Transcendent Sound Forum
Latest Topics
 
 
 


Reply
  Author   Comment   Page 5 of 6     «   Prev   2   3   4   5   6   Next
Kelvin Tyler
Reply with quote  #61 
W. Nice video. I never doubted the logical sequence of your circuit. Just speculating about transient noise origin. Interesting to see if switching disturbances persist when triggering is done properly with VU meter output.
K.
Kelvin Tyler
Reply with quote  #62 
W. Yes, the k4306 VU meter looks to be exactly what you need. Basically a precision half-wave rectifier feeding a number of comparators each with its own precise reference voltage. Flexible input sensitivities. Very modestly priced here. Ideal. [smile]
K.
Wolfgang
Reply with quote  #63 
Kelvin, I also think this VU has everything I need and should be precise enough. I hope the switching transients come from the different turn ON/OFF times of the SSRs I have used(Omron g3vm_61ar, IXYS LCA715,Toshiba TLP241A). That would be an easy fix.

Wolfgang
Kelvin Tyler
Reply with quote  #64 
A back-of-envelope calculation, making allowance for the variation in gain with output impedance, points to some amelioration of the power drop when switching in the A-T.
If a 300B SE OTL amp is nearing the limit of 0.5W with a 4 ohm load, I estimate that switching to a load of 8 ohms at constant input level  incurs a drop to 0.32W or 1.9 dB. Changing from 8 to 16 ohms would give a further drop to 0.18W or 2.5 dB.
If the gain is considered to be constant then the power drop would be 3dB in each case, so the above suggests that the 'switch change' could be less noticeable.
Kelvin
Wolfgang
Reply with quote  #65 

Kelvin,

I am a bit frustrated because using the output of the LM3916 as trigger doesn’t work as I thought it would. The voltages are not really useable when operated with music. With levels set by sine generator/sine wave it would give me precise voltages but with music it’s just a mess of unusable up and down (dot and bar configuration alike).  It would be a lot easier to simply use the DC signal at D1/D3. Any idea what could solve this problem or how to connect the outputs of the LM3916 to the 4093  for using them as threshold levels? I use one power supply with 12V/5V regulators, common ground.

On the topic of power drop/switching the AT I was just thinking along similar lines. But my idea was to rather use loudness meters in my music editing program for figuring out the relative level increase (after the power drop caused by impedance switching). After all it's about subjectively perceived increase in loudness (undistorted peak at max possible output of the amp minus the power drops you have calculated) which is possible if the amp can drive the peaks and doesn't compress/distort anymore and the average loudness can be set to a higher level because of that.

 

Wolfgang

Attached Images
Click image for larger version - Name: VU meter circuit.png, Views: 8, Size: 119.39 KB 

Kelvin Tyler
Reply with quote  #66 
Wolfgang,
Sorry to hear of your problems with the LM3916 based Velleman kit. I do not have access to this kit or an LM3916 so what I say now is just off the top of my head.
I do not know how you have taken the output from this circuit, but I would think it is imperative to use the bar mode setting. Then I would choose one led connection, say 0 dB at pin 10 on IC3 and IC4, disconnect their +ve feeds and reconnect these leds with their own 470 ohm pull-up resistors to +V. To my mind the outputs taken from pin 10 should then not 'wobble'. The common pull-ups,R21 and R22, are the problem here I think. You will need to measure the voltage change at pin 10; I imagine it will probably be +12 to 0V.
I note that the rectifier time constant is about 0.35sec. You will obviously have to adjust RV1 and RV2.
Hope this is not all wrong. It is 7am here!
Good luck.
Kelvin
Wolfgang
Reply with quote  #67 
Thank you, Kelvin. I will try a few things today including what you have proposed. The adjustable "hold" function of the more expensive VU meter would come in handy now.

Wolfgang
Wolfgang
Reply with quote  #68 
I tested a few things and finished one channel. Using DC for triggering  from the VU meter doesn't work. Real AC peaks from the music are necessary in order to trigger the CT accurately and fast enough exactly when the peaks happen. It works better using the output of the amp together with a pot for fine tuning instead of using the line signal from the pre.

There is a difference between 3 and 5 switches. The switching transients sound different (brighter and sharper with 3, darker and softer and less annoying with 5 switches). The sound also seems kind of a little  less "veiled" in direct mode  using 5 switches.

The truly amazing part is that even in my main system the switching at peaks is not noticeable (with 3 switches it is noticeable) regarding loudness changes of any kind and does exactly this: making the peak distortion completely disappear. This part works absolutely seamlessly.

However, one problem still persists: there are these switching transients like little vinyl crackle. It must come from the AT switching as they changed  their sound character when going from 3 to 5 switches. I cannot explain why. I also cannot explain why these switching transients weren't audible with regular relays and slower switching by hand. Could it come from using the ATs with  SSRs?

Of course I could  also switch the ground connection of the AT and try it with 6 switches as a last effort (illogical though) to get rid of the switching transients.

Any ideas what could be the cause of these switching transients and how to get rid of them?
Kelvin Tyler
Reply with quote  #69 
You have certainly made some interesting observations, Wolfgang. Some of these, it seems to me, are entering into the Donald Rumsfeld area of 'unknown unknowns'. [smile]
I guess that the Velleman VU meter response is just too slow for this application with the long time constant at the rectifier.
I must say I had assumed that the symmetrical Zero A-T winding pattern was intended just to accommodate Circlotron balanced-output type amplifiers. I would never have thought of using an A-T like your 5-switch implementation with an unbalanced, SE output stage. Must be missing something.
Your 5-switch solution presumably leaves two SSRs in circuit for 'direct' OTL; the 3-switch solution only leaves one. As I observe no audible effect when connecting any single lead of the A-T to the ground speaker lead, either at the speaker or the amp end, we have to conclude that the 'veiling' effect you hear with the 3-switch 'direct' setup arises from the absence of the extra SSR. Very weird.
With regard to transient switching effects. I would have to implement your scheme to usefully comment. Always held some reservations here. I suppose you might try different SSR types, if you have not already done so. On the plus side you have at least proved the increased 'headroom' concept to be valid.
On a lighter note I attach a couple of pictures of a prototype of my 'speaker protector' which I made yesterday evening sitting in an armchair! I actually used a couple of 74HCT132 Schmitt nand chips and the LCA715 MOSFET switch . It works rather well. I have set the trigger points at + or - 26V.
IMG_2490.jpg IMG_2491.jpg 
Kelvin

Wolfgang
Reply with quote  #70 
Kelvin,

yes, the "unknown unknowns" which make this case even more interesting and challenging.
The "veiling" effect comes in my opinion from not disconnecting the AT fully as it is the case in the 3 switch version. Also don't forget the audible loudness change when switching with the 3 switch version. No change in the 5 switch version. Both versions double the impedance at least in the same way.
I guess I will have to try a few more things.

Wolfgang


Kelvin Tyler
Reply with quote  #71 
Wolfgang. Not a big deal, but if I understand you correctly, you seem to be saying that the A-T in the 3-switch setup for direct OTL is somehow less disconnected than in the 5-switch case. In what way is this so? Am I missing something here?
Kelvin
Wolfgang
Reply with quote  #72 

Kelvin, if you look at your drawing in post #51 you can see that if the AT is ON, the ground connection is different compared to 5 switches. That must be the reason why there is an audible loudness difference when switching with 3 switches.  This is not the case with the 5 switch version. That's simply a fact and you can try it yourself if you don't believe me. I also checked the impedance curve while the circuit was actively switching because I was surprised myself. With PN you get an instantaneous readout on the screen and it goes from normal to double impedance back and forth with each switching. So the CL and the AT work as intended.

If the AT is OFF there is no difference between the 3 and the 5 switch version because in both cases all connections to the AT "positive" are interrupted but it still sounds less veiled with OTL direct and using the 5 switch version. That's what I can hear despite all logic that tells me (and you) it shouldn't be the case. But I think that’s the least of the problems.

It looks like the cause for the "switching transient" is nothing more than a very short  interruption of the signal as the switches don't connect/disconnect exactly at the same time or in a way so that the connection  is not broken for some ms. Same thing as if one would use a "break before make" rotary switch for a stepped attenuator. Is there an easy fix for this? At least it has nothing to do with amp or AT reactance.

Ideally all SSRs would have a few ms delay before switching OFF and would  turn ON without delay.

Wolfgang

Attached Images
Click image for larger version - Name: 100us_div sine.png, Views: 4, Size: 24.00 KB  Click image for larger version - Name: 200us_div 1kHz sine.png, Views: 5, Size: 25.66 KB 

Kelvin Tyler
Reply with quote  #73 
Wolfgang. Indeed it is clearly better to use the 5-switch solution given that you need a 2:1 impedance ratio and you are then using the whole number of turns. Apart from  the turns ratio, the 3-switch connection scheme will have different inductance and DC resistance characteristics. I did mention  the volume drop in post No.51. Pity there is not a tap that allows the 3-switch method with a 2:1 impedance ratio. The closest I can see, using the whole winding, is 1.78:1.
I do not doubt your observations in the slightest. I was just surprised by the veiling effect you found with the 3-switch, direct OTL situation.
One last thought. I take it there is no question that the CD4093 can fully activate the led in the SSR via a 2K resistor and a 5V supply. With the 74HCT132 I used 1K with the LCA715.
Kelvin
Achim
Reply with quote  #74 
Wolfgang, do you maybe need to get a multiple SSR, or do you need to calibrate the sensitivity of each single SSR?
Kelvin Tyler
Reply with quote  #75 
Achim and Wolfgang,

From the IXYS data sheet for the LCA715 device The typical switch-on time and switch-off times are given as 0.7 and 0.115 ms respectively. The maximum values are 2.5 and 0.25 ms in each case. As the A-T switching involves both turning on and turning off for either direction of switching, it would appear to be impossible to connect or disconnect all connections at precisely the same time. Faster acting devices would be required to reduce this effect.

Kelvin

Previous Topic | Next Topic
Print
Reply

Quick Navigation:

Easily create a Forum Website with Website Toolbox.