diyAudio (
-   Solid State (
-   -   Overload considerations in RIAA preamps (

ashok 8th March 2004 03:05 AM

Overload considerations in RIAA preamps
There were a couple of posts in the "NE5532......misunderstood" thread, regarding overload in RIAA stages. So we started a new thread for this.

I will just copy some bits from the posts to show where it was headed. Will put up a circuit a little while later.

.............I just completed a two stage passive equalised RIAA phono amp. All opamps are on sockets.
The input opamp is configured for a gain of about 31 .
The passive 75uS eq follows and is buffered by a unity gain opamp. Then comes the passive 3180uS and 318uS filter and followed by an opamp with a gain of about 10.

The last opamp is a Burr Brown OPA2134.

......................don't you think you'll get better S/N if the mid stage has some gain

.......................I was looking at the overload factor when I picked the configuration. I might be wrong of course.
The input stage is designed to accept about 400mV peak that seems to be the max output one can expect in the worst case ( noisy ) scenario from a MM cartridge.
The second opamp will have to handle this full signal if it occurs at anything less than about 2kHz. But I am not sure of the noise spectrum of surface noise . I was just trying to be extra careful.
In practice , with average albums , it might be possible to have gain here without encountering any overload problems on spiky noise. Spiky noise basically must be HF trash. But what about LF content in it ? Something like a drum . Large transient with decaying LF signals.
On second thought , that sounds wrong!
Can you and other RIAA experts comment on this - overload due to non musical signals ?

...............Hi Ashok

This is way out of topic and we should probably rather move it to a new thread if anyone else shows some interest.

Spiky noise basically must be HF trash. But what about LF content in it ? Something like a drum . Large transient with decaying LF signals

But the LF is still some 20db below your 12v limit, right? Anyway, i didn't realise you expect 400mV peaks, that's rather cautious, isn't it? How do you estimate the S/N in everyday use?

I haven't tried a split riaa with opamps for almost 20 years, maybe the time has finally come One potential issue is the need for separate regulators for each opamp - i find this really determines the absolute transparency of the cascade.

Have to run. More later.

ashok 8th March 2004 05:35 AM

Here is the circuit diagram
2 Attachment(s)
The Circuit diagram is attached.

I have concluded that the maximum possible signal that can come out of a MM cartridge under very noisy conditions is 400mV peak. That is about 283mV rms. This figure is based on figures that I have seen on the web and the overload limit on many commercial preamps being 250mV. We are assuming that the cartridge produces 1mV/cm/sec.

So the input stage gain was set at 31. The value of R3 is 30K and not 27K as in the ckt diagram. This permits the input stage to handle the worst case input signal without clipping. In a tube stage this overload limit is 'much' higher and possibly why they sound cleaner on noisy surfaces.
Using discrete components it is probably easier to design an input stage that will outperform a chip implementation.

Considering 'noise' and its spectra. I vaguely remember reading that very high amplitudes produced on playback of old discs lies around the 5Khz region.
Can this under any condition extend to below 2kHz ?
What is the spectrum of this noise? How high are the harmonics ?

Like Analog_sa suggested the other frequencies must be quite low . Looking at a noise spike , it might look like it could be less than
the main spike by at least 5 times ( 20%). Is that a realistic value and what would be the lowest frequency?

Listening to some loud clicks off the disc it does seem to have some reasonably low frequency content.
Has anyone ever looked at this ?

It would be interesting to know how much gain the second opamp can have without danger of clipping noise spikes. The 75uS filter should be cutting down the amplitude of the spike .
Can the phono experts shed light on this issue.

sam9 8th March 2004 06:36 AM

"The input stage is designed to accept about 400mV peak that seems to be the max output one can expect in the worst case ( noisy ) scenario from a MM cartridge."

Per J.L.Hood "Art of Linear Electronics" Ch 9, the maximum out pou available from a MM cartridge is 19mV RMS. Unless I misunderstood something, you are amply covered if you can handle 400mV. Hood goes on to say that the resulting output to the power amp needs to be .77V RMS with the result that +/-15V rails will be adequate.

Don't know if this helps.

ashok 8th March 2004 07:41 AM

Noise spikes.
Hi Sam,
You are right about the musical signal levels. We are refering to signals generated when you have bad clicks and pops due to dust and surface imperfections .
You will find that people have studied this and found noise spikes to be several times higher than musical signals. I even came across a graph showing this , somewhere on the net. I will search it out again.

To quote approximately : ' loud noise pulses clip in the input stage causing audible artifacts to be generated within the audio range and degrading the sound that one hears. This is why some preamps sound fine with noisy surfaces and others make the same recordings un-listenable.'

So I am trying to find if someone has done an analysis on this and how it can be used to optimise the circuit gain.

ashok 8th March 2004 07:52 AM

NAD-PP1 phono preamp specs
Here are the NAD-PP1 phono stage specs:

Tech specs
Input impedance: 47 kohm + 220 pF
Input sensitivity for 200 mV out: 2,5 mV
Signal/noise ratio: 78 dB (weighted A) - 72 (unweighted)
Input overload at 20 Hz: 55 mV
Input overload at 1 kHz: 63 mV
Input overload at 20 kHz: 580 mV
THD (20-20k Hz): 0.04 %
RIAA accuracy: +/- 0.5 dB

As you can see the HF overload is very high .
The actual overload wrt to freq also depends on the kind of equalisation we use.

The kind of overload figures you see here are usually in preamps with the equalisation networks in the feedback loop. The circuit shown in an earlier post has no frequency dependant feedback.
Thus it's overload characteristic is different.

There are pro's and con's for the different schemes that we can implement. Picking one and perfecting it is tough! Personally I would like to try out all the different schemes at my place so that the ONLY variable is the circuit . Before I start doing that it would be good to discuss this and see what others have to say. I can't obviously try all the schemes .

ashok 8th March 2004 02:42 PM

Noise etc.

Yes the LF will be quite safe even if stage two had gain. In the absence of any info regarding the spectrum of the noise , we can assume that the spike does have a fundamental of over 5Khz with LF content that is at least -10 db wrt to the peak.
Additionally the mid band typically needs less overload margin and 100mV seems to be quite good. So making a rough guess we could get away with a gain of about x4 in the second amp.

The input stage should have a S/N of about 67.5db with a 5mv input signal. All noise calculations are for flat bandwidth .
At the second amp the signal to Noise ratio should drop slightly. My rough calculation shows less than 0.5db.
So the S/N in stage two is now 67db .
The gain in the second stage will only affect the SN in the following stage. Here the current noise will have an affect due to the high source impedance. But I am using an OPA2134 here which has a current noise of 3fA(Hz)^0.5. So it turns out to be quite low. Additionally it also has a voltage noise that is 5.5 db below the AD826.
AD826 voltage noise is 15nV(Hz)^0.5
OPA2134 voltage noise is 8nV(Hz)^0.5
NE5532 voltag enoise is 5nV(Hz)^0.5

So we end up with about the same S/N ratio ( 67.3db) in stage 3.
By having gain in stage two we will hardly improve upon this as the S/N seems to be largely determined by the first stage.
Taking the equilization into consideration we will get some more improvement in noise. I read something about this recently and it said that the RIAA equilization only improved S/N by less than 3db. So we are close to 70db S/N.

With the NE5532 we will get a 9.5db improvement in noise - but what about the sound quality we heard with the AD826 ?
i practice I can hear his only when I put my ear against the speaker. From a meter away I can't hear any hiss. So practically the S/N is OK. On my measured noise spectrum the noise above 1Khz drops away pretty rapidly.

An OPA2134 will give a better S/N by about 5.5db. I must try this combination. The current noise is also very low. The last time I tried it , the AD826 sounded better than the OPA2134 in a straight forward buffer amp.

Does all this make sense? My head is spinning. Must settle down with a drink and have dinner and go to sleep.


analog_sa 8th March 2004 03:59 PM

Hi Ashok

Apparently there are other opinions as well. If you look at the Hagerman Buggle you may notice the gain structure is x11,x11,x10 and the low-frequency correction comes first.

No idea if this sounds better or not but it's certainly easy to test. What worries me though is that in your curcuit the mid opamp is only there to allow a split RIAA. The question is then if splitting warrants the inclusion of a dummy stage.


ThorstenL 8th March 2004 04:50 PM

Re: Here is the circuit diagram


Originally posted by ashok
We are assuming that the cartridge produces 1mV/cm/sec.

This assumption may be reasonable as average, but not as maximu value. The "loudest" MM Pickups I know come in around 2mV/cm/sec and can trace at 300Hz (HiFi News Test Record) a track with +18db above 5cm/sec. This kind of level covers more or less the 300Hz to 3KHz band.

With MC Catridges there is much lower level, but this may be stepped up using (for example) a wideband transformer. Let's use an example of 0.1mV/cm/sec and a +26db stepup, giving also 2mV/cm/sec. Unlike MM Pickups, MC pickups can genertae high output way into the supersonic range. usually tracking is not an issue at higher frequencies, so very sharp spikes (dustparticles) may produce very high levels with a very substantial slew rate.

Any which way, we can assume around 80mV (RMS) @ 300Hz...3KHz from the cartridge, AT least and the same sort of level and more via an MC Pickup and suitable stepup with a lot more high frequency content.


Originally posted by ashok
It would be interesting to know how much gain the second opamp can have without danger of clipping noise spikes.

If we amplify these 80mV by 30db we get around 2.5V RMS after the first Op-Amp. A normal Op-Amp on +/-12V rails can produce around 7V RMS into easy loads. If we have only the 75uS timeconstant ahead of the second stage the gain should be limited to under 10db to make sure no clipping occurs under normal conditions.

I personally prefer by far the use of an active, feedback RIAA EQ with solid state components, simply becuase headroom problems become near irelevant as well as the reduction of slew rate issues.

For MC pickups it may be interesting to use a pair of nice Op-Amp's in Shunt Feedback mode, say a AD797 or AD811 in the input with the 3180/318uS EQ in the feedbacl loop, inverting connection and then have the 75/3.18uS EQ in the second Op-Amp's feedback loop, maybe a nice FET type or another AD811/AD797....


jcx 8th March 2004 05:48 PM

care to explain your logic in choosing passive eq?

ChocoHolic 9th March 2004 08:47 PM

S/N ratio
9.5 db improvement by using the NE5532???
If I remember right, then the NE5532 is a fully BJT based
design and has remarkable input currents and with this
also higher input noise currents..... can't find that xxxx data sheet right now... So you will only get real low noise with that OP amp if you use very low impedances. But then NE 5532/NE5534 is hard to beat in noise.

I would prefer the AD8065 (or dual AD8066) for the first stage.
It offers:
AND !!
0.6 fA/Hz^0.5

If you want to use some cheap standard which is easily available everywhere then you might better use the NE 5534 (instead of NE5532) in the first stage. At this gain you will not have issues with stability.
With such BJT OP amps I would propose to reduce the resistors of the feedback network in the first stage by factor ten for this OP amp. The OP amp can easily drive this.
Input resistor of 47kOhm should not be reduced, otherwise
the load impedance for the MM would not be what they are designed for. But at this position the high resistance will not cause that much noise, because the internal MM-resistance is in parallel.


All times are GMT. The time now is 11:03 PM.

vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2014 DragonByte Technologies Ltd.
Copyright 1999-2014 diyAudio

Content Relevant URLs by vBSEO 3.3.2