Dry farts, I am so sad...

By Eugeny_Brychkov

Paragon (1127)

Аватар пользователя Eugeny_Brychkov

25-02-2020, 16:23

Some time ago Meits coined this expression in relation to the PSG percussive sounds. I am currently performing analysis of the circuit related to the MSX audio, and the findings make me sad.
My conclusion is that MSX is not standardized in its audio part, each manufacturer tries making their own audio circuit in belief their circuit is better than others'.
First of all, the mixer.

Yamaha YIS503

CX5 music computer

Pioneer PX7

Yamaha YIS805

All of them are using classical passive additive mixer with "virtual ground". The bad is that there's still a crosstalk between the devices, and devices' output characteristics are not properly defined.
The only parameter I found in the book called "MSX Technical Data Book" which says

Quote:

49 | SOUNDIN | Sound input (-5dbm)

What this -5dBm is meant here? This article states that

Quote:

In audio and telephony, dBm is typically referenced relative to a 600-ohm impedance, ...

but I have never seen audio impedance being mentioned in MSX literature, and no even near value on the circuit diagrams.
According to the table in Wikipedia article, -5 dBm corresponds to 316 uW (0.000316 W). Well, I know that power equals current multiplied by the voltage.

Let's look at the mixer circuit for YIS805. Consider no crosstalk, and impedance equal to the series resistors given. PSG (SSGSND) outputs 0.5-1.5 V (1Vpp), IC353-1 is having gain of 1, and output of the opamp is the same Vpp. How many dBm is on opamp output pin 7? P=U*I=1*(1/3300)=0.000303 W (303 uW). Ok, within the table's value for -5 dBm.

Now let's look at the circuit of YIS503. Same 1Vpp fed into the opamp with gain of 4.6, giving output range of -8.46 to -2.8 V (5.6 Vpp). Calculating P=U*I=5.6*(5.6/10000)=0.003136 W (3136 uW!) which is 10x the mixer's input power requirement. Of course, when PSG is on volume 15 on all 3 channels it will kill all other devices on the line.

From my point of view, the following things must have been standardized:
1. mixer input voltage over specific input impedance. It is already defined by dBm in the MSX Technical Data Book, but lacks clarification and lacks additional definitions (like impedance);
2. no spectral analysis had been explicitly done, what filter must be used prior and after the mixer. I have done the simulation, and it is not strange for the coined "dry fart" term:

3. output impedance of the SOUNDIN must be given too so that developers of the add-on devices (cartridges) will be able to match the values properly;
4. How to connect PSG outputs: there're versions for just direct connection, through resistors (like for CX5), using more complex resistor networks to create pseudo-stereo;
5. and the worst... the coupling of the SOUNDIN lines. We have seen machines which just have these lines tied together for both slots, and the unprotected outputs of carts installed are potentially in dander of being fried.

In general I believe that each external cart must have had operational amplifier at its output to lower output impedance, having series capacitor (for DC decoupling) and resistor (for current limitation). Then, each cart must output specific defined level of the audio signal - defined either by dBm, or Vpp, or Vrms or whatever making clear message what this means and how to adhere to it.

Для того, чтобы оставить комментарий, необходимо регистрация или !login

By NYYRIKKI

Enlighted (5483)

Аватар пользователя NYYRIKKI

25-02-2020, 18:02

If you want some good examples of bad implementations, then Phillips NMS-8250 is known to have a very bad mixer that may ie. mute the device connected to it or cause some other very hard to predict effects... Also ie. sound of Spectravideo SVI-738 is known to be off by a mile... The PSG is so loud that it cuts off all the time and SOUNDIN is so quiet that you practically can't hear it at all... It also seems to be quite a well optimized to pick up all possible noises ie. from CPU and VDP. Original FM-PAC tried to solve the balance problem by having 3 step volume control switch, but it was also way too fine adjustment to get even close on certain machines.

By Wlcracks

Champion (411)

Аватар пользователя Wlcracks

25-02-2020, 18:08

If brands can save a penny they will. If the standard doesn't set the rule, the problems start. Maybe ASCII and Microsoft just forgot to write it down. GR8NET has decoupling of the internal psg with a switch. A cheap audiomixer with the internal and external sound (i made my own sum amplifier) will solve the problems. Cannot go back to change it. NJM4885d is a good audio opamp though...
Maybe Philips oversaw this problem already with the musicmodule and made dedicated outputs.

By sd_snatcher

Prophet (3246)

Аватар пользователя sd_snatcher

25-02-2020, 19:44

Eugeny_Brychkov wrote:

The only parameter I found in the book called "MSX Technical Data Book" which says

Quote:

49 | SOUNDIN | Sound input (-5dbm)

What this -5dBm is meant here? This article states that

Quote:

In audio and telephony, dBm is typically referenced relative to a 600-ohm impedance, ...

but I have never seen audio impedance being mentioned in MSX literature, and no even near value on the circuit diagrams.

Sadly, this is pretty common in Japanese documentation. They let a lot of information subjective, to be deduced by the reader.

And these days we have Internet and can quickly search for info. Imagine that in the 80s, where the guy would have to go to a library. No wonder so many MSX mixer circuits had design shortcuts. Probably the overloaded engineer had no time to keep searching for that obscure 600 Ohm impedance info. (*1)

BTW, I'll add that impedance info to the MSX-Wiki, since it's very important.

*1: This excludes the guy who designed the audio circuitry of the Philips NMS-8250/55/80 series. This one had absolutely no idea of what he was doing. LOL!
These are Sanyo machines, and the other models from that brand are just marginally better in this aspect. The audio circuitry of Sanyo MSX computers always varied from horrible to dysfunctional. (The other parts of the machines are not bad though).

By spacemoai1973

Master (165)

Аватар пользователя spacemoai1973

25-02-2020, 19:55

sd_snatcher wrote:

The audio circuitry of Sanyo MSX computers always varied from horrible to dysfunctional.

Don't generalize please. At least the Sanyo MSX2+ models do not have any audio problems with external carts.

By sd_snatcher

Prophet (3246)

Аватар пользователя sd_snatcher

25-02-2020, 20:11

spacemoai1973 wrote:
sd_snatcher wrote:

The audio circuitry of Sanyo MSX computers always varied from horrible to dysfunctional.

Don't generalize please. At least the Sanyo MSX2+ models do not have any audio problems with external carts.

Definitely they do! The plain-vanilla SCC just won't output any sound on them. It's unbelievable that they didn't test this. :)

The owners who complained had their machines "fixed" by the consumer service by just soldering resistors over the old ones of the mixer circuitry. A very QND job that solved only the problem for the SCC, but other cartridges still got muted.

I have two of such machines. One came with the QND fix, the other was still original.

Even with the QND fix, the mixer volumes are still totally wrong. A real fix is published at this page.

By spacemoai1973

Master (165)

Аватар пользователя spacemoai1973

26-02-2020, 02:01

A quick glance at that thread shows it's about FAKE SCC. I have no problems with Konami SCC games or SD-Snatcher Sound Cartridge on my Sanyo MSX2+. I suppose it's not impossible that my machine has an audio fix, but I bought it in 1990.

By sd_snatcher

Prophet (3246)

Аватар пользователя sd_snatcher

26-02-2020, 04:03

Probably. There were various recipes of fixes published in Japanese sites.

In my case, the one had the "fix" worked only with original SCCs. Other sound cartridges (no matter the chip) were a lottery.

The other that was still original played the SCC at absurdly low volume.

By Parn

Champion (487)

Аватар пользователя Parn

26-02-2020, 13:34

My view is that unfortunately audio on MSX tends to be an afterthought, even by some people who make or used to make sound extension cartridges. Which is a shame, considering the MSX's vocation for great music. I also think most people just don't really care about this at all.

I would love to have an "all-in-one" sound cartridge which had its own sound output, low levels of noise, shadowing of standard devices like PSG and FM, some extra chips like SCC+, SN76489 (for Coleco and SG-1000 emulation) and YM2151 (everything could be implemented in FPGA for good measure), plus good, controllable mixing between those devices (perhaps even with channel panning). I would buy this in a heartbeat, but I'm fairly sure no one really wants something like this.

By the way, which are considered the best sound mixer designs? I mean, which MSX models have the best ones?

By Eugeny_Brychkov

Paragon (1127)

Аватар пользователя Eugeny_Brychkov

26-02-2020, 14:30

Parn wrote:

I would love to have an "all-in-one" sound cartridge which had its own sound output, low levels of noise, shadowing of standard devices like PSG and FM, some extra chips like SCC+

GR8NET? Actually I have started to research the MSX mixers in hope to have an universal way to have GR8NET putting its audio onto the MSX machine mixer and having proper balance with its internal PSG (and possibly other devices installed in the system).
Now I think it is mission impossible, or, I can consider some "average" mixer and tune default GR8NET performance to it. I even thought to make "sound profiles" in it to have shortcuts to set for the specific machines.

Parn wrote:

By the way, which are considered the best sound mixer designs? I mean, which MSX models have the best ones?

I would expect Yamaha or Pioneer to have best mixer (as they were focusing on audio those days). IMHO the best from the worst I've seen is YIS805's one. The better mixer must have some more active components in it (signal buffers at least).

By Parn

Champion (487)

Аватар пользователя Parn

26-02-2020, 21:24

To be honest I still haven't tinkered with GR8NET as much as I wanted. It is truly a marvel.