Simple Op Amp RIAA preamp 1st lash up

[bikerhifinut]

OK so I made the easiest one to get a flavour.
Based on the Velleman kit I made a couple of years ago for a mate.

Note I had to parallel up a couple of capacitors as I had no 3.3nF in the toybox.
I used some 2.2nF polyesters and a pair of 1.2nF which I managed to get a matched set of measuring as close to 3.3nF as I figured was practical with my DMM. I "Calibrated" the DMM with some samples of 0.5% and 1% capacitors I had so am reasonably confident Iam in the ball park, and as I only used standard 5% carbon film resistors from my stores it was never going to be that close.
Running it from a 15V AA battery pack.
It's got a nice sound, I haven't run it up on the scope yet, but I figure I need to make an inverse RIAA network if I want any sensible information.
I put 47ohm "Stoppers" after the output capacitors along with 100k "Bleed resistors" across the output.
Input and output coupling capacitors are poly film, 1uF on input and I reduced the output capacitors to 0.47 uF as I figured a bit extra bass roll off wouldn't be undesirable, feeding a 25k input Z on the preamp.
My only practical criticism so far is that it hasn't enough gain, I have the Volume control wound to 1 o clock to get a reasonable level and that's with the amp overall gain set to its 15dB maximum.
So perhaps an additional op amp gain stage of say 12 dB (x4) might be desirable and I guess after the preamp rather than in front, would possibly be handy rather than tweak up the gain around it as if memory serves me right, this circuit gets a bit twitchy if a 5532 is used rather than the TL072 its designed around.
I tried a TL072 and tgh I think a fair amount is from the carbon film resistors and there's an audible difference in the hiss produced, although a good amount of noise is probably from the carbon film resistors in the circuit.
Nevertheless it shows how a few hours in the den and some bits out of the scrapbox can produce a very acceptable result. Also there's no nasty hum pickup (using a battery) and even outside of a screened box and with simple twisted wire to the input and outputs I have had no dramas.

Photos attached, its built on perforated board and hard wired underneath, its the board where each hole is plated round so at least the components are anchored when soldered.

Andy.

[bikerhifinut]

link to the circuit, thanks to Mark Hennessy for the annotations etc.
It's flawed, and no doubt the flaws to the EQ will be picked up and explained by others better qualified than me.
And bear in mind I refined it a bit with "stopper resistors" and "bleed" resistors on the output.

https://www.vintage-radio.net/forum/...7&d=1595848466

Andy.

[Craig Sawyers]

I wouldn't use polyester in the RIAA. Polystyrene or polypropylene is what to use.

Even Baxandall mentions this in his 1981 paper "It has been found that polyester dielectric capacitors are used [in RIAA] networks, then the fall in specific permittivity with rising frequency, and the large loss angle which is characteristic of this material, produce a detectable high frequency error...". Self also finds that polyester gives rise to harmonic distortion when used in feedback networks.

I haven't done the sums on RIAA accuracy yet.

Share and enjoy

Craig

[bikerhifinut]

Thanks Craig, The Polyesters are there in the EQ because that's what I had at hand, I didn't have a 3.3nF of any description and these were the only things I had that got me close. The 680 is a polyprop 5%, so that stays.
This is just a starting point to see what the most basic of circuits and parts can get me.
I knew there was a reason why polyester isn't ideal for the EQ and if I'd read up a bit on capacitors again I'd have perhaps held fire and ordered a few up.
I'll not be fiddling for a couple of weeks due to an unavoidable commitment and then I'll be doing some more Head scratching around the classic circuits and another slightly off piste one that I'll have the parts for anyway.
By the way Craig, I assume silver mica capacitors are ok for EQ and trimming values?
You wouldn't believe the original EQ caps in the velleman kits are ceramics! I measured the originals that I still had in a box, the 680pF weren't far away but still high but the 3.3nF measured well in excess of 4nF for one and 3.9nF for the other which is a it of a joke but normal for ceramics I guess...…………..

Andy.

[Craig Sawyers]

C0G/NP0 ceramics are just fine for audio because they are low-K. Silver mica certainly has its followers in the audio world; I haven't used them in practice though.

I've just pulled a 1nF silver mica out and put it on my GR1650A bridge. This particular one measures 1.011(00)nF with a dissipation factor of 0.0004 at 1kHz. Which in my book means it looks just fine for audio/RIAA use. 1.1% error in value and tiny dissipation factor.

Craig

[bikerhifinut]

I suspect these ceramics are your normal cooking jobs Craig. And the tolerance is rubbish.
That's always good for folks to know about silver mica jobs. I was thinking more for trimming a value when its critical rather than the whole thing. I think they are getting very expensive now due to the world supply of the right sort of mica sheets being very close to exhausted? So maybe in the interests of the planet we should only use them if absolutely necessary?
One thing I have noted is that the difference in price from a 5% polyprop to a 1% seems to be orders of magnitude.
That's when it seems to make sense to adjust the resistors in a RC circuit as I reckon they have to be lots cheaper. If its that important to get the EQ absolutely on the button. Myself I wonder if other distortion factors in an LP replay system swamp small differences in RIAA EQ.
Ho hum, I have Eric Bibb strumming and singing in front of me via this lash up and I have to say even taking into consideration the 235% pleasure improvement factor from RW's DIY formula, that I think its a very pleasant little device and its got absolutely no special parts made by magical beings, it's made from the most mundane of bits. This was deliberate on my part, even down to eschewing metal film resistors in favour of bog basic cheaper than chips carbon things.
Things can only get better, but the question is, by how much, and for how much?

I got to thinking, the parts will have cost less than a fiver and providing you can get a 9V or better supply to the board (I'd just use a couple of PP3 in series and a switch) theres a solution in part to the age old problem of replacing unobtanium piezo cartridges providing the arm you have can track one of the heavier tracking magnetic cartridges (Think Disco and juke box here, even the shure M75 range had a stylus that tracked at more than 3.5g) and some of the better domestic record players had turntables that could take a magnetic.
A.

Andy.

[Radio Wrangler]

Silver mica capacitors are excellent. They work well right up to quite high RF frequencies where failings tend to get exaggerated. Intermodulation effects are low and this translates into any other way of looking at non-linearity. At RF they are limited by physical size dictating stray inductance, and by cost reasons as Andy mentioned. Mica is a limited and dwindling resource. Would you believe some of the right sort was ground up to do a job the small-grain stuff was used for - it went into Kaolin poultices?

Anyway, at audio, they're well nigh perfect.

'Ceramic' as a word covers a multitude of materials with orders of magnitude difference in their various properties. The dielectric constant of the material, K, is the chief one.

Several low-K materials have been concocted. They are stable and linear.... and performance continues to extremely high RF. These materials seem to have become best known by their temperature coefficient codes NP0 and C0G. These aren't actually names of unique materials, though it turns out pretty much that way.

The ceramics are made by sticking a mixture of minerals in a grinding mill, and grinding until a fine slurry is formed. The slurry is dried to make a paste, called the 'green' ceramic. This gets layered-up like multilayer cake with stuff that will turn into metal film when fired. It gets rolled to the chosen thickness. The whole lot goes through a furnace and it sinters into alternating layers of metal and rock-hard ceramic. The art is in arrangeing it so connections can be made to alternate metal layers. Thickness of the ceramic layer sets the capacitance for a given set of outside dimensions. This gives you an MLCC, Multi-Layer Ceramic Capacitor. It has a cousin made with the same technology where transmission line structures are made within a ceramic medium, with ground plane metallisation around it all. Transmission line tricks like directional couplers, delay lines etc can be made. They come under the heading of LTCC Low temperature, Co-fired Ceramic Components.

Back to that milling... obviously mixtures of stuff that can turn into different ceramics can be made

The next more capacitive ceramics like X7R fit in a lot more capacitance into the same size dimensions through having higher dielectric constant. In ye olde days, these went by the name hi-K. You get useful amounts of C for decoupleing use at RF, but the temperature coefficients are large. Linearity suffers. Capacitance measured varies with voltage applied to them somewhat. This means non-linearity. OK for decouplers where this doesn't matter.

More recently there has been the development and marketing of extremely high K materials, cramming enough microfarads into tiny SMT packages to replace small electrolytics. Now electrolytics have several bad habits, as do their ceramic alternatives. Well, the ceramics don't have the water loss lifetime limitation. but they do have nasty surprises. They have large temperature coefficients (forget any precision uses) They are very highly non lnear, capacitance diminishes rapidly with bias voltage. And on top of that they are very microphonic.

Let's go back to basics and make a capacitor. Two flat metal plates, spaced apart with a gap between them. The gap is filled with air or vacuum. We add so many electrons to one plate and nick the same number from the other by charging it up by running a current for a while, powered by abattery or something. This leaves a voltage difference between the plates.... a charged capacitor.

Now, we slide in a bit of insulating material between the plates... a dielectric. It needn't touch the plates but it's OK if it does. Dielectrics are insulators, usually good ones.

The electric field between the plates affects the dielectric. THe orbits of electrons within the material are distorted. Electrons think the positive plate is sexy and try to hang around nearer it for longer. The negative plate is comparably repellant. Protons feel the opposite influences, but they are more definitely located.

Electrons don't leave their atoms unless the forces are too much and lead to breakdown.

But just the distortion of their normal movements produces an electric field in response to the one which caused it. The created field acts in opposition to the original field and acts to partially cancel it. The voltage across the plates is affected by the response field. It is decreased. Now we haven't changed the numbers of electrons we squirted ont and stole from the plates, so the coulombs of charging we did haven't changed. But the voltage went down. So the capacitance must have gone up!

So the key to a good dielectric is hoe easy the electron positions can be distorted, AND how linear the response is to the applied field. Remember valence electrons are swapping between neighbouring atoms to hold chemical compounds together.

It begins to look like good dielectrics are rather special.

You can also get hysteresis effects which lead to nasty non'linearity effects like dielectric absorption. In ceramics, these get worse with higher K materials. Some plastic films are OK, some are terrible. Never usa a Mylar capacitor (PET) in a precision integrator like a dual-slope integrator DVM!

These are factors that explain things that audio people express preferences over, and seeing what's going on, I'm not surprised. But they are measurable and they've been known about for yonks.

Some applications for capacitors don't care about these factors, others do. You can get away with them for decouplers, up to a point, so it isn't a licence to kill.

In the RF world, for capacitances up to maybe pF to hundreds of pF there is an interesting ceramic: Porcelain. Tough isn't the word for it. These are capacitors for use in transmitters. Would you believe I put 350 Watt pulses at 1GHz through a tiny surface mount capacitor '0805' size? There are two companies, ATC American Technical Ceramics inc. and Dielectric Labs inc. known to make the ones which don't turn to slag in an instant flash of light. Quite expensive.

In the audio world, for small value capacitors whose value and linearity are important G0G/NP0 or silver-mica are fine.

For bigger values where they see signal voltages across them, you have to get choosy about plastic films. Polyester and mylar are best avoided on linearity grounds. High-K ceramics are right out. (shades of the Holy Handgrenade of Antioch)

David

[Craig Sawyers]

3.19 am David?

Anyway, thanks for such an superb post that covers all bases. There is also an excellent series of articles called Capsound by Cyril Bateman (RIP). He looked at linear distortion (at 1kHz) of capacitors. Bateman spent his working life designing capacitors at Erie, so he knew a thing or two about what went into the things.

He started off by designing an exceptionally low distortion 1kHz oscillator, and included the ability to add a DC bias. Then he built a twin-T notch filter to get rid of enough fundamental so he could look at distortion on a spectrum analyzer. He shows his schematics, circuit board layouts, and photos.

He worked his way through ceramic, plastic film, tantalum (don't, was the conclusion), and aluminium electrolytic.

For larger value electrolytics he had to drop the frequency to 100Hz and deliver sufficient current to get measurable distortions.

His papers are not that easy to read - they tend to ramble somewhat. But they can be downloaded from https://linearaudio.nl/cyril-bateman...sound-articles .

Self's rule of thumb for electrolytic capacitor distortion is don't get more than 0.3V ac across one in the audio band if you want to avoid harmonic and intermodulation distortion. That means that you need a value that is ten times, or more, what you calculated was necessary.

On a completely different topic, the y in ye is actually pronounded th. Reason is the the y in this context is called the thorn, and was part of English going way back to Old English. It had its own symbol, a bit like a Greek small case rho. At the time of Caxton, he standardised the alphabet, and decided not to include the thorn. So he picked a letter that was not often used - the y - and used it to represent the thorn.

So "ye" is actually (and always was since the time of Caxton) pronounced "the".

Icelandic is one of the few remaining languages to keep the thorn as a unique symbol.

Craig

[wd40addict]

If you'd like a FLAC of inverse RIAA test tones please PM me, it's too large to post here.

Came in handy recently whilst resurrecting an Amtron RIAA preamp kit I built when I was 15. Over the intervening 40 years a capacitor had failed, but much more concerning was awful RIAA accuracy found by using the above file (+6dB error at high frequencies). A bit of SPICE confirmed a design error. A bit more SPICE gave some improved capacitor values. Response now better than +/-0.3dB over most of the range with slight fall off at bottom end due to finite gain of the two transistor circuit. FLAC confirmed simulation results, all good now.

[Craig Sawyers]

OK - I've built a perfect RIAA preemphasis circuit in Spice using values to four places of decimals, according to the recipe on http://www.hagtech.com/pdf/riaa.pdf . Not his practical values as if you were building one, but the theoretical values.

What the above circuit is doing is attached. Basically from 20Hz to 20kHz there is +/-2dB, which is not so good. Actually, the topology of this circuit, from Lipschiz's paper, is probably the toughest to design. But the ratio of capacitors should be 2.916 and the ratio of resistors 12.403.

So without doing anything more than adopting those ratios - which means that R2 becomes 82k (the nearest sensible value to 1M/12.4) and C3 becomes 1.2nF (the nearest to 3.3/2.9) improves the response to +/-0.5dB. Also attached at the same scale as the first.

It is of course possible to do even better by using a complete analysis, selecting capacitor values on a meter, picking E96 value resistors etc - but it would only be worth the hassle for a full-blown no compromise stage.

And going from +/-2dB to +/-0.5dB is a rather good improvement for changing two values.

Craig

[Craig Sawyers]

For completeness, this is the response of the Baxandall RIAA circuit (for which he used the 5534A). The overall response variation is +/-0.25dB.

I've changed the vertical scale to show this small range in detail.

Craig

[Radio Wrangler]

3:19? Well, seeing how I seem to have involuntarily retired, I thought I'd treat myself to a present... A Nikon D6 and a copy of Capture-One. So I was up late reading and playing.

All of these approaches to RIAA are compromises, but where do you stop?

For people who cannot, for ideological reasons, admit to any limitations on their hearing, there can be no end. They remain expensively committed to the endless pursuit of diminishing returns.

Those last 0.5dB wobbles could be reduced.

In theory, with RIAA defined as a set of time-constants or pole positions, the required response should be realisable, but stray effects get in on the act as well as component tolerances.

We tend to analyse feedback systems as if they were idealised and either had infinite loop gain, or else the open loop gain was linear and flat. Spice does a good job of including realistic open loop behaviour including non-linearity.

If you really do have tons of OLG, then the series feedback circuit should hit the nail on the head.

The shunt feedback circuit is not proportionately set by the ratio of impedance of feedback impedances, there is that pesky +1 term. In a very high gain implementation, it's tempting to sweep it under the carpet as trivial, but that cuts no ice with audiophiles. they know they are certain to hear it, once they know it's there. The added pole can reduce this a lot, but not quite perfectly.... Don't tell the audiophiles!

Another thing, is that discs recorded with RIAA have the mid-range kink in the frequency response applied to them. When played with a piezo cartridge (which is position/force-sensitive as opposed to the magnetic cartridge's velocity sensitivity) the reproduction gets a 1-pole rolloff which is a sooth response across the range... this compensated for the overall slope of RIAA, but does nothing about that kink. People seem to listen to them without worrying about this. Maybe if they knew, they'd worry?
Maybe it explains various preferences expressed over the years?

A sense of proportion and when to stop is the only thing which stops all tasks being infinite.

David

[Craig Sawyers]

The tolerance question is a good one. Since 1% tolerance resistors are common and cheap, 47n and 100n polypropylene 63V (for Baxandall) are available in 1% from RS, there is no excuse for significant variation from the target response.

Craig

[bikerhifinut]

I think I need to take the tablet (computer) away with me and get some studying in if I can.
All that David has said about the various properties of dielectrics is good stuff, I first got an inkling of what the actual physics etc was about from a copy of Morgan Jones' valve amplifiers book. useful reading even if you aren't into valves. He has a very readable for a layman like me, chapter on the construction and materials used for capacitors.
Now given that its almost universally accepted that Polypropylene Film capacitors are superior to polyester film for Audio coupling and that film/foil is supposed to be superior to metallised films, I have to hold my hand up and admit that I am perfectly happy with the results I get from my default coupling jobs which are the standard polyester metallised film jobs like the yellow axials from the BVWS shop or the "orange drop" radials which are about the cheapest decent capacitors available for me.
I have witnessed some quite lively exchanges on audio fora over this touchy subject and I find solace in my apparent deafness to the differences that others so readily hear.
OK that's capacitors put to bed for now. I'm going to have a go at the VSPS single chip circuit, using polyprop tubulars and 1% metal films. Of course I am thinking about complicating the job...…….. so the next question in another thread.
A.

[Boulevardier]

Originally Posted by Craig Sawyers

On a completely different topic, the y in ye is actually pronounded th. Reason is the the y in this context is called the thorn, and was part of English going way back to Old English. It had its own symbol, a bit like a Greek small case rho. At the time of Caxton, he standardised the alphabet, and decided not to include the thorn. So he picked a letter that was not often used - the y - and used it to represent the thorn.

So "ye" is actually (and always was since the time of Caxton) pronounced "the".

Icelandic is one of the few remaining languages to keep the thorn as a unique symbol.

Craig

From what I've read, the reason for Caxton's change is also interesting. At that time, all typesets were produced in continental Europe, and that was where Caxton got his supply from. Thorn was simply not a feature of any other European languages, so was not available to him from those sources. So, he had to improvise and substituted the letters "th" for thorn in all his printing.

Little problems and their random solutions can have huge consequences - had the lead type-piece for thorn been available to Caxton, the English-speaking world would probably be using it to this day instead of "th"! That's if that story isn't just apocryphal...

Mike

[bikerhifinut]

Ok so why must I use a dual rail supply for the VSPS circuit, and will it materially effect the AUDIBLE result if I tweak it around to use a single supply rail, I do have a practical reason for asking and it looks easy enough for the cost of a couple of resistors and capacitors.
I was going to use a DC blocking capacitor on the cartridge input anyway so any offset issues shouldn't arise.
I'm not sure what value of input capacitor is optimum, the default seems to be a 1uF and I have some, but I've got even more 470 nF and 220nF in the parts bin.
I'd argue that a bit of LF roll off isn't always a bad thing here.
A.

The reasons for single rail are that a finished board could end up in a project I'm working on for a great nephew to build himself a decent stereo amplifier that won't disgrace itself and that I'll have an 18 to 24V DC supply in place for the other elements in the circuit. It's not a deal breaker though.

[daviddeakin]

Originally Posted by bikerhifinut

Ok so why must I use a dual rail supply for the VSPS circuit, and will it materially effect the AUDIBLE result if I tweak it around to use a single supply rail

The circuit you linked to is already single rail... In fact, if you load the output of the opamp with a resistor to ground then it will operate more in class A, so maybe you can claim it sounds even better!

Incidentally, since no one's mentioned it, you *can* use polyester caps in audio circuits without sacrificing performance. The trick is to use a voltage rating that is way larger than the signal voltages in circuit (i.e. thick dielectric). A 250V polyester cap is just as good as a 50V polystyrene, for audio at least.

[daviddeakin]

Here's a classic old Penfold RIAA circuit (also single rail), just for reference.

[Radio Wrangler]

That Penfold circuit is also a victim of fashion as so much of audio is.

That CA3130 was the newest innovation at the time, offering negligible input currents from its MOSFET input stage, as well as quite a lot of gain.bandwidth product.

It shows, however, how poorly noise was understood outside of specialist groups. It is terribly unsuitable for a low-Z input like a magnetic cartridge and the low voltage signals.

But the trend was to take the latest devices and showcase them in every possible application. This was THE FUTURE!

The CA3130 was indeed a dramatic step in opamp development. It's just that it was in the opposite direction to what best fits this application.

David

[Craig Sawyers]

Leaving aside single vs dual rail for the moment (and as David says, the choice of opamp!), the RIAA values are not too far off for E12 series parts.

For that circuit:

R6C4 = 3808us (target 3180) 16% error,
R7C5 = 70.5us (target 75us) 6.4% error
R6/R7 = 14.5 (target 11.8) 18.6% error
C4/C5 = 3.7 (target 3.6) 2.7% error

Since the capacitor ratio is pretty close to target, and more accurate resistors are cheap, changing R6 to 560k and R7 to 51.1k gives:

R6C4 = 3140us (1.26% error)
R7C5 = 76.7us (2.3% error)
R6/R7 = 11.0 (6.6% error)

Even just changing R6 to 560k and leaving R7 at 47k gives

R6C4 = 3140us (as above, 1.26% error)
R7C5 = 70.5us (6.4% error)
R6/R7 = 11.9 (0.85% error)


Craig