Yes, it's particularly been a problem with three phase motor drives and appears to be related to motor winding inductance or in my case, since I always use transformers between the amplifier output and the motor input, to the inductance reflected into the primary of the transformers.
I find that adding a Boucherot cell to to each output phase is often the easiest way to stabilise the whole thing, say 10 ohms / 50 uF.
If that doesn't work you could try an RC filter. Typically 1 ohm / 50uF is all I need, placed after the Boucherot cell. If your amps are running in BTL it is often useful to split the filter and tie the centre point to ground (eg 2 x 0.47 R and two x 100 uF)
I've now discovered more: The problem is not the L165 amps themselves, or their specific power supplies, or their interaction with the motor - it happens even with no motor present. If I cut the input to the amps - not just setting it to zero, but actually sever the input resistor, then all if fine - even though the input still has a path to ground via a dividing resistor.
I suspect some sort of grounding interaction with the preceeding circuitry. More testing in order this weekend!
I suspect some sort of grounding interaction with the preceeding circuitry. More testing in order this weekend!
This project/thread has been going on for exactly 2 years now.
With all the problems I really don't understand why not use these chips? If there is a problem with replacement later, why not just buy extra chips beforehand...problem solved.
To me, it doesn't make sense to not use these chips if they will finish the project. What is the point of chasing our tails with the assumption that certain parts will not be available in the future, when the project never gets finished anyways?
Just purchase the extra parts, build the controller/power supply and be able to use the damn thing!
Then, if you wish, keep working on a version with universal parts.
I have been here since the first posts of this thread, and after 2 years it just doesnt seem like it is ever going to go anywhere.
Just venting....and wishing I had the capability to design something myself...
Rick
Last edited:
Whoa, I didn't mean to start an argument. 
OK, Rick, I understand you are just venting, hopefully in a pleasant sort of way!Steerpike has done tremendous work on understanding what is needed for a good implementation of a replacement controller! Since it is his project, he is free to do as he feels given his time and interest.
I know that "free-time" is valuable and there is never enough of it. Some of my projects have taken many more than two years, some have yet to be completed and some never will be. So unless we are going to pay for development costs, we can't really complain too loudly about the pace of development. However, in spite of the time taken so far, the project appears to be progressing.That said, there is nothing stopping one of us from starting an alternate design. I would hope that Steerpike would not object to someone using the information he has so graciously posted to explore an alternate concept.
Personally, although I may be interested in pursuing an alternate design, given my to-do-list and the fact that my SP10 is currently in working order, it will probably be several years before I begin marching down that path.
Rick, although the various DSP chips will reduce the amount of circuit design required, there will be more software (or should I say firmware) development. Note that microchip has a low-voltage motor control demonstration board that might (repeat might) be applicable with minimal modifications. However, to do it right will require an amount of software development. There are a two circuits that will need to be built in addition. One is the 50 kHz exciter for the phase detection and then three detectors that can be interfaced to the DSP A/D inputs. (That is unless a totally totally different approach than that originally used by Technics is taken.) Hence it might be easier to start from a scratch design. That said, you select your poison, hardware, software or both.
My post was not aimed at anyone in particular, as this thread was started in an open context in an effort to get a new controller/power supply designed to replace the tiring older versions.
My post was certainly not aimed at steerpike, as I very much appreciate the work he has done, but my post was a general statement to everyone involved in this thread.
I know that replacement controllers/power supplies have been built for the SP10 MKII, I have seen examples on the web, but for the life of me cannot now find where I seen them....as I wanted of course to either purchase or find out where they were made.
I have heard rumour that in Japan there are many who have done just that...but, you know how hard it is to get information from there...
My post was not meant to be taken negatively, just a general venting and wishing I had the knowledge to design my own so that I can contribute to this thread.
Rick
My post was certainly not aimed at steerpike, as I very much appreciate the work he has done, but my post was a general statement to everyone involved in this thread.
I know that replacement controllers/power supplies have been built for the SP10 MKII, I have seen examples on the web, but for the life of me cannot now find where I seen them....as I wanted of course to either purchase or find out where they were made.
I have heard rumour that in Japan there are many who have done just that...but, you know how hard it is to get information from there...
My post was not meant to be taken negatively, just a general venting and wishing I had the knowledge to design my own so that I can contribute to this thread.
Rick
It was clear to me that your post was not meant to be offensive.
I hope that I wasn't taken as being too harsh.
That was the reason I reiterated your statement that you were just venting! (and hopefully in a nice way.)BTW, Rick, I know you had posted at some point that you wanted to get your SP10 in a new plinth with a custom circuit. I was unclear as to whether you have a currently working set of original boards?
Last edited:
Oscillation problem fixed: a 100n cap from +V to -V and removed the ones from gnd to +V and gnd to -V, and removed the op-amp output compensation cap.
Not that much left to do in fact - supply a timing reference and see if it will phase lock to it, and if not, tweak the parameters of the lead/lag compensator until it does.
I measure a fair bit of jitter in the tacho signal that is not noise induced by anything I've created. It is inherent in the tacho itself - I suspect the mechanical accuracy with which the pole-piece teeth have been stamped. (I should at some stage log the signal & see if it is cyclical)
I have no idea if my motor is a bad example or if that is what all the motors are like. It will be greatly swamped by the platter inertia, and it does make it unecessary to bother creating a really fast and accurate PLL. That's a 'relief' but also an annoyance from a theoretical point. Certainly there is now the option of a 'tweak' by building/adding ones own highly accurate tacho.
Not that much left to do in fact - supply a timing reference and see if it will phase lock to it, and if not, tweak the parameters of the lead/lag compensator until it does.
I measure a fair bit of jitter in the tacho signal that is not noise induced by anything I've created. It is inherent in the tacho itself - I suspect the mechanical accuracy with which the pole-piece teeth have been stamped. (I should at some stage log the signal & see if it is cyclical)
I have no idea if my motor is a bad example or if that is what all the motors are like. It will be greatly swamped by the platter inertia, and it does make it unecessary to bother creating a really fast and accurate PLL. That's a 'relief' but also an annoyance from a theoretical point. Certainly there is now the option of a 'tweak' by building/adding ones own highly accurate tacho.
When I was looking at this problem I came across a fascinating method of attacking jitter.
Basically the circuit converts the pulse train to a sawtooth wave using a charge pump and a capacitor with CCS drain. It then level shifts this so average DC = ground.
Because the charge pump gives a constant relative height to each rising edge and the CCS gives a constant downward slope, a line drawn through the waveform will cross the downslopes at a constant interval. Thus a a pulse generator triggered by a comparator which is itself triggered by the downslope level crossing will generate a pulse train with vastly reduced jitter.
The circuit is due to MJ Underhill and has been patented, licensed to Toric and seems to have been adopted by Fujitsu: FUJITSU EMEA.
The original article is in IEEE proceedings 2004, I had a copy somewhere but I can't seem to find it.
Basically the circuit converts the pulse train to a sawtooth wave using a charge pump and a capacitor with CCS drain. It then level shifts this so average DC = ground.
Because the charge pump gives a constant relative height to each rising edge and the CCS gives a constant downward slope, a line drawn through the waveform will cross the downslopes at a constant interval. Thus a a pulse generator triggered by a comparator which is itself triggered by the downslope level crossing will generate a pulse train with vastly reduced jitter.
The circuit is due to MJ Underhill and has been patented, licensed to Toric and seems to have been adopted by Fujitsu: FUJITSU EMEA.
The original article is in IEEE proceedings 2004, I had a copy somewhere but I can't seem to find it.
An interesting method! However, removing this 'jitter' is in fact introducing an error. Because the jitter is a result of non-uniformity in the tacho itself, if the jitter is removed, that really means the servo ceases to follow what the tacho is telling it to do. With the limited torque of the motor and the high inertia platter, the servo could never accurately follow these small & fast 'wobbles' in the tacho - which is the whole point of the heavy platter.
At preset, they aren't causing me any troubles - other quirks of the loop filter are giving bigger errors - which will hopefully soon be eliminated.
At preset, they aren't causing me any troubles - other quirks of the loop filter are giving bigger errors - which will hopefully soon be eliminated.
I disagree that this introduces an error, conceptually it is correcting the misplacement of the timing pulses. Of course it does this by removing the difference between each pulse interval and the average pulse interval but what I find fascinating is that it appears to do this without introducing a lag and thus slowing the loop.
The dynamic models I made of various TT configurations indicated that overall loop speed is quite important for optimal DD control.
The dynamic models I made of various TT configurations indicated that overall loop speed is quite important for optimal DD control.
You are both correct, of course; and there are different ways to correct the problem. The bottom line is how much of the instantaneous speed monitoring is lost, and whether it is significant.
I have the servo system now working, but it is slow and tends to hunt - not achieving a fixed steady-state phase lag as desired. I.e., non-optimal loop dynamics, but not too surprising since I didn't pay great attention to the values when choosing component values. I now have to go back to the maths & see what sort of numbers come out for ideal values.
Monday June 28th, 03h40am...
Done it! Stable quartz lock!! Let's put gin in our tea!
SP10 quartz PLL drive
Done it! Stable quartz lock!! Let's put gin in our tea!
SP10 quartz PLL drive
That sounds a great way to get UN-synched!!
Seriously though congratulations on this work.
Last edited: