Filtering out distortion from a function generator

What is the frequency accuracy/stability of record cutting machines? Practically speaking If it's X and Y PPM it makes no sense to lock a turntable to anything better. Perhaps it just is "because I can" -- and I get that it's one of those DIYer things. I indulge in it too....
 
Assuming you don't need frequency trimming, only accuracy, the proposal of post #18 with one second-order peaking low-pass filter section with a Q of about five should do. I guess you can still make that with a Sallen and Key or an MFB circuit, no need to go for a state variable filter.
 
If you don't want frequency adjustment, not sure you are getting anything by not using your power company. Mine shows the freq over time and it is between 60.05 and 59.95 always and almost always meets 60.025 to 59.975. 0.05/60 * 100 = .08%. So my utility is nominally +/- 0.04% and worst case seems to be 0.08%. The XR chip is not that good I don't think. Mark's crystal would be though.
 
What sort of distortion levels are you after?

Or do you want the THD to be simply 'inaudible'?
Here's what I want: Years ago, I just grabbed a random choke and a random capacitor off the shelf to make a power line cleanerizer for my turntable. It did make the turntable sound better, very noticeably, so much that I enjoyed listening less without the cleanerizer than with.

Until the day I measured the unloaded voltage coming off the thing and it was over 160V ac.

That kinda freaked me, so I decided to build something to manufacture my own turntable power.

All this stuff about fanciful technical jive may be fun for the kinds of people who live here, and I'm happy to have kicked off a fun conversation among you.

But really all I want is to make my turntable sound that good again.

Oh, and as for saying this at the outset, I deliberately keep things as open as possible when I start a conversation, so as to ignite the largest variety of takes. We funnel it down as we go. I always do this. It annoys some people but there are a lot of people here who really enjoy being annoyed. You're welcome.

Throwing another cherry bomb into the room, if series-filtering a square wave produces an adequate sine wave, adding that filtering to a rough shaper-generated sine wave could only be better, right?
.
 
What is your turntable? Does it use a synchronous motor?

What is the loaded voltage from your random choke and capacitor? If it is within spec for your mains, you don't really have to worry what the unloaded voltage is.

Was your original idea to use a cheapo oscillator to drive a power amp to drive a reverse connected transformer to generate a clean 60Hz 120V? If so, you can easily trim a function generator at 60Hz by ear to <1% THD provided your speaker is <1% at that frequency and the output of your power amp.

You could also use your power amp as part of a 60Hz State Variable oscillator which would be very low THD and stable. IIRC, there's a 1970s HFN & RR article on just such a beast for your exact application which is also how Thorens TD125 is done.
 
I agree with that the turntable application had to be mentioned first. This relaxes the THD requirements and opens up more possibilities for obtaining the sinusoidal waveform. For turntable, you'd need accuracy and variable speed for the different record speeds that exist.

For driving the turntable motor, you could obtain your "sinewave" from a square waveform of same frequency. You could also perform some simple low order harmonic removal without filters for example, the using delay / add method (single phase motor) or 12-pulse method (2 or 3-phase). Both methods work at almost any speed set by the user without using any filters.

The 12-pulse method is generally superior and eliminates all harmonics upto order 10 (and some >10 also), starting from a square waveform. There're also 18-pulse and 24-pulse methods that eliminate more harmonics but they're unlikely to yield significantly better results due to the 'law of diminishing marginal utility'.
 
Last edited:
When you use a crystal oscillator and a divider chain, you can also add a couple of EXOR gates to generate Walsh functions and use a weighted sum of those to get rid of the lower harmonics. In any case, you still have to filter out the higher harmonics then, so it just complicates things.
 
Best mileage comes with a resonant band-pass filter plus some notch filters for H2 and H3 - simple low-pass filters don't have much purchase on the low harmonics. Or simply use a sound-card?

For a given Q, a peaking low-pass filter with transfer function

H(s) = Klp/(s2 + (omegan/Q) s + omegan2)

suppresses the kth harmonic k times better than a true bandpass filter with transfer

H(s) = Kbp s/(s2 + (omegan/Q) s + omegan2)

In either case, you can add notches to improve it further. It's only of academic interest, because 1 % distortion will do according to Bamalama.