LCR riaa's

the road not taken.
Andrew
Posts: 201
Joined: Wed Apr 26, 2006 8:06 am

Post by Andrew »

Well, about time for another update.

I managed to create a test rig to check my implemented RIAA curve against the "perfect-world" Spice design. I used the export function in Spice to output the curve generated by an AC sim and then imported that data into Excel.

Using a sig gen on log sweep mode and a DVM to capture the results I then measured the implemented unit, just one side for the moment. The data was captured over the RS232 port of the DVM using hyperterm and cut'n'pasted into Excel.

Doesn't look to bad. The slight variation could be my crap experimental skills. It looks a bit up in the mid and treble to me. I'll have another bash later, now I that the rig is running and I know its not way out. Any other ideas/thoughts?

PS. I would never make a metrologist.

cheers,

-- Andrew
Attachments
riaa.jpg
riaa.jpg (79.44 KiB) Viewed 28577 times
dave slagle
Posts: 2085
Joined: Sat Jan 22, 2005 3:54 am
Location: NYC
Contact:

Post by dave slagle »

looks like the gain is off a bit to me... can you attach the whole file???

dave
Andrew
Posts: 201
Joined: Wed Apr 26, 2006 8:06 am

Post by Andrew »

Hi Dave,

Yes, I agree, off a bit.

However, after I'd posted it occurred to me that I have no way, at the moment, of knowing what points the sig gen has chosen as its sweep. I'm not sure...are they close, or the same, as Spice uses for the same start and end freq? Also, I may be better doing the sweep manually,

On this run the sig gen sweep had 141 points and Spice 140 so that would throw things out. I fudged a 141st result from Spice so I do need to go back over things and make sure they're set up the same.

Here's the file, but please bear in mind its only a first pass. I was pleased it was even close as you know this is a bit more complex than a typical three resistors and two caps RIAA.

How did you do this?

cheers,

-- Andrew
Attachments
LCR_results27June.xls
(54 KiB) Downloaded 970 times
Andrew
Posts: 201
Joined: Wed Apr 26, 2006 8:06 am

Post by Andrew »

Hey Dave & co,

How do you get Spice to label the graph?

I added a param list to cycle through some values and end up with several lines, but don't know which one is which :oops:

cheers,

-- Andrew
sbench
Posts: 296
Joined: Fri Jan 28, 2005 3:45 pm

Post by sbench »

In tools menu, there is a color preferences tab. Go to the WAVEFORM setting and the order is displayed (and can be edited). This gets a little tricky when you have multiple waveforms present. When you're stepping waveforms, the color displayed for the selected parameter (for instance V(output), or V(n0001)) will be the FIRST step. The colors then go in LTSpices order. For instance, let's assume that color was light green. Then the next step would be light plue, then light red etc. You can edit this and change it to your preference so, for instance, you could set up the colors like the resistor color code (brown=1, red=2 etc) and that would help you see which plot is which step.

-Steve
Andrew
Posts: 201
Joined: Wed Apr 26, 2006 8:06 am

Post by Andrew »

Hi Steve,

Thanks, I'll look into that, I want to plot my ideal values agains the real values of L and C I have! The R should be pretty good as they are 0.1% Holcos

Hey Dave,

I was right, Spice and the sig gen were using different sweep points hence it looked low on gain, the whole graph was moved to the left, rather than down, but it had the same effect.

New version using the same plot points is attached, much better. It looks pretty consistent until 5k, about 50mdB up all along, then the treble starts to go up, or rather not attenuate enough, peaking at about 1/4 db up. What do you think?

cheers,

I'll look later and report back on what values I have set for L and C.

-- Andrew
Attachments
LCR_results28June_test.xls
(40 KiB) Downloaded 717 times
dave slagle
Posts: 2085
Joined: Sat Jan 22, 2005 3:54 am
Location: NYC
Contact:

Post by dave slagle »

i'll peek at the new plots later, but try to tweak the spice model to match the actual model to match the two and then you can do the inverse to your network to match.

dave
Andrew
Posts: 201
Joined: Wed Apr 26, 2006 8:06 am

Post by Andrew »

The RIAA modules are still not tested but I'm developing some VB scripts using Excel to drive some test gear at work. Having done one module manually made me realize some automation was in order. The scripts also follow Nick G's suggestion of testing not only freq resp against ideal but also at varying amplitudes, which is the hard bit to get right for the chokes.

Can I resist not hooking it up later this weekend and seeing what happens? Well I'm still a bit uncomfortable with the solid state supplies, I need some bigger wattage resistors on the HT protection circuit. I'll see what time I get spare from DIY duty on the bathroom and time with the kids etc.

Onwards...slowly onwards...

cheers,

-- Andrew
Andrew
Posts: 201
Joined: Wed Apr 26, 2006 8:06 am

Post by Andrew »

VB scripts done and ready to roll.

I'm using a sig gen to create a sine wave that is read by a DMM with true RMS mode. The DMM is run in db mode with 1Khz set an the NULL point. The instruments are driven over GPIB using Excel's VB macros with embedded SCPI commands, roll on LXI.

Anyway, I should be able to test the units using a specified freq sweep that matches the Spice ouput of Dave and my design/spec/model for 1k5 RIAA.

I have also programmed things so they are run at varying amplitudes so hopefully this should be a pretty comprehensive test of the completed modules with the chokes in situ.

Hoping to post results soon.

cheers,

-- Andrew
Andrew
Posts: 201
Joined: Wed Apr 26, 2006 8:06 am

Post by Andrew »

Here's my 1k5 measuremnets, these are raw out of the can au naturel before I start tweaking.

The first image is the control, the important bit to note is the measurement error at low amplitude and high freq. The measuremnet kit can't quite resolve down at 30mV in -> 0.3mV out, this accounts for the slight drop in attenuation seen in the two measurements at high freq low amplitude.

cheers,

-- Andrew
Attachments
A channel.
A channel.
riaa_29_Aug.2_6882_image001.gif (21.51 KiB) Viewed 27979 times
B channel.
B channel.
riaa_30_Aug.3_13235_image001.gif (21.56 KiB) Viewed 27979 times
This is run without the RIAA but with 40db loss, the control reference.
This is run without the RIAA but with 40db loss, the control reference.
correction.0_10822_image001.gif (26.12 KiB) Viewed 27979 times
Andrew
Posts: 201
Joined: Wed Apr 26, 2006 8:06 am

Post by Andrew »

By the way Dave, I tried really hard to break these, I really did, but respect to you, to the best of what I can measure, at present, they really are OK, no actually they are very very good. Now, I know my 50nF caps are off on my LCR meter so that probably accounts for some of the deviation. But I'd need to get better kit, or better understanding, to improve on +- 0.1 db, which is about where I seem to be at.

Nick is sending me his S&B 600R to measure, it will be an interesting comparison.

Its all about to go together now, this is where I get it all wrong.......

cheers,

-- Andrew
Andrew
Posts: 201
Joined: Wed Apr 26, 2006 8:06 am

Post by Andrew »

Hey Dave,

Just wondering to myself and thought well I might as well ask. If I use a E810F -> 1k5 -> D3A (triode) I have about 10db gain over the magic 40db, what's the options for a step down, say 3:1, to load the D3A, block the DC, and reduce Zout? Just thinking.........

cheers,

-- Andrew
reVintage
Posts: 365
Joined: Fri Sep 07, 2007 9:54 am
Location: Sweden

Post by reVintage »

Hi Andrew, check my solution in the 1200ohm.... thread. I really believe in an output tranny. Why not go for 4:1(12dB loss) to get a even lower Zout of 100-250ohm+RDC of the secondary.
Brgds
Lars
Andrew
Posts: 201
Joined: Wed Apr 26, 2006 8:06 am

Post by Andrew »

Hi Lars and Dave,

Well, so I've had the thing running and then taken it apart to re-build the PSU and sort a serious hum problem, I'm in the process of putting it all back together and when I'm done........well I wanted to talk about a next step.

Lars you inspired me to think about transformer coupling, and what you call the 1200R but with a twist :wink:

Actually, I was thinking of transformer coupling between the input stage and the LCR. So, a high gm low noise pentode input, with a higher load than before, say 5k6, this goes into a cathode follower where the leg of the follower contains a 1:1 'interstage'. The cathode follower would be 5842 (or some such) to drive into the LCR. If I read Dave correctly a 1:1 bifilar wound interstage that DC blocked would be possible. Dave please feel free to chip in here, if I'm off by miles. I think you said about 1k DCR so that the Zout of the 5842 plus the 1k DCR and some added resistance to make 1k5 would give me a matched impedance to the 1k5 LCR.

OK so shoot me down....what say?

cheers,

-- Andrew
Attachments
Sort of a bit like this.....
Sort of a bit like this.....
untitled.JPG (23.61 KiB) Viewed 27154 times
dave slagle
Posts: 2085
Joined: Sat Jan 22, 2005 3:54 am
Location: NYC
Contact:

Post by dave slagle »

hey andrew,

a 1:1 bifilar would work great here, but wouldn't the LCR's reflected 1K5 in series with that 8K2 form a voltage divider tossing away bout 15 dB of gain?

dave
Andrew
Posts: 201
Joined: Wed Apr 26, 2006 8:06 am

Post by Andrew »

Hey Dave,

Could be?

I wondered if it all wouldn't be dominated by the 1k5 load to ground of the LCR?

I'll do a sim or two? A before and an after. :D

What's the likely bandwidth of a bifilar and what's the likely inductance?

thanks,

-- Andrew
dave slagle
Posts: 2085
Joined: Sat Jan 22, 2005 3:54 am
Location: NYC
Contact:

Post by dave slagle »

the bandwidth would be pretty high , but due to the 1K5 load the copper losses (DCR) would need to be kept down in order to keep the shift from magnetic to capacitive coupling in check. I would guess 60hy @ 20ma would be in the realm of reason.

dave
Andrew
Posts: 201
Joined: Wed Apr 26, 2006 8:06 am

Post by Andrew »

Any ideas on a realistic value for the DCR?

It seems that, as far as the sim/maths is concerned, the DCR must be in the region of 800R-900R for optimal power transfer and balanced freq response; that's assuming we terminate the LCR with 1Meg. Does that make sense?

That's with a 6k8 cathode and 14mA on the 5842, I suspect I need the DCR in series with the load resistor (6k8) then in parallel to 1700R (Ra of the 5842) or something like that....to match the Z on the network.

cheers,

-- Andrew
dave slagle
Posts: 2085
Joined: Sat Jan 22, 2005 3:54 am
Location: NYC
Contact:

Post by dave slagle »

Andrew wrote: It seems that, as far as the sim/maths is concerned, the DCR must be in the region of 800R-900R for optimal power transfer and balanced freq response; that's assuming we terminate the LCR with 1Meg. Does that make sense?
yes and no... as lars pointed out, the LCR will still provide a 1K5 load and if you place that load across the secondary of a bifilar you will draw current causing a voltage drop based on the DCR. The capacitively coupled nature of the bifilar doesn't have those losses so you see an increase in response at some frequency.

here is some more info on previous experiments

That's with a 6k8 cathode and 14mA on the 5842, I suspect I need the DCR in series with the load resistor (6k8) then in parallel to 1700R (Ra of the 5842) or something like that....to match the Z on the network.
now you are really losing me. the previous sketch had the transformer in the cathode of the 5842.

dave[/url]
Andrew
Posts: 201
Joined: Wed Apr 26, 2006 8:06 am

Post by Andrew »

Hi Dave thanks for the info, lots to digest there...

As I understand it, the 5842 is acting as a cathode follower so we still have to consider the 8k2 as load resistor, as you said there was 1k5 plus the DCR is series with that and then there's the anode resistance of the 5842.

I'll be back........

thanks again,

-- Andrew
dave slagle
Posts: 2085
Joined: Sat Jan 22, 2005 3:54 am
Location: NYC
Contact:

Post by dave slagle »

I see where you are going..

ideally we would want the secondary to be open so we get the inductive cathode load. Since it is terminated by 1K5 we need more of a load to keep things happy hence the additional series resistor. I am 99% sure we can consider the reflected load from the 1K5 as resistive so we have 1K5 in series with the additional added resistive load and these two form a voltage divider with only the voltage across the 1K5 being sent on down the line.

I'm not sure where the anode impedance works into things... From the perspective of what is driving the 1K5 LCR wouldn't it be the DCR in series with the parallel combo of the 6K8 and the cathode impedance?

dave
Andrew
Posts: 201
Joined: Wed Apr 26, 2006 8:06 am

Post by Andrew »

Yes, seems were on the same wavelength...regarding the cathode resistor etc.

I'm intrigued by the thread you pointed me too, and the measurements, it seems the difficult load of the 1k5 could cause problems.

Do I read the thread correctly, was it remarked that the bifilar was less "ideal" into a difficult load? I noted 18k load was mentioned. What about a into a 1k5 load? My guess is that's worse....

I'm wondering, if I put the cap after the LCR, as I have done to date, then I only need a 47nF cap, and could even drop that to 33nF or even 22nF in the name of a "warp filter".

Is the issue here that a 1k5 load is too much when we consider that I'm not replacing some huge multi uF cap but rather a quite a small one?

Am I barking up the wrong tree?

All input greatly appreciated, this sim looks OK, but I suspect its not telling me much of the real story at all.

thanks,

-- Andrew
dave slagle
Posts: 2085
Joined: Sat Jan 22, 2005 3:54 am
Location: NYC
Contact:

Post by dave slagle »

hey,

a few things about the referenced thread. The inductance was really high (200hy @ 20ma) so the wire needed to get small and up shot the DCR. It is the DCR that becomes the issue when loaded since the current drawn by the load causes a voltage drop in the magnetically coupled circuit. At some point the capacitance shows a lower impedance path and the circuit essentially operates as a LCL (with the L's coupled!) and the coper losses go away.

IIRC the DCR of these coils was of the order of 2K8 for each coil and if the DCR's of the coils you want are in the 900 ohm range i would expect the same behavior. If we can keep the DCR down the 1500 ohm load shouldn't be to much of an issue (see the end of sheldone's 801 thread) Of course keeping the DCR low enough might prove problematic but we won't know until we try. Actually if the coil behavior is known th eLCR could be tweaked to correct :-)

WRT the caps after the LCR. This will put some DC current through the chokes which may or may not cause issues depending on how much current we are talking about.

dave
Andrew
Posts: 201
Joined: Wed Apr 26, 2006 8:06 am

Post by Andrew »


WRT the caps after the LCR. This will put some DC current through the chokes which may or may not cause issues depending on how much current we are talking about.

dave
About 2mA through the first choke and 0.5mA through the second, do you reckon this could be an issue?

What range would be a low DCR? I guess I'm asking how much lower than 900R?

I played with adding some 'parasitic' parallel capacitance to the inductors in the sim I'm working on, that makes a significant difference to the bandwidth behaviour.

Yes, good idea, its always possible to tweak the LCR.

I'll re-read the various threads, still much to learn.

cheers,

-- Andrew
dave slagle
Posts: 2085
Joined: Sat Jan 22, 2005 3:54 am
Location: NYC
Contact:

Post by dave slagle »

in order to see the behavior in spice you need to put a cap across the two windings.

dave
Andrew
Posts: 201
Joined: Wed Apr 26, 2006 8:06 am

Post by Andrew »

I'll try the cap...

Do you think I need to worry about the DC across the chokes?

cheers,

-- Andrew
dave slagle
Posts: 2085
Joined: Sat Jan 22, 2005 3:54 am
Location: NYC
Contact:

Post by dave slagle »

hey andrew,

typically for the designs you have the DC currents we are talking about should not be an issue. The easiest thing to do is to try it and measure to see how your inductance value changes. Just to be safe, I'd run the plots at two distinctly different levels of excitation which should show any saturation effects.

dave
Andrew
Posts: 201
Joined: Wed Apr 26, 2006 8:06 am

Post by Andrew »

Hi Dave,

I noticed Lars was able to work in extra caps to make sure the LCR was AC coupled to ground, so DC will only pass the chokes when the caps charge, might give this idea a go.

cheers,

-- Andrew
reVintage
Posts: 365
Joined: Fri Sep 07, 2007 9:54 am
Location: Sweden

Post by reVintage »

Hi guys,
Unfortunately I have not had the chance to try it yet. But will do.....
Brgds
Lars
dave slagle
Posts: 2085
Joined: Sat Jan 22, 2005 3:54 am
Location: NYC
Contact:

Post by dave slagle »

i'm not sure which version allowed lars to work in caps. iirc, lars was suggesting that the source would define the LCR value so the terminating resistor wasn't needed. It is that resistor that draws the bulk of the current through the LCR if you attempt to direct couple.

Lars also didn't use the parallel resistors to the shunt caps and again from memory the original LCR didn't have the resistors in parallel to the caps like the tango / pultec does.

dave
Post Reply