should have worked out the gains and reference levels of every stage to get a "real" measurement. This gives only the noise of the complete setup. But I know that the noise of the RME is at least 10dB lower than the cartridge/preamp, so can be discarded. As said, the gain could be 6dB higher without problems. This would just shift everything 6dB higher.

A better -more correct- result would be to connect my HP3400A, but then I would have to make a filter to exclude the MHz band and calculate the influence etc etc. Far to complicated. And it tells nothing about the frequency response of the noise. And this is far more important IMHO! I don't care about a couple of dB with a LP, you get 50dB at most once it is playing. But having 50Hz coupled in is to be avoided. It also shows nicely when the rectifiers of the psu are coming through (or the TV or the fridge).


I could put the RME in 96kHz sampling, but this would only raise the noise floor of the convertor, while the contribution of the preamp goes down. And the frequency response of a LP/cartridge is not to write home about once you are at 20kHz, so it doesn't matter after all. One of the basics in low noise design is not using more bandwidth than absolutly needed.

I don't trust scopes for this sort of measurement, certainly not digital ones. Their input sensitivity is not good enough, and they start from only 8bit if you are lucky. Not that I wouldn't like one around..... :)

Ashok: if you do the calculations, then you will see that the difference can be neglected. I don't know how to get a formula into this, but the only difference between the 2 is the placement of the 100 Ohm impedance of the cartridge.

In the non-inverting, it sits at the + input, and contributes to the current noise. But 100 Ohm is so small that even with the 0.7pA/sqrt(Hz) it can be neglected.

In the inverting case, it sits in series with the 47 kOhm. But that is so large that the difference is just as neglectable.

Agreed, this is just "back of envlope" stuff as you would need to take into acount the riaa correction, Z(f) of the cartridge etc etc.

Best result would be to use a NE5534A and just drop the compensation capacitor. He would gain a bit in noise figure and a lot in slew rate.
 
I pretty much have to limit the gain of one of the stages to 10 so the RIAA step filter can be combined with it. That means that if I only used one other gain stage, its gain would be 40dB, with resistor values like 10k and 1M. Sure this would save the noise generated by an extra stage, but at the cost of extra noise picked up by the high resistance value. I think there could be more noise overall.

Havoc mentioned slew rates. This could only an issue if the gain per opamp stage is too great, and in fact I think a low slew-rate would be better because the opamp then inherently filters out RF and reduces the need for external capacitors.

I'm still undecided whether inverting or non-inverting is better in this case. It's not just the inherent noise of a given topology that matters but it also depends on the op-amp. Incidentally I have quite a few NE5532ANs but NE5534s so they're not an option.

CM
 
Ex-Moderator
Joined 2003
The trouble is that there are lots of ways of measuring noise, and they all give different results. For instance, Havoc's method of measuring noise is completely non-standard but extremely revealing of problems on his set-up. The bandwidth of the measurement system affects the results because white noise is proportional to the square root of bandwidth. As an extreme example, applying "A" weighting allows super figures because it not only restricts the bandwidth, but removes hum and 1/f noise.

You really do want to change you input stage to non-inverting configuration. The NE5534 is a quiet op-amp in terms of voltage noise, but (like all bipolar op-amps) generates a lot of current noise, making it sensitive to the resistance to ground. That 47k resistor not only adds Johnson noise of its own, but converts the op-amp's current noise into a voltage noise source in series with the cartridge...
 
For instance, Havoc's method of measuring noise is completely non-standard but extremely revealing of problems on his set-up.

Agreed, like I said in last post. I only tought about the problem of translating it to other setups later. The issue is that noise is in most setups not the problem, but hum and grounding is the main problem. And this can best be seen by something that gives you the frequency response of the noise.

its gain would be 40dB, with resistor values like 10k and 1M.

You can just as well use 100 Ohm and 10 kOhm. Or even 10 Ohm and 1 kOhm!

I think a low slew-rate would be better because the opamp then inherently filters out RF and reduces the need for external capacitors

While this may seem so, it means that you will introduce TIM distortion. Do not rely on the inherent bandwidth to limit the final bandwidth. Use external roll-off to define the BW as you need it.

I still think you need to check how your noise floor looks in terms of frequency content. You may have other problems besides the inverting input.
 
Havoc said:
...You can just as well use 100 Ohm and 10 kOhm. Or even 10 Ohm and 1 kOhm!...
Good idea! Now, where did I put those 1500uF caps? :D

...it means that you will introduce TIM distortion.
Didn't think of that. Wouldn't be surprised if some of the audible noise will probably be a result of IM distortion of ultrasonic noise.
...Do not rely on the inherent bandwidth to limit the final bandwidth. Use external roll-off to define the BW as you need it...
Op-amps tend to have a much wider bandwidth at a lower gain anyway. At a gain of 10 the NE5532 has a bandwidth of around 2MHz anyway, as opposed to 200kHz at a gain of 100.

Does anyone have a nice and easy way of accurately calculating the theoretical noise performance of an inverting vs non-inverting phono input stage (as opposed to rules of thumb where "this is better than that")? Here is some useful data:
50V/mV open-loop gain
300k ohm input resistance
5nV/sqrt Hz noise voltage (@1kHz)
0.7pA/sqrt Hz noise current (@1kHz)

In any case, I still need to put in a regulated voltage reference as opposed to that RC network. That might make the biggest difference of all.

CM
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.