Castles_Saloon

to view the frequency responses of the tweet, front door mid, rear door mid and the rear deck... click the link below

http://picasaweb.google.com/castles....08629869328850

The curves are far from flat. I'm not sure whether this was intentional to account for the cabin response or if it just not was engineered for hi-fi. This could explain weak midrange and boomy bass. The xover point for the tweeter is a little high (3db ~5k).

Also, when viewing sine waves, the waveforms showed distortion with spikes at 100kHz and 625Khz. Although, these shouldn't be audible, they can distort or add incorrect energy to the sound you hear through the speakers. I think there is a moderately easy fix. I'll let you know if this is why the hi frequencies have sounded so harsh.

jlukja

^^ How did you measure those?

Castles_Saloon

have plenty of toys at work...
-12V power supply to the amp
-fluke 5500A calibrator as source (frequency generator)
-fluke 199 scope

after applying freqs to the various amp inputs, i recorded the amp output in Vac rms value on the scope. It doesn't provide a whole lot of resolution, but its plenty accurate for dB land.

Note: this data was measured with a hi z(impedance) load, 10Mohm so there may be some error, but curves should be close.

Castles_Saloon

i mentioned earlier the tweeter amp output waveform looked distorted with a lot of improper high freq components (100kHz & 625kHz). Disassembling the stock amp reveals a circuit looking like...

tweeter path
input->signal conditioning (xover)->class D amp IC (ST microelectronics) ->capacitor-> connector->speaker

if the capacitor is replaced with an LC network, the waveform cleans up significantly.
------inductor---------->connector
|
|
capacitor
|
|
ground

i'm still experimenting with values, but i think L=10uH and C=1uF might just do the trick without rolling off treble too much (~1dB at 20kHz)

st class d amp (this link may not be the exact one, but very close)
http://www.st.com/stonline/products/...e/ds/12882.pdf

Castles_Saloon

my LC in the previous post drawing failed...

L should replace the cap already there. the new cap should tap the point where the connector and the L meet and the other terminal should go to ground. Sorry about that.

jlukja

Can you tell the value of the internal capacitor?

I ask because I replaced my tweeter with the avincar tweeter which called for an additional 4.7uF capacitor in series with the speaker. Just wondering what the resultant x-over frequency would be.

Castles_Saloon

2uF 50V elna audio grade cap in series with the tweet. Its basically just a dc blocking cap to save the tweet from low freqs. you can open up the amp and find it on the bottom pcb easily (in the middle towards the socket). Remember you cannot add caps in series to increase the value... paralleling adds. unfortunately you'll need to open up the amp to change it.

if you replace it the 2u with a 4u, you may slightly lower the 3db of your system and get a little more midrange with you tweet, but probably not night and day difference.

The main xover is before the IC amp so the 2uF has a smaller relative affect.

mercman

The output device is a pioneer PAL006A that I’m 99% sure crosses over to a TDA7382 four channel class A/B (analog, not class D) amp. It is rated at 18.5 watts at 1Khz with 10 % THD (not a real usable number).

All the crossover (EQ) takes place before the audio output amps, so placing caps in line with the tweeter will not change the xover point. It will change the tweeter response somewhat by creating two or more “poles”. How good or bad that would sound depends on a lot of factors. The crossovers are probably active (op amp), based on the amount of op amps on the daughter card. It would be hard to mod them with out a schematic.

The poor response (and hi freq noise) is likely because of the 10m ohm impedance (lack of a proper load) on the amp. The amp is designed for 4 ohms. Try retesting with a 4 ohm non inductive load on each channel. Also don’t forget that the inputs are 600 ohm balanced. If your test generator is unbalance this could also explain the hi freq junk.

jeff

vidgamer

I think the idea behind the additional cap with the Avincar tweets was to raise the xover value. If it creates "two poles" where the end result is that further low frequencies are cut, then I think that was the goal as I understand it.

18.5@10% distortion? Like when I crank it with bass heavy music? Very usable, at least for bass.

The high frequencies were a bit too bright for me, so I made an L-pad circuit, and wired that in right before the tweeter, so who knows what that does when combined with the rest of it.

Castles_Saloon

mercman is right about the amp ic being a Pioneer Pal006A, but i'm not quite sure what ST part pioneer chose to put its label on. The reason i thought it was a class d part is due to the high frequency components within the waveform. High f components usually indicate high frequency switching which is what class d's do. after the high speed switching, they filter the snot out of it to come up with the audio band instead of an ugly pwm (varying widthed pulses) waveform. unfortunately, this implementation doesn't use adequate filtering.

if it is the class a/b part, then load instability could make sense to create the high f noise. However later tests, with a 10 Mohm and a 10ohm resistor exhibited the same waveform (only tested the filtered tweeter path). i'll check out other paths with different loads later.

The tweeter frequency response was consistent with a 10Mohm and a 10 ohm load. I'll confirm the mids and rear maintain the same response as i posted earlier.

the test setup used earlier to measure freq response did not use a balanced signal source. Although this can cause pickup of noise, I don't think this was the noise i saw on the output. the input waveform was quite clean. I'll confirm circuit stages prior to the amp IC were ok too.

BTW, the LC circuit leading to the tweeter that i added, cleaned up a lot of distortion in the waveform. listening tests indicate a noticeable decrease in treble output, but clean. not sure if the overall f-response is well balanced between the drivers.
I think part of the harsh sound is the tweeter response itself. another theory to explain the perception of tweeter harshness is the imbalance of mids to highs (mids<high in dB). With mids mounted by your feet, there is a lot of sound absorbed by your legs killing the midrange and low treble. a tweeter with a really low xover point might be better suited for this type of speaker layout.
enough ranting

Castles_Saloon

changed the amp loading to 8ohms instead of 10M and the frequency plots are the same as before.

the waveforms look a whole better after the LC networks are in place

Now lets see if i can flatten out the midrange f-response...

junktionfet

You know, it really does sound like a switching amplifier, but I formerly attributed that to peaky tweeter response and perhaps some cheap electrolytic coupling caps... I'd be curious to open it up and see exactly what's going on. I've never seen the amp module... does the heat-sinking indicate a class AB or class D design? What about quiescent current/dissipation?

vidgamer

Castles: I'm not following the LC network bit. What is it for?

I left the stock cap alone (not wanting to mess with the stock system any more than I have to), but, like Jlukja mentioned, added a capacitor to go along with the different tweeter. Then, to dampen the overall volume of the tweet (which seemed a tad on the harsh side), I built an l-pad using this design: http://www.lalena.com/Audio/Calculator/LPad/. (I think I used 4ohms and 2.5 db damping to obtain the resistor values from this calculator.)

Do you have any comments on this approach? Do you have any other suggestions if one leaves the stock capacitor in place?

junktionfet

Think of the LC network as a higher-order passive crossover. It's purpose is to provide a sharper rolloff at whatever frequency it is tuned for.

He said he noticed a lot of high frequency garbage (not unlike that seen from a switiching amplifier design). The use of this type of filter will reduce that.

If you wanted to discriminate even more without affecting the audible frequency response, you could construct an even higher order network (using more LC pairs), but that generally produces a more reactive load that might upset the stability of the amp.

Does the stock amp have some kind of zobel network (snubber) after the output? If not, that may explain the garbage you see in the output. Even if it does have such a network, I wonder if it is tuned correctly or departs from the manufacturer's application circuit.

Castles_Saloon

To answer a few ?'s

They are using decent audio grade Elna caps... i think this is the one.
http://www.elna.co.jp/en/ct/pdf/rbde.pdf

Both topologies need heat sinking, but class D should need less due to the efficiency gain. The quiescent current was about 0.43A (unloaded). I guess if your batteries weak, the radio off


For post amplified eq adjustments an L-pad is a decent way to go, easily customizing your sound. The quality of resistive element in the L-pad can affect fidelity, but likely won't amount to much in an audible difference.

When you say "added a cap" is it in series or parallel with the speaker? If in series, the added cap, say 4uF, is in series with the 2uF in the stock amp. Capacitors in series work like resistors paralleled mathematically. Ctotal = (1/4uF+1/2uF)^-1 = 1.33uF. If in parallel, then the new capacitor creates an ac voltage divider with the stock amp's 2uF (reducing signal to 1/3 of original)

I doubt the added cap is needed since there's already 2uF in the stock amp.
The stock amp has nothing between the IC amps and the speaker except for 2uF in the tweets. The manufacturer's datasheet does not show a snubber (RC) or LC on the output, but their application notes do! The design of the stock amp either ignored these recommendations or removed the costly inductor and C in the design for good enough performance. I would have added the snubber, but got lazy considering 8 channels worth of work. I'll be doing a little amp design myself soon. nothing fancy, but a slightly better version of the stock amp.

vidgamer

Right, the cap is added in series, and is a value as recommended by Elduderino for the new tweeter, taking the stock cap into account. Why was this recommendation made? Difference in tweeter design? Matching the sound from the door speakers? Adding the cap should... um, lower or raise the cutoff? Must be lower. I bet this is to obtain more midrange, to raise the soundstage a bit. Hmm, that sounds right...

I got the parts from Madisound, IIRC, so the quality should be OK.

Castles_Saloon

I'm not sure if the stock tweet cap was taken into account. If so then it reduces the signal level just a tad (maybe not that noticeable). As Cap value goes down, impedance goes up creating greater signal drop across the cap than originally. Also the 3dB shifts up in frequency. This could be intentional to eq the tweets closer to the mids, but I don't know for sure. It may be just a generic precaution to make sure low frequency signals are reduced before they reach the tweet to prevent damage.

You may get more mids through your tweet if you remove the added 4uF. I'd double check with Elduderino to see if the tweet is ok with just the stock 2uF in series.

jaspervic

Those are some interesting plots. I just ordered some Polk Audio MMC6500 component speakers for the fronts and was considering the best way to wire them up. I was originally going to stick with the existing wiring/amp for now, but the crossover on the tweeter seems pretty high. On the other hand, using the Polk's passive crossover would waste some of the watts available from the built-in amp.

Is there any way to determine the number of watts dedicated to each channel? I would assume that there are more watts going towards the 6x9's than the tweeters, for example.

BTW, I'm running CDT Audio CL-69 in the rear deck and probably going to disconnect the rear speakers next. What would be the easiest & least invasive way to disconnect them (I don't really want to cut them)? I tried removing them from the connector with no luck (I've gotten this to work on similar connectors. Am I going to have to remove the doors?

Castles_Saloon

the voltage signal output from the stock amp didn't vary significantly from channel to channel. The number of watts you get to your 6x9's will be a function of their impedance.

Assuming 10Vpeak output and a 4 ohm impedance, P=(10V)^2/(4ohm)=25Watts pk

Assuming 10Vpeak output and a 8 ohm impedance, P=(10V)^2/(4ohm)=12.5Watts pk

I've never disconnected the 6x9's, but i would have guessed the connectors should pop off. Make sure there isn't a hook that needs to be pressed down before tugging on the connector

jaspervic

Sorry, to clarify, I meant disconnecting the rear _door_ speakers. I tried removing the wires from the connector at the back of the amp to no avail.

Jeff_B

Take off the door panels and unplug them... it's really easy. There are write-ups in the FAQ showing how to take off the door panels.

Castles_Saloon

Removing wire harness from the amp...
Although i've scratched my blue connector up a bit, I use a large flat head screwdriver to press down the flexible hook and exert some force on the back ridge of the connector. My other hand moves the connector side to side. be careful to avoid letting the screwdriver head slipping. Its not perfect, but it may help.

feuss2

iTrader: (1)

Are you willing to e-mail the excel spreadsheet?

feuss2

iTrader: (1)

I found this on a Saab forum:

http://www.saabcentral.com/forums/sh...9&page=2&pp=35

Looks like the only amp IC's on here are 2 of these 4-channel amps. There are 2 caps on the board that could be used to change the tweeter x-over freq. One of the filtering caps could definitely come off and the thru-holes could be used to connect wires to a larger, external capacitor. Might be possible to insert some "bass-blocker" circuits at the preamp stage. If you've got a sub, it may be desirable to roll off freq's below 100~250Hz to some of those stock speakers. I bet they would sound alot better if they weren't trying to reproduce those lows!

Castles_Saloon

yes

feuss2, your idea to remove the large elna caps, short the through hole pads, and add an external cap for the tweets is very good. That would allow aftermarket tweets their recommended circuit.
Bass blocking - If you reference the amp IC datashet, a simple CR, highpass filter, circuit could be added to the two tweeter inputs channels. This is probably what you meant, but i wanted to clarify.

It would be handy if the active (opamp) based xover was traced out. Then we could tailor the circuit to whatever frequency response we'd like per channel. Otherwise low voltage passive components (resistors, capacitors, inductors) could be added to the AMP IC inputs to roll off the low freqs to door speakers.

I'm not convinced the Saab forums assumption of Pioneer's PAL006A = TDA7385 or TDA7386. They may be correct, which is what mercman was thinking too.... some sort of class AB. The reason i doubt its a class AB is all of the 100kHz and >600kHz components to the output waveform. This is usually indicative of class D, unless the class AB was unstable and oscillating. then we really have a problem.

feuss2

iTrader: (1)

actually, i was looking at the two 3000uF caps, thinking that they were used for power supply filtering and that one could be left in place and the other could be removed/replaced with leads going to an external capacitor. this might ease clipping some, and/or smooth out the bass. C237 and C238 set the freq for the right and left tweeters. this i found unusual, as most tweeters actually behave like capacitors in a circuit, if they were simply in parallel with the front door speakers, there would be frequencies at which their impedence would be less than the door speakers and they would reproduce more of the highs. instead, it appears they have a full range input, and have the lower frequencies rolled off with a series capacitor. seems like a waste of a channel. in the saab application, they had 2 of the same/similar IC's and only used 6 of the available 8 channels. if C237 and C238 were replaced with jumpers, a full range output would be available. for at least the front door speakers, i would think a high-pass filter (set somewhere between 100~250Hz) would make those speakers sound alot better. they just can't hit the lows anyway, so it may clean up the rest of the range of those are filtered out. I think this would be better than having a full-range preamp in, then using some passive components to filter after the fact. I would be pretty sure that these amps are not class D, as they can produce freqencies through the entire audio range. the few class D's that can do that would make the amp far more expensive...

Castles_Saloon

3mF would be the ps caps. What kind of external cap were you thinking of connecting to? Keep in mind, that large peak currents travel through both leads of the cap so any additional wire in series with the cap raises the impedance seen by the current forcing some of it to travel elsewhere. 6mF should be plenty to handle peak power needs unless its a really powerful amp.

Are c237 and c238 the 2uF caps? For higher freqs, I'd expect tweeters to appear inductive due to the voice coil. Although, there are some exotic tweets, like piezoelectrics which are basically a capacitor. Sorry if i'm confusing your words, but i think the tweet amp channel sees only high freqs at its input and outputs the same at a larger amplitude. The output cap is in series with the amp output affecting the freq response, but i think (i may be wrong here) it mainly blocks dc and low freq from accidentally reaching the tweet.

The door speakers have the their own amp channel which is roughly full range. Rear door speakers seem to have a high pass. If c237/8 are the 2uf caps, removing these will have no effect on the door speakers.

I think you are right though.... if the door speakers had a high pass ~80-100Hz that would be best. Playing freqs close to their resonant frequency is always a bad idea. This amp's xover does seem to have room for improvement.

Class D was my guess and seems to be the minority vote. Perhaps my testing was flawed introducing the high F distortion. My added LC did tame the problem though and the tweets sound much less harsh, but not perfect.